Migrations
/
pytorch-image-models
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

📝 add hfdocs documentation

Browse Source
pull/1473/head
nateraw 2 years ago
parent 5dc4343308
commit e4c99d2bd6
76 changed files with 18398 additions and 0 deletions
Show Stats Download Patch File Download Diff File

16
.github/workflows/build_pr_documentation.yml vendored
Unescape View File

@ -0,0 +1,16 @@
name: Build PR Documentation
on:
pull_request:
concurrency:
group: ${{ github.workflow }}-${{ github.head_ref || github.run_id }}
cancel-in-progress: true
jobs:
build:
uses: huggingface/doc-builder/.github/workflows/build_pr_documentation.yml@main
with:
commit_sha: ${{ github.event.pull_request.head.sha }}
pr_number: ${{ github.event.number }}
package: timm

13
.github/workflows/delete_doc_comment.yml vendored
Unescape View File

@ -0,0 +1,13 @@
name: Delete dev documentation
on:
pull_request:
types: [ closed ]
jobs:
delete:
uses: huggingface/doc-builder/.github/workflows/delete_doc_comment.yml@main
with:
pr_number: ${{ github.event.number }}
package: timm

1
hfdocs/source/_conf.yml
Unescape View File

@ -0,0 +1 @@
default_branch_name = "master"

149
hfdocs/source/_toctree.yml
Unescape View File

@ -0,0 +1,149 @@
- sections:
- local: index
title: Pytorch Image Models (timm)
- local: models
title: Model Summaries
- local: results
title: Results
- local: scripts
title: Scripts
- local: training_hparam_examples
title: Training Examples
- local: feature_extraction
title: Feature Extraction
- local: changes
title: Recent Changes
- local: archived_changes
title: Archived Changes
- local: model_pages
title: Model Pages
isExpanded: false
sections:
- local: models/adversarial-inception-v3
title: Adversarial Inception v3
- local: models/advprop
title: AdvProp (EfficientNet)
- local: models/big-transfer
title: Big Transfer (BiT)
- local: models/csp-darknet
title: CSP-DarkNet
- local: models/csp-resnet
title: CSP-ResNet
- local: models/csp-resnext
title: CSP-ResNeXt
- local: models/densenet
title: DenseNet
- local: models/dla
title: Deep Layer Aggregation
- local: models/dpn
title: Dual Path Network (DPN)
- local: models/ecaresnet
title: ECA-ResNet
- local: models/efficientnet
title: EfficientNet
- local: models/efficientnet-pruned
title: EfficientNet (Knapsack Pruned)
- local: models/ensemble-adversarial
title: Ensemble Adversarial Inception ResNet v2
- local: models/ese-vovnet
title: ESE-VoVNet
- local: models/fbnet
title: FBNet
- local: models/gloun-inception-v3
title: (Gluon) Inception v3
- local: models/gloun-resnet
title: (Gluon) ResNet
- local: models/gloun-resnext
title: (Gluon) ResNeXt
- local: models/gloun-senet
title: (Gluon) SENet
- local: models/gloun-seresnext
title: (Gluon) SE-ResNeXt
- local: models/gloun-xception
title: (Gluon) Xception
- local: models/hrnet
title: HRNet
- local: models/ig-resnext
title: Instagram ResNeXt WSL
- local: models/inception-resnet-v2
title: Inception ResNet v2
- local: models/inception-v3
title: Inception v3
- local: models/inception-v4
title: Inception v4
- local: models/legacy-se-resnet
title: (Legacy) SE-ResNet
- local: models/legacy-se-resnext
title: (Legacy) SE-ResNeXt
- local: models/legacy-senet
title: (Legacy) SENet
- local: models/mixnet
title: MixNet
- local: models/mnasnet
title: MnasNet
- local: models/mobilenet-v2
title: MobileNet v2
- local: models/mobilenet-v3
title: MobileNet v3
- local: models/nasnet
title: NASNet
- local: models/noisy-student
title: Noisy Student (EfficientNet)
- local: models/pnasnet
title: PNASNet
- local: models/regnetx
title: RegNetX
- local: models/regnety
title: RegNetY
- local: models/res2net
title: Res2Net
- local: models/res2next
title: Res2NeXt
- local: models/resnest
title: ResNeSt
- local: models/resnet
title: ResNet
- local: models/resnet-d
title: ResNet-D
- local: models/resnext
title: ResNeXt
- local: models/rexnet
title: RexNet
- local: models/se-resnet
title: SE-ResNet
- local: models/selecsls
title: SelecSLS
- local: models/seresnext
title: SE-ResNeXt
- local: models/skresnet
title: SK-ResNet
- local: models/skresnext
title: SK-ResNeXt
- local: models/spnasnet
title: SPNASNet
- local: models/ssl-resnet
title: SSL ResNet
- local: models/swsl-resnet
title: SWSL ResNet
- local: models/swsl-resnext
title: SWSL ResNeXt
- local: models/tf-efficientnet
title: (Tensorflow) EfficientNet
- local: models/tf-efficientnet-condconv
title: (Tensorflow) EfficientNet CondConv
- local: models/tf-efficientnet-lite
title: (Tensorflow) EfficientNet Lite
- local: models/tf-inception-v3
title: (Tensorflow) Inception v3
- local: models/tf-mixnet
title: (Tensorflow) MixNet
- local: models/tf-mobilenet-v3
title: (Tensorflow) MobileNet v3
- local: models/tresnet
title: TResNet
- local: models/wide-resnet
title: Wide ResNet
- local: models/xception
title: Xception
title: Get started

418
hfdocs/source/archived_changes.mdx
Unescape View File

@ -0,0 +1,418 @@
# Archived Changes
### July 12, 2021
* Add XCiT models from [official facebook impl](https://github.com/facebookresearch/xcit). Contributed by [Alexander Soare](https://github.com/alexander-soare)
### July 5-9, 2021
* Add `efficientnetv2_rw_t` weights, a custom 'tiny' 13.6M param variant that is a bit better than (non NoisyStudent) B3 models. Both faster and better accuracy (at same or lower res)
* top-1 82.34 @ 288x288 and 82.54 @ 320x320
* Add [SAM pretrained](https://arxiv.org/abs/2106.01548) in1k weight for ViT B/16 (`vit_base_patch16_sam_224`) and B/32 (`vit_base_patch32_sam_224`) models.
* Add 'Aggregating Nested Transformer' (NesT) w/ weights converted from official [Flax impl](https://github.com/google-research/nested-transformer). Contributed by [Alexander Soare](https://github.com/alexander-soare).
* `jx_nest_base` - 83.534, `jx_nest_small` - 83.120, `jx_nest_tiny` - 81.426
### June 23, 2021
* Reproduce gMLP model training, `gmlp_s16_224` trained to 79.6 top-1, matching [paper](https://arxiv.org/abs/2105.08050). Hparams for this and other recent MLP training [here](https://gist.github.com/rwightman/d6c264a9001f9167e06c209f630b2cc6)
### June 20, 2021
* Release Vision Transformer 'AugReg' weights from [How to train your ViT? Data, Augmentation, and Regularization in Vision Transformers](https://arxiv.org/abs/2106.10270)
* .npz weight loading support added, can load any of the 50K+ weights from the [AugReg series](https://console.cloud.google.com/storage/browser/vit_models/augreg)
* See [example notebook](https://colab.research.google.com/github/google-research/vision_transformer/blob/master/vit_jax_augreg.ipynb) from [official impl](https://github.com/google-research/vision_transformer/) for navigating the augreg weights
* Replaced all default weights w/ best AugReg variant (if possible). All AugReg 21k classifiers work.
* Highlights: `vit_large_patch16_384` (87.1 top-1), `vit_large_r50_s32_384` (86.2 top-1), `vit_base_patch16_384` (86.0 top-1)
* `vit_deit_*` renamed to just `deit_*`
* Remove my old small model, replace with DeiT compatible small w/ AugReg weights
* Add 1st training of my `gmixer_24_224` MLP /w GLU, 78.1 top-1 w/ 25M params.
* Add weights from official ResMLP release (https://github.com/facebookresearch/deit)
* Add `eca_nfnet_l2` weights from my 'lightweight' series. 84.7 top-1 at 384x384.
* Add distilled BiT 50x1 student and 152x2 Teacher weights from [Knowledge distillation: A good teacher is patient and consistent](https://arxiv.org/abs/2106.05237)
* NFNets and ResNetV2-BiT models work w/ Pytorch XLA now
* weight standardization uses F.batch_norm instead of std_mean (std_mean wasn't lowered)
* eps values adjusted, will be slight differences but should be quite close
* Improve test coverage and classifier interface of non-conv (vision transformer and mlp) models
* Cleanup a few classifier / flatten details for models w/ conv classifiers or early global pool
* Please report any regressions, this PR touched quite a few models.
### June 8, 2021
* Add first ResMLP weights, trained in PyTorch XLA on TPU-VM w/ my XLA branch. 24 block variant, 79.2 top-1.
* Add ResNet51-Q model w/ pretrained weights at 82.36 top-1.
* NFNet inspired block layout with quad layer stem and no maxpool
* Same param count (35.7M) and throughput as ResNetRS-50 but +1.5 top-1 @ 224x224 and +2.5 top-1 at 288x288
### May 25, 2021
* Add LeViT, Visformer, Convit (PR by Aman Arora), Twins (PR by paper authors) transformer models
* Cleanup input_size/img_size override handling and testing for all vision transformer models
* Add `efficientnetv2_rw_m` model and weights (started training before official code). 84.8 top-1, 53M params.
### May 14, 2021
* Add EfficientNet-V2 official model defs w/ ported weights from official [Tensorflow/Keras](https://github.com/google/automl/tree/master/efficientnetv2) impl.
* 1k trained variants: `tf_efficientnetv2_s/m/l`
* 21k trained variants: `tf_efficientnetv2_s/m/l_in21k`
* 21k pretrained -> 1k fine-tuned: `tf_efficientnetv2_s/m/l_in21ft1k`
* v2 models w/ v1 scaling: `tf_efficientnetv2_b0` through `b3`
* Rename my prev V2 guess `efficientnet_v2s` -> `efficientnetv2_rw_s`
* Some blank `efficientnetv2_*` models in-place for future native PyTorch training
### May 5, 2021
* Add MLP-Mixer models and port pretrained weights from [Google JAX impl](https://github.com/google-research/vision_transformer/tree/linen)
* Add CaiT models and pretrained weights from [FB](https://github.com/facebookresearch/deit)
* Add ResNet-RS models and weights from [TF](https://github.com/tensorflow/tpu/tree/master/models/official/resnet/resnet_rs). Thanks [Aman Arora](https://github.com/amaarora)
* Add CoaT models and weights. Thanks [Mohammed Rizin](https://github.com/morizin)
* Add new ImageNet-21k weights & finetuned weights for TResNet, MobileNet-V3, ViT models. Thanks [mrT](https://github.com/mrT23)
* Add GhostNet models and weights. Thanks [Kai Han](https://github.com/iamhankai)
* Update ByoaNet attention modles
* Improve SA module inits
* Hack together experimental stand-alone Swin based attn module and `swinnet`
* Consistent '26t' model defs for experiments.
* Add improved Efficientnet-V2S (prelim model def) weights. 83.8 top-1.
* WandB logging support
### April 13, 2021
* Add Swin Transformer models and weights from https://github.com/microsoft/Swin-Transformer
### April 12, 2021
* Add ECA-NFNet-L1 (slimmed down F1 w/ SiLU, 41M params) trained with this code. 84% top-1 @ 320x320. Trained at 256x256.
* Add EfficientNet-V2S model (unverified model definition) weights. 83.3 top-1 @ 288x288. Only trained single res 224. Working on progressive training.
* Add ByoaNet model definition (Bring-your-own-attention) w/ SelfAttention block and corresponding SA/SA-like modules and model defs
* Lambda Networks - https://arxiv.org/abs/2102.08602
* Bottleneck Transformers - https://arxiv.org/abs/2101.11605
* Halo Nets - https://arxiv.org/abs/2103.12731
* Adabelief optimizer contributed by Juntang Zhuang
### April 1, 2021
* Add snazzy `benchmark.py` script for bulk `timm` model benchmarking of train and/or inference
* Add Pooling-based Vision Transformer (PiT) models (from https://github.com/naver-ai/pit)
* Merged distilled variant into main for torchscript compatibility
* Some `timm` cleanup/style tweaks and weights have hub download support
* Cleanup Vision Transformer (ViT) models
* Merge distilled (DeiT) model into main so that torchscript can work
* Support updated weight init (defaults to old still) that closer matches original JAX impl (possibly better training from scratch)
* Separate hybrid model defs into different file and add several new model defs to fiddle with, support patch_size != 1 for hybrids
* Fix fine-tuning num_class changes (PiT and ViT) and pos_embed resizing (Vit) with distilled variants
* nn.Sequential for block stack (does not break downstream compat)
* TnT (Transformer-in-Transformer) models contributed by author (from https://gitee.com/mindspore/mindspore/tree/master/model_zoo/research/cv/TNT)
* Add RegNetY-160 weights from DeiT teacher model
* Add new NFNet-L0 w/ SE attn (rename `nfnet_l0b`->`nfnet_l0`) weights 82.75 top-1 @ 288x288
* Some fixes/improvements for TFDS dataset wrapper
### March 7, 2021
* First 0.4.x PyPi release w/ NFNets (& related), ByoB (GPU-Efficient, RepVGG, etc).
* Change feature extraction for pre-activation nets (NFNets, ResNetV2) to return features before activation.
### Feb 18, 2021
* Add pretrained weights and model variants for NFNet-F* models from [DeepMind Haiku impl](https://github.com/deepmind/deepmind-research/tree/master/nfnets).
* Models are prefixed with `dm_`. They require SAME padding conv, skipinit enabled, and activation gains applied in act fn.
* These models are big, expect to run out of GPU memory. With the GELU activiation + other options, they are roughly 1/2 the inference speed of my SiLU PyTorch optimized `s` variants.
* Original model results are based on pre-processing that is not the same as all other models so you'll see different results in the results csv (once updated).
* Matching the original pre-processing as closely as possible I get these results:
* `dm_nfnet_f6` - 86.352
* `dm_nfnet_f5` - 86.100
* `dm_nfnet_f4` - 85.834
* `dm_nfnet_f3` - 85.676
* `dm_nfnet_f2` - 85.178
* `dm_nfnet_f1` - 84.696
* `dm_nfnet_f0` - 83.464
### Feb 16, 2021
* Add Adaptive Gradient Clipping (AGC) as per https://arxiv.org/abs/2102.06171. Integrated w/ PyTorch gradient clipping via mode arg that defaults to prev 'norm' mode. For backward arg compat, clip-grad arg must be specified to enable when using train.py.
* AGC w/ default clipping factor `--clip-grad .01 --clip-mode agc`
* PyTorch global norm of 1.0 (old behaviour, always norm), `--clip-grad 1.0`
* PyTorch value clipping of 10, `--clip-grad 10. --clip-mode value`
* AGC performance is definitely sensitive to the clipping factor. More experimentation needed to determine good values for smaller batch sizes and optimizers besides those in paper. So far I've found .001-.005 is necessary for stable RMSProp training w/ NFNet/NF-ResNet.
### Feb 12, 2021
* Update Normalization-Free nets to include new NFNet-F (https://arxiv.org/abs/2102.06171) model defs
### Feb 10, 2021
* More model archs, incl a flexible ByobNet backbone ('Bring-your-own-blocks')
* GPU-Efficient-Networks (https://github.com/idstcv/GPU-Efficient-Networks), impl in `byobnet.py`
* RepVGG (https://github.com/DingXiaoH/RepVGG), impl in `byobnet.py`
* classic VGG (from torchvision, impl in `vgg`)
* Refinements to normalizer layer arg handling and normalizer+act layer handling in some models
* Default AMP mode changed to native PyTorch AMP instead of APEX. Issues not being fixed with APEX. Native works with `--channels-last` and `--torchscript` model training, APEX does not.
* Fix a few bugs introduced since last pypi release
### Feb 8, 2021
* Add several ResNet weights with ECA attention. 26t & 50t trained @ 256, test @ 320. 269d train @ 256, fine-tune @320, test @ 352.
* `ecaresnet26t` - 79.88 top-1 @ 320x320, 79.08 @ 256x256
* `ecaresnet50t` - 82.35 top-1 @ 320x320, 81.52 @ 256x256
* `ecaresnet269d` - 84.93 top-1 @ 352x352, 84.87 @ 320x320
* Remove separate tiered (`t`) vs tiered_narrow (`tn`) ResNet model defs, all `tn` changed to `t` and `t` models removed (`seresnext26t_32x4d` only model w/ weights that was removed).
* Support model default_cfgs with separate train vs test resolution `test_input_size` and remove extra `_320` suffix ResNet model defs that were just for test.
### Jan 30, 2021
* Add initial "Normalization Free" NF-RegNet-B* and NF-ResNet model definitions based on [paper](https://arxiv.org/abs/2101.08692)
### Jan 25, 2021
* Add ResNetV2 Big Transfer (BiT) models w/ ImageNet-1k and 21k weights from https://github.com/google-research/big_transfer
* Add official R50+ViT-B/16 hybrid models + weights from https://github.com/google-research/vision_transformer
* ImageNet-21k ViT weights are added w/ model defs and representation layer (pre logits) support
* NOTE: ImageNet-21k classifier heads were zero'd in original weights, they are only useful for transfer learning
* Add model defs and weights for DeiT Vision Transformer models from https://github.com/facebookresearch/deit
* Refactor dataset classes into ImageDataset/IterableImageDataset + dataset specific parser classes
* Add Tensorflow-Datasets (TFDS) wrapper to allow use of TFDS image classification sets with train script
* Ex: `train.py /data/tfds --dataset tfds/oxford_iiit_pet --val-split test --model resnet50 -b 256 --amp --num-classes 37 --opt adamw --lr 3e-4 --weight-decay .001 --pretrained -j 2`
* Add improved .tar dataset parser that reads images from .tar, folder of .tar files, or .tar within .tar
* Run validation on full ImageNet-21k directly from tar w/ BiT model: `validate.py /data/fall11_whole.tar --model resnetv2_50x1_bitm_in21k --amp`
* Models in this update should be stable w/ possible exception of ViT/BiT, possibility of some regressions with train/val scripts and dataset handling
### Jan 3, 2021
* Add SE-ResNet-152D weights
* 256x256 val, 0.94 crop top-1 - 83.75
* 320x320 val, 1.0 crop - 84.36
* Update results files
### Dec 18, 2020
* Add ResNet-101D, ResNet-152D, and ResNet-200D weights trained @ 256x256
* 256x256 val, 0.94 crop (top-1) - 101D (82.33), 152D (83.08), 200D (83.25)
* 288x288 val, 1.0 crop - 101D (82.64), 152D (83.48), 200D (83.76)
* 320x320 val, 1.0 crop - 101D (83.00), 152D (83.66), 200D (84.01)
### Dec 7, 2020
* Simplify EMA module (ModelEmaV2), compatible with fully torchscripted models
* Misc fixes for SiLU ONNX export, default_cfg missing from Feature extraction models, Linear layer w/ AMP + torchscript
* PyPi release @ 0.3.2 (needed by EfficientDet)
### Oct 30, 2020
* Test with PyTorch 1.7 and fix a small top-n metric view vs reshape issue.
* Convert newly added 224x224 Vision Transformer weights from official JAX repo. 81.8 top-1 for B/16, 83.1 L/16.
* Support PyTorch 1.7 optimized, native SiLU (aka Swish) activation. Add mapping to 'silu' name, custom swish will eventually be deprecated.
* Fix regression for loading pretrained classifier via direct model entrypoint functions. Didn't impact create_model() factory usage.
* PyPi release @ 0.3.0 version!
### Oct 26, 2020
* Update Vision Transformer models to be compatible with official code release at https://github.com/google-research/vision_transformer
* Add Vision Transformer weights (ImageNet-21k pretrain) for 384x384 base and large models converted from official jax impl
* ViT-B/16 - 84.2
* ViT-B/32 - 81.7
* ViT-L/16 - 85.2
* ViT-L/32 - 81.5
### Oct 21, 2020
* Weights added for Vision Transformer (ViT) models. 77.86 top-1 for 'small' and 79.35 for 'base'. Thanks to [Christof](https://www.kaggle.com/christofhenkel) for training the base model w/ lots of GPUs.
### Oct 13, 2020
* Initial impl of Vision Transformer models. Both patch and hybrid (CNN backbone) variants. Currently trying to train...
* Adafactor and AdaHessian (FP32 only, no AMP) optimizers
* EdgeTPU-M (`efficientnet_em`) model trained in PyTorch, 79.3 top-1
* Pip release, doc updates pending a few more changes...
### Sept 18, 2020
* New ResNet 'D' weights. 72.7 (top-1) ResNet-18-D, 77.1 ResNet-34-D, 80.5 ResNet-50-D
* Added a few untrained defs for other ResNet models (66D, 101D, 152D, 200/200D)
### Sept 3, 2020
* New weights
* Wide-ResNet50 - 81.5 top-1 (vs 78.5 torchvision)
* SEResNeXt50-32x4d - 81.3 top-1 (vs 79.1 cadene)
* Support for native Torch AMP and channels_last memory format added to train/validate scripts (`--channels-last`, `--native-amp` vs `--apex-amp`)
* Models tested with channels_last on latest NGC 20.08 container. AdaptiveAvgPool in attn layers changed to mean((2,3)) to work around bug with NHWC kernel.
### Aug 12, 2020
* New/updated weights from training experiments
* EfficientNet-B3 - 82.1 top-1 (vs 81.6 for official with AA and 81.9 for AdvProp)
* RegNetY-3.2GF - 82.0 top-1 (78.9 from official ver)
* CSPResNet50 - 79.6 top-1 (76.6 from official ver)
* Add CutMix integrated w/ Mixup. See [pull request](https://github.com/rwightman/pytorch-image-models/pull/218) for some usage examples
* Some fixes for using pretrained weights with `in_chans` != 3 on several models.
### Aug 5, 2020
Universal feature extraction, new models, new weights, new test sets.
* All models support the `features_only=True` argument for `create_model` call to return a network that extracts feature maps from the deepest layer at each stride.
* New models
* CSPResNet, CSPResNeXt, CSPDarkNet, DarkNet
* ReXNet
* (Modified Aligned) Xception41/65/71 (a proper port of TF models)
* New trained weights
* SEResNet50 - 80.3 top-1
* CSPDarkNet53 - 80.1 top-1
* CSPResNeXt50 - 80.0 top-1
* DPN68b - 79.2 top-1
* EfficientNet-Lite0 (non-TF ver) - 75.5 (submitted by [@hal-314](https://github.com/hal-314))
* Add 'real' labels for ImageNet and ImageNet-Renditions test set, see [`results/README.md`](results/README.md)
* Test set ranking/top-n diff script by [@KushajveerSingh](https://github.com/KushajveerSingh)
* Train script and loader/transform tweaks to punch through more aug arguments
* README and documentation overhaul. See initial (WIP) documentation at https://rwightman.github.io/pytorch-image-models/
* adamp and sgdp optimizers added by [@hellbell](https://github.com/hellbell)
### June 11, 2020
Bunch of changes:
* DenseNet models updated with memory efficient addition from torchvision (fixed a bug), blur pooling and deep stem additions
* VoVNet V1 and V2 models added, 39 V2 variant (ese_vovnet_39b) trained to 79.3 top-1
* Activation factory added along with new activations:
* select act at model creation time for more flexibility in using activations compatible with scripting or tracing (ONNX export)
* hard_mish (experimental) added with memory-efficient grad, along with ME hard_swish
* context mgr for setting exportable/scriptable/no_jit states
* Norm + Activation combo layers added with initial trial support in DenseNet and VoVNet along with impl of EvoNorm and InplaceAbn wrapper that fit the interface
* Torchscript works for all but two of the model types as long as using Pytorch 1.5+, tests added for this
* Some import cleanup and classifier reset changes, all models will have classifier reset to nn.Identity on reset_classifer(0) call
* Prep for 0.1.28 pip release
### May 12, 2020
* Add ResNeSt models (code adapted from https://github.com/zhanghang1989/ResNeSt, paper https://arxiv.org/abs/2004.08955))
### May 3, 2020
* Pruned EfficientNet B1, B2, and B3 (https://arxiv.org/abs/2002.08258) contributed by [Yonathan Aflalo](https://github.com/yoniaflalo)
### May 1, 2020
* Merged a number of execellent contributions in the ResNet model family over the past month
* BlurPool2D and resnetblur models initiated by [Chris Ha](https://github.com/VRandme), I trained resnetblur50 to 79.3.
* TResNet models and SpaceToDepth, AntiAliasDownsampleLayer layers by [mrT23](https://github.com/mrT23)
* ecaresnet (50d, 101d, light) models and two pruned variants using pruning as per (https://arxiv.org/abs/2002.08258) by [Yonathan Aflalo](https://github.com/yoniaflalo)
* 200 pretrained models in total now with updated results csv in results folder
### April 5, 2020
* Add some newly trained MobileNet-V2 models trained with latest h-params, rand augment. They compare quite favourably to EfficientNet-Lite
* 3.5M param MobileNet-V2 100 @ 73%
* 4.5M param MobileNet-V2 110d @ 75%
* 6.1M param MobileNet-V2 140 @ 76.5%
* 5.8M param MobileNet-V2 120d @ 77.3%
### March 18, 2020
* Add EfficientNet-Lite models w/ weights ported from [Tensorflow TPU](https://github.com/tensorflow/tpu/tree/master/models/official/efficientnet/lite)
* Add RandAugment trained ResNeXt-50 32x4d weights with 79.8 top-1. Trained by [Andrew Lavin](https://github.com/andravin) (see Training section for hparams)
### April 5, 2020
* Add some newly trained MobileNet-V2 models trained with latest h-params, rand augment. They compare quite favourably to EfficientNet-Lite
* 3.5M param MobileNet-V2 100 @ 73%
* 4.5M param MobileNet-V2 110d @ 75%
* 6.1M param MobileNet-V2 140 @ 76.5%
* 5.8M param MobileNet-V2 120d @ 77.3%
### March 18, 2020
* Add EfficientNet-Lite models w/ weights ported from [Tensorflow TPU](https://github.com/tensorflow/tpu/tree/master/models/official/efficientnet/lite)
* Add RandAugment trained ResNeXt-50 32x4d weights with 79.8 top-1. Trained by [Andrew Lavin](https://github.com/andravin) (see Training section for hparams)
### Feb 29, 2020
* New MobileNet-V3 Large weights trained from stratch with this code to 75.77% top-1
* IMPORTANT CHANGE - default weight init changed for all MobilenetV3 / EfficientNet / related models
* overall results similar to a bit better training from scratch on a few smaller models tried
* performance early in training seems consistently improved but less difference by end
* set `fix_group_fanout=False` in `_init_weight_goog` fn if you need to reproducte past behaviour
* Experimental LR noise feature added applies a random perturbation to LR each epoch in specified range of training
### Feb 18, 2020
* Big refactor of model layers and addition of several attention mechanisms. Several additions motivated by 'Compounding the Performance Improvements...' (https://arxiv.org/abs/2001.06268):
* Move layer/module impl into `layers` subfolder/module of `models` and organize in a more granular fashion
* ResNet downsample paths now properly support dilation (output stride != 32) for avg_pool ('D' variant) and 3x3 (SENets) networks
* Add Selective Kernel Nets on top of ResNet base, pretrained weights
* skresnet18 - 73% top-1
* skresnet34 - 76.9% top-1
* skresnext50_32x4d (equiv to SKNet50) - 80.2% top-1
* ECA and CECA (circular padding) attention layer contributed by [Chris Ha](https://github.com/VRandme)
* CBAM attention experiment (not the best results so far, may remove)
* Attention factory to allow dynamically selecting one of SE, ECA, CBAM in the `.se` position for all ResNets
* Add DropBlock and DropPath (formerly DropConnect for EfficientNet/MobileNetv3) support to all ResNet variants
* Full dataset results updated that incl NoisyStudent weights and 2 of the 3 SK weights
### Feb 12, 2020
* Add EfficientNet-L2 and B0-B7 NoisyStudent weights ported from [Tensorflow TPU](https://github.com/tensorflow/tpu/tree/master/models/official/efficientnet)
### Feb 6, 2020
* Add RandAugment trained EfficientNet-ES (EdgeTPU-Small) weights with 78.1 top-1. Trained by [Andrew Lavin](https://github.com/andravin) (see Training section for hparams)
### Feb 1/2, 2020
* Port new EfficientNet-B8 (RandAugment) weights, these are different than the B8 AdvProp, different input normalization.
* Update results csv files on all models for ImageNet validation and three other test sets
* Push PyPi package update
### Jan 31, 2020
* Update ResNet50 weights with a new 79.038 result from further JSD / AugMix experiments. Full command line for reproduction in training section below.
### Jan 11/12, 2020
* Master may be a bit unstable wrt to training, these changes have been tested but not all combos
* Implementations of AugMix added to existing RA and AA. Including numerous supporting pieces like JSD loss (Jensen-Shannon divergence + CE), and AugMixDataset
* SplitBatchNorm adaptation layer added for implementing Auxiliary BN as per AdvProp paper
* ResNet-50 AugMix trained model w/ 79% top-1 added
* `seresnext26tn_32x4d` - 77.99 top-1, 93.75 top-5 added to tiered experiment, higher img/s than 't' and 'd'
### Jan 3, 2020
* Add RandAugment trained EfficientNet-B0 weight with 77.7 top-1. Trained by [Michael Klachko](https://github.com/michaelklachko) with this code and recent hparams (see Training section)
* Add `avg_checkpoints.py` script for post training weight averaging and update all scripts with header docstrings and shebangs.
### Dec 30, 2019
* Merge [Dushyant Mehta's](https://github.com/mehtadushy) PR for SelecSLS (Selective Short and Long Range Skip Connections) networks. Good GPU memory consumption and throughput. Original: https://github.com/mehtadushy/SelecSLS-Pytorch
### Dec 28, 2019
* Add new model weights and training hparams (see Training Hparams section)
* `efficientnet_b3` - 81.5 top-1, 95.7 top-5 at default res/crop, 81.9, 95.8 at 320x320 1.0 crop-pct
* trained with RandAugment, ended up with an interesting but less than perfect result (see training section)
* `seresnext26d_32x4d`- 77.6 top-1, 93.6 top-5
* deep stem (32, 32, 64), avgpool downsample
* stem/dowsample from bag-of-tricks paper
* `seresnext26t_32x4d`- 78.0 top-1, 93.7 top-5
* deep tiered stem (24, 48, 64), avgpool downsample (a modified 'D' variant)
* stem sizing mods from Jeremy Howard and fastai devs discussing ResNet architecture experiments
### Dec 23, 2019
* Add RandAugment trained MixNet-XL weights with 80.48 top-1.
* `--dist-bn` argument added to train.py, will distribute BN stats between nodes after each train epoch, before eval
### Dec 4, 2019
* Added weights from the first training from scratch of an EfficientNet (B2) with my new RandAugment implementation. Much better than my previous B2 and very close to the official AdvProp ones (80.4 top-1, 95.08 top-5).
### Nov 29, 2019
* Brought EfficientNet and MobileNetV3 up to date with my https://github.com/rwightman/gen-efficientnet-pytorch code. Torchscript and ONNX export compat excluded.
* AdvProp weights added
* Official TF MobileNetv3 weights added
* EfficientNet and MobileNetV3 hook based 'feature extraction' classes added. Will serve as basis for using models as backbones in obj detection/segmentation tasks. Lots more to be done here...
* HRNet classification models and weights added from https://github.com/HRNet/HRNet-Image-Classification
* Consistency in global pooling, `reset_classifer`, and `forward_features` across models
* `forward_features` always returns unpooled feature maps now
* Reasonable chance I broke something... let me know
### Nov 22, 2019
* Add ImageNet training RandAugment implementation alongside AutoAugment. PyTorch Transform compatible format, using PIL. Currently training two EfficientNet models from scratch with promising results... will update.
* `drop-connect` cmd line arg finally added to `train.py`, no need to hack model fns. Works for efficientnet/mobilenetv3 based models, ignored otherwise.

187
hfdocs/source/changes.mdx
Unescape View File

@ -0,0 +1,187 @@
# Recent Changes
### July 27, 2022
* All runtime benchmark and validation result csv files are up-to-date!
* A few more weights & model defs added:
* `darknetaa53` - 79.8 @ 256, 80.5 @ 288
* `convnext_nano` - 80.8 @ 224, 81.5 @ 288
* `cs3sedarknet_l` - 81.2 @ 256, 81.8 @ 288
* `cs3darknet_x` - 81.8 @ 256, 82.2 @ 288
* `cs3sedarknet_x` - 82.2 @ 256, 82.7 @ 288
* `cs3edgenet_x` - 82.2 @ 256, 82.7 @ 288
* `cs3se_edgenet_x` - 82.8 @ 256, 83.5 @ 320
* `cs3*` weights above all trained on TPU w/ `bits_and_tpu` branch. Thanks to TRC program!
* Add output_stride=8 and 16 support to ConvNeXt (dilation)
* deit3 models not being able to resize pos_emb fixed
* Version 0.6.7 PyPi release (/w above bug fixes and new weighs since 0.6.5)
### July 8, 2022
More models, more fixes
* Official research models (w/ weights) added:
* EdgeNeXt from (https://github.com/mmaaz60/EdgeNeXt)
* MobileViT-V2 from (https://github.com/apple/ml-cvnets)
* DeiT III (Revenge of the ViT) from (https://github.com/facebookresearch/deit)
* My own models:
* Small `ResNet` defs added by request with 1 block repeats for both basic and bottleneck (resnet10 and resnet14)
* `CspNet` refactored with dataclass config, simplified CrossStage3 (`cs3`) option. These are closer to YOLO-v5+ backbone defs.
* More relative position vit fiddling. Two `srelpos` (shared relative position) models trained, and a medium w/ class token.
* Add an alternate downsample mode to EdgeNeXt and train a `small` model. Better than original small, but not their new USI trained weights.
* My own model weight results (all ImageNet-1k training)
* `resnet10t` - 66.5 @ 176, 68.3 @ 224
* `resnet14t` - 71.3 @ 176, 72.3 @ 224
* `resnetaa50` - 80.6 @ 224 , 81.6 @ 288
* `darknet53` - 80.0 @ 256, 80.5 @ 288
* `cs3darknet_m` - 77.0 @ 256, 77.6 @ 288
* `cs3darknet_focus_m` - 76.7 @ 256, 77.3 @ 288
* `cs3darknet_l` - 80.4 @ 256, 80.9 @ 288
* `cs3darknet_focus_l` - 80.3 @ 256, 80.9 @ 288
* `vit_srelpos_small_patch16_224` - 81.1 @ 224, 82.1 @ 320
* `vit_srelpos_medium_patch16_224` - 82.3 @ 224, 83.1 @ 320
* `vit_relpos_small_patch16_cls_224` - 82.6 @ 224, 83.6 @ 320
* `edgnext_small_rw` - 79.6 @ 224, 80.4 @ 320
* `cs3`, `darknet`, and `vit_*relpos` weights above all trained on TPU thanks to TRC program! Rest trained on overheating GPUs.
* Hugging Face Hub support fixes verified, demo notebook TBA
* Pretrained weights / configs can be loaded externally (ie from local disk) w/ support for head adaptation.
* Add support to change image extensions scanned by `timm` datasets/parsers. See (https://github.com/rwightman/pytorch-image-models/pull/1274#issuecomment-1178303103)
* Default ConvNeXt LayerNorm impl to use `F.layer_norm(x.permute(0, 2, 3, 1), ...).permute(0, 3, 1, 2)` via `LayerNorm2d` in all cases.
* a bit slower than previous custom impl on some hardware (ie Ampere w/ CL), but overall fewer regressions across wider HW / PyTorch version ranges.
* previous impl exists as `LayerNormExp2d` in `models/layers/norm.py`
* Numerous bug fixes
* Currently testing for imminent PyPi 0.6.x release
* LeViT pretraining of larger models still a WIP, they don't train well / easily without distillation. Time to add distill support (finally)?
* ImageNet-22k weight training + finetune ongoing, work on multi-weight support (slowly) chugging along (there are a LOT of weights, sigh) ...
### May 13, 2022
* Official Swin-V2 models and weights added from (https://github.com/microsoft/Swin-Transformer). Cleaned up to support torchscript.
* Some refactoring for existing `timm` Swin-V2-CR impl, will likely do a bit more to bring parts closer to official and decide whether to merge some aspects.
* More Vision Transformer relative position / residual post-norm experiments (all trained on TPU thanks to TRC program)
* `vit_relpos_small_patch16_224` - 81.5 @ 224, 82.5 @ 320 -- rel pos, layer scale, no class token, avg pool
* `vit_relpos_medium_patch16_rpn_224` - 82.3 @ 224, 83.1 @ 320 -- rel pos + res-post-norm, no class token, avg pool
* `vit_relpos_medium_patch16_224` - 82.5 @ 224, 83.3 @ 320 -- rel pos, layer scale, no class token, avg pool
* `vit_relpos_base_patch16_gapcls_224` - 82.8 @ 224, 83.9 @ 320 -- rel pos, layer scale, class token, avg pool (by mistake)
* Bring 512 dim, 8-head 'medium' ViT model variant back to life (after using in a pre DeiT 'small' model for first ViT impl back in 2020)
* Add ViT relative position support for switching btw existing impl and some additions in official Swin-V2 impl for future trials
* Sequencer2D impl (https://arxiv.org/abs/2205.01972), added via PR from author (https://github.com/okojoalg)
### May 2, 2022
* Vision Transformer experiments adding Relative Position (Swin-V2 log-coord) (`vision_transformer_relpos.py`) and Residual Post-Norm branches (from Swin-V2) (`vision_transformer*.py`)
* `vit_relpos_base_patch32_plus_rpn_256` - 79.5 @ 256, 80.6 @ 320 -- rel pos + extended width + res-post-norm, no class token, avg pool
* `vit_relpos_base_patch16_224` - 82.5 @ 224, 83.6 @ 320 -- rel pos, layer scale, no class token, avg pool
* `vit_base_patch16_rpn_224` - 82.3 @ 224 -- rel pos + res-post-norm, no class token, avg pool
* Vision Transformer refactor to remove representation layer that was only used in initial vit and rarely used since with newer pretrain (ie `How to Train Your ViT`)
* `vit_*` models support removal of class token, use of global average pool, use of fc_norm (ala beit, mae).
### April 22, 2022
* `timm` models are now officially supported in [fast.ai](https://www.fast.ai/)! Just in time for the new Practical Deep Learning course. `timmdocs` documentation link updated to [timm.fast.ai](http://timm.fast.ai/).
* Two more model weights added in the TPU trained [series](https://github.com/rwightman/pytorch-image-models/releases/tag/v0.1-tpu-weights). Some In22k pretrain still in progress.
* `seresnext101d_32x8d` - 83.69 @ 224, 84.35 @ 288
* `seresnextaa101d_32x8d` (anti-aliased w/ AvgPool2d) - 83.85 @ 224, 84.57 @ 288
### March 23, 2022
* Add `ParallelBlock` and `LayerScale` option to base vit models to support model configs in [Three things everyone should know about ViT](https://arxiv.org/abs/2203.09795)
* `convnext_tiny_hnf` (head norm first) weights trained with (close to) A2 recipe, 82.2% top-1, could do better with more epochs.
### March 21, 2022
* Merge `norm_norm_norm`. **IMPORTANT** this update for a coming 0.6.x release will likely de-stabilize the master branch for a while. Branch [`0.5.x`](https://github.com/rwightman/pytorch-image-models/tree/0.5.x) or a previous 0.5.x release can be used if stability is required.
* Significant weights update (all TPU trained) as described in this [release](https://github.com/rwightman/pytorch-image-models/releases/tag/v0.1-tpu-weights)
* `regnety_040` - 82.3 @ 224, 82.96 @ 288
* `regnety_064` - 83.0 @ 224, 83.65 @ 288
* `regnety_080` - 83.17 @ 224, 83.86 @ 288
* `regnetv_040` - 82.44 @ 224, 83.18 @ 288 (timm pre-act)
* `regnetv_064` - 83.1 @ 224, 83.71 @ 288 (timm pre-act)
* `regnetz_040` - 83.67 @ 256, 84.25 @ 320
* `regnetz_040h` - 83.77 @ 256, 84.5 @ 320 (w/ extra fc in head)
* `resnetv2_50d_gn` - 80.8 @ 224, 81.96 @ 288 (pre-act GroupNorm)
* `resnetv2_50d_evos` 80.77 @ 224, 82.04 @ 288 (pre-act EvoNormS)
* `regnetz_c16_evos` - 81.9 @ 256, 82.64 @ 320 (EvoNormS)
* `regnetz_d8_evos` - 83.42 @ 256, 84.04 @ 320 (EvoNormS)
* `xception41p` - 82 @ 299 (timm pre-act)
* `xception65` - 83.17 @ 299
* `xception65p` - 83.14 @ 299 (timm pre-act)
* `resnext101_64x4d` - 82.46 @ 224, 83.16 @ 288
* `seresnext101_32x8d` - 83.57 @ 224, 84.270 @ 288
* `resnetrs200` - 83.85 @ 256, 84.44 @ 320
* HuggingFace hub support fixed w/ initial groundwork for allowing alternative 'config sources' for pretrained model definitions and weights (generic local file / remote url support soon)
* SwinTransformer-V2 implementation added. Submitted by [Christoph Reich](https://github.com/ChristophReich1996). Training experiments and model changes by myself are ongoing so expect compat breaks.
* Swin-S3 (AutoFormerV2) models / weights added from https://github.com/microsoft/Cream/tree/main/AutoFormerV2
* MobileViT models w/ weights adapted from https://github.com/apple/ml-cvnets
* PoolFormer models w/ weights adapted from https://github.com/sail-sg/poolformer
* VOLO models w/ weights adapted from https://github.com/sail-sg/volo
* Significant work experimenting with non-BatchNorm norm layers such as EvoNorm, FilterResponseNorm, GroupNorm, etc
* Enhance support for alternate norm + act ('NormAct') layers added to a number of models, esp EfficientNet/MobileNetV3, RegNet, and aligned Xception
* Grouped conv support added to EfficientNet family
* Add 'group matching' API to all models to allow grouping model parameters for application of 'layer-wise' LR decay, lr scale added to LR scheduler
* Gradient checkpointing support added to many models
* `forward_head(x, pre_logits=False)` fn added to all models to allow separate calls of `forward_features` + `forward_head`
* All vision transformer and vision MLP models update to return non-pooled / non-token selected features from `foward_features`, for consistency with CNN models, token selection or pooling now applied in `forward_head`
### Feb 2, 2022
* [Chris Hughes](https://github.com/Chris-hughes10) posted an exhaustive run through of `timm` on his blog yesterday. Well worth a read. [Getting Started with PyTorch Image Models (timm): A Practitioner’s Guide](https://towardsdatascience.com/getting-started-with-pytorch-image-models-timm-a-practitioners-guide-4e77b4bf9055)
* I'm currently prepping to merge the `norm_norm_norm` branch back to master (ver 0.6.x) in next week or so.
* The changes are more extensive than usual and may destabilize and break some model API use (aiming for full backwards compat). So, beware `pip install git+https://github.com/rwightman/pytorch-image-models` installs!
* `0.5.x` releases and a `0.5.x` branch will remain stable with a cherry pick or two until dust clears. Recommend sticking to pypi install for a bit if you want stable.
### Jan 14, 2022
* Version 0.5.4 w/ release to be pushed to pypi. It's been a while since last pypi update and riskier changes will be merged to main branch soon....
* Add ConvNeXT models /w weights from official impl (https://github.com/facebookresearch/ConvNeXt), a few perf tweaks, compatible with timm features
* Tried training a few small (~1.8-3M param) / mobile optimized models, a few are good so far, more on the way...
* `mnasnet_small` - 65.6 top-1
* `mobilenetv2_050` - 65.9
* `lcnet_100/075/050` - 72.1 / 68.8 / 63.1
* `semnasnet_075` - 73
* `fbnetv3_b/d/g` - 79.1 / 79.7 / 82.0
* TinyNet models added by [rsomani95](https://github.com/rsomani95)
* LCNet added via MobileNetV3 architecture
### Nov 22, 2021
* A number of updated weights anew new model defs
* `eca_halonext26ts` - 79.5 @ 256
* `resnet50_gn` (new) - 80.1 @ 224, 81.3 @ 288
* `resnet50` - 80.7 @ 224, 80.9 @ 288 (trained at 176, not replacing current a1 weights as default since these don't scale as well to higher res, [weights](https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-rsb-weights/resnet50_a1h2_176-001a1197.pth))
* `resnext50_32x4d` - 81.1 @ 224, 82.0 @ 288
* `sebotnet33ts_256` (new) - 81.2 @ 224
* `lamhalobotnet50ts_256` - 81.5 @ 256
* `halonet50ts` - 81.7 @ 256
* `halo2botnet50ts_256` - 82.0 @ 256
* `resnet101` - 82.0 @ 224, 82.8 @ 288
* `resnetv2_101` (new) - 82.1 @ 224, 83.0 @ 288
* `resnet152` - 82.8 @ 224, 83.5 @ 288
* `regnetz_d8` (new) - 83.5 @ 256, 84.0 @ 320
* `regnetz_e8` (new) - 84.5 @ 256, 85.0 @ 320
* `vit_base_patch8_224` (85.8 top-1) & `in21k` variant weights added thanks [Martins Bruveris](https://github.com/martinsbruveris)
* Groundwork in for FX feature extraction thanks to [Alexander Soare](https://github.com/alexander-soare)
* models updated for tracing compatibility (almost full support with some distlled transformer exceptions)
### Oct 19, 2021
* ResNet strikes back (https://arxiv.org/abs/2110.00476) weights added, plus any extra training components used. Model weights and some more details here (https://github.com/rwightman/pytorch-image-models/releases/tag/v0.1-rsb-weights)
* BCE loss and Repeated Augmentation support for RSB paper
* 4 series of ResNet based attention model experiments being added (implemented across byobnet.py/byoanet.py). These include all sorts of attention, from channel attn like SE, ECA to 2D QKV self-attention layers such as Halo, Bottlneck, Lambda. Details here (https://github.com/rwightman/pytorch-image-models/releases/tag/v0.1-attn-weights)
* Working implementations of the following 2D self-attention modules (likely to be differences from paper or eventual official impl):
* Halo (https://arxiv.org/abs/2103.12731)
* Bottleneck Transformer (https://arxiv.org/abs/2101.11605)
* LambdaNetworks (https://arxiv.org/abs/2102.08602)
* A RegNetZ series of models with some attention experiments (being added to). These do not follow the paper (https://arxiv.org/abs/2103.06877) in any way other than block architecture, details of official models are not available. See more here (https://github.com/rwightman/pytorch-image-models/releases/tag/v0.1-attn-weights)
* ConvMixer (https://openreview.net/forum?id=TVHS5Y4dNvM), CrossVit (https://arxiv.org/abs/2103.14899), and BeiT (https://arxiv.org/abs/2106.08254) architectures + weights added
* freeze/unfreeze helpers by [Alexander Soare](https://github.com/alexander-soare)
### Aug 18, 2021
* Optimizer bonanza!
* Add LAMB and LARS optimizers, incl trust ratio clipping options. Tweaked to work properly in PyTorch XLA (tested on TPUs w/ `timm bits` [branch](https://github.com/rwightman/pytorch-image-models/tree/bits_and_tpu/timm/bits))
* Add MADGRAD from FB research w/ a few tweaks (decoupled decay option, step handling that works with PyTorch XLA)
* Some cleanup on all optimizers and factory. No more `.data`, a bit more consistency, unit tests for all!
* SGDP and AdamP still won't work with PyTorch XLA but others should (have yet to test Adabelief, Adafactor, Adahessian myself).
* EfficientNet-V2 XL TF ported weights added, but they don't validate well in PyTorch (L is better). The pre-processing for the V2 TF training is a bit diff and the fine-tuned 21k -> 1k weights are very sensitive and less robust than the 1k weights.
* Added PyTorch trained EfficientNet-V2 'Tiny' w/ GlobalContext attn weights. Only .1-.2 top-1 better than the SE so more of a curiosity for those interested.

196
hfdocs/source/feature_extraction.mdx
Unescape View File

@ -0,0 +1,196 @@
# Feature Extraction
All of the models in `timm` have consistent mechanisms for obtaining various types of features from the model for tasks besides classification.
## Penultimate Layer Features (Pre-Classifier Features)
The features from the penultimate model layer can be obtained in several ways without requiring model surgery (although feel free to do surgery). One must first decide if they want pooled or un-pooled features.
### Unpooled
There are three ways to obtain unpooled features.
Without modifying the network, one can call `model.forward_features(input)` on any model instead of the usual `model(input)`. This will bypass the head classifier and global pooling for networks.
If one wants to explicitly modify the network to return unpooled features, they can either create the model without a classifier and pooling, or remove it later. Both paths remove the parameters associated with the classifier from the network.
#### forward_features()
```py
>>> import torch
>>> import timm
>>> m = timm.create_model('xception41', pretrained=True)
>>> o = m(torch.randn(2, 3, 299, 299))
>>> print(f'Original shape: {o.shape}')
>>> o = m.forward_features(torch.randn(2, 3, 299, 299))
>>> print(f'Unpooled shape: {o.shape}')
```
Output:
```text
Original shape: torch.Size([2, 1000])
Unpooled shape: torch.Size([2, 2048, 10, 10])
```
#### Create with no classifier and pooling
```py
>>> import torch
>>> import timm
>>> m = timm.create_model('resnet50', pretrained=True, num_classes=0, global_pool='')
>>> o = m(torch.randn(2, 3, 224, 224))
>>> print(f'Unpooled shape: {o.shape}')
```
Output:
```text
Unpooled shape: torch.Size([2, 2048, 7, 7])
```
#### Remove it later
```py
>>> import torch
>>> import timm
>>> m = timm.create_model('densenet121', pretrained=True)
>>> o = m(torch.randn(2, 3, 224, 224))
>>> print(f'Original shape: {o.shape}')
>>> m.reset_classifier(0, '')
>>> o = m(torch.randn(2, 3, 224, 224))
>>> print(f'Unpooled shape: {o.shape}')
```
Output:
```text
Original shape: torch.Size([2, 1000])
Unpooled shape: torch.Size([2, 1024, 7, 7])
```
### Pooled
To modify the network to return pooled features, one can use `forward_features()` and pool/flatten the result themselves, or modify the network like above but keep pooling intact.
#### Create with no classifier
```py
>>> import torch
>>> import timm
>>> m = timm.create_model('resnet50', pretrained=True, num_classes=0)
>>> o = m(torch.randn(2, 3, 224, 224))
>>> print(f'Pooled shape: {o.shape}')
```
Output:
```text
Pooled shape: torch.Size([2, 2048])
```
#### Remove it later
```py
>>> import torch
>>> import timm
>>> m = timm.create_model('ese_vovnet19b_dw', pretrained=True)
>>> o = m(torch.randn(2, 3, 224, 224))
>>> print(f'Original shape: {o.shape}')
>>> m.reset_classifier(0)
>>> o = m(torch.randn(2, 3, 224, 224))
>>> print(f'Pooled shape: {o.shape}')
```
Output:
```text
Original shape: torch.Size([2, 1000])
Pooled shape: torch.Size([2, 1024])
```
## Multi-scale Feature Maps (Feature Pyramid)
Object detection, segmentation, keypoint, and a variety of dense pixel tasks require access to feature maps from the backbone network at multiple scales. This is often done by modifying the original classification network. Since each network varies quite a bit in structure, it's not uncommon to see only a few backbones supported in any given obj detection or segmentation library.
`timm` allows a consistent interface for creating any of the included models as feature backbones that output feature maps for selected levels.
A feature backbone can be created by adding the argument `features_only=True` to any `create_model` call. By default 5 strides will be output from most models (not all have that many), with the first starting at 2 (some start at 1 or 4).
### Create a feature map extraction model
```py
>>> import torch
>>> import timm
>>> m = timm.create_model('resnest26d', features_only=True, pretrained=True)
>>> o = m(torch.randn(2, 3, 224, 224))
>>> for x in o:
... print(x.shape)
```
Output:
```text
torch.Size([2, 64, 112, 112])
torch.Size([2, 256, 56, 56])
torch.Size([2, 512, 28, 28])
torch.Size([2, 1024, 14, 14])
torch.Size([2, 2048, 7, 7])
```
### Query the feature information
After a feature backbone has been created, it can be queried to provide channel or resolution reduction information to the downstream heads without requiring static config or hardcoded constants. The `.feature_info` attribute is a class encapsulating the information about the feature extraction points.
```py
>>> import torch
>>> import timm
>>> m = timm.create_model('regnety_032', features_only=True, pretrained=True)
>>> print(f'Feature channels: {m.feature_info.channels()}')
>>> o = m(torch.randn(2, 3, 224, 224))
>>> for x in o:
... print(x.shape)
```
Output:
```text
Feature channels: [32, 72, 216, 576, 1512]
torch.Size([2, 32, 112, 112])
torch.Size([2, 72, 56, 56])
torch.Size([2, 216, 28, 28])
torch.Size([2, 576, 14, 14])
torch.Size([2, 1512, 7, 7])
```
### Select specific feature levels or limit the stride
There are two additional creation arguments impacting the output features.
* `out_indices` selects which indices to output
* `output_stride` limits the feature output stride of the network (also works in classification mode BTW)
`out_indices` is supported by all models, but not all models have the same index to feature stride mapping. Look at the code or check feature_info to compare. The out indices generally correspond to the `C(i+1)th` feature level (a `2^(i+1)` reduction). For most models, index 0 is the stride 2 features, and index 4 is stride 32.
`output_stride` is achieved by converting layers to use dilated convolutions. Doing so is not always straightforward, some networks only support `output_stride=32`.
```py
>>> import torch
>>> import timm
>>> m = timm.create_model('ecaresnet101d', features_only=True, output_stride=8, out_indices=(2, 4), pretrained=True)
>>> print(f'Feature channels: {m.feature_info.channels()}')
>>> print(f'Feature reduction: {m.feature_info.reduction()}')
>>> o = m(torch.randn(2, 3, 320, 320))
>>> for x in o:
... print(x.shape)
```
Output:
```text
Feature channels: [512, 2048]
Feature reduction: [8, 8]
torch.Size([2, 512, 40, 40])
torch.Size([2, 2048, 40, 40])
```

89
hfdocs/source/index.mdx
Unescape View File

@ -0,0 +1,89 @@
# Getting Started
## Welcome
Welcome to the `timm` documentation, a lean set of docs that covers the basics of `timm`.
For a more comprehensive set of docs (currently under development), please visit [timmdocs](http://timm.fast.ai) by [Aman Arora](https://github.com/amaarora).
## Install
The library can be installed with pip:
```
pip install timm
```
I update the PyPi (pip) packages when I'm confident there are no significant model regressions from previous releases. If you want to pip install the bleeding edge from GitHub, use:
```
pip install git+https://github.com/rwightman/pytorch-image-models.git
```
### Conda Environment
<Tip>
- All development and testing has been done in Conda Python 3 environments on Linux x86-64 systems, specifically 3.7, 3.8, 3.9, 3.10
- Little to no care has been taken to be Python 2.x friendly and will not support it. If you run into any challenges running on Windows, or other OS, I'm definitely open to looking into those issues so long as it's in a reproducible (read Conda) environment.
- PyTorch versions 1.9, 1.10, 1.11 have been tested with the latest versions of this code.
</Tip>
I've tried to keep the dependencies minimal, the setup is as per the PyTorch default install instructions for Conda:
```bash
conda create -n torch-env
conda activate torch-env
conda install pytorch torchvision cudatoolkit=11.3 -c pytorch
conda install pyyaml
```
## Load a Pretrained Model
Pretrained models can be loaded using `timm.create_model`
```py
>>> import timm
>>> m = timm.create_model('mobilenetv3_large_100', pretrained=True)
>>> m.eval()
```
## List Models with Pretrained Weights
```py
>>> import timm
>>> from pprint import pprint
>>> model_names = timm.list_models(pretrained=True)
>>> pprint(model_names)
[
'adv_inception_v3',
'cspdarknet53',
'cspresnext50',
'densenet121',
'densenet161',
'densenet169',
'densenet201',
'densenetblur121d',
'dla34',
'dla46_c',
]
```
## List Model Architectures by Wildcard
```py
>>> import timm
>>> from pprint import pprint
>>> model_names = timm.list_models('*resne*t*')
>>> pprint(model_names)
[
'cspresnet50',
'cspresnet50d',
'cspresnet50w',
'cspresnext50',
...
]
```

5
hfdocs/source/model_pages.mdx
Unescape View File

@ -0,0 +1,5 @@
# Available Models
`timm` comes bundled with a number of model architectures and corresponding pretrained models.
In these pages, you will find the models available in the `timm` library, as well as information on how to use them.

230
hfdocs/source/models.mdx
Unescape View File

@ -0,0 +1,230 @@
# Model Summaries
The model architectures included come from a wide variety of sources. Sources, including papers, original impl ("reference code") that I rewrote / adapted, and PyTorch impl that I leveraged directly ("code") are listed below.
Most included models have pretrained weights. The weights are either:
1. from their original sources
2. ported by myself from their original impl in a different framework (e.g. Tensorflow models)
3. trained from scratch using the included training script
The validation results for the pretrained weights are [here](results)
A more exciting view (with pretty pictures) of the models within `timm` can be found at [paperswithcode](https://paperswithcode.com/lib/timm).
## Big Transfer ResNetV2 (BiT)
* Implementation: [resnetv2.py](https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/resnetv2.py)
* Paper: `Big Transfer (BiT): General Visual Representation Learning` - https://arxiv.org/abs/1912.11370
* Reference code: https://github.com/google-research/big_transfer
## Cross-Stage Partial Networks
* Implementation: [cspnet.py](https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/cspnet.py)
* Paper: `CSPNet: A New Backbone that can Enhance Learning Capability of CNN` - https://arxiv.org/abs/1911.11929
* Reference impl: https://github.com/WongKinYiu/CrossStagePartialNetworks
## DenseNet
* Implementation: [densenet.py](https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/densenet.py)
* Paper: `Densely Connected Convolutional Networks` - https://arxiv.org/abs/1608.06993
* Code: https://github.com/pytorch/vision/tree/master/torchvision/models
## DLA
* Implementation: [dla.py](https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/dla.py)
* Paper: https://arxiv.org/abs/1707.06484
* Code: https://github.com/ucbdrive/dla
## Dual-Path Networks
* Implementation: [dpn.py](https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/dpn.py)
* Paper: `Dual Path Networks` - https://arxiv.org/abs/1707.01629
* My PyTorch code: https://github.com/rwightman/pytorch-dpn-pretrained
* Reference code: https://github.com/cypw/DPNs
## GPU-Efficient Networks
* Implementation: [byobnet.py](https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/byobnet.py)
* Paper: `Neural Architecture Design for GPU-Efficient Networks` - https://arxiv.org/abs/2006.14090
* Reference code: https://github.com/idstcv/GPU-Efficient-Networks
## HRNet
* Implementation: [hrnet.py](https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/hrnet.py)
* Paper: `Deep High-Resolution Representation Learning for Visual Recognition` - https://arxiv.org/abs/1908.07919
* Code: https://github.com/HRNet/HRNet-Image-Classification
## Inception-V3
* Implementation: [inception_v3.py](https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/inception_v3.py)
* Paper: `Rethinking the Inception Architecture for Computer Vision` - https://arxiv.org/abs/1512.00567
* Code: https://github.com/pytorch/vision/tree/master/torchvision/models
## Inception-V4
* Implementation: [inception_v4.py](https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/inception_v4.py)
* Paper: `Inception-v4, Inception-ResNet and the Impact of Residual Connections on Learning` - https://arxiv.org/abs/1602.07261
* Code: https://github.com/Cadene/pretrained-models.pytorch
* Reference code: https://github.com/tensorflow/models/tree/master/research/slim/nets
## Inception-ResNet-V2
* Implementation: [inception_resnet_v2.py](https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/inception_resnet_v2.py)
* Paper: `Inception-v4, Inception-ResNet and the Impact of Residual Connections on Learning` - https://arxiv.org/abs/1602.07261
* Code: https://github.com/Cadene/pretrained-models.pytorch
* Reference code: https://github.com/tensorflow/models/tree/master/research/slim/nets
## NASNet-A
* Implementation: [nasnet.py](https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/nasnet.py)
* Papers: `Learning Transferable Architectures for Scalable Image Recognition` - https://arxiv.org/abs/1707.07012
* Code: https://github.com/Cadene/pretrained-models.pytorch
* Reference code: https://github.com/tensorflow/models/tree/master/research/slim/nets/nasnet
## PNasNet-5
* Implementation: [pnasnet.py](https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/pnasnet.py)
* Papers: `Progressive Neural Architecture Search` - https://arxiv.org/abs/1712.00559
* Code: https://github.com/Cadene/pretrained-models.pytorch
* Reference code: https://github.com/tensorflow/models/tree/master/research/slim/nets/nasnet
## EfficientNet
* Implementation: [efficientnet.py](https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/efficientnet.py)
* Papers:
* EfficientNet NoisyStudent (B0-B7, L2) - https://arxiv.org/abs/1911.04252
* EfficientNet AdvProp (B0-B8) - https://arxiv.org/abs/1911.09665
* EfficientNet (B0-B7) - https://arxiv.org/abs/1905.11946
* EfficientNet-EdgeTPU (S, M, L) - https://ai.googleblog.com/2019/08/efficientnet-edgetpu-creating.html
* MixNet - https://arxiv.org/abs/1907.09595
* MNASNet B1, A1 (Squeeze-Excite), and Small - https://arxiv.org/abs/1807.11626
* MobileNet-V2 - https://arxiv.org/abs/1801.04381
* FBNet-C - https://arxiv.org/abs/1812.03443
* Single-Path NAS - https://arxiv.org/abs/1904.02877
* My PyTorch code: https://github.com/rwightman/gen-efficientnet-pytorch
* Reference code: https://github.com/tensorflow/tpu/tree/master/models/official/efficientnet
## MobileNet-V3
* Implementation: [mobilenetv3.py](https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/mobilenetv3.py)
* Paper: `Searching for MobileNetV3` - https://arxiv.org/abs/1905.02244
* Reference code: https://github.com/tensorflow/models/tree/master/research/slim/nets/mobilenet
## RegNet
* Implementation: [regnet.py](https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/regnet.py)
* Paper: `Designing Network Design Spaces` - https://arxiv.org/abs/2003.13678
* Reference code: https://github.com/facebookresearch/pycls/blob/master/pycls/models/regnet.py
## RepVGG
* Implementation: [byobnet.py](https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/byobnet.py)
* Paper: `Making VGG-style ConvNets Great Again` - https://arxiv.org/abs/2101.03697
* Reference code: https://github.com/DingXiaoH/RepVGG
## ResNet, ResNeXt
* Implementation: [resnet.py](https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/resnet.py)
* ResNet (V1B)
* Paper: `Deep Residual Learning for Image Recognition` - https://arxiv.org/abs/1512.03385
* Code: https://github.com/pytorch/vision/tree/master/torchvision/models
* ResNeXt
* Paper: `Aggregated Residual Transformations for Deep Neural Networks` - https://arxiv.org/abs/1611.05431
* Code: https://github.com/pytorch/vision/tree/master/torchvision/models
* 'Bag of Tricks' / Gluon C, D, E, S ResNet variants
* Paper: `Bag of Tricks for Image Classification with CNNs` - https://arxiv.org/abs/1812.01187
* Code: https://github.com/dmlc/gluon-cv/blob/master/gluoncv/model_zoo/resnetv1b.py
* Instagram pretrained / ImageNet tuned ResNeXt101
* Paper: `Exploring the Limits of Weakly Supervised Pretraining` - https://arxiv.org/abs/1805.00932
* Weights: https://pytorch.org/hub/facebookresearch_WSL-Images_resnext (NOTE: CC BY-NC 4.0 License, NOT commercial friendly)
* Semi-supervised (SSL) / Semi-weakly Supervised (SWSL) ResNet and ResNeXts
* Paper: `Billion-scale semi-supervised learning for image classification` - https://arxiv.org/abs/1905.00546
* Weights: https://github.com/facebookresearch/semi-supervised-ImageNet1K-models (NOTE: CC BY-NC 4.0 License, NOT commercial friendly)
* Squeeze-and-Excitation Networks
* Paper: `Squeeze-and-Excitation Networks` - https://arxiv.org/abs/1709.01507
* Code: Added to ResNet base, this is current version going forward, old `senet.py` is being deprecated
* ECAResNet (ECA-Net)
* Paper: `ECA-Net: Efficient Channel Attention for Deep CNN` - https://arxiv.org/abs/1910.03151v4
* Code: Added to ResNet base, ECA module contributed by @VRandme, reference https://github.com/BangguWu/ECANet
## Res2Net
* Implementation: [res2net.py](https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/res2net.py)
* Paper: `Res2Net: A New Multi-scale Backbone Architecture` - https://arxiv.org/abs/1904.01169
* Code: https://github.com/gasvn/Res2Net
## ResNeSt
* Implementation: [resnest.py](https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/resnest.py)
* Paper: `ResNeSt: Split-Attention Networks` - https://arxiv.org/abs/2004.08955
* Code: https://github.com/zhanghang1989/ResNeSt
## ReXNet
* Implementation: [rexnet.py](https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/rexnet.py)
* Paper: `ReXNet: Diminishing Representational Bottleneck on CNN` - https://arxiv.org/abs/2007.00992
* Code: https://github.com/clovaai/rexnet
## Selective-Kernel Networks
* Implementation: [sknet.py](https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/sknet.py)
* Paper: `Selective-Kernel Networks` - https://arxiv.org/abs/1903.06586
* Code: https://github.com/implus/SKNet, https://github.com/clovaai/assembled-cnn
## SelecSLS
* Implementation: [selecsls.py](https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/selecsls.py)
* Paper: `XNect: Real-time Multi-Person 3D Motion Capture with a Single RGB Camera` - https://arxiv.org/abs/1907.00837
* Code: https://github.com/mehtadushy/SelecSLS-Pytorch
## Squeeze-and-Excitation Networks
* Implementation: [senet.py](https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/senet.py)
NOTE: I am deprecating this version of the networks, the new ones are part of `resnet.py`
* Paper: `Squeeze-and-Excitation Networks` - https://arxiv.org/abs/1709.01507
* Code: https://github.com/Cadene/pretrained-models.pytorch
## TResNet
* Implementation: [tresnet.py](https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/tresnet.py)
* Paper: `TResNet: High Performance GPU-Dedicated Architecture` - https://arxiv.org/abs/2003.13630
* Code: https://github.com/mrT23/TResNet
## VGG
* Implementation: [vgg.py](https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/vgg.py)
* Paper: `Very Deep Convolutional Networks For Large-Scale Image Recognition` - https://arxiv.org/pdf/1409.1556.pdf
* Reference code: https://github.com/pytorch/vision/blob/master/torchvision/models/vgg.py
## Vision Transformer
* Implementation: [vision_transformer.py](https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/vision_transformer.py)
* Paper: `An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale` - https://arxiv.org/abs/2010.11929
* Reference code and pretrained weights: https://github.com/google-research/vision_transformer
## VovNet V2 and V1
* Implementation: [vovnet.py](https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/vovnet.py)
* Paper: `CenterMask : Real-Time Anchor-Free Instance Segmentation` - https://arxiv.org/abs/1911.06667
* Reference code: https://github.com/youngwanLEE/vovnet-detectron2
## Xception
* Implementation: [xception.py](https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/xception.py)
* Paper: `Xception: Deep Learning with Depthwise Separable Convolutions` - https://arxiv.org/abs/1610.02357
* Code: https://github.com/Cadene/pretrained-models.pytorch
## Xception (Modified Aligned, Gluon)
* Implementation: [gluon_xception.py](https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/gluon_xception.py)
* Paper: `Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation` - https://arxiv.org/abs/1802.02611
* Reference code: https://github.com/dmlc/gluon-cv/tree/master/gluoncv/model_zoo, https://github.com/jfzhang95/pytorch-deeplab-xception/
## Xception (Modified Aligned, TF)
* Implementation: [aligned_xception.py](https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/aligned_xception.py)
* Paper: `Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation` - https://arxiv.org/abs/1802.02611
* Reference code: https://github.com/tensorflow/models/tree/master/research/deeplab

165
hfdocs/source/models/adversarial-inception-v3.mdx
Unescape View File

@ -0,0 +1,165 @@
# Adversarial Inception v3
**Inception v3** is a convolutional neural network architecture from the Inception family that makes several improvements including using [Label Smoothing](https://paperswithcode.com/method/label-smoothing), Factorized 7 x 7 convolutions, and the use of an [auxiliary classifer](https://paperswithcode.com/method/auxiliary-classifier) to propagate label information lower down the network (along with the use of batch normalization for layers in the sidehead). The key building block is an [Inception Module](https://paperswithcode.com/method/inception-v3-module).
This particular model was trained for study of adversarial examples (adversarial training).
The weights from this model were ported from [Tensorflow/Models](https://github.com/tensorflow/models).
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('adv_inception_v3', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `adv_inception_v3`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('adv_inception_v3', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@article{DBLP:journals/corr/abs-1804-00097,
author = {Alexey Kurakin and
Ian J. Goodfellow and
Samy Bengio and
Yinpeng Dong and
Fangzhou Liao and
Ming Liang and
Tianyu Pang and
Jun Zhu and
Xiaolin Hu and
Cihang Xie and
Jianyu Wang and
Zhishuai Zhang and
Zhou Ren and
Alan L. Yuille and
Sangxia Huang and
Yao Zhao and
Yuzhe Zhao and
Zhonglin Han and
Junjiajia Long and
Yerkebulan Berdibekov and
Takuya Akiba and
Seiya Tokui and
Motoki Abe},
title = {Adversarial Attacks and Defences Competition},
journal = {CoRR},
volume = {abs/1804.00097},
year = {2018},
url = {http://arxiv.org/abs/1804.00097},
archivePrefix = {arXiv},
eprint = {1804.00097},
timestamp = {Thu, 31 Oct 2019 16:31:22 +0100},
biburl = {https://dblp.org/rec/journals/corr/abs-1804-00097.bib},
bibsource = {dblp computer science bibliography, https://dblp.org}
}
```
<!--
Type: model-index
Collections:
- Name: Adversarial Inception v3
Paper:
Title: Adversarial Attacks and Defences Competition
URL: https://paperswithcode.com/paper/adversarial-attacks-and-defences-competition
Models:
- Name: adv_inception_v3
In Collection: Adversarial Inception v3
Metadata:
FLOPs: 7352418880
Parameters: 23830000
File Size: 95549439
Architecture:
- 1x1 Convolution
- Auxiliary Classifier
- Average Pooling
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inception-v3 Module
- Max Pooling
- ReLU
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: adv_inception_v3
Crop Pct: '0.875'
Image Size: '299'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/inception_v3.py#L456
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/adv_inception_v3-9e27bd63.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 77.58%
Top 5 Accuracy: 93.74%
-->

524
hfdocs/source/models/advprop.mdx
Unescape View File

@ -0,0 +1,524 @@
# AdvProp (EfficientNet)
**AdvProp** is an adversarial training scheme which treats adversarial examples as additional examples, to prevent overfitting. Key to the method is the usage of a separate auxiliary batch norm for adversarial examples, as they have different underlying distributions to normal examples.
The weights from this model were ported from [Tensorflow/TPU](https://github.com/tensorflow/tpu).
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('tf_efficientnet_b0_ap', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `tf_efficientnet_b0_ap`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('tf_efficientnet_b0_ap', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{xie2020adversarial,
title={Adversarial Examples Improve Image Recognition},
author={Cihang Xie and Mingxing Tan and Boqing Gong and Jiang Wang and Alan Yuille and Quoc V. Le},
year={2020},
eprint={1911.09665},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
<!--
Type: model-index
Collections:
- Name: AdvProp
Paper:
Title: Adversarial Examples Improve Image Recognition
URL: https://paperswithcode.com/paper/adversarial-examples-improve-image
Models:
- Name: tf_efficientnet_b0_ap
In Collection: AdvProp
Metadata:
FLOPs: 488688572
Parameters: 5290000
File Size: 21385973
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Techniques:
- AdvProp
- AutoAugment
- Label Smoothing
- RMSProp
- Stochastic Depth
- Weight Decay
Training Data:
- ImageNet
ID: tf_efficientnet_b0_ap
LR: 0.256
Epochs: 350
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 2048
Image Size: '224'
Weight Decay: 1.0e-05
Interpolation: bicubic
RMSProp Decay: 0.9
Label Smoothing: 0.1
BatchNorm Momentum: 0.99
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1334
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_b0_ap-f262efe1.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 77.1%
Top 5 Accuracy: 93.26%
- Name: tf_efficientnet_b1_ap
In Collection: AdvProp
Metadata:
FLOPs: 883633200
Parameters: 7790000
File Size: 31515350
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Techniques:
- AdvProp
- AutoAugment
- Label Smoothing
- RMSProp
- Stochastic Depth
- Weight Decay
Training Data:
- ImageNet
ID: tf_efficientnet_b1_ap
LR: 0.256
Epochs: 350
Crop Pct: '0.882'
Momentum: 0.9
Batch Size: 2048
Image Size: '240'
Weight Decay: 1.0e-05
Interpolation: bicubic
RMSProp Decay: 0.9
Label Smoothing: 0.1
BatchNorm Momentum: 0.99
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1344
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_b1_ap-44ef0a3d.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.28%
Top 5 Accuracy: 94.3%
- Name: tf_efficientnet_b2_ap
In Collection: AdvProp
Metadata:
FLOPs: 1234321170
Parameters: 9110000
File Size: 36800745
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Techniques:
- AdvProp
- AutoAugment
- Label Smoothing
- RMSProp
- Stochastic Depth
- Weight Decay
Training Data:
- ImageNet
ID: tf_efficientnet_b2_ap
LR: 0.256
Epochs: 350
Crop Pct: '0.89'
Momentum: 0.9
Batch Size: 2048
Image Size: '260'
Weight Decay: 1.0e-05
Interpolation: bicubic
RMSProp Decay: 0.9
Label Smoothing: 0.1
BatchNorm Momentum: 0.99
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1354
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_b2_ap-2f8e7636.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 80.3%
Top 5 Accuracy: 95.03%
- Name: tf_efficientnet_b3_ap
In Collection: AdvProp
Metadata:
FLOPs: 2275247568
Parameters: 12230000
File Size: 49384538
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Techniques:
- AdvProp
- AutoAugment
- Label Smoothing
- RMSProp
- Stochastic Depth
- Weight Decay
Training Data:
- ImageNet
ID: tf_efficientnet_b3_ap
LR: 0.256
Epochs: 350
Crop Pct: '0.904'
Momentum: 0.9
Batch Size: 2048
Image Size: '300'
Weight Decay: 1.0e-05
Interpolation: bicubic
RMSProp Decay: 0.9
Label Smoothing: 0.1
BatchNorm Momentum: 0.99
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1364
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_b3_ap-aad25bdd.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 81.82%
Top 5 Accuracy: 95.62%
- Name: tf_efficientnet_b4_ap
In Collection: AdvProp
Metadata:
FLOPs: 5749638672
Parameters: 19340000
File Size: 77993585
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Techniques:
- AdvProp
- AutoAugment
- Label Smoothing
- RMSProp
- Stochastic Depth
- Weight Decay
Training Data:
- ImageNet
ID: tf_efficientnet_b4_ap
LR: 0.256
Epochs: 350
Crop Pct: '0.922'
Momentum: 0.9
Batch Size: 2048
Image Size: '380'
Weight Decay: 1.0e-05
Interpolation: bicubic
RMSProp Decay: 0.9
Label Smoothing: 0.1
BatchNorm Momentum: 0.99
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1374
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_b4_ap-dedb23e6.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 83.26%
Top 5 Accuracy: 96.39%
- Name: tf_efficientnet_b5_ap
In Collection: AdvProp
Metadata:
FLOPs: 13176501888
Parameters: 30390000
File Size: 122403150
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Techniques:
- AdvProp
- AutoAugment
- Label Smoothing
- RMSProp
- Stochastic Depth
- Weight Decay
Training Data:
- ImageNet
ID: tf_efficientnet_b5_ap
LR: 0.256
Epochs: 350
Crop Pct: '0.934'
Momentum: 0.9
Batch Size: 2048
Image Size: '456'
Weight Decay: 1.0e-05
Interpolation: bicubic
RMSProp Decay: 0.9
Label Smoothing: 0.1
BatchNorm Momentum: 0.99
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1384
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_b5_ap-9e82fae8.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 84.25%
Top 5 Accuracy: 96.97%
- Name: tf_efficientnet_b6_ap
In Collection: AdvProp
Metadata:
FLOPs: 24180518488
Parameters: 43040000
File Size: 173237466
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Techniques:
- AdvProp
- AutoAugment
- Label Smoothing
- RMSProp
- Stochastic Depth
- Weight Decay
Training Data:
- ImageNet
ID: tf_efficientnet_b6_ap
LR: 0.256
Epochs: 350
Crop Pct: '0.942'
Momentum: 0.9
Batch Size: 2048
Image Size: '528'
Weight Decay: 1.0e-05
Interpolation: bicubic
RMSProp Decay: 0.9
Label Smoothing: 0.1
BatchNorm Momentum: 0.99
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1394
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_b6_ap-4ffb161f.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 84.79%
Top 5 Accuracy: 97.14%
- Name: tf_efficientnet_b7_ap
In Collection: AdvProp
Metadata:
FLOPs: 48205304880
Parameters: 66349999
File Size: 266850607
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Techniques:
- AdvProp
- AutoAugment
- Label Smoothing
- RMSProp
- Stochastic Depth
- Weight Decay
Training Data:
- ImageNet
ID: tf_efficientnet_b7_ap
LR: 0.256
Epochs: 350
Crop Pct: '0.949'
Momentum: 0.9
Batch Size: 2048
Image Size: '600'
Weight Decay: 1.0e-05
Interpolation: bicubic
RMSProp Decay: 0.9
Label Smoothing: 0.1
BatchNorm Momentum: 0.99
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1405
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_b7_ap-ddb28fec.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 85.12%
Top 5 Accuracy: 97.25%
- Name: tf_efficientnet_b8_ap
In Collection: AdvProp
Metadata:
FLOPs: 80962956270
Parameters: 87410000
File Size: 351412563
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Techniques:
- AdvProp
- AutoAugment
- Label Smoothing
- RMSProp
- Stochastic Depth
- Weight Decay
Training Data:
- ImageNet
ID: tf_efficientnet_b8_ap
LR: 0.128
Epochs: 350
Crop Pct: '0.954'
Momentum: 0.9
Batch Size: 2048
Image Size: '672'
Weight Decay: 1.0e-05
Interpolation: bicubic
RMSProp Decay: 0.9
Label Smoothing: 0.1
BatchNorm Momentum: 0.99
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1416
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_b8_ap-00e169fa.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 85.37%
Top 5 Accuracy: 97.3%
-->

362
hfdocs/source/models/big-transfer.mdx
Unescape View File

@ -0,0 +1,362 @@
# Big Transfer (BiT)
**Big Transfer (BiT)** is a type of pretraining recipe that pre-trains on a large supervised source dataset, and fine-tunes the weights on the target task. Models are trained on the JFT-300M dataset. The finetuned models contained in this collection are finetuned on ImageNet.
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('resnetv2_101x1_bitm', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `resnetv2_101x1_bitm`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('resnetv2_101x1_bitm', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{kolesnikov2020big,
title={Big Transfer (BiT): General Visual Representation Learning},
author={Alexander Kolesnikov and Lucas Beyer and Xiaohua Zhai and Joan Puigcerver and Jessica Yung and Sylvain Gelly and Neil Houlsby},
year={2020},
eprint={1912.11370},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
<!--
Type: model-index
Collections:
- Name: Big Transfer
Paper:
Title: 'Big Transfer (BiT): General Visual Representation Learning'
URL: https://paperswithcode.com/paper/large-scale-learning-of-general-visual
Models:
- Name: resnetv2_101x1_bitm
In Collection: Big Transfer
Metadata:
FLOPs: 5330896
Parameters: 44540000
File Size: 178256468
Architecture:
- 1x1 Convolution
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Group Normalization
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
- Weight Standardization
Tasks:
- Image Classification
Training Techniques:
- Mixup
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
- JFT-300M
Training Resources: Cloud TPUv3-512
ID: resnetv2_101x1_bitm
LR: 0.03
Epochs: 90
Layers: 101
Crop Pct: '1.0'
Momentum: 0.9
Batch Size: 4096
Image Size: '480'
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/b9843f954b0457af2db4f9dea41a8538f51f5d78/timm/models/resnetv2.py#L444
Weights: https://storage.googleapis.com/bit_models/BiT-M-R101x1-ILSVRC2012.npz
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 82.21%
Top 5 Accuracy: 96.47%
- Name: resnetv2_101x3_bitm
In Collection: Big Transfer
Metadata:
FLOPs: 15988688
Parameters: 387930000
File Size: 1551830100
Architecture:
- 1x1 Convolution
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Group Normalization
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
- Weight Standardization
Tasks:
- Image Classification
Training Techniques:
- Mixup
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
- JFT-300M
Training Resources: Cloud TPUv3-512
ID: resnetv2_101x3_bitm
LR: 0.03
Epochs: 90
Layers: 101
Crop Pct: '1.0'
Momentum: 0.9
Batch Size: 4096
Image Size: '480'
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/b9843f954b0457af2db4f9dea41a8538f51f5d78/timm/models/resnetv2.py#L451
Weights: https://storage.googleapis.com/bit_models/BiT-M-R101x3-ILSVRC2012.npz
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 84.38%
Top 5 Accuracy: 97.37%
- Name: resnetv2_152x2_bitm
In Collection: Big Transfer
Metadata:
FLOPs: 10659792
Parameters: 236340000
File Size: 945476668
Architecture:
- 1x1 Convolution
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Group Normalization
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
- Weight Standardization
Tasks:
- Image Classification
Training Techniques:
- Mixup
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
- JFT-300M
ID: resnetv2_152x2_bitm
Crop Pct: '1.0'
Image Size: '480'
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/b9843f954b0457af2db4f9dea41a8538f51f5d78/timm/models/resnetv2.py#L458
Weights: https://storage.googleapis.com/bit_models/BiT-M-R152x2-ILSVRC2012.npz
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 84.4%
Top 5 Accuracy: 97.43%
- Name: resnetv2_152x4_bitm
In Collection: Big Transfer
Metadata:
FLOPs: 21317584
Parameters: 936530000
File Size: 3746270104
Architecture:
- 1x1 Convolution
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Group Normalization
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
- Weight Standardization
Tasks:
- Image Classification
Training Techniques:
- Mixup
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
- JFT-300M
Training Resources: Cloud TPUv3-512
ID: resnetv2_152x4_bitm
Crop Pct: '1.0'
Image Size: '480'
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/b9843f954b0457af2db4f9dea41a8538f51f5d78/timm/models/resnetv2.py#L465
Weights: https://storage.googleapis.com/bit_models/BiT-M-R152x4-ILSVRC2012.npz
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 84.95%
Top 5 Accuracy: 97.45%
- Name: resnetv2_50x1_bitm
In Collection: Big Transfer
Metadata:
FLOPs: 5330896
Parameters: 25550000
File Size: 102242668
Architecture:
- 1x1 Convolution
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Group Normalization
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
- Weight Standardization
Tasks:
- Image Classification
Training Techniques:
- Mixup
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
- JFT-300M
Training Resources: Cloud TPUv3-512
ID: resnetv2_50x1_bitm
LR: 0.03
Epochs: 90
Layers: 50
Crop Pct: '1.0'
Momentum: 0.9
Batch Size: 4096
Image Size: '480'
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/b9843f954b0457af2db4f9dea41a8538f51f5d78/timm/models/resnetv2.py#L430
Weights: https://storage.googleapis.com/bit_models/BiT-M-R50x1-ILSVRC2012.npz
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 80.19%
Top 5 Accuracy: 95.63%
- Name: resnetv2_50x3_bitm
In Collection: Big Transfer
Metadata:
FLOPs: 15988688
Parameters: 217320000
File Size: 869321580
Architecture:
- 1x1 Convolution
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Group Normalization
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
- Weight Standardization
Tasks:
- Image Classification
Training Techniques:
- Mixup
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
- JFT-300M
Training Resources: Cloud TPUv3-512
ID: resnetv2_50x3_bitm
LR: 0.03
Epochs: 90
Layers: 50
Crop Pct: '1.0'
Momentum: 0.9
Batch Size: 4096
Image Size: '480'
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/b9843f954b0457af2db4f9dea41a8538f51f5d78/timm/models/resnetv2.py#L437
Weights: https://storage.googleapis.com/bit_models/BiT-M-R50x3-ILSVRC2012.npz
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 83.75%
Top 5 Accuracy: 97.12%
-->

148
hfdocs/source/models/csp-darknet.mdx
Unescape View File

@ -0,0 +1,148 @@
# CSP-DarkNet
**CSPDarknet53** is a convolutional neural network and backbone for object detection that uses [DarkNet-53](https://paperswithcode.com/method/darknet-53). It employs a CSPNet strategy to partition the feature map of the base layer into two parts and then merges them through a cross-stage hierarchy. The use of a split and merge strategy allows for more gradient flow through the network.
This CNN is used as the backbone for [YOLOv4](https://paperswithcode.com/method/yolov4).
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('cspdarknet53', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `cspdarknet53`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('cspdarknet53', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{bochkovskiy2020yolov4,
title={YOLOv4: Optimal Speed and Accuracy of Object Detection},
author={Alexey Bochkovskiy and Chien-Yao Wang and Hong-Yuan Mark Liao},
year={2020},
eprint={2004.10934},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
<!--
Type: model-index
Collections:
- Name: CSP DarkNet
Paper:
Title: 'YOLOv4: Optimal Speed and Accuracy of Object Detection'
URL: https://paperswithcode.com/paper/yolov4-optimal-speed-and-accuracy-of-object
Models:
- Name: cspdarknet53
In Collection: CSP DarkNet
Metadata:
FLOPs: 8545018880
Parameters: 27640000
File Size: 110775135
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Global Average Pooling
- Mish
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Techniques:
- CutMix
- Label Smoothing
- Mosaic
- Polynomial Learning Rate Decay
- SGD with Momentum
- Self-Adversarial Training
- Weight Decay
Training Data:
- ImageNet
Training Resources: 1x NVIDIA RTX 2070 GPU
ID: cspdarknet53
LR: 0.1
Layers: 53
Crop Pct: '0.887'
Momentum: 0.9
Batch Size: 128
Image Size: '256'
Warmup Steps: 1000
Weight Decay: 0.0005
Interpolation: bilinear
Training Steps: 8000000
FPS (GPU RTX 2070): 66
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/cspnet.py#L441
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/cspdarknet53_ra_256-d05c7c21.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 80.05%
Top 5 Accuracy: 95.09%
-->

143
hfdocs/source/models/csp-resnet.mdx
Unescape View File

@ -0,0 +1,143 @@
# CSP-ResNet
**CSPResNet** is a convolutional neural network where we apply the Cross Stage Partial Network (CSPNet) approach to [ResNet](https://paperswithcode.com/method/resnet). The CSPNet partitions the feature map of the base layer into two parts and then merges them through a cross-stage hierarchy. The use of a split and merge strategy allows for more gradient flow through the network.
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('cspresnet50', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `cspresnet50`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('cspresnet50', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{wang2019cspnet,
title={CSPNet: A New Backbone that can Enhance Learning Capability of CNN},
author={Chien-Yao Wang and Hong-Yuan Mark Liao and I-Hau Yeh and Yueh-Hua Wu and Ping-Yang Chen and Jun-Wei Hsieh},
year={2019},
eprint={1911.11929},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
<!--
Type: model-index
Collections:
- Name: CSP ResNet
Paper:
Title: 'CSPNet: A New Backbone that can Enhance Learning Capability of CNN'
URL: https://paperswithcode.com/paper/cspnet-a-new-backbone-that-can-enhance
Models:
- Name: cspresnet50
In Collection: CSP ResNet
Metadata:
FLOPs: 5924992000
Parameters: 21620000
File Size: 86679303
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Techniques:
- Label Smoothing
- Polynomial Learning Rate Decay
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
ID: cspresnet50
LR: 0.1
Layers: 50
Crop Pct: '0.887'
Momentum: 0.9
Batch Size: 128
Image Size: '256'
Weight Decay: 0.005
Interpolation: bilinear
Training Steps: 8000000
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/cspnet.py#L415
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/cspresnet50_ra-d3e8d487.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.57%
Top 5 Accuracy: 94.71%
-->

144
hfdocs/source/models/csp-resnext.mdx
Unescape View File

@ -0,0 +1,144 @@
# CSP-ResNeXt
**CSPResNeXt** is a convolutional neural network where we apply the Cross Stage Partial Network (CSPNet) approach to [ResNeXt](https://paperswithcode.com/method/resnext). The CSPNet partitions the feature map of the base layer into two parts and then merges them through a cross-stage hierarchy. The use of a split and merge strategy allows for more gradient flow through the network.
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('cspresnext50', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `cspresnext50`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('cspresnext50', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{wang2019cspnet,
title={CSPNet: A New Backbone that can Enhance Learning Capability of CNN},
author={Chien-Yao Wang and Hong-Yuan Mark Liao and I-Hau Yeh and Yueh-Hua Wu and Ping-Yang Chen and Jun-Wei Hsieh},
year={2019},
eprint={1911.11929},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
<!--
Type: model-index
Collections:
- Name: CSP ResNeXt
Paper:
Title: 'CSPNet: A New Backbone that can Enhance Learning Capability of CNN'
URL: https://paperswithcode.com/paper/cspnet-a-new-backbone-that-can-enhance
Models:
- Name: cspresnext50
In Collection: CSP ResNeXt
Metadata:
FLOPs: 3962945536
Parameters: 20570000
File Size: 82562887
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Global Average Pooling
- Grouped Convolution
- Max Pooling
- ReLU
- ResNeXt Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Techniques:
- Label Smoothing
- Polynomial Learning Rate Decay
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 1x GPU
ID: cspresnext50
LR: 0.1
Layers: 50
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 128
Image Size: '224'
Weight Decay: 0.005
Interpolation: bilinear
Training Steps: 8000000
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/cspnet.py#L430
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/cspresnext50_ra_224-648b4713.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 80.05%
Top 5 Accuracy: 94.94%
-->

372
hfdocs/source/models/densenet.mdx
Unescape View File

@ -0,0 +1,372 @@
# DenseNet
**DenseNet** is a type of convolutional neural network that utilises dense connections between layers, through [Dense Blocks](http://www.paperswithcode.com/method/dense-block), where we connect *all layers* (with matching feature-map sizes) directly with each other. To preserve the feed-forward nature, each layer obtains additional inputs from all preceding layers and passes on its own feature-maps to all subsequent layers.
The **DenseNet Blur** variant in this collection by Ross Wightman employs [Blur Pooling](http://www.paperswithcode.com/method/blur-pooling)
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('densenet121', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `densenet121`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('densenet121', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@article{DBLP:journals/corr/HuangLW16a,
author = {Gao Huang and
Zhuang Liu and
Kilian Q. Weinberger},
title = {Densely Connected Convolutional Networks},
journal = {CoRR},
volume = {abs/1608.06993},
year = {2016},
url = {http://arxiv.org/abs/1608.06993},
archivePrefix = {arXiv},
eprint = {1608.06993},
timestamp = {Mon, 10 Sep 2018 15:49:32 +0200},
biburl = {https://dblp.org/rec/journals/corr/HuangLW16a.bib},
bibsource = {dblp computer science bibliography, https://dblp.org}
}
```
```
@misc{rw2019timm,
author = {Ross Wightman},
title = {PyTorch Image Models},
year = {2019},
publisher = {GitHub},
journal = {GitHub repository},
doi = {10.5281/zenodo.4414861},
howpublished = {\url{https://github.com/rwightman/pytorch-image-models}}
}
```
<!--
Type: model-index
Collections:
- Name: DenseNet
Paper:
Title: Densely Connected Convolutional Networks
URL: https://paperswithcode.com/paper/densely-connected-convolutional-networks
Models:
- Name: densenet121
In Collection: DenseNet
Metadata:
FLOPs: 3641843200
Parameters: 7980000
File Size: 32376726
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Block
- Dense Connections
- Dropout
- Max Pooling
- ReLU
- Softmax
Tasks:
- Image Classification
Training Techniques:
- Kaiming Initialization
- Nesterov Accelerated Gradient
- Weight Decay
Training Data:
- ImageNet
ID: densenet121
LR: 0.1
Epochs: 90
Layers: 121
Dropout: 0.2
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 256
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/densenet.py#L295
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/densenet121_ra-50efcf5c.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 75.56%
Top 5 Accuracy: 92.65%
- Name: densenet161
In Collection: DenseNet
Metadata:
FLOPs: 9931959264
Parameters: 28680000
File Size: 115730790
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Block
- Dense Connections
- Dropout
- Max Pooling
- ReLU
- Softmax
Tasks:
- Image Classification
Training Techniques:
- Kaiming Initialization
- Nesterov Accelerated Gradient
- Weight Decay
Training Data:
- ImageNet
ID: densenet161
LR: 0.1
Epochs: 90
Layers: 161
Dropout: 0.2
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 256
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/densenet.py#L347
Weights: https://download.pytorch.org/models/densenet161-8d451a50.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 77.36%
Top 5 Accuracy: 93.63%
- Name: densenet169
In Collection: DenseNet
Metadata:
FLOPs: 4316945792
Parameters: 14150000
File Size: 57365526
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Block
- Dense Connections
- Dropout
- Max Pooling
- ReLU
- Softmax
Tasks:
- Image Classification
Training Techniques:
- Kaiming Initialization
- Nesterov Accelerated Gradient
- Weight Decay
Training Data:
- ImageNet
ID: densenet169
LR: 0.1
Epochs: 90
Layers: 169
Dropout: 0.2
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 256
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/densenet.py#L327
Weights: https://download.pytorch.org/models/densenet169-b2777c0a.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 75.9%
Top 5 Accuracy: 93.02%
- Name: densenet201
In Collection: DenseNet
Metadata:
FLOPs: 5514321024
Parameters: 20010000
File Size: 81131730
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Block
- Dense Connections
- Dropout
- Max Pooling
- ReLU
- Softmax
Tasks:
- Image Classification
Training Techniques:
- Kaiming Initialization
- Nesterov Accelerated Gradient
- Weight Decay
Training Data:
- ImageNet
ID: densenet201
LR: 0.1
Epochs: 90
Layers: 201
Dropout: 0.2
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 256
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/densenet.py#L337
Weights: https://download.pytorch.org/models/densenet201-c1103571.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 77.29%
Top 5 Accuracy: 93.48%
- Name: densenetblur121d
In Collection: DenseNet
Metadata:
FLOPs: 3947812864
Parameters: 8000000
File Size: 32456500
Architecture:
- 1x1 Convolution
- Batch Normalization
- Blur Pooling
- Convolution
- Dense Block
- Dense Connections
- Dropout
- Max Pooling
- ReLU
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: densenetblur121d
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/densenet.py#L305
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/densenetblur121d_ra-100dcfbc.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 76.59%
Top 5 Accuracy: 93.2%
- Name: tv_densenet121
In Collection: DenseNet
Metadata:
FLOPs: 3641843200
Parameters: 7980000
File Size: 32342954
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Block
- Dense Connections
- Dropout
- Max Pooling
- ReLU
- Softmax
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
ID: tv_densenet121
LR: 0.1
Epochs: 90
Crop Pct: '0.875'
LR Gamma: 0.1
Momentum: 0.9
Batch Size: 32
Image Size: '224'
LR Step Size: 30
Weight Decay: 0.0001
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/densenet.py#L379
Weights: https://download.pytorch.org/models/densenet121-a639ec97.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 74.74%
Top 5 Accuracy: 92.15%
-->

612
hfdocs/source/models/dla.mdx
Unescape View File

@ -0,0 +1,612 @@
# Deep Layer Aggregation
Extending “shallow” skip connections, **Dense Layer Aggregation (DLA)** incorporates more depth and sharing. The authors introduce two structures for deep layer aggregation (DLA): iterative deep aggregation (IDA) and hierarchical deep aggregation (HDA). These structures are expressed through an architectural framework, independent of the choice of backbone, for compatibility with current and future networks.
IDA focuses on fusing resolutions and scales while HDA focuses on merging features from all modules and channels. IDA follows the base hierarchy to refine resolution and aggregate scale stage-bystage. HDA assembles its own hierarchy of tree-structured connections that cross and merge stages to aggregate different levels of representation.
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('dla102', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `dla102`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('dla102', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{yu2019deep,
title={Deep Layer Aggregation},
author={Fisher Yu and Dequan Wang and Evan Shelhamer and Trevor Darrell},
year={2019},
eprint={1707.06484},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
<!--
Type: model-index
Collections:
- Name: DLA
Paper:
Title: Deep Layer Aggregation
URL: https://paperswithcode.com/paper/deep-layer-aggregation
Models:
- Name: dla102
In Collection: DLA
Metadata:
FLOPs: 7192952808
Parameters: 33270000
File Size: 135290579
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- DLA Bottleneck Residual Block
- DLA Residual Block
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x GPUs
ID: dla102
LR: 0.1
Epochs: 120
Layers: 102
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 256
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/dla.py#L410
Weights: http://dl.yf.io/dla/models/imagenet/dla102-d94d9790.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 78.03%
Top 5 Accuracy: 93.95%
- Name: dla102x
In Collection: DLA
Metadata:
FLOPs: 5886821352
Parameters: 26310000
File Size: 107552695
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- DLA Bottleneck Residual Block
- DLA Residual Block
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x GPUs
ID: dla102x
LR: 0.1
Epochs: 120
Layers: 102
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 256
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/dla.py#L418
Weights: http://dl.yf.io/dla/models/imagenet/dla102x-ad62be81.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 78.51%
Top 5 Accuracy: 94.23%
- Name: dla102x2
In Collection: DLA
Metadata:
FLOPs: 9343847400
Parameters: 41280000
File Size: 167645295
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- DLA Bottleneck Residual Block
- DLA Residual Block
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x GPUs
ID: dla102x2
LR: 0.1
Epochs: 120
Layers: 102
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 256
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/dla.py#L426
Weights: http://dl.yf.io/dla/models/imagenet/dla102x2-262837b6.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.44%
Top 5 Accuracy: 94.65%
- Name: dla169
In Collection: DLA
Metadata:
FLOPs: 11598004200
Parameters: 53390000
File Size: 216547113
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- DLA Bottleneck Residual Block
- DLA Residual Block
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x GPUs
ID: dla169
LR: 0.1
Epochs: 120
Layers: 169
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 256
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/dla.py#L434
Weights: http://dl.yf.io/dla/models/imagenet/dla169-0914e092.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 78.69%
Top 5 Accuracy: 94.33%
- Name: dla34
In Collection: DLA
Metadata:
FLOPs: 3070105576
Parameters: 15740000
File Size: 63228658
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- DLA Bottleneck Residual Block
- DLA Residual Block
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
ID: dla34
LR: 0.1
Epochs: 120
Layers: 32
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 256
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/dla.py#L362
Weights: http://dl.yf.io/dla/models/imagenet/dla34-ba72cf86.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 74.62%
Top 5 Accuracy: 92.06%
- Name: dla46_c
In Collection: DLA
Metadata:
FLOPs: 583277288
Parameters: 1300000
File Size: 5307963
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- DLA Bottleneck Residual Block
- DLA Residual Block
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
ID: dla46_c
LR: 0.1
Epochs: 120
Layers: 46
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 256
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/dla.py#L369
Weights: http://dl.yf.io/dla/models/imagenet/dla46_c-2bfd52c3.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 64.87%
Top 5 Accuracy: 86.29%
- Name: dla46x_c
In Collection: DLA
Metadata:
FLOPs: 544052200
Parameters: 1070000
File Size: 4387641
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- DLA Bottleneck Residual Block
- DLA Residual Block
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
ID: dla46x_c
LR: 0.1
Epochs: 120
Layers: 46
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 256
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/dla.py#L378
Weights: http://dl.yf.io/dla/models/imagenet/dla46x_c-d761bae7.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 65.98%
Top 5 Accuracy: 86.99%
- Name: dla60
In Collection: DLA
Metadata:
FLOPs: 4256251880
Parameters: 22040000
File Size: 89560235
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- DLA Bottleneck Residual Block
- DLA Residual Block
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
ID: dla60
LR: 0.1
Epochs: 120
Layers: 60
Dropout: 0.2
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 256
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/dla.py#L394
Weights: http://dl.yf.io/dla/models/imagenet/dla60-24839fc4.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 77.04%
Top 5 Accuracy: 93.32%
- Name: dla60_res2net
In Collection: DLA
Metadata:
FLOPs: 4147578504
Parameters: 20850000
File Size: 84886593
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- DLA Bottleneck Residual Block
- DLA Residual Block
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
ID: dla60_res2net
Layers: 60
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/dla.py#L346
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-res2net/res2net_dla60_4s-d88db7f9.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 78.46%
Top 5 Accuracy: 94.21%
- Name: dla60_res2next
In Collection: DLA
Metadata:
FLOPs: 3485335272
Parameters: 17030000
File Size: 69639245
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- DLA Bottleneck Residual Block
- DLA Residual Block
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
ID: dla60_res2next
Layers: 60
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/dla.py#L354
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-res2net/res2next_dla60_4s-d327927b.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 78.44%
Top 5 Accuracy: 94.16%
- Name: dla60x
In Collection: DLA
Metadata:
FLOPs: 3544204264
Parameters: 17350000
File Size: 70883139
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- DLA Bottleneck Residual Block
- DLA Residual Block
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
ID: dla60x
LR: 0.1
Epochs: 120
Layers: 60
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 256
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/dla.py#L402
Weights: http://dl.yf.io/dla/models/imagenet/dla60x-d15cacda.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 78.25%
Top 5 Accuracy: 94.02%
- Name: dla60x_c
In Collection: DLA
Metadata:
FLOPs: 593325032
Parameters: 1320000
File Size: 5454396
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- DLA Bottleneck Residual Block
- DLA Residual Block
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
ID: dla60x_c
LR: 0.1
Epochs: 120
Layers: 60
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 256
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/dla.py#L386
Weights: http://dl.yf.io/dla/models/imagenet/dla60x_c-b870c45c.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 67.91%
Top 5 Accuracy: 88.42%
-->

323
hfdocs/source/models/dpn.mdx
Unescape View File

@ -0,0 +1,323 @@
# Dual Path Network (DPN)
A **Dual Path Network (DPN)** is a convolutional neural network which presents a new topology of connection paths internally. The intuition is that [ResNets](https://paperswithcode.com/method/resnet) enables feature re-usage while DenseNet enables new feature exploration, and both are important for learning good representations. To enjoy the benefits from both path topologies, Dual Path Networks share common features while maintaining the flexibility to explore new features through dual path architectures.
The principal building block is an [DPN Block](https://paperswithcode.com/method/dpn-block).
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('dpn107', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `dpn107`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('dpn107', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{chen2017dual,
title={Dual Path Networks},
author={Yunpeng Chen and Jianan Li and Huaxin Xiao and Xiaojie Jin and Shuicheng Yan and Jiashi Feng},
year={2017},
eprint={1707.01629},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
<!--
Type: model-index
Collections:
- Name: DPN
Paper:
Title: Dual Path Networks
URL: https://paperswithcode.com/paper/dual-path-networks
Models:
- Name: dpn107
In Collection: DPN
Metadata:
FLOPs: 23524280296
Parameters: 86920000
File Size: 348612331
Architecture:
- Batch Normalization
- Convolution
- DPN Block
- Dense Connections
- Global Average Pooling
- Max Pooling
- Softmax
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 40x K80 GPUs
ID: dpn107
LR: 0.316
Layers: 107
Crop Pct: '0.875'
Batch Size: 1280
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/dpn.py#L310
Weights: https://github.com/rwightman/pytorch-dpn-pretrained/releases/download/v0.1/dpn107_extra-1ac7121e2.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 80.16%
Top 5 Accuracy: 94.91%
- Name: dpn131
In Collection: DPN
Metadata:
FLOPs: 20586274792
Parameters: 79250000
File Size: 318016207
Architecture:
- Batch Normalization
- Convolution
- DPN Block
- Dense Connections
- Global Average Pooling
- Max Pooling
- Softmax
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 40x K80 GPUs
ID: dpn131
LR: 0.316
Layers: 131
Crop Pct: '0.875'
Batch Size: 960
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/dpn.py#L302
Weights: https://github.com/rwightman/pytorch-dpn-pretrained/releases/download/v0.1/dpn131-71dfe43e0.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.83%
Top 5 Accuracy: 94.71%
- Name: dpn68
In Collection: DPN
Metadata:
FLOPs: 2990567880
Parameters: 12610000
File Size: 50761994
Architecture:
- Batch Normalization
- Convolution
- DPN Block
- Dense Connections
- Global Average Pooling
- Max Pooling
- Softmax
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 40x K80 GPUs
ID: dpn68
LR: 0.316
Layers: 68
Crop Pct: '0.875'
Batch Size: 1280
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/dpn.py#L270
Weights: https://github.com/rwightman/pytorch-dpn-pretrained/releases/download/v0.1/dpn68-66bebafa7.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 76.31%
Top 5 Accuracy: 92.97%
- Name: dpn68b
In Collection: DPN
Metadata:
FLOPs: 2990567880
Parameters: 12610000
File Size: 50781025
Architecture:
- Batch Normalization
- Convolution
- DPN Block
- Dense Connections
- Global Average Pooling
- Max Pooling
- Softmax
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 40x K80 GPUs
ID: dpn68b
LR: 0.316
Layers: 68
Crop Pct: '0.875'
Batch Size: 1280
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/dpn.py#L278
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/dpn68b_ra-a31ca160.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.21%
Top 5 Accuracy: 94.42%
- Name: dpn92
In Collection: DPN
Metadata:
FLOPs: 8357659624
Parameters: 37670000
File Size: 151248422
Architecture:
- Batch Normalization
- Convolution
- DPN Block
- Dense Connections
- Global Average Pooling
- Max Pooling
- Softmax
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 40x K80 GPUs
ID: dpn92
LR: 0.316
Layers: 92
Crop Pct: '0.875'
Batch Size: 1280
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/dpn.py#L286
Weights: https://github.com/rwightman/pytorch-dpn-pretrained/releases/download/v0.1/dpn92_extra-b040e4a9b.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.99%
Top 5 Accuracy: 94.84%
- Name: dpn98
In Collection: DPN
Metadata:
FLOPs: 15003675112
Parameters: 61570000
File Size: 247021307
Architecture:
- Batch Normalization
- Convolution
- DPN Block
- Dense Connections
- Global Average Pooling
- Max Pooling
- Softmax
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 40x K80 GPUs
ID: dpn98
LR: 0.4
Layers: 98
Crop Pct: '0.875'
Batch Size: 1280
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/dpn.py#L294
Weights: https://github.com/rwightman/pytorch-dpn-pretrained/releases/download/v0.1/dpn98-5b90dec4d.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.65%
Top 5 Accuracy: 94.61%
-->

303
hfdocs/source/models/ecaresnet.mdx
Unescape View File

@ -0,0 +1,303 @@
# ECA-ResNet
An **ECA ResNet** is a variant on a [ResNet](https://paperswithcode.com/method/resnet) that utilises an [Efficient Channel Attention module](https://paperswithcode.com/method/efficient-channel-attention). Efficient Channel Attention is an architectural unit based on [squeeze-and-excitation blocks](https://paperswithcode.com/method/squeeze-and-excitation-block) that reduces model complexity without dimensionality reduction.
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('ecaresnet101d', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `ecaresnet101d`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('ecaresnet101d', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{wang2020ecanet,
title={ECA-Net: Efficient Channel Attention for Deep Convolutional Neural Networks},
author={Qilong Wang and Banggu Wu and Pengfei Zhu and Peihua Li and Wangmeng Zuo and Qinghua Hu},
year={2020},
eprint={1910.03151},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
<!--
Type: model-index
Collections:
- Name: ECAResNet
Paper:
Title: 'ECA-Net: Efficient Channel Attention for Deep Convolutional Neural Networks'
URL: https://paperswithcode.com/paper/eca-net-efficient-channel-attention-for-deep
Models:
- Name: ecaresnet101d
In Collection: ECAResNet
Metadata:
FLOPs: 10377193728
Parameters: 44570000
File Size: 178815067
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Efficient Channel Attention
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 4x RTX 2080Ti GPUs
ID: ecaresnet101d
LR: 0.1
Epochs: 100
Layers: 101
Crop Pct: '0.875'
Batch Size: 256
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/a7f95818e44b281137503bcf4b3e3e94d8ffa52f/timm/models/resnet.py#L1087
Weights: https://imvl-automl-sh.oss-cn-shanghai.aliyuncs.com/darts/hyperml/hyperml/job_45402/outputs/ECAResNet101D_281c5844.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 82.18%
Top 5 Accuracy: 96.06%
- Name: ecaresnet101d_pruned
In Collection: ECAResNet
Metadata:
FLOPs: 4463972081
Parameters: 24880000
File Size: 99852736
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Efficient Channel Attention
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
ID: ecaresnet101d_pruned
Layers: 101
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/a7f95818e44b281137503bcf4b3e3e94d8ffa52f/timm/models/resnet.py#L1097
Weights: https://imvl-automl-sh.oss-cn-shanghai.aliyuncs.com/darts/hyperml/hyperml/job_45610/outputs/ECAResNet101D_P_75a3370e.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 80.82%
Top 5 Accuracy: 95.64%
- Name: ecaresnet50d
In Collection: ECAResNet
Metadata:
FLOPs: 5591090432
Parameters: 25580000
File Size: 102579290
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Efficient Channel Attention
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 4x RTX 2080Ti GPUs
ID: ecaresnet50d
LR: 0.1
Epochs: 100
Layers: 50
Crop Pct: '0.875'
Batch Size: 256
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/a7f95818e44b281137503bcf4b3e3e94d8ffa52f/timm/models/resnet.py#L1045
Weights: https://imvl-automl-sh.oss-cn-shanghai.aliyuncs.com/darts/hyperml/hyperml/job_45402/outputs/ECAResNet50D_833caf58.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 80.61%
Top 5 Accuracy: 95.31%
- Name: ecaresnet50d_pruned
In Collection: ECAResNet
Metadata:
FLOPs: 3250730657
Parameters: 19940000
File Size: 79990436
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Efficient Channel Attention
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
ID: ecaresnet50d_pruned
Layers: 50
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/a7f95818e44b281137503bcf4b3e3e94d8ffa52f/timm/models/resnet.py#L1055
Weights: https://imvl-automl-sh.oss-cn-shanghai.aliyuncs.com/darts/hyperml/hyperml/job_45899/outputs/ECAResNet50D_P_9c67f710.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.71%
Top 5 Accuracy: 94.88%
- Name: ecaresnetlight
In Collection: ECAResNet
Metadata:
FLOPs: 5276118784
Parameters: 30160000
File Size: 120956612
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Efficient Channel Attention
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
ID: ecaresnetlight
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/a7f95818e44b281137503bcf4b3e3e94d8ffa52f/timm/models/resnet.py#L1077
Weights: https://imvl-automl-sh.oss-cn-shanghai.aliyuncs.com/darts/hyperml/hyperml/job_45402/outputs/ECAResNetLight_4f34b35b.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 80.46%
Top 5 Accuracy: 95.25%
-->

212
hfdocs/source/models/efficientnet-pruned.mdx
Unescape View File

@ -0,0 +1,212 @@
# EfficientNet (Knapsack Pruned)
**EfficientNet** is a convolutional neural network architecture and scaling method that uniformly scales all dimensions of depth/width/resolution using a *compound coefficient*. Unlike conventional practice that arbitrary scales these factors, the EfficientNet scaling method uniformly scales network width, depth, and resolution with a set of fixed scaling coefficients. For example, if we want to use $2^N$ times more computational resources, then we can simply increase the network depth by $\alpha ^ N$, width by $\beta ^ N$, and image size by $\gamma ^ N$, where $\alpha, \beta, \gamma$ are constant coefficients determined by a small grid search on the original small model. EfficientNet uses a compound coefficient $\phi$ to uniformly scales network width, depth, and resolution in a principled way.
The compound scaling method is justified by the intuition that if the input image is bigger, then the network needs more layers to increase the receptive field and more channels to capture more fine-grained patterns on the bigger image.
The base EfficientNet-B0 network is based on the inverted bottleneck residual blocks of [MobileNetV2](https://paperswithcode.com/method/mobilenetv2), in addition to [squeeze-and-excitation blocks](https://paperswithcode.com/method/squeeze-and-excitation-block).
This collection consists of pruned EfficientNet models.
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('efficientnet_b1_pruned', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `efficientnet_b1_pruned`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('efficientnet_b1_pruned', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{tan2020efficientnet,
title={EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks},
author={Mingxing Tan and Quoc V. Le},
year={2020},
eprint={1905.11946},
archivePrefix={arXiv},
primaryClass={cs.LG}
}
```
```
@misc{aflalo2020knapsack,
title={Knapsack Pruning with Inner Distillation},
author={Yonathan Aflalo and Asaf Noy and Ming Lin and Itamar Friedman and Lihi Zelnik},
year={2020},
eprint={2002.08258},
archivePrefix={arXiv},
primaryClass={cs.LG}
}
```
<!--
Type: model-index
Collections:
- Name: EfficientNet Pruned
Paper:
Title: Knapsack Pruning with Inner Distillation
URL: https://paperswithcode.com/paper/knapsack-pruning-with-inner-distillation
Models:
- Name: efficientnet_b1_pruned
In Collection: EfficientNet Pruned
Metadata:
FLOPs: 489653114
Parameters: 6330000
File Size: 25595162
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: efficientnet_b1_pruned
Crop Pct: '0.882'
Image Size: '240'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/a7f95818e44b281137503bcf4b3e3e94d8ffa52f/timm/models/efficientnet.py#L1208
Weights: https://imvl-automl-sh.oss-cn-shanghai.aliyuncs.com/darts/hyperml/hyperml/job_45403/outputs/effnetb1_pruned_9ebb3fe6.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 78.25%
Top 5 Accuracy: 93.84%
- Name: efficientnet_b2_pruned
In Collection: EfficientNet Pruned
Metadata:
FLOPs: 878133915
Parameters: 8310000
File Size: 33555005
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: efficientnet_b2_pruned
Crop Pct: '0.89'
Image Size: '260'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/a7f95818e44b281137503bcf4b3e3e94d8ffa52f/timm/models/efficientnet.py#L1219
Weights: https://imvl-automl-sh.oss-cn-shanghai.aliyuncs.com/darts/hyperml/hyperml/job_45403/outputs/effnetb2_pruned_203f55bc.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.91%
Top 5 Accuracy: 94.86%
- Name: efficientnet_b3_pruned
In Collection: EfficientNet Pruned
Metadata:
FLOPs: 1239590641
Parameters: 9860000
File Size: 39770812
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: efficientnet_b3_pruned
Crop Pct: '0.904'
Image Size: '300'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/a7f95818e44b281137503bcf4b3e3e94d8ffa52f/timm/models/efficientnet.py#L1230
Weights: https://imvl-automl-sh.oss-cn-shanghai.aliyuncs.com/darts/hyperml/hyperml/job_45403/outputs/effnetb3_pruned_5abcc29f.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 80.86%
Top 5 Accuracy: 95.24%
-->

392
hfdocs/source/models/efficientnet.mdx
Unescape View File

@ -0,0 +1,392 @@
# EfficientNet
**EfficientNet** is a convolutional neural network architecture and scaling method that uniformly scales all dimensions of depth/width/resolution using a *compound coefficient*. Unlike conventional practice that arbitrary scales these factors, the EfficientNet scaling method uniformly scales network width, depth, and resolution with a set of fixed scaling coefficients. For example, if we want to use $2^N$ times more computational resources, then we can simply increase the network depth by $\alpha ^ N$, width by $\beta ^ N$, and image size by $\gamma ^ N$, where $\alpha, \beta, \gamma$ are constant coefficients determined by a small grid search on the original small model. EfficientNet uses a compound coefficient $\phi$ to uniformly scales network width, depth, and resolution in a principled way.
The compound scaling method is justified by the intuition that if the input image is bigger, then the network needs more layers to increase the receptive field and more channels to capture more fine-grained patterns on the bigger image.
The base EfficientNet-B0 network is based on the inverted bottleneck residual blocks of [MobileNetV2](https://paperswithcode.com/method/mobilenetv2), in addition to [squeeze-and-excitation blocks](https://paperswithcode.com/method/squeeze-and-excitation-block).
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('efficientnet_b0', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `efficientnet_b0`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('efficientnet_b0', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{tan2020efficientnet,
title={EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks},
author={Mingxing Tan and Quoc V. Le},
year={2020},
eprint={1905.11946},
archivePrefix={arXiv},
primaryClass={cs.LG}
}
```
<!--
Type: model-index
Collections:
- Name: EfficientNet
Paper:
Title: 'EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks'
URL: https://paperswithcode.com/paper/efficientnet-rethinking-model-scaling-for
Models:
- Name: efficientnet_b0
In Collection: EfficientNet
Metadata:
FLOPs: 511241564
Parameters: 5290000
File Size: 21376743
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: efficientnet_b0
Layers: 18
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/a7f95818e44b281137503bcf4b3e3e94d8ffa52f/timm/models/efficientnet.py#L1002
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/efficientnet_b0_ra-3dd342df.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 77.71%
Top 5 Accuracy: 93.52%
- Name: efficientnet_b1
In Collection: EfficientNet
Metadata:
FLOPs: 909691920
Parameters: 7790000
File Size: 31502706
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: efficientnet_b1
Crop Pct: '0.875'
Image Size: '240'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/a7f95818e44b281137503bcf4b3e3e94d8ffa52f/timm/models/efficientnet.py#L1011
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/efficientnet_b1-533bc792.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 78.71%
Top 5 Accuracy: 94.15%
- Name: efficientnet_b2
In Collection: EfficientNet
Metadata:
FLOPs: 1265324514
Parameters: 9110000
File Size: 36788104
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: efficientnet_b2
Crop Pct: '0.875'
Image Size: '260'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/a7f95818e44b281137503bcf4b3e3e94d8ffa52f/timm/models/efficientnet.py#L1020
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/efficientnet_b2_ra-bcdf34b7.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 80.38%
Top 5 Accuracy: 95.08%
- Name: efficientnet_b2a
In Collection: EfficientNet
Metadata:
FLOPs: 1452041554
Parameters: 9110000
File Size: 49369973
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: efficientnet_b2a
Crop Pct: '1.0'
Image Size: '288'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/a7f95818e44b281137503bcf4b3e3e94d8ffa52f/timm/models/efficientnet.py#L1029
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/efficientnet_b3_ra2-cf984f9c.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 80.61%
Top 5 Accuracy: 95.32%
- Name: efficientnet_b3
In Collection: EfficientNet
Metadata:
FLOPs: 2327905920
Parameters: 12230000
File Size: 49369973
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: efficientnet_b3
Crop Pct: '0.904'
Image Size: '300'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/a7f95818e44b281137503bcf4b3e3e94d8ffa52f/timm/models/efficientnet.py#L1038
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/efficientnet_b3_ra2-cf984f9c.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 82.08%
Top 5 Accuracy: 96.03%
- Name: efficientnet_b3a
In Collection: EfficientNet
Metadata:
FLOPs: 2600628304
Parameters: 12230000
File Size: 49369973
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: efficientnet_b3a
Crop Pct: '1.0'
Image Size: '320'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/a7f95818e44b281137503bcf4b3e3e94d8ffa52f/timm/models/efficientnet.py#L1047
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/efficientnet_b3_ra2-cf984f9c.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 82.25%
Top 5 Accuracy: 96.11%
- Name: efficientnet_em
In Collection: EfficientNet
Metadata:
FLOPs: 3935516480
Parameters: 6900000
File Size: 27927309
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: efficientnet_em
Crop Pct: '0.882'
Image Size: '240'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/a7f95818e44b281137503bcf4b3e3e94d8ffa52f/timm/models/efficientnet.py#L1118
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/efficientnet_em_ra2-66250f76.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.26%
Top 5 Accuracy: 94.79%
- Name: efficientnet_es
In Collection: EfficientNet
Metadata:
FLOPs: 2317181824
Parameters: 5440000
File Size: 22003339
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: efficientnet_es
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/a7f95818e44b281137503bcf4b3e3e94d8ffa52f/timm/models/efficientnet.py#L1110
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/efficientnet_es_ra-f111e99c.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 78.09%
Top 5 Accuracy: 93.93%
- Name: efficientnet_lite0
In Collection: EfficientNet
Metadata:
FLOPs: 510605024
Parameters: 4650000
File Size: 18820005
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: efficientnet_lite0
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/a7f95818e44b281137503bcf4b3e3e94d8ffa52f/timm/models/efficientnet.py#L1163
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/efficientnet_lite0_ra-37913777.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 75.5%
Top 5 Accuracy: 92.51%
-->

165
hfdocs/source/models/ensemble-adversarial.mdx
Unescape View File

@ -0,0 +1,165 @@
# # Ensemble Adversarial Inception ResNet v2
**Inception-ResNet-v2** is a convolutional neural architecture that builds on the Inception family of architectures but incorporates [residual connections](https://paperswithcode.com/method/residual-connection) (replacing the filter concatenation stage of the Inception architecture).
This particular model was trained for study of adversarial examples (adversarial training).
The weights from this model were ported from [Tensorflow/Models](https://github.com/tensorflow/models).
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('ens_adv_inception_resnet_v2', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `ens_adv_inception_resnet_v2`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('ens_adv_inception_resnet_v2', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@article{DBLP:journals/corr/abs-1804-00097,
author = {Alexey Kurakin and
Ian J. Goodfellow and
Samy Bengio and
Yinpeng Dong and
Fangzhou Liao and
Ming Liang and
Tianyu Pang and
Jun Zhu and
Xiaolin Hu and
Cihang Xie and
Jianyu Wang and
Zhishuai Zhang and
Zhou Ren and
Alan L. Yuille and
Sangxia Huang and
Yao Zhao and
Yuzhe Zhao and
Zhonglin Han and
Junjiajia Long and
Yerkebulan Berdibekov and
Takuya Akiba and
Seiya Tokui and
Motoki Abe},
title = {Adversarial Attacks and Defences Competition},
journal = {CoRR},
volume = {abs/1804.00097},
year = {2018},
url = {http://arxiv.org/abs/1804.00097},
archivePrefix = {arXiv},
eprint = {1804.00097},
timestamp = {Thu, 31 Oct 2019 16:31:22 +0100},
biburl = {https://dblp.org/rec/journals/corr/abs-1804-00097.bib},
bibsource = {dblp computer science bibliography, https://dblp.org}
}
```
<!--
Type: model-index
Collections:
- Name: Ensemble Adversarial
Paper:
Title: Adversarial Attacks and Defences Competition
URL: https://paperswithcode.com/paper/adversarial-attacks-and-defences-competition
Models:
- Name: ens_adv_inception_resnet_v2
In Collection: Ensemble Adversarial
Metadata:
FLOPs: 16959133120
Parameters: 55850000
File Size: 223774238
Architecture:
- 1x1 Convolution
- Auxiliary Classifier
- Average Pooling
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inception-v3 Module
- Max Pooling
- ReLU
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: ens_adv_inception_resnet_v2
Crop Pct: '0.897'
Image Size: '299'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/inception_resnet_v2.py#L351
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/ens_adv_inception_resnet_v2-2592a550.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 1.0%
Top 5 Accuracy: 17.32%
-->

159
hfdocs/source/models/ese-vovnet.mdx
Unescape View File

@ -0,0 +1,159 @@
# ESE-VoVNet
**VoVNet** is a convolutional neural network that seeks to make [DenseNet](https://paperswithcode.com/method/densenet) more efficient by concatenating all features only once in the last feature map, which makes input size constant and enables enlarging new output channel.
Read about [one-shot aggregation here](https://paperswithcode.com/method/one-shot-aggregation).
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('ese_vovnet19b_dw', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `ese_vovnet19b_dw`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('ese_vovnet19b_dw', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{lee2019energy,
title={An Energy and GPU-Computation Efficient Backbone Network for Real-Time Object Detection},
author={Youngwan Lee and Joong-won Hwang and Sangrok Lee and Yuseok Bae and Jongyoul Park},
year={2019},
eprint={1904.09730},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
<!--
Type: model-index
Collections:
- Name: ESE VovNet
Paper:
Title: 'CenterMask : Real-Time Anchor-Free Instance Segmentation'
URL: https://paperswithcode.com/paper/centermask-real-time-anchor-free-instance-1
Models:
- Name: ese_vovnet19b_dw
In Collection: ESE VovNet
Metadata:
FLOPs: 1711959904
Parameters: 6540000
File Size: 26243175
Architecture:
- Batch Normalization
- Convolution
- Max Pooling
- One-Shot Aggregation
- ReLU
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: ese_vovnet19b_dw
Layers: 19
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/vovnet.py#L361
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/ese_vovnet19b_dw-a8741004.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 76.82%
Top 5 Accuracy: 93.28%
- Name: ese_vovnet39b
In Collection: ESE VovNet
Metadata:
FLOPs: 9089259008
Parameters: 24570000
File Size: 98397138
Architecture:
- Batch Normalization
- Convolution
- Max Pooling
- One-Shot Aggregation
- ReLU
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: ese_vovnet39b
Layers: 39
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/vovnet.py#L371
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/ese_vovnet39b-f912fe73.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.31%
Top 5 Accuracy: 94.72%
-->

143
hfdocs/source/models/fbnet.mdx
Unescape View File

@ -0,0 +1,143 @@
# FBNet
**FBNet** is a type of convolutional neural architectures discovered through [DNAS](https://paperswithcode.com/method/dnas) neural architecture search. It utilises a basic type of image model block inspired by [MobileNetv2](https://paperswithcode.com/method/mobilenetv2) that utilises depthwise convolutions and an inverted residual structure (see components).
The principal building block is the [FBNet Block](https://paperswithcode.com/method/fbnet-block).
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('fbnetc_100', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `fbnetc_100`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('fbnetc_100', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{wu2019fbnet,
title={FBNet: Hardware-Aware Efficient ConvNet Design via Differentiable Neural Architecture Search},
author={Bichen Wu and Xiaoliang Dai and Peizhao Zhang and Yanghan Wang and Fei Sun and Yiming Wu and Yuandong Tian and Peter Vajda and Yangqing Jia and Kurt Keutzer},
year={2019},
eprint={1812.03443},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
<!--
Type: model-index
Collections:
- Name: FBNet
Paper:
Title: 'FBNet: Hardware-Aware Efficient ConvNet Design via Differentiable Neural
Architecture Search'
URL: https://paperswithcode.com/paper/fbnet-hardware-aware-efficient-convnet-design
Models:
- Name: fbnetc_100
In Collection: FBNet
Metadata:
FLOPs: 508940064
Parameters: 5570000
File Size: 22525094
Architecture:
- 1x1 Convolution
- Convolution
- Dense Connections
- Dropout
- FBNet Block
- Global Average Pooling
- Softmax
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x GPUs
ID: fbnetc_100
LR: 0.1
Epochs: 360
Layers: 22
Dropout: 0.2
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 256
Image Size: '224'
Weight Decay: 0.0005
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L985
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/fbnetc_100-c345b898.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 75.12%
Top 5 Accuracy: 92.37%
-->

145
hfdocs/source/models/gloun-inception-v3.mdx
Unescape View File

@ -0,0 +1,145 @@
# (Gluon) Inception v3
**Inception v3** is a convolutional neural network architecture from the Inception family that makes several improvements including using [Label Smoothing](https://paperswithcode.com/method/label-smoothing), Factorized 7 x 7 convolutions, and the use of an [auxiliary classifer](https://paperswithcode.com/method/auxiliary-classifier) to propagate label information lower down the network (along with the use of batch normalization for layers in the sidehead). The key building block is an [Inception Module](https://paperswithcode.com/method/inception-v3-module).
The weights from this model were ported from [Gluon](https://cv.gluon.ai/model_zoo/classification.html).
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('gluon_inception_v3', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `gluon_inception_v3`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('gluon_inception_v3', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@article{DBLP:journals/corr/SzegedyVISW15,
author = {Christian Szegedy and
Vincent Vanhoucke and
Sergey Ioffe and
Jonathon Shlens and
Zbigniew Wojna},
title = {Rethinking the Inception Architecture for Computer Vision},
journal = {CoRR},
volume = {abs/1512.00567},
year = {2015},
url = {http://arxiv.org/abs/1512.00567},
archivePrefix = {arXiv},
eprint = {1512.00567},
timestamp = {Mon, 13 Aug 2018 16:49:07 +0200},
biburl = {https://dblp.org/rec/journals/corr/SzegedyVISW15.bib},
bibsource = {dblp computer science bibliography, https://dblp.org}
}
```
<!--
Type: model-index
Collections:
- Name: Gloun Inception v3
Paper:
Title: Rethinking the Inception Architecture for Computer Vision
URL: https://paperswithcode.com/paper/rethinking-the-inception-architecture-for
Models:
- Name: gluon_inception_v3
In Collection: Gloun Inception v3
Metadata:
FLOPs: 7352418880
Parameters: 23830000
File Size: 95567055
Architecture:
- 1x1 Convolution
- Auxiliary Classifier
- Average Pooling
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inception-v3 Module
- Max Pooling
- ReLU
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: gluon_inception_v3
Crop Pct: '0.875'
Image Size: '299'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/inception_v3.py#L464
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/gluon_inception_v3-9f746940.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 78.8%
Top 5 Accuracy: 94.38%
-->

571
hfdocs/source/models/gloun-resnet.mdx
Unescape View File

@ -0,0 +1,571 @@
# (Gluon) ResNet
**Residual Networks**, or **ResNets**, learn residual functions with reference to the layer inputs, instead of learning unreferenced functions. Instead of hoping each few stacked layers directly fit a desired underlying mapping, residual nets let these layers fit a residual mapping. They stack [residual blocks](https://paperswithcode.com/method/residual-block) ontop of each other to form network: e.g. a ResNet-50 has fifty layers using these blocks.
The weights from this model were ported from [Gluon](https://cv.gluon.ai/model_zoo/classification.html).
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('gluon_resnet101_v1b', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `gluon_resnet101_v1b`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('gluon_resnet101_v1b', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@article{DBLP:journals/corr/HeZRS15,
author = {Kaiming He and
Xiangyu Zhang and
Shaoqing Ren and
Jian Sun},
title = {Deep Residual Learning for Image Recognition},
journal = {CoRR},
volume = {abs/1512.03385},
year = {2015},
url = {http://arxiv.org/abs/1512.03385},
archivePrefix = {arXiv},
eprint = {1512.03385},
timestamp = {Wed, 17 Apr 2019 17:23:45 +0200},
biburl = {https://dblp.org/rec/journals/corr/HeZRS15.bib},
bibsource = {dblp computer science bibliography, https://dblp.org}
}
```
<!--
Type: model-index
Collections:
- Name: Gloun ResNet
Paper:
Title: Deep Residual Learning for Image Recognition
URL: https://paperswithcode.com/paper/deep-residual-learning-for-image-recognition
Models:
- Name: gluon_resnet101_v1b
In Collection: Gloun ResNet
Metadata:
FLOPs: 10068547584
Parameters: 44550000
File Size: 178723172
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: gluon_resnet101_v1b
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/gluon_resnet.py#L89
Weights: https://github.com/rwightman/pytorch-pretrained-gluonresnet/releases/download/v0.1/gluon_resnet101_v1b-3b017079.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.3%
Top 5 Accuracy: 94.53%
- Name: gluon_resnet101_v1c
In Collection: Gloun ResNet
Metadata:
FLOPs: 10376567296
Parameters: 44570000
File Size: 178802575
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: gluon_resnet101_v1c
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/gluon_resnet.py#L113
Weights: https://github.com/rwightman/pytorch-pretrained-gluonresnet/releases/download/v0.1/gluon_resnet101_v1c-1f26822a.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.53%
Top 5 Accuracy: 94.59%
- Name: gluon_resnet101_v1d
In Collection: Gloun ResNet
Metadata:
FLOPs: 10377018880
Parameters: 44570000
File Size: 178802755
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: gluon_resnet101_v1d
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/gluon_resnet.py#L138
Weights: https://github.com/rwightman/pytorch-pretrained-gluonresnet/releases/download/v0.1/gluon_resnet101_v1d-0f9c8644.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 80.4%
Top 5 Accuracy: 95.02%
- Name: gluon_resnet101_v1s
In Collection: Gloun ResNet
Metadata:
FLOPs: 11805511680
Parameters: 44670000
File Size: 179221777
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: gluon_resnet101_v1s
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/gluon_resnet.py#L166
Weights: https://github.com/rwightman/pytorch-pretrained-gluonresnet/releases/download/v0.1/gluon_resnet101_v1s-60fe0cc1.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 80.29%
Top 5 Accuracy: 95.16%
- Name: gluon_resnet152_v1b
In Collection: Gloun ResNet
Metadata:
FLOPs: 14857660416
Parameters: 60190000
File Size: 241534001
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: gluon_resnet152_v1b
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/gluon_resnet.py#L97
Weights: https://github.com/rwightman/pytorch-pretrained-gluonresnet/releases/download/v0.1/gluon_resnet152_v1b-c1edb0dd.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.69%
Top 5 Accuracy: 94.73%
- Name: gluon_resnet152_v1c
In Collection: Gloun ResNet
Metadata:
FLOPs: 15165680128
Parameters: 60210000
File Size: 241613404
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: gluon_resnet152_v1c
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/gluon_resnet.py#L121
Weights: https://github.com/rwightman/pytorch-pretrained-gluonresnet/releases/download/v0.1/gluon_resnet152_v1c-a3bb0b98.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.91%
Top 5 Accuracy: 94.85%
- Name: gluon_resnet152_v1d
In Collection: Gloun ResNet
Metadata:
FLOPs: 15166131712
Parameters: 60210000
File Size: 241613584
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: gluon_resnet152_v1d
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/gluon_resnet.py#L147
Weights: https://github.com/rwightman/pytorch-pretrained-gluonresnet/releases/download/v0.1/gluon_resnet152_v1d-bd354e12.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 80.48%
Top 5 Accuracy: 95.2%
- Name: gluon_resnet152_v1s
In Collection: Gloun ResNet
Metadata:
FLOPs: 16594624512
Parameters: 60320000
File Size: 242032606
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: gluon_resnet152_v1s
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/gluon_resnet.py#L175
Weights: https://github.com/rwightman/pytorch-pretrained-gluonresnet/releases/download/v0.1/gluon_resnet152_v1s-dcc41b81.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 81.02%
Top 5 Accuracy: 95.42%
- Name: gluon_resnet18_v1b
In Collection: Gloun ResNet
Metadata:
FLOPs: 2337073152
Parameters: 11690000
File Size: 46816736
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: gluon_resnet18_v1b
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/gluon_resnet.py#L65
Weights: https://github.com/rwightman/pytorch-pretrained-gluonresnet/releases/download/v0.1/gluon_resnet18_v1b-0757602b.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 70.84%
Top 5 Accuracy: 89.76%
- Name: gluon_resnet34_v1b
In Collection: Gloun ResNet
Metadata:
FLOPs: 4718469120
Parameters: 21800000
File Size: 87295112
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: gluon_resnet34_v1b
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/gluon_resnet.py#L73
Weights: https://github.com/rwightman/pytorch-pretrained-gluonresnet/releases/download/v0.1/gluon_resnet34_v1b-c6d82d59.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 74.59%
Top 5 Accuracy: 92.0%
- Name: gluon_resnet50_v1b
In Collection: Gloun ResNet
Metadata:
FLOPs: 5282531328
Parameters: 25560000
File Size: 102493763
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: gluon_resnet50_v1b
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/gluon_resnet.py#L81
Weights: https://github.com/rwightman/pytorch-pretrained-gluonresnet/releases/download/v0.1/gluon_resnet50_v1b-0ebe02e2.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 77.58%
Top 5 Accuracy: 93.72%
- Name: gluon_resnet50_v1c
In Collection: Gloun ResNet
Metadata:
FLOPs: 5590551040
Parameters: 25580000
File Size: 102573166
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: gluon_resnet50_v1c
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/gluon_resnet.py#L105
Weights: https://github.com/rwightman/pytorch-pretrained-gluonresnet/releases/download/v0.1/gluon_resnet50_v1c-48092f55.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 78.01%
Top 5 Accuracy: 93.99%
- Name: gluon_resnet50_v1d
In Collection: Gloun ResNet
Metadata:
FLOPs: 5591002624
Parameters: 25580000
File Size: 102573346
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: gluon_resnet50_v1d
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/gluon_resnet.py#L129
Weights: https://github.com/rwightman/pytorch-pretrained-gluonresnet/releases/download/v0.1/gluon_resnet50_v1d-818a1b1b.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.06%
Top 5 Accuracy: 94.46%
- Name: gluon_resnet50_v1s
In Collection: Gloun ResNet
Metadata:
FLOPs: 7019495424
Parameters: 25680000
File Size: 102992368
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: gluon_resnet50_v1s
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/gluon_resnet.py#L156
Weights: https://github.com/rwightman/pytorch-pretrained-gluonresnet/releases/download/v0.1/gluon_resnet50_v1s-1762acc0.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 78.7%
Top 5 Accuracy: 94.25%
-->

209
hfdocs/source/models/gloun-resnext.mdx
Unescape View File

@ -0,0 +1,209 @@
# (Gluon) ResNeXt
A **ResNeXt** repeats a [building block](https://paperswithcode.com/method/resnext-block) that aggregates a set of transformations with the same topology. Compared to a [ResNet](https://paperswithcode.com/method/resnet), it exposes a new dimension, *cardinality* (the size of the set of transformations) $C$, as an essential factor in addition to the dimensions of depth and width.
The weights from this model were ported from [Gluon](https://cv.gluon.ai/model_zoo/classification.html).
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('gluon_resnext101_32x4d', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `gluon_resnext101_32x4d`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('gluon_resnext101_32x4d', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@article{DBLP:journals/corr/XieGDTH16,
author = {Saining Xie and
Ross B. Girshick and
Piotr Doll{\'{a}}r and
Zhuowen Tu and
Kaiming He},
title = {Aggregated Residual Transformations for Deep Neural Networks},
journal = {CoRR},
volume = {abs/1611.05431},
year = {2016},
url = {http://arxiv.org/abs/1611.05431},
archivePrefix = {arXiv},
eprint = {1611.05431},
timestamp = {Mon, 13 Aug 2018 16:45:58 +0200},
biburl = {https://dblp.org/rec/journals/corr/XieGDTH16.bib},
bibsource = {dblp computer science bibliography, https://dblp.org}
}
```
<!--
Type: model-index
Collections:
- Name: Gloun ResNeXt
Paper:
Title: Aggregated Residual Transformations for Deep Neural Networks
URL: https://paperswithcode.com/paper/aggregated-residual-transformations-for-deep
Models:
- Name: gluon_resnext101_32x4d
In Collection: Gloun ResNeXt
Metadata:
FLOPs: 10298145792
Parameters: 44180000
File Size: 177367414
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Global Average Pooling
- Grouped Convolution
- Max Pooling
- ReLU
- ResNeXt Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: gluon_resnext101_32x4d
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/gluon_resnet.py#L193
Weights: https://github.com/rwightman/pytorch-pretrained-gluonresnet/releases/download/v0.1/gluon_resnext101_32x4d-b253c8c4.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 80.33%
Top 5 Accuracy: 94.91%
- Name: gluon_resnext101_64x4d
In Collection: Gloun ResNeXt
Metadata:
FLOPs: 19954172928
Parameters: 83460000
File Size: 334737852
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Global Average Pooling
- Grouped Convolution
- Max Pooling
- ReLU
- ResNeXt Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: gluon_resnext101_64x4d
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/gluon_resnet.py#L201
Weights: https://github.com/rwightman/pytorch-pretrained-gluonresnet/releases/download/v0.1/gluon_resnext101_64x4d-f9a8e184.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 80.63%
Top 5 Accuracy: 95.0%
- Name: gluon_resnext50_32x4d
In Collection: Gloun ResNeXt
Metadata:
FLOPs: 5472648192
Parameters: 25030000
File Size: 100441719
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Global Average Pooling
- Grouped Convolution
- Max Pooling
- ReLU
- ResNeXt Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: gluon_resnext50_32x4d
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/gluon_resnet.py#L185
Weights: https://github.com/rwightman/pytorch-pretrained-gluonresnet/releases/download/v0.1/gluon_resnext50_32x4d-e6a097c1.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.35%
Top 5 Accuracy: 94.42%
-->

130
hfdocs/source/models/gloun-senet.mdx
Unescape View File

@ -0,0 +1,130 @@
# (Gluon) SENet
A **SENet** is a convolutional neural network architecture that employs [squeeze-and-excitation blocks](https://paperswithcode.com/method/squeeze-and-excitation-block) to enable the network to perform dynamic channel-wise feature recalibration.
The weights from this model were ported from [Gluon](https://cv.gluon.ai/model_zoo/classification.html).
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('gluon_senet154', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `gluon_senet154`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('gluon_senet154', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{hu2019squeezeandexcitation,
title={Squeeze-and-Excitation Networks},
author={Jie Hu and Li Shen and Samuel Albanie and Gang Sun and Enhua Wu},
year={2019},
eprint={1709.01507},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
<!--
Type: model-index
Collections:
- Name: Gloun SENet
Paper:
Title: Squeeze-and-Excitation Networks
URL: https://paperswithcode.com/paper/squeeze-and-excitation-networks
Models:
- Name: gluon_senet154
In Collection: Gloun SENet
Metadata:
FLOPs: 26681705136
Parameters: 115090000
File Size: 461546622
Architecture:
- Convolution
- Dense Connections
- Global Average Pooling
- Max Pooling
- Softmax
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: gluon_senet154
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/gluon_resnet.py#L239
Weights: https://github.com/rwightman/pytorch-pretrained-gluonresnet/releases/download/v0.1/gluon_senet154-70a1a3c0.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 81.23%
Top 5 Accuracy: 95.35%
-->

203
hfdocs/source/models/gloun-seresnext.mdx
Unescape View File

@ -0,0 +1,203 @@
# (Gluon) SE-ResNeXt
**SE ResNeXt** is a variant of a [ResNext](https://www.paperswithcode.com/method/resnext) that employs [squeeze-and-excitation blocks](https://paperswithcode.com/method/squeeze-and-excitation-block) to enable the network to perform dynamic channel-wise feature recalibration.
The weights from this model were ported from [Gluon](https://cv.gluon.ai/model_zoo/classification.html).
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('gluon_seresnext101_32x4d', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `gluon_seresnext101_32x4d`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('gluon_seresnext101_32x4d', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{hu2019squeezeandexcitation,
title={Squeeze-and-Excitation Networks},
author={Jie Hu and Li Shen and Samuel Albanie and Gang Sun and Enhua Wu},
year={2019},
eprint={1709.01507},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
<!--
Type: model-index
Collections:
- Name: Gloun SEResNeXt
Paper:
Title: Squeeze-and-Excitation Networks
URL: https://paperswithcode.com/paper/squeeze-and-excitation-networks
Models:
- Name: gluon_seresnext101_32x4d
In Collection: Gloun SEResNeXt
Metadata:
FLOPs: 10302923504
Parameters: 48960000
File Size: 196505510
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Global Average Pooling
- Grouped Convolution
- Max Pooling
- ReLU
- ResNeXt Block
- Residual Connection
- Softmax
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: gluon_seresnext101_32x4d
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/gluon_resnet.py#L219
Weights: https://github.com/rwightman/pytorch-pretrained-gluonresnet/releases/download/v0.1/gluon_seresnext101_32x4d-cf52900d.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 80.87%
Top 5 Accuracy: 95.29%
- Name: gluon_seresnext101_64x4d
In Collection: Gloun SEResNeXt
Metadata:
FLOPs: 19958950640
Parameters: 88230000
File Size: 353875948
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Global Average Pooling
- Grouped Convolution
- Max Pooling
- ReLU
- ResNeXt Block
- Residual Connection
- Softmax
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: gluon_seresnext101_64x4d
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/gluon_resnet.py#L229
Weights: https://github.com/rwightman/pytorch-pretrained-gluonresnet/releases/download/v0.1/gluon_seresnext101_64x4d-f9926f93.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 80.88%
Top 5 Accuracy: 95.31%
- Name: gluon_seresnext50_32x4d
In Collection: Gloun SEResNeXt
Metadata:
FLOPs: 5475179184
Parameters: 27560000
File Size: 110578827
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Global Average Pooling
- Grouped Convolution
- Max Pooling
- ReLU
- ResNeXt Block
- Residual Connection
- Softmax
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: gluon_seresnext50_32x4d
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/gluon_resnet.py#L209
Weights: https://github.com/rwightman/pytorch-pretrained-gluonresnet/releases/download/v0.1/gluon_seresnext50_32x4d-90cf2d6e.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.92%
Top 5 Accuracy: 94.82%
-->

133
hfdocs/source/models/gloun-xception.mdx
Unescape View File

@ -0,0 +1,133 @@
# (Gluon) Xception
**Xception** is a convolutional neural network architecture that relies solely on [depthwise separable convolution](https://paperswithcode.com/method/depthwise-separable-convolution) layers.
The weights from this model were ported from [Gluon](https://cv.gluon.ai/model_zoo/classification.html).
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('gluon_xception65', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `gluon_xception65`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('gluon_xception65', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{chollet2017xception,
title={Xception: Deep Learning with Depthwise Separable Convolutions},
author={François Chollet},
year={2017},
eprint={1610.02357},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
<!--
Type: model-index
Collections:
- Name: Gloun Xception
Paper:
Title: 'Xception: Deep Learning with Depthwise Separable Convolutions'
URL: https://paperswithcode.com/paper/xception-deep-learning-with-depthwise
Models:
- Name: gluon_xception65
In Collection: Gloun Xception
Metadata:
FLOPs: 17594889728
Parameters: 39920000
File Size: 160551306
Architecture:
- 1x1 Convolution
- Convolution
- Dense Connections
- Depthwise Separable Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: gluon_xception65
Crop Pct: '0.903'
Image Size: '299'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/gluon_xception.py#L241
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/gluon_xception-7015a15c.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.7%
Top 5 Accuracy: 94.87%
-->

425
hfdocs/source/models/hrnet.mdx
Unescape View File

@ -0,0 +1,425 @@
# HRNet
**HRNet**, or **High-Resolution Net**, is a general purpose convolutional neural network for tasks like semantic segmentation, object detection and image classification. It is able to maintain high resolution representations through the whole process. We start from a high-resolution convolution stream, gradually add high-to-low resolution convolution streams one by one, and connect the multi-resolution streams in parallel. The resulting network consists of several ($4$ in the paper) stages and the $n$th stage contains $n$ streams corresponding to $n$ resolutions. The authors conduct repeated multi-resolution fusions by exchanging the information across the parallel streams over and over.
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('hrnet_w18', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `hrnet_w18`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('hrnet_w18', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{sun2019highresolution,
title={High-Resolution Representations for Labeling Pixels and Regions},
author={Ke Sun and Yang Zhao and Borui Jiang and Tianheng Cheng and Bin Xiao and Dong Liu and Yadong Mu and Xinggang Wang and Wenyu Liu and Jingdong Wang},
year={2019},
eprint={1904.04514},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
<!--
Type: model-index
Collections:
- Name: HRNet
Paper:
Title: Deep High-Resolution Representation Learning for Visual Recognition
URL: https://paperswithcode.com/paper/190807919
Models:
- Name: hrnet_w18
In Collection: HRNet
Metadata:
FLOPs: 5547205500
Parameters: 21300000
File Size: 85718883
Architecture:
- Batch Normalization
- Convolution
- ReLU
- Residual Connection
Tasks:
- Image Classification
Training Techniques:
- Nesterov Accelerated Gradient
- Weight Decay
Training Data:
- ImageNet
Training Resources: 4x NVIDIA V100 GPUs
ID: hrnet_w18
Epochs: 100
Layers: 18
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 256
Image Size: '224'
Weight Decay: 0.001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/hrnet.py#L800
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-hrnet/hrnetv2_w18-8cb57bb9.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 76.76%
Top 5 Accuracy: 93.44%
- Name: hrnet_w18_small
In Collection: HRNet
Metadata:
FLOPs: 2071651488
Parameters: 13190000
File Size: 52934302
Architecture:
- Batch Normalization
- Convolution
- ReLU
- Residual Connection
Tasks:
- Image Classification
Training Techniques:
- Nesterov Accelerated Gradient
- Weight Decay
Training Data:
- ImageNet
Training Resources: 4x NVIDIA V100 GPUs
ID: hrnet_w18_small
Epochs: 100
Layers: 18
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 256
Image Size: '224'
Weight Decay: 0.001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/hrnet.py#L790
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-hrnet/hrnet_w18_small_v1-f460c6bc.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 72.34%
Top 5 Accuracy: 90.68%
- Name: hrnet_w18_small_v2
In Collection: HRNet
Metadata:
FLOPs: 3360023160
Parameters: 15600000
File Size: 62682879
Architecture:
- Batch Normalization
- Convolution
- ReLU
- Residual Connection
Tasks:
- Image Classification
Training Techniques:
- Nesterov Accelerated Gradient
- Weight Decay
Training Data:
- ImageNet
Training Resources: 4x NVIDIA V100 GPUs
ID: hrnet_w18_small_v2
Epochs: 100
Layers: 18
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 256
Image Size: '224'
Weight Decay: 0.001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/hrnet.py#L795
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-hrnet/hrnet_w18_small_v2-4c50a8cb.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 75.11%
Top 5 Accuracy: 92.41%
- Name: hrnet_w30
In Collection: HRNet
Metadata:
FLOPs: 10474119492
Parameters: 37710000
File Size: 151452218
Architecture:
- Batch Normalization
- Convolution
- ReLU
- Residual Connection
Tasks:
- Image Classification
Training Techniques:
- Nesterov Accelerated Gradient
- Weight Decay
Training Data:
- ImageNet
Training Resources: 4x NVIDIA V100 GPUs
ID: hrnet_w30
Epochs: 100
Layers: 30
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 256
Image Size: '224'
Weight Decay: 0.001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/hrnet.py#L805
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-hrnet/hrnetv2_w30-8d7f8dab.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 78.21%
Top 5 Accuracy: 94.22%
- Name: hrnet_w32
In Collection: HRNet
Metadata:
FLOPs: 11524528320
Parameters: 41230000
File Size: 165547812
Architecture:
- Batch Normalization
- Convolution
- ReLU
- Residual Connection
Tasks:
- Image Classification
Training Techniques:
- Nesterov Accelerated Gradient
- Weight Decay
Training Data:
- ImageNet
Training Resources: 4x NVIDIA V100 GPUs
Training Time: 60 hours
ID: hrnet_w32
Epochs: 100
Layers: 32
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 256
Image Size: '224'
Weight Decay: 0.001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/hrnet.py#L810
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-hrnet/hrnetv2_w32-90d8c5fb.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 78.45%
Top 5 Accuracy: 94.19%
- Name: hrnet_w40
In Collection: HRNet
Metadata:
FLOPs: 16381182192
Parameters: 57560000
File Size: 230899236
Architecture:
- Batch Normalization
- Convolution
- ReLU
- Residual Connection
Tasks:
- Image Classification
Training Techniques:
- Nesterov Accelerated Gradient
- Weight Decay
Training Data:
- ImageNet
Training Resources: 4x NVIDIA V100 GPUs
ID: hrnet_w40
Epochs: 100
Layers: 40
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 256
Image Size: '224'
Weight Decay: 0.001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/hrnet.py#L815
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-hrnet/hrnetv2_w40-7cd397a4.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 78.93%
Top 5 Accuracy: 94.48%
- Name: hrnet_w44
In Collection: HRNet
Metadata:
FLOPs: 19202520264
Parameters: 67060000
File Size: 268957432
Architecture:
- Batch Normalization
- Convolution
- ReLU
- Residual Connection
Tasks:
- Image Classification
Training Techniques:
- Nesterov Accelerated Gradient
- Weight Decay
Training Data:
- ImageNet
Training Resources: 4x NVIDIA V100 GPUs
ID: hrnet_w44
Epochs: 100
Layers: 44
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 256
Image Size: '224'
Weight Decay: 0.001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/hrnet.py#L820
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-hrnet/hrnetv2_w44-c9ac8c18.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 78.89%
Top 5 Accuracy: 94.37%
- Name: hrnet_w48
In Collection: HRNet
Metadata:
FLOPs: 22285865760
Parameters: 77470000
File Size: 310603710
Architecture:
- Batch Normalization
- Convolution
- ReLU
- Residual Connection
Tasks:
- Image Classification
Training Techniques:
- Nesterov Accelerated Gradient
- Weight Decay
Training Data:
- ImageNet
Training Resources: 4x NVIDIA V100 GPUs
Training Time: 80 hours
ID: hrnet_w48
Epochs: 100
Layers: 48
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 256
Image Size: '224'
Weight Decay: 0.001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/hrnet.py#L825
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-hrnet/hrnetv2_w48-abd2e6ab.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.32%
Top 5 Accuracy: 94.51%
- Name: hrnet_w64
In Collection: HRNet
Metadata:
FLOPs: 37239321984
Parameters: 128060000
File Size: 513071818
Architecture:
- Batch Normalization
- Convolution
- ReLU
- Residual Connection
Tasks:
- Image Classification
Training Techniques:
- Nesterov Accelerated Gradient
- Weight Decay
Training Data:
- ImageNet
Training Resources: 4x NVIDIA V100 GPUs
ID: hrnet_w64
Epochs: 100
Layers: 64
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 256
Image Size: '224'
Weight Decay: 0.001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/hrnet.py#L830
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-hrnet/hrnetv2_w64-b47cc881.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.46%
Top 5 Accuracy: 94.65%
-->

276
hfdocs/source/models/ig-resnext.mdx
Unescape View File

@ -0,0 +1,276 @@
# Instagram ResNeXt WSL
A **ResNeXt** repeats a [building block](https://paperswithcode.com/method/resnext-block) that aggregates a set of transformations with the same topology. Compared to a [ResNet](https://paperswithcode.com/method/resnet), it exposes a new dimension, *cardinality* (the size of the set of transformations) $C$, as an essential factor in addition to the dimensions of depth and width.
This model was trained on billions of Instagram images using thousands of distinct hashtags as labels exhibit excellent transfer learning performance.
Please note the CC-BY-NC 4.0 license on theses weights, non-commercial use only.
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('ig_resnext101_32x16d', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `ig_resnext101_32x16d`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('ig_resnext101_32x16d', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{mahajan2018exploring,
title={Exploring the Limits of Weakly Supervised Pretraining},
author={Dhruv Mahajan and Ross Girshick and Vignesh Ramanathan and Kaiming He and Manohar Paluri and Yixuan Li and Ashwin Bharambe and Laurens van der Maaten},
year={2018},
eprint={1805.00932},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
<!--
Type: model-index
Collections:
- Name: IG ResNeXt
Paper:
Title: Exploring the Limits of Weakly Supervised Pretraining
URL: https://paperswithcode.com/paper/exploring-the-limits-of-weakly-supervised
Models:
- Name: ig_resnext101_32x16d
In Collection: IG ResNeXt
Metadata:
FLOPs: 46623691776
Parameters: 194030000
File Size: 777518664
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Global Average Pooling
- Grouped Convolution
- Max Pooling
- ReLU
- ResNeXt Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Techniques:
- Nesterov Accelerated Gradient
- Weight Decay
Training Data:
- IG-3.5B-17k
- ImageNet
Training Resources: 336x GPUs
ID: ig_resnext101_32x16d
Epochs: 100
Layers: 101
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 8064
Image Size: '224'
Weight Decay: 0.001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/resnet.py#L874
Weights: https://download.pytorch.org/models/ig_resnext101_32x16-c6f796b0.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 84.16%
Top 5 Accuracy: 97.19%
- Name: ig_resnext101_32x32d
In Collection: IG ResNeXt
Metadata:
FLOPs: 112225170432
Parameters: 468530000
File Size: 1876573776
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Global Average Pooling
- Grouped Convolution
- Max Pooling
- ReLU
- ResNeXt Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Techniques:
- Nesterov Accelerated Gradient
- Weight Decay
Training Data:
- IG-3.5B-17k
- ImageNet
Training Resources: 336x GPUs
ID: ig_resnext101_32x32d
Epochs: 100
Layers: 101
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 8064
Image Size: '224'
Weight Decay: 0.001
Interpolation: bilinear
Minibatch Size: 8064
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/resnet.py#L885
Weights: https://download.pytorch.org/models/ig_resnext101_32x32-e4b90b00.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 85.09%
Top 5 Accuracy: 97.44%
- Name: ig_resnext101_32x48d
In Collection: IG ResNeXt
Metadata:
FLOPs: 197446554624
Parameters: 828410000
File Size: 3317136976
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Global Average Pooling
- Grouped Convolution
- Max Pooling
- ReLU
- ResNeXt Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Techniques:
- Nesterov Accelerated Gradient
- Weight Decay
Training Data:
- IG-3.5B-17k
- ImageNet
Training Resources: 336x GPUs
ID: ig_resnext101_32x48d
Epochs: 100
Layers: 101
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 8064
Image Size: '224'
Weight Decay: 0.001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/resnet.py#L896
Weights: https://download.pytorch.org/models/ig_resnext101_32x48-3e41cc8a.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 85.42%
Top 5 Accuracy: 97.58%
- Name: ig_resnext101_32x8d
In Collection: IG ResNeXt
Metadata:
FLOPs: 21180417024
Parameters: 88790000
File Size: 356056638
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Global Average Pooling
- Grouped Convolution
- Max Pooling
- ReLU
- ResNeXt Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Techniques:
- Nesterov Accelerated Gradient
- Weight Decay
Training Data:
- IG-3.5B-17k
- ImageNet
Training Resources: 336x GPUs
ID: ig_resnext101_32x8d
Epochs: 100
Layers: 101
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 8064
Image Size: '224'
Weight Decay: 0.001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/resnet.py#L863
Weights: https://download.pytorch.org/models/ig_resnext101_32x8-c38310e5.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 82.7%
Top 5 Accuracy: 96.64%
-->

139
hfdocs/source/models/inception-resnet-v2.mdx
Unescape View File

@ -0,0 +1,139 @@
# Inception ResNet v2
**Inception-ResNet-v2** is a convolutional neural architecture that builds on the Inception family of architectures but incorporates [residual connections](https://paperswithcode.com/method/residual-connection) (replacing the filter concatenation stage of the Inception architecture).
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('inception_resnet_v2', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `inception_resnet_v2`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('inception_resnet_v2', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{szegedy2016inceptionv4,
title={Inception-v4, Inception-ResNet and the Impact of Residual Connections on Learning},
author={Christian Szegedy and Sergey Ioffe and Vincent Vanhoucke and Alex Alemi},
year={2016},
eprint={1602.07261},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
<!--
Type: model-index
Collections:
- Name: Inception ResNet v2
Paper:
Title: Inception-v4, Inception-ResNet and the Impact of Residual Connections on
Learning
URL: https://paperswithcode.com/paper/inception-v4-inception-resnet-and-the-impact
Models:
- Name: inception_resnet_v2
In Collection: Inception ResNet v2
Metadata:
FLOPs: 16959133120
Parameters: 55850000
File Size: 223774238
Architecture:
- Average Pooling
- Dropout
- Inception-ResNet-v2 Reduction-B
- Inception-ResNet-v2-A
- Inception-ResNet-v2-B
- Inception-ResNet-v2-C
- Reduction-A
- Softmax
Tasks:
- Image Classification
Training Techniques:
- Label Smoothing
- RMSProp
- Weight Decay
Training Data:
- ImageNet
Training Resources: 20x NVIDIA Kepler GPUs
ID: inception_resnet_v2
LR: 0.045
Dropout: 0.2
Crop Pct: '0.897'
Momentum: 0.9
Image Size: '299'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/inception_resnet_v2.py#L343
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/inception_resnet_v2-940b1cd6.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 0.95%
Top 5 Accuracy: 17.29%
-->

152
hfdocs/source/models/inception-v3.mdx
Unescape View File

@ -0,0 +1,152 @@
# Inception v3
**Inception v3** is a convolutional neural network architecture from the Inception family that makes several improvements including using [Label Smoothing](https://paperswithcode.com/method/label-smoothing), Factorized 7 x 7 convolutions, and the use of an [auxiliary classifer](https://paperswithcode.com/method/auxiliary-classifier) to propagate label information lower down the network (along with the use of batch normalization for layers in the sidehead). The key building block is an [Inception Module](https://paperswithcode.com/method/inception-v3-module).
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('inception_v3', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `inception_v3`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('inception_v3', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@article{DBLP:journals/corr/SzegedyVISW15,
author = {Christian Szegedy and
Vincent Vanhoucke and
Sergey Ioffe and
Jonathon Shlens and
Zbigniew Wojna},
title = {Rethinking the Inception Architecture for Computer Vision},
journal = {CoRR},
volume = {abs/1512.00567},
year = {2015},
url = {http://arxiv.org/abs/1512.00567},
archivePrefix = {arXiv},
eprint = {1512.00567},
timestamp = {Mon, 13 Aug 2018 16:49:07 +0200},
biburl = {https://dblp.org/rec/journals/corr/SzegedyVISW15.bib},
bibsource = {dblp computer science bibliography, https://dblp.org}
}
```
<!--
Type: model-index
Collections:
- Name: Inception v3
Paper:
Title: Rethinking the Inception Architecture for Computer Vision
URL: https://paperswithcode.com/paper/rethinking-the-inception-architecture-for
Models:
- Name: inception_v3
In Collection: Inception v3
Metadata:
FLOPs: 7352418880
Parameters: 23830000
File Size: 108857766
Architecture:
- 1x1 Convolution
- Auxiliary Classifier
- Average Pooling
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inception-v3 Module
- Max Pooling
- ReLU
- Softmax
Tasks:
- Image Classification
Training Techniques:
- Gradient Clipping
- Label Smoothing
- RMSProp
- Weight Decay
Training Data:
- ImageNet
Training Resources: 50x NVIDIA Kepler GPUs
ID: inception_v3
LR: 0.045
Dropout: 0.2
Crop Pct: '0.875'
Momentum: 0.9
Image Size: '299'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/inception_v3.py#L442
Weights: https://download.pytorch.org/models/inception_v3_google-1a9a5a14.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 77.46%
Top 5 Accuracy: 93.48%
-->

138
hfdocs/source/models/inception-v4.mdx
Unescape View File

@ -0,0 +1,138 @@
# Inception v4
**Inception-v4** is a convolutional neural network architecture that builds on previous iterations of the Inception family by simplifying the architecture and using more inception modules than [Inception-v3](https://paperswithcode.com/method/inception-v3).
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('inception_v4', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `inception_v4`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('inception_v4', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{szegedy2016inceptionv4,
title={Inception-v4, Inception-ResNet and the Impact of Residual Connections on Learning},
author={Christian Szegedy and Sergey Ioffe and Vincent Vanhoucke and Alex Alemi},
year={2016},
eprint={1602.07261},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
<!--
Type: model-index
Collections:
- Name: Inception v4
Paper:
Title: Inception-v4, Inception-ResNet and the Impact of Residual Connections on
Learning
URL: https://paperswithcode.com/paper/inception-v4-inception-resnet-and-the-impact
Models:
- Name: inception_v4
In Collection: Inception v4
Metadata:
FLOPs: 15806527936
Parameters: 42680000
File Size: 171082495
Architecture:
- Average Pooling
- Dropout
- Inception-A
- Inception-B
- Inception-C
- Reduction-A
- Reduction-B
- Softmax
Tasks:
- Image Classification
Training Techniques:
- Label Smoothing
- RMSProp
- Weight Decay
Training Data:
- ImageNet
Training Resources: 20x NVIDIA Kepler GPUs
ID: inception_v4
LR: 0.045
Dropout: 0.2
Crop Pct: '0.875'
Momentum: 0.9
Image Size: '299'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/inception_v4.py#L313
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-cadene/inceptionv4-8e4777a0.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 1.01%
Top 5 Accuracy: 16.85%
-->

324
hfdocs/source/models/legacy-se-resnet.mdx
Unescape View File

@ -0,0 +1,324 @@
# (Legacy) SE-ResNet
**SE ResNet** is a variant of a [ResNet](https://www.paperswithcode.com/method/resnet) that employs [squeeze-and-excitation blocks](https://paperswithcode.com/method/squeeze-and-excitation-block) to enable the network to perform dynamic channel-wise feature recalibration.
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('legacy_seresnet101', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `legacy_seresnet101`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('legacy_seresnet101', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{hu2019squeezeandexcitation,
title={Squeeze-and-Excitation Networks},
author={Jie Hu and Li Shen and Samuel Albanie and Gang Sun and Enhua Wu},
year={2019},
eprint={1709.01507},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
<!--
Type: model-index
Collections:
- Name: Legacy SE ResNet
Paper:
Title: Squeeze-and-Excitation Networks
URL: https://paperswithcode.com/paper/squeeze-and-excitation-networks
Models:
- Name: legacy_seresnet101
In Collection: Legacy SE ResNet
Metadata:
FLOPs: 9762614000
Parameters: 49330000
File Size: 197822624
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- Label Smoothing
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA Titan X GPUs
ID: legacy_seresnet101
LR: 0.6
Epochs: 100
Layers: 101
Dropout: 0.2
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 1024
Image Size: '224'
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/senet.py#L426
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-cadene/se_resnet101-7e38fcc6.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 78.38%
Top 5 Accuracy: 94.26%
- Name: legacy_seresnet152
In Collection: Legacy SE ResNet
Metadata:
FLOPs: 14553578160
Parameters: 66819999
File Size: 268033864
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- Label Smoothing
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA Titan X GPUs
ID: legacy_seresnet152
LR: 0.6
Epochs: 100
Layers: 152
Dropout: 0.2
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 1024
Image Size: '224'
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/senet.py#L433
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-cadene/se_resnet152-d17c99b7.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 78.67%
Top 5 Accuracy: 94.38%
- Name: legacy_seresnet18
In Collection: Legacy SE ResNet
Metadata:
FLOPs: 2328876024
Parameters: 11780000
File Size: 47175663
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- Label Smoothing
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA Titan X GPUs
ID: legacy_seresnet18
LR: 0.6
Epochs: 100
Layers: 18
Dropout: 0.2
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 1024
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/senet.py#L405
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/seresnet18-4bb0ce65.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 71.74%
Top 5 Accuracy: 90.34%
- Name: legacy_seresnet34
In Collection: Legacy SE ResNet
Metadata:
FLOPs: 4706201004
Parameters: 21960000
File Size: 87958697
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- Label Smoothing
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA Titan X GPUs
ID: legacy_seresnet34
LR: 0.6
Epochs: 100
Layers: 34
Dropout: 0.2
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 1024
Image Size: '224'
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/senet.py#L412
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/seresnet34-a4004e63.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 74.79%
Top 5 Accuracy: 92.13%
- Name: legacy_seresnet50
In Collection: Legacy SE ResNet
Metadata:
FLOPs: 4974351024
Parameters: 28090000
File Size: 112611220
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- Label Smoothing
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA Titan X GPUs
ID: legacy_seresnet50
LR: 0.6
Epochs: 100
Layers: 50
Dropout: 0.2
Crop Pct: '0.875'
Momentum: 0.9
Image Size: '224'
Interpolation: bilinear
Minibatch Size: 1024
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/senet.py#L419
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-cadene/se_resnet50-ce0d4300.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 77.64%
Top 5 Accuracy: 93.74%
-->

234
hfdocs/source/models/legacy-se-resnext.mdx
Unescape View File

@ -0,0 +1,234 @@
# (Legacy) SE-ResNeXt
**SE ResNeXt** is a variant of a [ResNeXt](https://www.paperswithcode.com/method/resnext) that employs [squeeze-and-excitation blocks](https://paperswithcode.com/method/squeeze-and-excitation-block) to enable the network to perform dynamic channel-wise feature recalibration.
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('legacy_seresnext101_32x4d', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `legacy_seresnext101_32x4d`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('legacy_seresnext101_32x4d', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{hu2019squeezeandexcitation,
title={Squeeze-and-Excitation Networks},
author={Jie Hu and Li Shen and Samuel Albanie and Gang Sun and Enhua Wu},
year={2019},
eprint={1709.01507},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
<!--
Type: model-index
Collections:
- Name: Legacy SE ResNeXt
Paper:
Title: Squeeze-and-Excitation Networks
URL: https://paperswithcode.com/paper/squeeze-and-excitation-networks
Models:
- Name: legacy_seresnext101_32x4d
In Collection: Legacy SE ResNeXt
Metadata:
FLOPs: 10287698672
Parameters: 48960000
File Size: 196466866
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Global Average Pooling
- Grouped Convolution
- Max Pooling
- ReLU
- ResNeXt Block
- Residual Connection
- Softmax
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- Label Smoothing
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA Titan X GPUs
ID: legacy_seresnext101_32x4d
LR: 0.6
Epochs: 100
Layers: 101
Dropout: 0.2
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 1024
Image Size: '224'
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/senet.py#L462
Weights: http://data.lip6.fr/cadene/pretrainedmodels/se_resnext101_32x4d-3b2fe3d8.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 80.23%
Top 5 Accuracy: 95.02%
- Name: legacy_seresnext26_32x4d
In Collection: Legacy SE ResNeXt
Metadata:
FLOPs: 3187342304
Parameters: 16790000
File Size: 67346327
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Global Average Pooling
- Grouped Convolution
- Max Pooling
- ReLU
- ResNeXt Block
- Residual Connection
- Softmax
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- Label Smoothing
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA Titan X GPUs
ID: legacy_seresnext26_32x4d
LR: 0.6
Epochs: 100
Layers: 26
Dropout: 0.2
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 1024
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/senet.py#L448
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/seresnext26_32x4d-65ebdb501.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 77.11%
Top 5 Accuracy: 93.31%
- Name: legacy_seresnext50_32x4d
In Collection: Legacy SE ResNeXt
Metadata:
FLOPs: 5459954352
Parameters: 27560000
File Size: 110559176
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Global Average Pooling
- Grouped Convolution
- Max Pooling
- ReLU
- ResNeXt Block
- Residual Connection
- Softmax
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- Label Smoothing
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA Titan X GPUs
ID: legacy_seresnext50_32x4d
LR: 0.6
Epochs: 100
Layers: 50
Dropout: 0.2
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 1024
Image Size: '224'
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/senet.py#L455
Weights: http://data.lip6.fr/cadene/pretrainedmodels/se_resnext50_32x4d-a260b3a4.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.08%
Top 5 Accuracy: 94.43%
-->

141
hfdocs/source/models/legacy-senet.mdx
Unescape View File

@ -0,0 +1,141 @@
# (Legacy) SENet
A **SENet** is a convolutional neural network architecture that employs [squeeze-and-excitation blocks](https://paperswithcode.com/method/squeeze-and-excitation-block) to enable the network to perform dynamic channel-wise feature recalibration.
The weights from this model were ported from Gluon.
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('legacy_senet154', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `legacy_senet154`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('legacy_senet154', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{hu2019squeezeandexcitation,
title={Squeeze-and-Excitation Networks},
author={Jie Hu and Li Shen and Samuel Albanie and Gang Sun and Enhua Wu},
year={2019},
eprint={1709.01507},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
<!--
Type: model-index
Collections:
- Name: Legacy SENet
Paper:
Title: Squeeze-and-Excitation Networks
URL: https://paperswithcode.com/paper/squeeze-and-excitation-networks
Models:
- Name: legacy_senet154
In Collection: Legacy SENet
Metadata:
FLOPs: 26659556016
Parameters: 115090000
File Size: 461488402
Architecture:
- Convolution
- Dense Connections
- Global Average Pooling
- Max Pooling
- Softmax
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- Label Smoothing
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA Titan X GPUs
ID: legacy_senet154
LR: 0.6
Epochs: 100
Layers: 154
Dropout: 0.2
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 1024
Image Size: '224'
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/senet.py#L440
Weights: http://data.lip6.fr/cadene/pretrainedmodels/senet154-c7b49a05.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 81.33%
Top 5 Accuracy: 95.51%
-->

231
hfdocs/source/models/mixnet.mdx
Unescape View File

@ -0,0 +1,231 @@
# MixNet
**MixNet** is a type of convolutional neural network discovered via AutoML that utilises [MixConvs](https://paperswithcode.com/method/mixconv) instead of regular [depthwise convolutions](https://paperswithcode.com/method/depthwise-convolution).
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('mixnet_l', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `mixnet_l`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('mixnet_l', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{tan2019mixconv,
title={MixConv: Mixed Depthwise Convolutional Kernels},
author={Mingxing Tan and Quoc V. Le},
year={2019},
eprint={1907.09595},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
<!--
Type: model-index
Collections:
- Name: MixNet
Paper:
Title: 'MixConv: Mixed Depthwise Convolutional Kernels'
URL: https://paperswithcode.com/paper/mixnet-mixed-depthwise-convolutional-kernels
Models:
- Name: mixnet_l
In Collection: MixNet
Metadata:
FLOPs: 738671316
Parameters: 7330000
File Size: 29608232
Architecture:
- Batch Normalization
- Dense Connections
- Dropout
- Global Average Pooling
- Grouped Convolution
- MixConv
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Techniques:
- MNAS
Training Data:
- ImageNet
ID: mixnet_l
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1669
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/mixnet_l-5a9a2ed8.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 78.98%
Top 5 Accuracy: 94.18%
- Name: mixnet_m
In Collection: MixNet
Metadata:
FLOPs: 454543374
Parameters: 5010000
File Size: 20298347
Architecture:
- Batch Normalization
- Dense Connections
- Dropout
- Global Average Pooling
- Grouped Convolution
- MixConv
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Techniques:
- MNAS
Training Data:
- ImageNet
ID: mixnet_m
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1660
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/mixnet_m-4647fc68.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 77.27%
Top 5 Accuracy: 93.42%
- Name: mixnet_s
In Collection: MixNet
Metadata:
FLOPs: 321264910
Parameters: 4130000
File Size: 16727982
Architecture:
- Batch Normalization
- Dense Connections
- Dropout
- Global Average Pooling
- Grouped Convolution
- MixConv
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Techniques:
- MNAS
Training Data:
- ImageNet
ID: mixnet_s
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1651
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/mixnet_s-a907afbc.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 75.99%
Top 5 Accuracy: 92.79%
- Name: mixnet_xl
In Collection: MixNet
Metadata:
FLOPs: 1195880424
Parameters: 11900000
File Size: 48001170
Architecture:
- Batch Normalization
- Dense Connections
- Dropout
- Global Average Pooling
- Grouped Convolution
- MixConv
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Techniques:
- MNAS
Training Data:
- ImageNet
ID: mixnet_xl
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1678
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/mixnet_xl_ra-aac3c00c.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 80.47%
Top 5 Accuracy: 94.93%
-->

176
hfdocs/source/models/mnasnet.mdx
Unescape View File

@ -0,0 +1,176 @@
# MnasNet
**MnasNet** is a type of convolutional neural network optimized for mobile devices that is discovered through mobile neural architecture search, which explicitly incorporates model latency into the main objective so that the search can identify a model that achieves a good trade-off between accuracy and latency. The main building block is an [inverted residual block](https://paperswithcode.com/method/inverted-residual-block) (from [MobileNetV2](https://paperswithcode.com/method/mobilenetv2)).
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('mnasnet_100', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `mnasnet_100`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('mnasnet_100', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{tan2019mnasnet,
title={MnasNet: Platform-Aware Neural Architecture Search for Mobile},
author={Mingxing Tan and Bo Chen and Ruoming Pang and Vijay Vasudevan and Mark Sandler and Andrew Howard and Quoc V. Le},
year={2019},
eprint={1807.11626},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
<!--
Type: model-index
Collections:
- Name: MNASNet
Paper:
Title: 'MnasNet: Platform-Aware Neural Architecture Search for Mobile'
URL: https://paperswithcode.com/paper/mnasnet-platform-aware-neural-architecture
Models:
- Name: mnasnet_100
In Collection: MNASNet
Metadata:
FLOPs: 416415488
Parameters: 4380000
File Size: 17731774
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Depthwise Separable Convolution
- Dropout
- Global Average Pooling
- Inverted Residual Block
- Max Pooling
- ReLU
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Techniques:
- RMSProp
- Weight Decay
Training Data:
- ImageNet
ID: mnasnet_100
Layers: 100
Dropout: 0.2
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 4000
Image Size: '224'
Interpolation: bicubic
RMSProp Decay: 0.9
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L894
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/mnasnet_b1-74cb7081.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 74.67%
Top 5 Accuracy: 92.1%
- Name: semnasnet_100
In Collection: MNASNet
Metadata:
FLOPs: 414570766
Parameters: 3890000
File Size: 15731489
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Depthwise Separable Convolution
- Dropout
- Global Average Pooling
- Inverted Residual Block
- Max Pooling
- ReLU
- Residual Connection
- Softmax
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: semnasnet_100
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L928
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/mnasnet_a1-d9418771.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 75.45%
Top 5 Accuracy: 92.61%
-->

277
hfdocs/source/models/mobilenet-v2.mdx
Unescape View File

@ -0,0 +1,277 @@
# MobileNet v2
**MobileNetV2** is a convolutional neural network architecture that seeks to perform well on mobile devices. It is based on an [inverted residual structure](https://paperswithcode.com/method/inverted-residual-block) where the residual connections are between the bottleneck layers. The intermediate expansion layer uses lightweight depthwise convolutions to filter features as a source of non-linearity. As a whole, the architecture of MobileNetV2 contains the initial fully convolution layer with 32 filters, followed by 19 residual bottleneck layers.
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('mobilenetv2_100', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `mobilenetv2_100`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('mobilenetv2_100', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@article{DBLP:journals/corr/abs-1801-04381,
author = {Mark Sandler and
Andrew G. Howard and
Menglong Zhu and
Andrey Zhmoginov and
Liang{-}Chieh Chen},
title = {Inverted Residuals and Linear Bottlenecks: Mobile Networks for Classification,
Detection and Segmentation},
journal = {CoRR},
volume = {abs/1801.04381},
year = {2018},
url = {http://arxiv.org/abs/1801.04381},
archivePrefix = {arXiv},
eprint = {1801.04381},
timestamp = {Tue, 12 Jan 2021 15:30:06 +0100},
biburl = {https://dblp.org/rec/journals/corr/abs-1801-04381.bib},
bibsource = {dblp computer science bibliography, https://dblp.org}
}
```
<!--
Type: model-index
Collections:
- Name: MobileNet V2
Paper:
Title: 'MobileNetV2: Inverted Residuals and Linear Bottlenecks'
URL: https://paperswithcode.com/paper/mobilenetv2-inverted-residuals-and-linear
Models:
- Name: mobilenetv2_100
In Collection: MobileNet V2
Metadata:
FLOPs: 401920448
Parameters: 3500000
File Size: 14202571
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Depthwise Separable Convolution
- Dropout
- Inverted Residual Block
- Max Pooling
- ReLU6
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Techniques:
- RMSProp
- Weight Decay
Training Data:
- ImageNet
Training Resources: 16x GPUs
ID: mobilenetv2_100
LR: 0.045
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 1536
Image Size: '224'
Weight Decay: 4.0e-05
Interpolation: bicubic
RMSProp Decay: 0.9
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L955
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/mobilenetv2_100_ra-b33bc2c4.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 72.95%
Top 5 Accuracy: 91.0%
- Name: mobilenetv2_110d
In Collection: MobileNet V2
Metadata:
FLOPs: 573958832
Parameters: 4520000
File Size: 18316431
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Depthwise Separable Convolution
- Dropout
- Inverted Residual Block
- Max Pooling
- ReLU6
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Techniques:
- RMSProp
- Weight Decay
Training Data:
- ImageNet
Training Resources: 16x GPUs
ID: mobilenetv2_110d
LR: 0.045
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 1536
Image Size: '224'
Weight Decay: 4.0e-05
Interpolation: bicubic
RMSProp Decay: 0.9
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L969
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/mobilenetv2_110d_ra-77090ade.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 75.05%
Top 5 Accuracy: 92.19%
- Name: mobilenetv2_120d
In Collection: MobileNet V2
Metadata:
FLOPs: 888510048
Parameters: 5830000
File Size: 23651121
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Depthwise Separable Convolution
- Dropout
- Inverted Residual Block
- Max Pooling
- ReLU6
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Techniques:
- RMSProp
- Weight Decay
Training Data:
- ImageNet
Training Resources: 16x GPUs
ID: mobilenetv2_120d
LR: 0.045
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 1536
Image Size: '224'
Weight Decay: 4.0e-05
Interpolation: bicubic
RMSProp Decay: 0.9
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L977
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/mobilenetv2_120d_ra-5987e2ed.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 77.28%
Top 5 Accuracy: 93.51%
- Name: mobilenetv2_140
In Collection: MobileNet V2
Metadata:
FLOPs: 770196784
Parameters: 6110000
File Size: 24673555
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Depthwise Separable Convolution
- Dropout
- Inverted Residual Block
- Max Pooling
- ReLU6
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Techniques:
- RMSProp
- Weight Decay
Training Data:
- ImageNet
Training Resources: 16x GPUs
ID: mobilenetv2_140
LR: 0.045
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 1536
Image Size: '224'
Weight Decay: 4.0e-05
Interpolation: bicubic
RMSProp Decay: 0.9
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L962
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/mobilenetv2_140_ra-21a4e913.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 76.51%
Top 5 Accuracy: 93.0%
-->

205
hfdocs/source/models/mobilenet-v3.mdx
Unescape View File

@ -0,0 +1,205 @@
# MobileNet v3
**MobileNetV3** is a convolutional neural network that is designed for mobile phone CPUs. The network design includes the use of a [hard swish activation](https://paperswithcode.com/method/hard-swish) and [squeeze-and-excitation](https://paperswithcode.com/method/squeeze-and-excitation-block) modules in the [MBConv blocks](https://paperswithcode.com/method/inverted-residual-block).
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('mobilenetv3_large_100', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `mobilenetv3_large_100`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('mobilenetv3_large_100', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@article{DBLP:journals/corr/abs-1905-02244,
author = {Andrew Howard and
Mark Sandler and
Grace Chu and
Liang{-}Chieh Chen and
Bo Chen and
Mingxing Tan and
Weijun Wang and
Yukun Zhu and
Ruoming Pang and
Vijay Vasudevan and
Quoc V. Le and
Hartwig Adam},
title = {Searching for MobileNetV3},
journal = {CoRR},
volume = {abs/1905.02244},
year = {2019},
url = {http://arxiv.org/abs/1905.02244},
archivePrefix = {arXiv},
eprint = {1905.02244},
timestamp = {Tue, 12 Jan 2021 15:30:06 +0100},
biburl = {https://dblp.org/rec/journals/corr/abs-1905-02244.bib},
bibsource = {dblp computer science bibliography, https://dblp.org}
}
```
<!--
Type: model-index
Collections:
- Name: MobileNet V3
Paper:
Title: Searching for MobileNetV3
URL: https://paperswithcode.com/paper/searching-for-mobilenetv3
Models:
- Name: mobilenetv3_large_100
In Collection: MobileNet V3
Metadata:
FLOPs: 287193752
Parameters: 5480000
File Size: 22076443
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Dense Connections
- Depthwise Separable Convolution
- Dropout
- Global Average Pooling
- Hard Swish
- Inverted Residual Block
- ReLU
- Residual Connection
- Softmax
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- RMSProp
- Weight Decay
Training Data:
- ImageNet
Training Resources: 4x4 TPU Pod
ID: mobilenetv3_large_100
LR: 0.1
Dropout: 0.8
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 4096
Image Size: '224'
Weight Decay: 1.0e-05
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/mobilenetv3.py#L363
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/mobilenetv3_large_100_ra-f55367f5.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 75.77%
Top 5 Accuracy: 92.54%
- Name: mobilenetv3_rw
In Collection: MobileNet V3
Metadata:
FLOPs: 287190638
Parameters: 5480000
File Size: 22064048
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Dense Connections
- Depthwise Separable Convolution
- Dropout
- Global Average Pooling
- Hard Swish
- Inverted Residual Block
- ReLU
- Residual Connection
- Softmax
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- RMSProp
- Weight Decay
Training Data:
- ImageNet
Training Resources: 4x4 TPU Pod
ID: mobilenetv3_rw
LR: 0.1
Dropout: 0.8
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 4096
Image Size: '224'
Weight Decay: 1.0e-05
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/mobilenetv3.py#L384
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/mobilenetv3_100-35495452.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 75.62%
Top 5 Accuracy: 92.71%
-->

137
hfdocs/source/models/nasnet.mdx
Unescape View File

@ -0,0 +1,137 @@
# NASNet
**NASNet** is a type of convolutional neural network discovered through neural architecture search. The building blocks consist of normal and reduction cells.
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('nasnetalarge', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `nasnetalarge`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('nasnetalarge', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{zoph2018learning,
title={Learning Transferable Architectures for Scalable Image Recognition},
author={Barret Zoph and Vijay Vasudevan and Jonathon Shlens and Quoc V. Le},
year={2018},
eprint={1707.07012},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
<!--
Type: model-index
Collections:
- Name: NASNet
Paper:
Title: Learning Transferable Architectures for Scalable Image Recognition
URL: https://paperswithcode.com/paper/learning-transferable-architectures-for
Models:
- Name: nasnetalarge
In Collection: NASNet
Metadata:
FLOPs: 30242402862
Parameters: 88750000
File Size: 356056626
Architecture:
- Average Pooling
- Batch Normalization
- Convolution
- Depthwise Separable Convolution
- Dropout
- ReLU
Tasks:
- Image Classification
Training Techniques:
- Label Smoothing
- RMSProp
- Weight Decay
Training Data:
- ImageNet
Training Resources: 50x Tesla K40 GPUs
ID: nasnetalarge
Dropout: 0.5
Crop Pct: '0.911'
Momentum: 0.9
Image Size: '331'
Interpolation: bicubic
Label Smoothing: 0.1
RMSProp $\epsilon$: 1.0
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/nasnet.py#L562
Weights: http://data.lip6.fr/cadene/pretrainedmodels/nasnetalarge-a1897284.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 82.63%
Top 5 Accuracy: 96.05%
-->

577
hfdocs/source/models/noisy-student.mdx
Unescape View File

@ -0,0 +1,577 @@
# Noisy Student (EfficientNet)
**Noisy Student Training** is a semi-supervised learning approach. It extends the idea of self-training
and distillation with the use of equal-or-larger student models and noise added to the student during learning. It has three main steps:
1. train a teacher model on labeled images
2. use the teacher to generate pseudo labels on unlabeled images
3. train a student model on the combination of labeled images and pseudo labeled images.
The algorithm is iterated a few times by treating the student as a teacher to relabel the unlabeled data and training a new student.
Noisy Student Training seeks to improve on self-training and distillation in two ways. First, it makes the student larger than, or at least equal to, the teacher so the student can better learn from a larger dataset. Second, it adds noise to the student so the noised student is forced to learn harder from the pseudo labels. To noise the student, it uses input noise such as RandAugment data augmentation, and model noise such as dropout and stochastic depth during training.
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('tf_efficientnet_b0_ns', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `tf_efficientnet_b0_ns`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('tf_efficientnet_b0_ns', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{xie2020selftraining,
title={Self-training with Noisy Student improves ImageNet classification},
author={Qizhe Xie and Minh-Thang Luong and Eduard Hovy and Quoc V. Le},
year={2020},
eprint={1911.04252},
archivePrefix={arXiv},
primaryClass={cs.LG}
}
```
<!--
Type: model-index
Collections:
- Name: Noisy Student
Paper:
Title: Self-training with Noisy Student improves ImageNet classification
URL: https://paperswithcode.com/paper/self-training-with-noisy-student-improves
Models:
- Name: tf_efficientnet_b0_ns
In Collection: Noisy Student
Metadata:
FLOPs: 488688572
Parameters: 5290000
File Size: 21386709
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Techniques:
- AutoAugment
- FixRes
- Label Smoothing
- Noisy Student
- RMSProp
- RandAugment
- Weight Decay
Training Data:
- ImageNet
- JFT-300M
Training Resources: Cloud TPU v3 Pod
ID: tf_efficientnet_b0_ns
LR: 0.128
Epochs: 700
Dropout: 0.5
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 2048
Image Size: '224'
Weight Decay: 1.0e-05
Interpolation: bicubic
RMSProp Decay: 0.9
Label Smoothing: 0.1
BatchNorm Momentum: 0.99
Stochastic Depth Survival: 0.8
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1427
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_b0_ns-c0e6a31c.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 78.66%
Top 5 Accuracy: 94.37%
- Name: tf_efficientnet_b1_ns
In Collection: Noisy Student
Metadata:
FLOPs: 883633200
Parameters: 7790000
File Size: 31516408
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Techniques:
- AutoAugment
- FixRes
- Label Smoothing
- Noisy Student
- RMSProp
- RandAugment
- Weight Decay
Training Data:
- ImageNet
- JFT-300M
Training Resources: Cloud TPU v3 Pod
ID: tf_efficientnet_b1_ns
LR: 0.128
Epochs: 700
Dropout: 0.5
Crop Pct: '0.882'
Momentum: 0.9
Batch Size: 2048
Image Size: '240'
Weight Decay: 1.0e-05
Interpolation: bicubic
RMSProp Decay: 0.9
Label Smoothing: 0.1
BatchNorm Momentum: 0.99
Stochastic Depth Survival: 0.8
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1437
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_b1_ns-99dd0c41.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 81.39%
Top 5 Accuracy: 95.74%
- Name: tf_efficientnet_b2_ns
In Collection: Noisy Student
Metadata:
FLOPs: 1234321170
Parameters: 9110000
File Size: 36801803
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Techniques:
- AutoAugment
- FixRes
- Label Smoothing
- Noisy Student
- RMSProp
- RandAugment
- Weight Decay
Training Data:
- ImageNet
- JFT-300M
Training Resources: Cloud TPU v3 Pod
ID: tf_efficientnet_b2_ns
LR: 0.128
Epochs: 700
Dropout: 0.5
Crop Pct: '0.89'
Momentum: 0.9
Batch Size: 2048
Image Size: '260'
Weight Decay: 1.0e-05
Interpolation: bicubic
RMSProp Decay: 0.9
Label Smoothing: 0.1
BatchNorm Momentum: 0.99
Stochastic Depth Survival: 0.8
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1447
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_b2_ns-00306e48.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 82.39%
Top 5 Accuracy: 96.24%
- Name: tf_efficientnet_b3_ns
In Collection: Noisy Student
Metadata:
FLOPs: 2275247568
Parameters: 12230000
File Size: 49385734
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Techniques:
- AutoAugment
- FixRes
- Label Smoothing
- Noisy Student
- RMSProp
- RandAugment
- Weight Decay
Training Data:
- ImageNet
- JFT-300M
Training Resources: Cloud TPU v3 Pod
ID: tf_efficientnet_b3_ns
LR: 0.128
Epochs: 700
Dropout: 0.5
Crop Pct: '0.904'
Momentum: 0.9
Batch Size: 2048
Image Size: '300'
Weight Decay: 1.0e-05
Interpolation: bicubic
RMSProp Decay: 0.9
Label Smoothing: 0.1
BatchNorm Momentum: 0.99
Stochastic Depth Survival: 0.8
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1457
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_b3_ns-9d44bf68.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 84.04%
Top 5 Accuracy: 96.91%
- Name: tf_efficientnet_b4_ns
In Collection: Noisy Student
Metadata:
FLOPs: 5749638672
Parameters: 19340000
File Size: 77995057
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Techniques:
- AutoAugment
- FixRes
- Label Smoothing
- Noisy Student
- RMSProp
- RandAugment
- Weight Decay
Training Data:
- ImageNet
- JFT-300M
Training Resources: Cloud TPU v3 Pod
ID: tf_efficientnet_b4_ns
LR: 0.128
Epochs: 700
Dropout: 0.5
Crop Pct: '0.922'
Momentum: 0.9
Batch Size: 2048
Image Size: '380'
Weight Decay: 1.0e-05
Interpolation: bicubic
RMSProp Decay: 0.9
Label Smoothing: 0.1
BatchNorm Momentum: 0.99
Stochastic Depth Survival: 0.8
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1467
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_b4_ns-d6313a46.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 85.15%
Top 5 Accuracy: 97.47%
- Name: tf_efficientnet_b5_ns
In Collection: Noisy Student
Metadata:
FLOPs: 13176501888
Parameters: 30390000
File Size: 122404944
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Techniques:
- AutoAugment
- FixRes
- Label Smoothing
- Noisy Student
- RMSProp
- RandAugment
- Weight Decay
Training Data:
- ImageNet
- JFT-300M
Training Resources: Cloud TPU v3 Pod
ID: tf_efficientnet_b5_ns
LR: 0.128
Epochs: 350
Dropout: 0.5
Crop Pct: '0.934'
Momentum: 0.9
Batch Size: 2048
Image Size: '456'
Weight Decay: 1.0e-05
Interpolation: bicubic
RMSProp Decay: 0.9
Label Smoothing: 0.1
BatchNorm Momentum: 0.99
Stochastic Depth Survival: 0.8
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1477
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_b5_ns-6f26d0cf.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 86.08%
Top 5 Accuracy: 97.75%
- Name: tf_efficientnet_b6_ns
In Collection: Noisy Student
Metadata:
FLOPs: 24180518488
Parameters: 43040000
File Size: 173239537
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Techniques:
- AutoAugment
- FixRes
- Label Smoothing
- Noisy Student
- RMSProp
- RandAugment
- Weight Decay
Training Data:
- ImageNet
- JFT-300M
Training Resources: Cloud TPU v3 Pod
ID: tf_efficientnet_b6_ns
LR: 0.128
Epochs: 350
Dropout: 0.5
Crop Pct: '0.942'
Momentum: 0.9
Batch Size: 2048
Image Size: '528'
Weight Decay: 1.0e-05
Interpolation: bicubic
RMSProp Decay: 0.9
Label Smoothing: 0.1
BatchNorm Momentum: 0.99
Stochastic Depth Survival: 0.8
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1487
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_b6_ns-51548356.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 86.45%
Top 5 Accuracy: 97.88%
- Name: tf_efficientnet_b7_ns
In Collection: Noisy Student
Metadata:
FLOPs: 48205304880
Parameters: 66349999
File Size: 266853140
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Techniques:
- AutoAugment
- FixRes
- Label Smoothing
- Noisy Student
- RMSProp
- RandAugment
- Weight Decay
Training Data:
- ImageNet
- JFT-300M
Training Resources: Cloud TPU v3 Pod
ID: tf_efficientnet_b7_ns
LR: 0.128
Epochs: 350
Dropout: 0.5
Crop Pct: '0.949'
Momentum: 0.9
Batch Size: 2048
Image Size: '600'
Weight Decay: 1.0e-05
Interpolation: bicubic
RMSProp Decay: 0.9
Label Smoothing: 0.1
BatchNorm Momentum: 0.99
Stochastic Depth Survival: 0.8
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1498
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_b7_ns-1dbc32de.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 86.83%
Top 5 Accuracy: 98.08%
- Name: tf_efficientnet_l2_ns
In Collection: Noisy Student
Metadata:
FLOPs: 611646113804
Parameters: 480310000
File Size: 1925950424
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Techniques:
- AutoAugment
- FixRes
- Label Smoothing
- Noisy Student
- RMSProp
- RandAugment
- Weight Decay
Training Data:
- ImageNet
- JFT-300M
Training Resources: Cloud TPU v3 Pod
Training Time: 6 days
ID: tf_efficientnet_l2_ns
LR: 0.128
Epochs: 350
Dropout: 0.5
Crop Pct: '0.96'
Momentum: 0.9
Batch Size: 2048
Image Size: '800'
Weight Decay: 1.0e-05
Interpolation: bicubic
RMSProp Decay: 0.9
Label Smoothing: 0.1
BatchNorm Momentum: 0.99
Stochastic Depth Survival: 0.8
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1520
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_l2_ns-df73bb44.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 88.35%
Top 5 Accuracy: 98.66%
-->

138
hfdocs/source/models/pnasnet.mdx
Unescape View File

@ -0,0 +1,138 @@
# PNASNet
**Progressive Neural Architecture Search**, or **PNAS**, is a method for learning the structure of convolutional neural networks (CNNs). It uses a sequential model-based optimization (SMBO) strategy, where we search the space of cell structures, starting with simple (shallow) models and progressing to complex ones, pruning out unpromising structures as we go.
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('pnasnet5large', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `pnasnet5large`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('pnasnet5large', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{liu2018progressive,
title={Progressive Neural Architecture Search},
author={Chenxi Liu and Barret Zoph and Maxim Neumann and Jonathon Shlens and Wei Hua and Li-Jia Li and Li Fei-Fei and Alan Yuille and Jonathan Huang and Kevin Murphy},
year={2018},
eprint={1712.00559},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
<!--
Type: model-index
Collections:
- Name: PNASNet
Paper:
Title: Progressive Neural Architecture Search
URL: https://paperswithcode.com/paper/progressive-neural-architecture-search
Models:
- Name: pnasnet5large
In Collection: PNASNet
Metadata:
FLOPs: 31458865950
Parameters: 86060000
File Size: 345153926
Architecture:
- Average Pooling
- Batch Normalization
- Convolution
- Depthwise Separable Convolution
- Dropout
- ReLU
Tasks:
- Image Classification
Training Techniques:
- Label Smoothing
- RMSProp
- Weight Decay
Training Data:
- ImageNet
Training Resources: 100x NVIDIA P100 GPUs
ID: pnasnet5large
LR: 0.015
Dropout: 0.5
Crop Pct: '0.911'
Momentum: 0.9
Batch Size: 1600
Image Size: '331'
Interpolation: bicubic
Label Smoothing: 0.1
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/pnasnet.py#L343
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-cadene/pnasnet5large-bf079911.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 0.98%
Top 5 Accuracy: 18.58%
-->

559
hfdocs/source/models/regnetx.mdx
Unescape View File

@ -0,0 +1,559 @@
# RegNetX
**RegNetX** is a convolutional network design space with simple, regular models with parameters: depth $d$, initial width $w\_{0} > 0$, and slope $w\_{a} > 0$, and generates a different block width $u\_{j}$ for each block $j < d$. The key restriction for the RegNet types of model is that there is a linear parameterisation of block widths (the design space only contains models with this linear structure):
$$ u\_{j} = w\_{0} + w\_{a}\cdot{j} $$
For **RegNetX** we have additional restrictions: we set $b = 1$ (the bottleneck ratio), $12 \leq d \leq 28$, and $w\_{m} \geq 2$ (the width multiplier).
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('regnetx_002', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `regnetx_002`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('regnetx_002', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{radosavovic2020designing,
title={Designing Network Design Spaces},
author={Ilija Radosavovic and Raj Prateek Kosaraju and Ross Girshick and Kaiming He and Piotr Dollár},
year={2020},
eprint={2003.13678},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
<!--
Type: model-index
Collections:
- Name: RegNetX
Paper:
Title: Designing Network Design Spaces
URL: https://paperswithcode.com/paper/designing-network-design-spaces
Models:
- Name: regnetx_002
In Collection: RegNetX
Metadata:
FLOPs: 255276032
Parameters: 2680000
File Size: 10862199
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Dense Connections
- Global Average Pooling
- Grouped Convolution
- ReLU
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA V100 GPUs
ID: regnetx_002
Epochs: 100
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 1024
Image Size: '224'
Weight Decay: 5.0e-05
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/regnet.py#L337
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-regnet/regnetx_002-e7e85e5c.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 68.75%
Top 5 Accuracy: 88.56%
- Name: regnetx_004
In Collection: RegNetX
Metadata:
FLOPs: 510619136
Parameters: 5160000
File Size: 20841309
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Dense Connections
- Global Average Pooling
- Grouped Convolution
- ReLU
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA V100 GPUs
ID: regnetx_004
Epochs: 100
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 1024
Image Size: '224'
Weight Decay: 5.0e-05
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/regnet.py#L343
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-regnet/regnetx_004-7d0e9424.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 72.39%
Top 5 Accuracy: 90.82%
- Name: regnetx_006
In Collection: RegNetX
Metadata:
FLOPs: 771659136
Parameters: 6200000
File Size: 24965172
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Dense Connections
- Global Average Pooling
- Grouped Convolution
- ReLU
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA V100 GPUs
ID: regnetx_006
Epochs: 100
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 1024
Image Size: '224'
Weight Decay: 5.0e-05
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/regnet.py#L349
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-regnet/regnetx_006-85ec1baa.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 73.84%
Top 5 Accuracy: 91.68%
- Name: regnetx_008
In Collection: RegNetX
Metadata:
FLOPs: 1027038208
Parameters: 7260000
File Size: 29235944
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Dense Connections
- Global Average Pooling
- Grouped Convolution
- ReLU
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA V100 GPUs
ID: regnetx_008
Epochs: 100
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 1024
Image Size: '224'
Weight Decay: 5.0e-05
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/regnet.py#L355
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-regnet/regnetx_008-d8b470eb.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 75.05%
Top 5 Accuracy: 92.34%
- Name: regnetx_016
In Collection: RegNetX
Metadata:
FLOPs: 2059337856
Parameters: 9190000
File Size: 36988158
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Dense Connections
- Global Average Pooling
- Grouped Convolution
- ReLU
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA V100 GPUs
ID: regnetx_016
Epochs: 100
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 1024
Image Size: '224'
Weight Decay: 5.0e-05
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/regnet.py#L361
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-regnet/regnetx_016-65ca972a.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 76.95%
Top 5 Accuracy: 93.43%
- Name: regnetx_032
In Collection: RegNetX
Metadata:
FLOPs: 4082555904
Parameters: 15300000
File Size: 61509573
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Dense Connections
- Global Average Pooling
- Grouped Convolution
- ReLU
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA V100 GPUs
ID: regnetx_032
Epochs: 100
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 512
Image Size: '224'
Weight Decay: 5.0e-05
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/regnet.py#L367
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-regnet/regnetx_032-ed0c7f7e.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 78.15%
Top 5 Accuracy: 94.09%
- Name: regnetx_040
In Collection: RegNetX
Metadata:
FLOPs: 5095167744
Parameters: 22120000
File Size: 88844824
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Dense Connections
- Global Average Pooling
- Grouped Convolution
- ReLU
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA V100 GPUs
ID: regnetx_040
Epochs: 100
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 512
Image Size: '224'
Weight Decay: 5.0e-05
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/regnet.py#L373
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-regnet/regnetx_040-73c2a654.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 78.48%
Top 5 Accuracy: 94.25%
- Name: regnetx_064
In Collection: RegNetX
Metadata:
FLOPs: 8303405824
Parameters: 26210000
File Size: 105184854
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Dense Connections
- Global Average Pooling
- Grouped Convolution
- ReLU
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA V100 GPUs
ID: regnetx_064
Epochs: 100
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 512
Image Size: '224'
Weight Decay: 5.0e-05
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/regnet.py#L379
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-regnet/regnetx_064-29278baa.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.06%
Top 5 Accuracy: 94.47%
- Name: regnetx_080
In Collection: RegNetX
Metadata:
FLOPs: 10276726784
Parameters: 39570000
File Size: 158720042
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Dense Connections
- Global Average Pooling
- Grouped Convolution
- ReLU
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA V100 GPUs
ID: regnetx_080
Epochs: 100
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 512
Image Size: '224'
Weight Decay: 5.0e-05
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/regnet.py#L385
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-regnet/regnetx_080-7c7fcab1.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.21%
Top 5 Accuracy: 94.55%
- Name: regnetx_120
In Collection: RegNetX
Metadata:
FLOPs: 15536378368
Parameters: 46110000
File Size: 184866342
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Dense Connections
- Global Average Pooling
- Grouped Convolution
- ReLU
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA V100 GPUs
ID: regnetx_120
Epochs: 100
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 512
Image Size: '224'
Weight Decay: 5.0e-05
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/regnet.py#L391
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-regnet/regnetx_120-65d5521e.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.61%
Top 5 Accuracy: 94.73%
- Name: regnetx_160
In Collection: RegNetX
Metadata:
FLOPs: 20491740672
Parameters: 54280000
File Size: 217623862
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Dense Connections
- Global Average Pooling
- Grouped Convolution
- ReLU
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA V100 GPUs
ID: regnetx_160
Epochs: 100
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 512
Image Size: '224'
Weight Decay: 5.0e-05
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/regnet.py#L397
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-regnet/regnetx_160-c98c4112.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.84%
Top 5 Accuracy: 94.82%
- Name: regnetx_320
In Collection: RegNetX
Metadata:
FLOPs: 40798958592
Parameters: 107810000
File Size: 431962133
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Dense Connections
- Global Average Pooling
- Grouped Convolution
- ReLU
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA V100 GPUs
ID: regnetx_320
Epochs: 100
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 256
Image Size: '224'
Weight Decay: 5.0e-05
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/regnet.py#L403
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-regnet/regnetx_320-8ea38b93.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 80.25%
Top 5 Accuracy: 95.03%
-->

573
hfdocs/source/models/regnety.mdx
Unescape View File

@ -0,0 +1,573 @@
# RegNetY
**RegNetY** is a convolutional network design space with simple, regular models with parameters: depth $d$, initial width $w\_{0} > 0$, and slope $w\_{a} > 0$, and generates a different block width $u\_{j}$ for each block $j < d$. The key restriction for the RegNet types of model is that there is a linear parameterisation of block widths (the design space only contains models with this linear structure):
$$ u\_{j} = w\_{0} + w\_{a}\cdot{j} $$
For **RegNetX** authors have additional restrictions: we set $b = 1$ (the bottleneck ratio), $12 \leq d \leq 28$, and $w\_{m} \geq 2$ (the width multiplier).
For **RegNetY** authors make one change, which is to include [Squeeze-and-Excitation blocks](https://paperswithcode.com/method/squeeze-and-excitation-block).
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('regnety_002', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `regnety_002`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('regnety_002', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{radosavovic2020designing,
title={Designing Network Design Spaces},
author={Ilija Radosavovic and Raj Prateek Kosaraju and Ross Girshick and Kaiming He and Piotr Dollár},
year={2020},
eprint={2003.13678},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
<!--
Type: model-index
Collections:
- Name: RegNetY
Paper:
Title: Designing Network Design Spaces
URL: https://paperswithcode.com/paper/designing-network-design-spaces
Models:
- Name: regnety_002
In Collection: RegNetY
Metadata:
FLOPs: 255754236
Parameters: 3160000
File Size: 12782926
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Dense Connections
- Global Average Pooling
- Grouped Convolution
- ReLU
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA V100 GPUs
ID: regnety_002
Epochs: 100
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 1024
Image Size: '224'
Weight Decay: 5.0e-05
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/regnet.py#L409
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-regnet/regnety_002-e68ca334.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 70.28%
Top 5 Accuracy: 89.55%
- Name: regnety_004
In Collection: RegNetY
Metadata:
FLOPs: 515664568
Parameters: 4340000
File Size: 17542753
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Dense Connections
- Global Average Pooling
- Grouped Convolution
- ReLU
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA V100 GPUs
ID: regnety_004
Epochs: 100
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 1024
Image Size: '224'
Weight Decay: 5.0e-05
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/regnet.py#L415
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-regnet/regnety_004-0db870e6.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 74.02%
Top 5 Accuracy: 91.76%
- Name: regnety_006
In Collection: RegNetY
Metadata:
FLOPs: 771746928
Parameters: 6060000
File Size: 24394127
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Dense Connections
- Global Average Pooling
- Grouped Convolution
- ReLU
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA V100 GPUs
ID: regnety_006
Epochs: 100
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 1024
Image Size: '224'
Weight Decay: 5.0e-05
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/regnet.py#L421
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-regnet/regnety_006-c67e57ec.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 75.27%
Top 5 Accuracy: 92.53%
- Name: regnety_008
In Collection: RegNetY
Metadata:
FLOPs: 1023448952
Parameters: 6260000
File Size: 25223268
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Dense Connections
- Global Average Pooling
- Grouped Convolution
- ReLU
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA V100 GPUs
ID: regnety_008
Epochs: 100
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 1024
Image Size: '224'
Weight Decay: 5.0e-05
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/regnet.py#L427
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-regnet/regnety_008-dc900dbe.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 76.32%
Top 5 Accuracy: 93.07%
- Name: regnety_016
In Collection: RegNetY
Metadata:
FLOPs: 2070895094
Parameters: 11200000
File Size: 45115589
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Dense Connections
- Global Average Pooling
- Grouped Convolution
- ReLU
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA V100 GPUs
ID: regnety_016
Epochs: 100
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 1024
Image Size: '224'
Weight Decay: 5.0e-05
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/regnet.py#L433
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-regnet/regnety_016-54367f74.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 77.87%
Top 5 Accuracy: 93.73%
- Name: regnety_032
In Collection: RegNetY
Metadata:
FLOPs: 4081118714
Parameters: 19440000
File Size: 78084523
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Dense Connections
- Global Average Pooling
- Grouped Convolution
- ReLU
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA V100 GPUs
ID: regnety_032
Epochs: 100
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 512
Image Size: '224'
Weight Decay: 5.0e-05
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/regnet.py#L439
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/regnety_032_ra-7f2439f9.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 82.01%
Top 5 Accuracy: 95.91%
- Name: regnety_040
In Collection: RegNetY
Metadata:
FLOPs: 5105933432
Parameters: 20650000
File Size: 82913909
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Dense Connections
- Global Average Pooling
- Grouped Convolution
- ReLU
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA V100 GPUs
ID: regnety_040
Epochs: 100
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 512
Image Size: '224'
Weight Decay: 5.0e-05
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/regnet.py#L445
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-regnet/regnety_040-f0d569f9.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.23%
Top 5 Accuracy: 94.64%
- Name: regnety_064
In Collection: RegNetY
Metadata:
FLOPs: 8167730444
Parameters: 30580000
File Size: 122751416
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Dense Connections
- Global Average Pooling
- Grouped Convolution
- ReLU
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA V100 GPUs
ID: regnety_064
Epochs: 100
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 512
Image Size: '224'
Weight Decay: 5.0e-05
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/regnet.py#L451
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-regnet/regnety_064-0a48325c.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.73%
Top 5 Accuracy: 94.76%
- Name: regnety_080
In Collection: RegNetY
Metadata:
FLOPs: 10233621420
Parameters: 39180000
File Size: 157124671
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Dense Connections
- Global Average Pooling
- Grouped Convolution
- ReLU
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA V100 GPUs
ID: regnety_080
Epochs: 100
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 512
Image Size: '224'
Weight Decay: 5.0e-05
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/regnet.py#L457
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-regnet/regnety_080-e7f3eb93.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.87%
Top 5 Accuracy: 94.83%
- Name: regnety_120
In Collection: RegNetY
Metadata:
FLOPs: 15542094856
Parameters: 51820000
File Size: 207743949
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Dense Connections
- Global Average Pooling
- Grouped Convolution
- ReLU
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA V100 GPUs
ID: regnety_120
Epochs: 100
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 512
Image Size: '224'
Weight Decay: 5.0e-05
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/regnet.py#L463
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-regnet/regnety_120-721ba79a.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 80.38%
Top 5 Accuracy: 95.12%
- Name: regnety_160
In Collection: RegNetY
Metadata:
FLOPs: 20450196852
Parameters: 83590000
File Size: 334916722
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Dense Connections
- Global Average Pooling
- Grouped Convolution
- ReLU
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA V100 GPUs
ID: regnety_160
Epochs: 100
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 512
Image Size: '224'
Weight Decay: 5.0e-05
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/regnet.py#L469
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-regnet/regnety_160-d64013cd.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 80.28%
Top 5 Accuracy: 94.97%
- Name: regnety_320
In Collection: RegNetY
Metadata:
FLOPs: 41492618394
Parameters: 145050000
File Size: 580891965
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Dense Connections
- Global Average Pooling
- Grouped Convolution
- ReLU
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA V100 GPUs
ID: regnety_320
Epochs: 100
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 256
Image Size: '224'
Weight Decay: 5.0e-05
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/regnet.py#L475
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-regnet/regnety_320-ba464b29.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 80.8%
Top 5 Accuracy: 95.25%
-->

327
hfdocs/source/models/res2net.mdx
Unescape View File

@ -0,0 +1,327 @@
# Res2Net
**Res2Net** is an image model that employs a variation on bottleneck residual blocks, [Res2Net Blocks](https://paperswithcode.com/method/res2net-block). The motivation is to be able to represent features at multiple scales. This is achieved through a novel building block for CNNs that constructs hierarchical residual-like connections within one single residual block. This represents multi-scale features at a granular level and increases the range of receptive fields for each network layer.
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('res2net101_26w_4s', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `res2net101_26w_4s`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('res2net101_26w_4s', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@article{Gao_2021,
title={Res2Net: A New Multi-Scale Backbone Architecture},
volume={43},
ISSN={1939-3539},
url={http://dx.doi.org/10.1109/TPAMI.2019.2938758},
DOI={10.1109/tpami.2019.2938758},
number={2},
journal={IEEE Transactions on Pattern Analysis and Machine Intelligence},
publisher={Institute of Electrical and Electronics Engineers (IEEE)},
author={Gao, Shang-Hua and Cheng, Ming-Ming and Zhao, Kai and Zhang, Xin-Yu and Yang, Ming-Hsuan and Torr, Philip},
year={2021},
month={Feb},
pages={652–662}
}
```
<!--
Type: model-index
Collections:
- Name: Res2Net
Paper:
Title: 'Res2Net: A New Multi-scale Backbone Architecture'
URL: https://paperswithcode.com/paper/res2net-a-new-multi-scale-backbone
Models:
- Name: res2net101_26w_4s
In Collection: Res2Net
Metadata:
FLOPs: 10415881200
Parameters: 45210000
File Size: 181456059
Architecture:
- Batch Normalization
- Convolution
- Global Average Pooling
- ReLU
- Res2Net Block
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 4x Titan Xp GPUs
ID: res2net101_26w_4s
LR: 0.1
Epochs: 100
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 256
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/res2net.py#L152
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-res2net/res2net101_26w_4s-02a759a1.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.19%
Top 5 Accuracy: 94.43%
- Name: res2net50_14w_8s
In Collection: Res2Net
Metadata:
FLOPs: 5403546768
Parameters: 25060000
File Size: 100638543
Architecture:
- Batch Normalization
- Convolution
- Global Average Pooling
- ReLU
- Res2Net Block
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 4x Titan Xp GPUs
ID: res2net50_14w_8s
LR: 0.1
Epochs: 100
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 256
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/res2net.py#L196
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-res2net/res2net50_14w_8s-6527dddc.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 78.14%
Top 5 Accuracy: 93.86%
- Name: res2net50_26w_4s
In Collection: Res2Net
Metadata:
FLOPs: 5499974064
Parameters: 25700000
File Size: 103110087
Architecture:
- Batch Normalization
- Convolution
- Global Average Pooling
- ReLU
- Res2Net Block
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 4x Titan Xp GPUs
ID: res2net50_26w_4s
LR: 0.1
Epochs: 100
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 256
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/res2net.py#L141
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-res2net/res2net50_26w_4s-06e79181.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 77.99%
Top 5 Accuracy: 93.85%
- Name: res2net50_26w_6s
In Collection: Res2Net
Metadata:
FLOPs: 8130156528
Parameters: 37050000
File Size: 148603239
Architecture:
- Batch Normalization
- Convolution
- Global Average Pooling
- ReLU
- Res2Net Block
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 4x Titan Xp GPUs
ID: res2net50_26w_6s
LR: 0.1
Epochs: 100
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 256
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/res2net.py#L163
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-res2net/res2net50_26w_6s-19041792.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 78.57%
Top 5 Accuracy: 94.12%
- Name: res2net50_26w_8s
In Collection: Res2Net
Metadata:
FLOPs: 10760338992
Parameters: 48400000
File Size: 194085165
Architecture:
- Batch Normalization
- Convolution
- Global Average Pooling
- ReLU
- Res2Net Block
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 4x Titan Xp GPUs
ID: res2net50_26w_8s
LR: 0.1
Epochs: 100
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 256
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/res2net.py#L174
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-res2net/res2net50_26w_8s-2c7c9f12.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.19%
Top 5 Accuracy: 94.37%
- Name: res2net50_48w_2s
In Collection: Res2Net
Metadata:
FLOPs: 5375291520
Parameters: 25290000
File Size: 101421406
Architecture:
- Batch Normalization
- Convolution
- Global Average Pooling
- ReLU
- Res2Net Block
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 4x Titan Xp GPUs
ID: res2net50_48w_2s
LR: 0.1
Epochs: 100
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 256
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/res2net.py#L185
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-res2net/res2net50_48w_2s-afed724a.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 77.53%
Top 5 Accuracy: 93.56%
-->

142
hfdocs/source/models/res2next.mdx
Unescape View File

@ -0,0 +1,142 @@
# Res2NeXt
**Res2NeXt** is an image model that employs a variation on [ResNeXt](https://paperswithcode.com/method/resnext) bottleneck residual blocks. The motivation is to be able to represent features at multiple scales. This is achieved through a novel building block for CNNs that constructs hierarchical residual-like connections within one single residual block. This represents multi-scale features at a granular level and increases the range of receptive fields for each network layer.
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('res2next50', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `res2next50`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('res2next50', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@article{Gao_2021,
title={Res2Net: A New Multi-Scale Backbone Architecture},
volume={43},
ISSN={1939-3539},
url={http://dx.doi.org/10.1109/TPAMI.2019.2938758},
DOI={10.1109/tpami.2019.2938758},
number={2},
journal={IEEE Transactions on Pattern Analysis and Machine Intelligence},
publisher={Institute of Electrical and Electronics Engineers (IEEE)},
author={Gao, Shang-Hua and Cheng, Ming-Ming and Zhao, Kai and Zhang, Xin-Yu and Yang, Ming-Hsuan and Torr, Philip},
year={2021},
month={Feb},
pages={652–662}
}
```
<!--
Type: model-index
Collections:
- Name: Res2NeXt
Paper:
Title: 'Res2Net: A New Multi-scale Backbone Architecture'
URL: https://paperswithcode.com/paper/res2net-a-new-multi-scale-backbone
Models:
- Name: res2next50
In Collection: Res2NeXt
Metadata:
FLOPs: 5396798208
Parameters: 24670000
File Size: 99019592
Architecture:
- Batch Normalization
- Convolution
- Global Average Pooling
- ReLU
- Res2NeXt Block
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 4x Titan Xp GPUs
ID: res2next50
LR: 0.1
Epochs: 100
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 256
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/res2net.py#L207
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-res2net/res2next50_4s-6ef7e7bf.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 78.24%
Top 5 Accuracy: 93.91%
-->

475
hfdocs/source/models/resnest.mdx
Unescape View File

@ -0,0 +1,475 @@
# ResNeSt
A **ResNeSt** is a variant on a [ResNet](https://paperswithcode.com/method/resnet), which instead stacks [Split-Attention blocks](https://paperswithcode.com/method/split-attention). The cardinal group representations are then concatenated along the channel dimension: $V = \text{Concat}${$V^{1},V^{2},\cdots{V}^{K}$}. As in standard residual blocks, the final output $Y$ of otheur Split-Attention block is produced using a shortcut connection: $Y=V+X$, if the input and output feature-map share the same shape. For blocks with a stride, an appropriate transformation $\mathcal{T}$ is applied to the shortcut connection to align the output shapes: $Y=V+\mathcal{T}(X)$. For example, $\mathcal{T}$ can be strided convolution or combined convolution-with-pooling.
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('resnest101e', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `resnest101e`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('resnest101e', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{zhang2020resnest,
title={ResNeSt: Split-Attention Networks},
author={Hang Zhang and Chongruo Wu and Zhongyue Zhang and Yi Zhu and Haibin Lin and Zhi Zhang and Yue Sun and Tong He and Jonas Mueller and R. Manmatha and Mu Li and Alexander Smola},
year={2020},
eprint={2004.08955},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
<!--
Type: model-index
Collections:
- Name: ResNeSt
Paper:
Title: 'ResNeSt: Split-Attention Networks'
URL: https://paperswithcode.com/paper/resnest-split-attention-networks
Models:
- Name: resnest101e
In Collection: ResNeSt
Metadata:
FLOPs: 17423183648
Parameters: 48280000
File Size: 193782911
Architecture:
- 1x1 Convolution
- Convolution
- Dense Connections
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Connection
- Softmax
- Split Attention
Tasks:
- Image Classification
Training Techniques:
- AutoAugment
- DropBlock
- Label Smoothing
- Mixup
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 64x NVIDIA V100 GPUs
ID: resnest101e
LR: 0.1
Epochs: 270
Layers: 101
Dropout: 0.2
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 4096
Image Size: '256'
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/resnest.py#L182
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-resnest/resnest101-22405ba7.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 82.88%
Top 5 Accuracy: 96.31%
- Name: resnest14d
In Collection: ResNeSt
Metadata:
FLOPs: 3548594464
Parameters: 10610000
File Size: 42562639
Architecture:
- 1x1 Convolution
- Convolution
- Dense Connections
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Connection
- Softmax
- Split Attention
Tasks:
- Image Classification
Training Techniques:
- AutoAugment
- DropBlock
- Label Smoothing
- Mixup
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 64x NVIDIA V100 GPUs
ID: resnest14d
LR: 0.1
Epochs: 270
Layers: 14
Dropout: 0.2
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 8192
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/resnest.py#L148
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/gluon_resnest14-9c8fe254.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 75.51%
Top 5 Accuracy: 92.52%
- Name: resnest200e
In Collection: ResNeSt
Metadata:
FLOPs: 45954387872
Parameters: 70200000
File Size: 193782911
Architecture:
- 1x1 Convolution
- Convolution
- Dense Connections
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Connection
- Softmax
- Split Attention
Tasks:
- Image Classification
Training Techniques:
- AutoAugment
- DropBlock
- Label Smoothing
- Mixup
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 64x NVIDIA V100 GPUs
ID: resnest200e
LR: 0.1
Epochs: 270
Layers: 200
Dropout: 0.2
Crop Pct: '0.909'
Momentum: 0.9
Batch Size: 2048
Image Size: '320'
Weight Decay: 0.0001
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/resnest.py#L194
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-resnest/resnest101-22405ba7.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 83.85%
Top 5 Accuracy: 96.89%
- Name: resnest269e
In Collection: ResNeSt
Metadata:
FLOPs: 100830307104
Parameters: 110930000
File Size: 445402691
Architecture:
- 1x1 Convolution
- Convolution
- Dense Connections
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Connection
- Softmax
- Split Attention
Tasks:
- Image Classification
Training Techniques:
- AutoAugment
- DropBlock
- Label Smoothing
- Mixup
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 64x NVIDIA V100 GPUs
ID: resnest269e
LR: 0.1
Epochs: 270
Layers: 269
Dropout: 0.2
Crop Pct: '0.928'
Momentum: 0.9
Batch Size: 2048
Image Size: '416'
Weight Decay: 0.0001
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/resnest.py#L206
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-resnest/resnest269-0cc87c48.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 84.53%
Top 5 Accuracy: 96.99%
- Name: resnest26d
In Collection: ResNeSt
Metadata:
FLOPs: 4678918720
Parameters: 17070000
File Size: 68470242
Architecture:
- 1x1 Convolution
- Convolution
- Dense Connections
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Connection
- Softmax
- Split Attention
Tasks:
- Image Classification
Training Techniques:
- AutoAugment
- DropBlock
- Label Smoothing
- Mixup
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 64x NVIDIA V100 GPUs
ID: resnest26d
LR: 0.1
Epochs: 270
Layers: 26
Dropout: 0.2
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 8192
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/resnest.py#L159
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/gluon_resnest26-50eb607c.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 78.48%
Top 5 Accuracy: 94.3%
- Name: resnest50d
In Collection: ResNeSt
Metadata:
FLOPs: 6937106336
Parameters: 27480000
File Size: 110273258
Architecture:
- 1x1 Convolution
- Convolution
- Dense Connections
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Connection
- Softmax
- Split Attention
Tasks:
- Image Classification
Training Techniques:
- AutoAugment
- DropBlock
- Label Smoothing
- Mixup
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 64x NVIDIA V100 GPUs
ID: resnest50d
LR: 0.1
Epochs: 270
Layers: 50
Dropout: 0.2
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 8192
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/resnest.py#L170
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-resnest/resnest50-528c19ca.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 80.96%
Top 5 Accuracy: 95.38%
- Name: resnest50d_1s4x24d
In Collection: ResNeSt
Metadata:
FLOPs: 5686764544
Parameters: 25680000
File Size: 103045531
Architecture:
- 1x1 Convolution
- Convolution
- Dense Connections
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Connection
- Softmax
- Split Attention
Tasks:
- Image Classification
Training Techniques:
- AutoAugment
- DropBlock
- Label Smoothing
- Mixup
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 64x NVIDIA V100 GPUs
ID: resnest50d_1s4x24d
LR: 0.1
Epochs: 270
Layers: 50
Dropout: 0.2
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 8192
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/resnest.py#L229
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-resnest/resnest50_fast_1s4x24d-d4a4f76f.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 81.0%
Top 5 Accuracy: 95.33%
- Name: resnest50d_4s2x40d
In Collection: ResNeSt
Metadata:
FLOPs: 5657064720
Parameters: 30420000
File Size: 122133282
Architecture:
- 1x1 Convolution
- Convolution
- Dense Connections
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Connection
- Softmax
- Split Attention
Tasks:
- Image Classification
Training Techniques:
- AutoAugment
- DropBlock
- Label Smoothing
- Mixup
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 64x NVIDIA V100 GPUs
ID: resnest50d_4s2x40d
LR: 0.1
Epochs: 270
Layers: 50
Dropout: 0.2
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 8192
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/resnest.py#L218
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-resnest/resnest50_fast_4s2x40d-41d14ed0.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 81.11%
Top 5 Accuracy: 95.55%
-->

330
hfdocs/source/models/resnet-d.mdx
Unescape View File

@ -0,0 +1,330 @@
# ResNet-D
**ResNet-D** is a modification on the [ResNet](https://paperswithcode.com/method/resnet) architecture that utilises an [average pooling](https://paperswithcode.com/method/average-pooling) tweak for downsampling. The motivation is that in the unmodified ResNet, the [1×1 convolution](https://paperswithcode.com/method/1x1-convolution) for the downsampling block ignores 3/4 of input feature maps, so this is modified so no information will be ignored
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('resnet101d', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `resnet101d`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('resnet101d', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{he2018bag,
title={Bag of Tricks for Image Classification with Convolutional Neural Networks},
author={Tong He and Zhi Zhang and Hang Zhang and Zhongyue Zhang and Junyuan Xie and Mu Li},
year={2018},
eprint={1812.01187},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
<!--
Type: model-index
Collections:
- Name: ResNet-D
Paper:
Title: Bag of Tricks for Image Classification with Convolutional Neural Networks
URL: https://paperswithcode.com/paper/bag-of-tricks-for-image-classification-with
Models:
- Name: resnet101d
In Collection: ResNet-D
Metadata:
FLOPs: 13805639680
Parameters: 44570000
File Size: 178791263
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: resnet101d
Crop Pct: '0.94'
Image Size: '256'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/resnet.py#L716
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/resnet101d_ra2-2803ffab.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 82.31%
Top 5 Accuracy: 96.06%
- Name: resnet152d
In Collection: ResNet-D
Metadata:
FLOPs: 20155275264
Parameters: 60210000
File Size: 241596837
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: resnet152d
Crop Pct: '0.94'
Image Size: '256'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/resnet.py#L724
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/resnet152d_ra2-5cac0439.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 83.13%
Top 5 Accuracy: 96.35%
- Name: resnet18d
In Collection: ResNet-D
Metadata:
FLOPs: 2645205760
Parameters: 11710000
File Size: 46893231
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: resnet18d
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/resnet.py#L649
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/resnet18d_ra2-48a79e06.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 72.27%
Top 5 Accuracy: 90.69%
- Name: resnet200d
In Collection: ResNet-D
Metadata:
FLOPs: 26034378752
Parameters: 64690000
File Size: 259662933
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: resnet200d
Crop Pct: '0.94'
Image Size: '256'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/resnet.py#L749
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/resnet200d_ra2-bdba9bf9.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 83.24%
Top 5 Accuracy: 96.49%
- Name: resnet26d
In Collection: ResNet-D
Metadata:
FLOPs: 3335276032
Parameters: 16010000
File Size: 64209122
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: resnet26d
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/resnet.py#L683
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/resnet26d-69e92c46.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 76.69%
Top 5 Accuracy: 93.15%
- Name: resnet34d
In Collection: ResNet-D
Metadata:
FLOPs: 5026601728
Parameters: 21820000
File Size: 87369807
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: resnet34d
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/resnet.py#L666
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/resnet34d_ra2-f8dcfcaf.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 77.11%
Top 5 Accuracy: 93.38%
- Name: resnet50d
In Collection: ResNet-D
Metadata:
FLOPs: 5591002624
Parameters: 25580000
File Size: 102567109
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: resnet50d
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/resnet.py#L699
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/resnet50d_ra2-464e36ba.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 80.55%
Top 5 Accuracy: 95.16%
-->

445
hfdocs/source/models/resnet.mdx
Unescape View File

@ -0,0 +1,445 @@
# ResNet
**Residual Networks**, or **ResNets**, learn residual functions with reference to the layer inputs, instead of learning unreferenced functions. Instead of hoping each few stacked layers directly fit a desired underlying mapping, residual nets let these layers fit a residual mapping. They stack [residual blocks](https://paperswithcode.com/method/residual-block) ontop of each other to form network: e.g. a ResNet-50 has fifty layers using these blocks.
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('resnet18', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `resnet18`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('resnet18', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@article{DBLP:journals/corr/HeZRS15,
author = {Kaiming He and
Xiangyu Zhang and
Shaoqing Ren and
Jian Sun},
title = {Deep Residual Learning for Image Recognition},
journal = {CoRR},
volume = {abs/1512.03385},
year = {2015},
url = {http://arxiv.org/abs/1512.03385},
archivePrefix = {arXiv},
eprint = {1512.03385},
timestamp = {Wed, 17 Apr 2019 17:23:45 +0200},
biburl = {https://dblp.org/rec/journals/corr/HeZRS15.bib},
bibsource = {dblp computer science bibliography, https://dblp.org}
}
```
<!--
Type: model-index
Collections:
- Name: ResNet
Paper:
Title: Deep Residual Learning for Image Recognition
URL: https://paperswithcode.com/paper/deep-residual-learning-for-image-recognition
Models:
- Name: resnet18
In Collection: ResNet
Metadata:
FLOPs: 2337073152
Parameters: 11690000
File Size: 46827520
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: resnet18
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/resnet.py#L641
Weights: https://download.pytorch.org/models/resnet18-5c106cde.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 69.74%
Top 5 Accuracy: 89.09%
- Name: resnet26
In Collection: ResNet
Metadata:
FLOPs: 3026804736
Parameters: 16000000
File Size: 64129972
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: resnet26
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/resnet.py#L675
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/resnet26-9aa10e23.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 75.29%
Top 5 Accuracy: 92.57%
- Name: resnet34
In Collection: ResNet
Metadata:
FLOPs: 4718469120
Parameters: 21800000
File Size: 87290831
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: resnet34
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/resnet.py#L658
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/resnet34-43635321.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 75.11%
Top 5 Accuracy: 92.28%
- Name: resnet50
In Collection: ResNet
Metadata:
FLOPs: 5282531328
Parameters: 25560000
File Size: 102488165
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: resnet50
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/resnet.py#L691
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/resnet50_ram-a26f946b.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.04%
Top 5 Accuracy: 94.39%
- Name: resnetblur50
In Collection: ResNet
Metadata:
FLOPs: 6621606912
Parameters: 25560000
File Size: 102488165
Architecture:
- 1x1 Convolution
- Batch Normalization
- Blur Pooling
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: resnetblur50
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/resnet.py#L1160
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/resnetblur50-84f4748f.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.29%
Top 5 Accuracy: 94.64%
- Name: tv_resnet101
In Collection: ResNet
Metadata:
FLOPs: 10068547584
Parameters: 44550000
File Size: 178728960
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
ID: tv_resnet101
LR: 0.1
Epochs: 90
Crop Pct: '0.875'
LR Gamma: 0.1
Momentum: 0.9
Batch Size: 32
Image Size: '224'
LR Step Size: 30
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/resnet.py#L761
Weights: https://download.pytorch.org/models/resnet101-5d3b4d8f.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 77.37%
Top 5 Accuracy: 93.56%
- Name: tv_resnet152
In Collection: ResNet
Metadata:
FLOPs: 14857660416
Parameters: 60190000
File Size: 241530880
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
ID: tv_resnet152
LR: 0.1
Epochs: 90
Crop Pct: '0.875'
LR Gamma: 0.1
Momentum: 0.9
Batch Size: 32
Image Size: '224'
LR Step Size: 30
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/resnet.py#L769
Weights: https://download.pytorch.org/models/resnet152-b121ed2d.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 78.32%
Top 5 Accuracy: 94.05%
- Name: tv_resnet34
In Collection: ResNet
Metadata:
FLOPs: 4718469120
Parameters: 21800000
File Size: 87306240
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
ID: tv_resnet34
LR: 0.1
Epochs: 90
Crop Pct: '0.875'
LR Gamma: 0.1
Momentum: 0.9
Batch Size: 32
Image Size: '224'
LR Step Size: 30
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/resnet.py#L745
Weights: https://download.pytorch.org/models/resnet34-333f7ec4.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 73.3%
Top 5 Accuracy: 91.42%
- Name: tv_resnet50
In Collection: ResNet
Metadata:
FLOPs: 5282531328
Parameters: 25560000
File Size: 102502400
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
ID: tv_resnet50
LR: 0.1
Epochs: 90
Crop Pct: '0.875'
LR Gamma: 0.1
Momentum: 0.9
Batch Size: 32
Image Size: '224'
LR Step Size: 30
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/resnet.py#L753
Weights: https://download.pytorch.org/models/resnet50-19c8e357.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 76.16%
Top 5 Accuracy: 92.88%
-->

250
hfdocs/source/models/resnext.mdx
Unescape View File

@ -0,0 +1,250 @@
# ResNeXt
A **ResNeXt** repeats a [building block](https://paperswithcode.com/method/resnext-block) that aggregates a set of transformations with the same topology. Compared to a [ResNet](https://paperswithcode.com/method/resnet), it exposes a new dimension, *cardinality* (the size of the set of transformations) $C$, as an essential factor in addition to the dimensions of depth and width.
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('resnext101_32x8d', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `resnext101_32x8d`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('resnext101_32x8d', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@article{DBLP:journals/corr/XieGDTH16,
author = {Saining Xie and
Ross B. Girshick and
Piotr Doll{\'{a}}r and
Zhuowen Tu and
Kaiming He},
title = {Aggregated Residual Transformations for Deep Neural Networks},
journal = {CoRR},
volume = {abs/1611.05431},
year = {2016},
url = {http://arxiv.org/abs/1611.05431},
archivePrefix = {arXiv},
eprint = {1611.05431},
timestamp = {Mon, 13 Aug 2018 16:45:58 +0200},
biburl = {https://dblp.org/rec/journals/corr/XieGDTH16.bib},
bibsource = {dblp computer science bibliography, https://dblp.org}
}
```
<!--
Type: model-index
Collections:
- Name: ResNeXt
Paper:
Title: Aggregated Residual Transformations for Deep Neural Networks
URL: https://paperswithcode.com/paper/aggregated-residual-transformations-for-deep
Models:
- Name: resnext101_32x8d
In Collection: ResNeXt
Metadata:
FLOPs: 21180417024
Parameters: 88790000
File Size: 356082095
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Global Average Pooling
- Grouped Convolution
- Max Pooling
- ReLU
- ResNeXt Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: resnext101_32x8d
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/b9843f954b0457af2db4f9dea41a8538f51f5d78/timm/models/resnet.py#L877
Weights: https://download.pytorch.org/models/resnext101_32x8d-8ba56ff5.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.3%
Top 5 Accuracy: 94.53%
- Name: resnext50_32x4d
In Collection: ResNeXt
Metadata:
FLOPs: 5472648192
Parameters: 25030000
File Size: 100435887
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Global Average Pooling
- Grouped Convolution
- Max Pooling
- ReLU
- ResNeXt Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: resnext50_32x4d
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/b9843f954b0457af2db4f9dea41a8538f51f5d78/timm/models/resnet.py#L851
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/resnext50_32x4d_ra-d733960d.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.79%
Top 5 Accuracy: 94.61%
- Name: resnext50d_32x4d
In Collection: ResNeXt
Metadata:
FLOPs: 5781119488
Parameters: 25050000
File Size: 100515304
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Global Average Pooling
- Grouped Convolution
- Max Pooling
- ReLU
- ResNeXt Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: resnext50d_32x4d
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/b9843f954b0457af2db4f9dea41a8538f51f5d78/timm/models/resnet.py#L869
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/resnext50d_32x4d-103e99f8.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.67%
Top 5 Accuracy: 94.87%
- Name: tv_resnext50_32x4d
In Collection: ResNeXt
Metadata:
FLOPs: 5472648192
Parameters: 25030000
File Size: 100441675
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Global Average Pooling
- Grouped Convolution
- Max Pooling
- ReLU
- ResNeXt Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
ID: tv_resnext50_32x4d
LR: 0.1
Epochs: 90
Crop Pct: '0.875'
LR Gamma: 0.1
Momentum: 0.9
Batch Size: 32
Image Size: '224'
LR Step Size: 30
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/resnet.py#L842
Weights: https://download.pytorch.org/models/resnext50_32x4d-7cdf4587.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 77.61%
Top 5 Accuracy: 93.68%
-->

264
hfdocs/source/models/rexnet.mdx
Unescape View File

@ -0,0 +1,264 @@
# RexNet
**Rank Expansion Networks** (ReXNets) follow a set of new design principles for designing bottlenecks in image classification models. Authors refine each layer by 1) expanding the input channel size of the convolution layer and 2) replacing the [ReLU6s](https://www.paperswithcode.com/method/relu6).
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('rexnet_100', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `rexnet_100`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('rexnet_100', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{han2020rexnet,
title={ReXNet: Diminishing Representational Bottleneck on Convolutional Neural Network},
author={Dongyoon Han and Sangdoo Yun and Byeongho Heo and YoungJoon Yoo},
year={2020},
eprint={2007.00992},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
<!--
Type: model-index
Collections:
- Name: RexNet
Paper:
Title: 'ReXNet: Diminishing Representational Bottleneck on Convolutional Neural
Network'
URL: https://paperswithcode.com/paper/rexnet-diminishing-representational
Models:
- Name: rexnet_100
In Collection: RexNet
Metadata:
FLOPs: 509989377
Parameters: 4800000
File Size: 19417552
Architecture:
- Batch Normalization
- Convolution
- Dropout
- ReLU6
- Residual Connection
Tasks:
- Image Classification
Training Techniques:
- Label Smoothing
- Linear Warmup With Cosine Annealing
- Nesterov Accelerated Gradient
- Weight Decay
Training Data:
- ImageNet
Training Resources: 4x NVIDIA V100 GPUs
ID: rexnet_100
LR: 0.5
Epochs: 400
Dropout: 0.2
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 512
Image Size: '224'
Weight Decay: 1.0e-05
Interpolation: bicubic
Label Smoothing: 0.1
Code: https://github.com/rwightman/pytorch-image-models/blob/b9843f954b0457af2db4f9dea41a8538f51f5d78/timm/models/rexnet.py#L212
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-rexnet/rexnetv1_100-1b4dddf4.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 77.86%
Top 5 Accuracy: 93.88%
- Name: rexnet_130
In Collection: RexNet
Metadata:
FLOPs: 848364461
Parameters: 7560000
File Size: 30508197
Architecture:
- Batch Normalization
- Convolution
- Dropout
- ReLU6
- Residual Connection
Tasks:
- Image Classification
Training Techniques:
- Label Smoothing
- Linear Warmup With Cosine Annealing
- Nesterov Accelerated Gradient
- Weight Decay
Training Data:
- ImageNet
Training Resources: 4x NVIDIA V100 GPUs
ID: rexnet_130
LR: 0.5
Epochs: 400
Dropout: 0.2
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 512
Image Size: '224'
Weight Decay: 1.0e-05
Interpolation: bicubic
Label Smoothing: 0.1
Code: https://github.com/rwightman/pytorch-image-models/blob/b9843f954b0457af2db4f9dea41a8538f51f5d78/timm/models/rexnet.py#L218
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-rexnet/rexnetv1_130-590d768e.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.49%
Top 5 Accuracy: 94.67%
- Name: rexnet_150
In Collection: RexNet
Metadata:
FLOPs: 1122374469
Parameters: 9730000
File Size: 39227315
Architecture:
- Batch Normalization
- Convolution
- Dropout
- ReLU6
- Residual Connection
Tasks:
- Image Classification
Training Techniques:
- Label Smoothing
- Linear Warmup With Cosine Annealing
- Nesterov Accelerated Gradient
- Weight Decay
Training Data:
- ImageNet
Training Resources: 4x NVIDIA V100 GPUs
ID: rexnet_150
LR: 0.5
Epochs: 400
Dropout: 0.2
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 512
Image Size: '224'
Weight Decay: 1.0e-05
Interpolation: bicubic
Label Smoothing: 0.1
Code: https://github.com/rwightman/pytorch-image-models/blob/b9843f954b0457af2db4f9dea41a8538f51f5d78/timm/models/rexnet.py#L224
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-rexnet/rexnetv1_150-bd1a6aa8.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 80.31%
Top 5 Accuracy: 95.16%
- Name: rexnet_200
In Collection: RexNet
Metadata:
FLOPs: 1960224938
Parameters: 16370000
File Size: 65862221
Architecture:
- Batch Normalization
- Convolution
- Dropout
- ReLU6
- Residual Connection
Tasks:
- Image Classification
Training Techniques:
- Label Smoothing
- Linear Warmup With Cosine Annealing
- Nesterov Accelerated Gradient
- Weight Decay
Training Data:
- ImageNet
Training Resources: 4x NVIDIA V100 GPUs
ID: rexnet_200
LR: 0.5
Epochs: 400
Dropout: 0.2
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 512
Image Size: '224'
Weight Decay: 1.0e-05
Interpolation: bicubic
Label Smoothing: 0.1
Code: https://github.com/rwightman/pytorch-image-models/blob/b9843f954b0457af2db4f9dea41a8538f51f5d78/timm/models/rexnet.py#L230
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-rexnet/rexnetv1_200-8c0b7f2d.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 81.63%
Top 5 Accuracy: 95.67%
-->

189
hfdocs/source/models/se-resnet.mdx
Unescape View File

@ -0,0 +1,189 @@
# SE-ResNet
**SE ResNet** is a variant of a [ResNet](https://www.paperswithcode.com/method/resnet) that employs [squeeze-and-excitation blocks](https://paperswithcode.com/method/squeeze-and-excitation-block) to enable the network to perform dynamic channel-wise feature recalibration.
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('seresnet152d', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `seresnet152d`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('seresnet152d', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{hu2019squeezeandexcitation,
title={Squeeze-and-Excitation Networks},
author={Jie Hu and Li Shen and Samuel Albanie and Gang Sun and Enhua Wu},
year={2019},
eprint={1709.01507},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
<!--
Type: model-index
Collections:
- Name: SE ResNet
Paper:
Title: Squeeze-and-Excitation Networks
URL: https://paperswithcode.com/paper/squeeze-and-excitation-networks
Models:
- Name: seresnet152d
In Collection: SE ResNet
Metadata:
FLOPs: 20161904304
Parameters: 66840000
File Size: 268144497
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- Label Smoothing
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA Titan X GPUs
ID: seresnet152d
LR: 0.6
Epochs: 100
Layers: 152
Dropout: 0.2
Crop Pct: '0.94'
Momentum: 0.9
Batch Size: 1024
Image Size: '256'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/a7f95818e44b281137503bcf4b3e3e94d8ffa52f/timm/models/resnet.py#L1206
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/seresnet152d_ra2-04464dd2.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 83.74%
Top 5 Accuracy: 96.77%
- Name: seresnet50
In Collection: SE ResNet
Metadata:
FLOPs: 5285062320
Parameters: 28090000
File Size: 112621903
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- Label Smoothing
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA Titan X GPUs
ID: seresnet50
LR: 0.6
Epochs: 100
Layers: 50
Dropout: 0.2
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 1024
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/a7f95818e44b281137503bcf4b3e3e94d8ffa52f/timm/models/resnet.py#L1180
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/seresnet50_ra_224-8efdb4bb.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 80.26%
Top 5 Accuracy: 95.07%
-->

203
hfdocs/source/models/selecsls.mdx
Unescape View File

@ -0,0 +1,203 @@
# SelecSLS
**SelecSLS** uses novel selective long and short range skip connections to improve the information flow allowing for a drastically faster network without compromising accuracy.
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('selecsls42b', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `selecsls42b`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('selecsls42b', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@article{Mehta_2020,
title={XNect},
volume={39},
ISSN={1557-7368},
url={http://dx.doi.org/10.1145/3386569.3392410},
DOI={10.1145/3386569.3392410},
number={4},
journal={ACM Transactions on Graphics},
publisher={Association for Computing Machinery (ACM)},
author={Mehta, Dushyant and Sotnychenko, Oleksandr and Mueller, Franziska and Xu, Weipeng and Elgharib, Mohamed and Fua, Pascal and Seidel, Hans-Peter and Rhodin, Helge and Pons-Moll, Gerard and Theobalt, Christian},
year={2020},
month={Jul}
}
```
<!--
Type: model-index
Collections:
- Name: SelecSLS
Paper:
Title: 'XNect: Real-time Multi-Person 3D Motion Capture with a Single RGB Camera'
URL: https://paperswithcode.com/paper/xnect-real-time-multi-person-3d-human-pose
Models:
- Name: selecsls42b
In Collection: SelecSLS
Metadata:
FLOPs: 3824022528
Parameters: 32460000
File Size: 129948954
Architecture:
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Global Average Pooling
- ReLU
- SelecSLS Block
Tasks:
- Image Classification
Training Techniques:
- Cosine Annealing
- Random Erasing
Training Data:
- ImageNet
ID: selecsls42b
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/b9843f954b0457af2db4f9dea41a8538f51f5d78/timm/models/selecsls.py#L335
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-selecsls/selecsls42b-8af30141.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 77.18%
Top 5 Accuracy: 93.39%
- Name: selecsls60
In Collection: SelecSLS
Metadata:
FLOPs: 4610472600
Parameters: 30670000
File Size: 122839714
Architecture:
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Global Average Pooling
- ReLU
- SelecSLS Block
Tasks:
- Image Classification
Training Techniques:
- Cosine Annealing
- Random Erasing
Training Data:
- ImageNet
ID: selecsls60
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/b9843f954b0457af2db4f9dea41a8538f51f5d78/timm/models/selecsls.py#L342
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-selecsls/selecsls60-bbf87526.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 77.99%
Top 5 Accuracy: 93.83%
- Name: selecsls60b
In Collection: SelecSLS
Metadata:
FLOPs: 4657653144
Parameters: 32770000
File Size: 131252898
Architecture:
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Global Average Pooling
- ReLU
- SelecSLS Block
Tasks:
- Image Classification
Training Techniques:
- Cosine Annealing
- Random Erasing
Training Data:
- ImageNet
ID: selecsls60b
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/b9843f954b0457af2db4f9dea41a8538f51f5d78/timm/models/selecsls.py#L349
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-selecsls/selecsls60b-94e619b5.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 78.41%
Top 5 Accuracy: 94.18%
-->

234
hfdocs/source/models/seresnext.mdx
Unescape View File

@ -0,0 +1,234 @@
# SE-ResNeXt
**SE ResNeXt** is a variant of a [ResNext](https://www.paperswithcode.com/method/resneXt) that employs [squeeze-and-excitation blocks](https://paperswithcode.com/method/squeeze-and-excitation-block) to enable the network to perform dynamic channel-wise feature recalibration.
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('seresnext26d_32x4d', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `seresnext26d_32x4d`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('seresnext26d_32x4d', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{hu2019squeezeandexcitation,
title={Squeeze-and-Excitation Networks},
author={Jie Hu and Li Shen and Samuel Albanie and Gang Sun and Enhua Wu},
year={2019},
eprint={1709.01507},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
<!--
Type: model-index
Collections:
- Name: SEResNeXt
Paper:
Title: Squeeze-and-Excitation Networks
URL: https://paperswithcode.com/paper/squeeze-and-excitation-networks
Models:
- Name: seresnext26d_32x4d
In Collection: SEResNeXt
Metadata:
FLOPs: 3507053024
Parameters: 16810000
File Size: 67425193
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Global Average Pooling
- Grouped Convolution
- Max Pooling
- ReLU
- ResNeXt Block
- Residual Connection
- Softmax
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- Label Smoothing
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA Titan X GPUs
ID: seresnext26d_32x4d
LR: 0.6
Epochs: 100
Layers: 26
Dropout: 0.2
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 1024
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/a7f95818e44b281137503bcf4b3e3e94d8ffa52f/timm/models/resnet.py#L1234
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/seresnext26d_32x4d-80fa48a3.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 77.59%
Top 5 Accuracy: 93.61%
- Name: seresnext26t_32x4d
In Collection: SEResNeXt
Metadata:
FLOPs: 3466436448
Parameters: 16820000
File Size: 67414838
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Global Average Pooling
- Grouped Convolution
- Max Pooling
- ReLU
- ResNeXt Block
- Residual Connection
- Softmax
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- Label Smoothing
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA Titan X GPUs
ID: seresnext26t_32x4d
LR: 0.6
Epochs: 100
Layers: 26
Dropout: 0.2
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 1024
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/a7f95818e44b281137503bcf4b3e3e94d8ffa52f/timm/models/resnet.py#L1246
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/seresnext26tn_32x4d-569cb627.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 77.99%
Top 5 Accuracy: 93.73%
- Name: seresnext50_32x4d
In Collection: SEResNeXt
Metadata:
FLOPs: 5475179184
Parameters: 27560000
File Size: 110569859
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Global Average Pooling
- Grouped Convolution
- Max Pooling
- ReLU
- ResNeXt Block
- Residual Connection
- Softmax
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- Label Smoothing
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA Titan X GPUs
ID: seresnext50_32x4d
LR: 0.6
Epochs: 100
Layers: 50
Dropout: 0.2
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 1024
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/a7f95818e44b281137503bcf4b3e3e94d8ffa52f/timm/models/resnet.py#L1267
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/seresnext50_32x4d_racm-a304a460.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 81.27%
Top 5 Accuracy: 95.62%
-->

179
hfdocs/source/models/skresnet.mdx
Unescape View File

@ -0,0 +1,179 @@
# SK-ResNet
**SK ResNet** is a variant of a [ResNet](https://www.paperswithcode.com/method/resnet) that employs a [Selective Kernel](https://paperswithcode.com/method/selective-kernel) unit. In general, all the large kernel convolutions in the original bottleneck blocks in ResNet are replaced by the proposed [SK convolutions](https://paperswithcode.com/method/selective-kernel-convolution), enabling the network to choose appropriate receptive field sizes in an adaptive manner.
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('skresnet18', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `skresnet18`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('skresnet18', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{li2019selective,
title={Selective Kernel Networks},
author={Xiang Li and Wenhai Wang and Xiaolin Hu and Jian Yang},
year={2019},
eprint={1903.06586},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
<!--
Type: model-index
Collections:
- Name: SKResNet
Paper:
Title: Selective Kernel Networks
URL: https://paperswithcode.com/paper/selective-kernel-networks
Models:
- Name: skresnet18
In Collection: SKResNet
Metadata:
FLOPs: 2333467136
Parameters: 11960000
File Size: 47923238
Architecture:
- Convolution
- Dense Connections
- Global Average Pooling
- Max Pooling
- Residual Connection
- Selective Kernel
- Softmax
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x GPUs
ID: skresnet18
LR: 0.1
Epochs: 100
Layers: 18
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 256
Image Size: '224'
Weight Decay: 4.0e-05
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/a7f95818e44b281137503bcf4b3e3e94d8ffa52f/timm/models/sknet.py#L148
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/skresnet18_ra-4eec2804.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 73.03%
Top 5 Accuracy: 91.17%
- Name: skresnet34
In Collection: SKResNet
Metadata:
FLOPs: 4711849952
Parameters: 22280000
File Size: 89299314
Architecture:
- Convolution
- Dense Connections
- Global Average Pooling
- Max Pooling
- Residual Connection
- Selective Kernel
- Softmax
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x GPUs
ID: skresnet34
LR: 0.1
Epochs: 100
Layers: 34
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 256
Image Size: '224'
Weight Decay: 4.0e-05
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/a7f95818e44b281137503bcf4b3e3e94d8ffa52f/timm/models/sknet.py#L165
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/skresnet34_ra-bdc0ccde.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 76.93%
Top 5 Accuracy: 93.32%
-->

137
hfdocs/source/models/skresnext.mdx
Unescape View File

@ -0,0 +1,137 @@
# SK-ResNeXt
**SK ResNeXt** is a variant of a [ResNeXt](https://www.paperswithcode.com/method/resnext) that employs a [Selective Kernel](https://paperswithcode.com/method/selective-kernel) unit. In general, all the large kernel convolutions in the original bottleneck blocks in ResNext are replaced by the proposed [SK convolutions](https://paperswithcode.com/method/selective-kernel-convolution), enabling the network to choose appropriate receptive field sizes in an adaptive manner.
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('skresnext50_32x4d', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `skresnext50_32x4d`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('skresnext50_32x4d', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{li2019selective,
title={Selective Kernel Networks},
author={Xiang Li and Wenhai Wang and Xiaolin Hu and Jian Yang},
year={2019},
eprint={1903.06586},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
<!--
Type: model-index
Collections:
- Name: SKResNeXt
Paper:
Title: Selective Kernel Networks
URL: https://paperswithcode.com/paper/selective-kernel-networks
Models:
- Name: skresnext50_32x4d
In Collection: SKResNeXt
Metadata:
FLOPs: 5739845824
Parameters: 27480000
File Size: 110340975
Architecture:
- Convolution
- Dense Connections
- Global Average Pooling
- Grouped Convolution
- Max Pooling
- Residual Connection
- Selective Kernel
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
Training Resources: 8x GPUs
ID: skresnext50_32x4d
LR: 0.1
Epochs: 100
Layers: 50
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 256
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/a7f95818e44b281137503bcf4b3e3e94d8ffa52f/timm/models/sknet.py#L210
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/skresnext50_ra-f40e40bf.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 80.15%
Top 5 Accuracy: 94.64%
-->

129
hfdocs/source/models/spnasnet.mdx
Unescape View File

@ -0,0 +1,129 @@
# SPNASNet
**Single-Path NAS** is a novel differentiable NAS method for designing hardware-efficient ConvNets in less than 4 hours.
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('spnasnet_100', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `spnasnet_100`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('spnasnet_100', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{stamoulis2019singlepath,
title={Single-Path NAS: Designing Hardware-Efficient ConvNets in less than 4 Hours},
author={Dimitrios Stamoulis and Ruizhou Ding and Di Wang and Dimitrios Lymberopoulos and Bodhi Priyantha and Jie Liu and Diana Marculescu},
year={2019},
eprint={1904.02877},
archivePrefix={arXiv},
primaryClass={cs.LG}
}
```
<!--
Type: model-index
Collections:
- Name: SPNASNet
Paper:
Title: 'Single-Path NAS: Designing Hardware-Efficient ConvNets in less than 4
Hours'
URL: https://paperswithcode.com/paper/single-path-nas-designing-hardware-efficient
Models:
- Name: spnasnet_100
In Collection: SPNASNet
Metadata:
FLOPs: 442385600
Parameters: 4420000
File Size: 17902337
Architecture:
- Average Pooling
- Batch Normalization
- Convolution
- Depthwise Separable Convolution
- Dropout
- ReLU
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: spnasnet_100
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L995
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/spnasnet_100-048bc3f4.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 74.08%
Top 5 Accuracy: 91.82%
-->

198
hfdocs/source/models/ssl-resnet.mdx
Unescape View File

@ -0,0 +1,198 @@
# SSL ResNet
**Residual Networks**, or **ResNets**, learn residual functions with reference to the layer inputs, instead of learning unreferenced functions. Instead of hoping each few stacked layers directly fit a desired underlying mapping, residual nets let these layers fit a residual mapping. They stack [residual blocks](https://paperswithcode.com/method/residual-block) ontop of each other to form network: e.g. a ResNet-50 has fifty layers using these blocks.
The model in this collection utilises semi-supervised learning to improve the performance of the model. The approach brings important gains to standard architectures for image, video and fine-grained classification.
Please note the CC-BY-NC 4.0 license on theses weights, non-commercial use only.
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('ssl_resnet18', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `ssl_resnet18`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('ssl_resnet18', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@article{DBLP:journals/corr/abs-1905-00546,
author = {I. Zeki Yalniz and
Herv{\'{e}} J{\'{e}}gou and
Kan Chen and
Manohar Paluri and
Dhruv Mahajan},
title = {Billion-scale semi-supervised learning for image classification},
journal = {CoRR},
volume = {abs/1905.00546},
year = {2019},
url = {http://arxiv.org/abs/1905.00546},
archivePrefix = {arXiv},
eprint = {1905.00546},
timestamp = {Mon, 28 Sep 2020 08:19:37 +0200},
biburl = {https://dblp.org/rec/journals/corr/abs-1905-00546.bib},
bibsource = {dblp computer science bibliography, https://dblp.org}
}
```
<!--
Type: model-index
Collections:
- Name: SSL ResNet
Paper:
Title: Billion-scale semi-supervised learning for image classification
URL: https://paperswithcode.com/paper/billion-scale-semi-supervised-learning-for
Models:
- Name: ssl_resnet18
In Collection: SSL ResNet
Metadata:
FLOPs: 2337073152
Parameters: 11690000
File Size: 46811375
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
- YFCC-100M
Training Resources: 64x GPUs
ID: ssl_resnet18
LR: 0.0015
Epochs: 30
Layers: 18
Crop Pct: '0.875'
Batch Size: 1536
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/resnet.py#L894
Weights: https://dl.fbaipublicfiles.com/semiweaksupervision/model_files/semi_supervised_resnet18-d92f0530.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 72.62%
Top 5 Accuracy: 91.42%
- Name: ssl_resnet50
In Collection: SSL ResNet
Metadata:
FLOPs: 5282531328
Parameters: 25560000
File Size: 102480594
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
- YFCC-100M
Training Resources: 64x GPUs
ID: ssl_resnet50
LR: 0.0015
Epochs: 30
Layers: 50
Crop Pct: '0.875'
Batch Size: 1536
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/resnet.py#L904
Weights: https://dl.fbaipublicfiles.com/semiweaksupervision/model_files/semi_supervised_resnet50-08389792.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.24%
Top 5 Accuracy: 94.83%
-->

198
hfdocs/source/models/swsl-resnet.mdx
Unescape View File

@ -0,0 +1,198 @@
# SWSL ResNet
**Residual Networks**, or **ResNets**, learn residual functions with reference to the layer inputs, instead of learning unreferenced functions. Instead of hoping each few stacked layers directly fit a desired underlying mapping, residual nets let these layers fit a residual mapping. They stack [residual blocks](https://paperswithcode.com/method/residual-block) ontop of each other to form network: e.g. a ResNet-50 has fifty layers using these blocks.
The models in this collection utilise semi-weakly supervised learning to improve the performance of the model. The approach brings important gains to standard architectures for image, video and fine-grained classification.
Please note the CC-BY-NC 4.0 license on theses weights, non-commercial use only.
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('swsl_resnet18', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `swsl_resnet18`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('swsl_resnet18', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@article{DBLP:journals/corr/abs-1905-00546,
author = {I. Zeki Yalniz and
Herv{\'{e}} J{\'{e}}gou and
Kan Chen and
Manohar Paluri and
Dhruv Mahajan},
title = {Billion-scale semi-supervised learning for image classification},
journal = {CoRR},
volume = {abs/1905.00546},
year = {2019},
url = {http://arxiv.org/abs/1905.00546},
archivePrefix = {arXiv},
eprint = {1905.00546},
timestamp = {Mon, 28 Sep 2020 08:19:37 +0200},
biburl = {https://dblp.org/rec/journals/corr/abs-1905-00546.bib},
bibsource = {dblp computer science bibliography, https://dblp.org}
}
```
<!--
Type: model-index
Collections:
- Name: SWSL ResNet
Paper:
Title: Billion-scale semi-supervised learning for image classification
URL: https://paperswithcode.com/paper/billion-scale-semi-supervised-learning-for
Models:
- Name: swsl_resnet18
In Collection: SWSL ResNet
Metadata:
FLOPs: 2337073152
Parameters: 11690000
File Size: 46811375
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- IG-1B-Targeted
- ImageNet
Training Resources: 64x GPUs
ID: swsl_resnet18
LR: 0.0015
Epochs: 30
Layers: 18
Crop Pct: '0.875'
Batch Size: 1536
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/resnet.py#L954
Weights: https://dl.fbaipublicfiles.com/semiweaksupervision/model_files/semi_weakly_supervised_resnet18-118f1556.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 73.28%
Top 5 Accuracy: 91.76%
- Name: swsl_resnet50
In Collection: SWSL ResNet
Metadata:
FLOPs: 5282531328
Parameters: 25560000
File Size: 102480594
Architecture:
- 1x1 Convolution
- Batch Normalization
- Bottleneck Residual Block
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- IG-1B-Targeted
- ImageNet
Training Resources: 64x GPUs
ID: swsl_resnet50
LR: 0.0015
Epochs: 30
Layers: 50
Crop Pct: '0.875'
Batch Size: 1536
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/resnet.py#L965
Weights: https://dl.fbaipublicfiles.com/semiweaksupervision/model_files/semi_weakly_supervised_resnet50-16a12f1b.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 81.14%
Top 5 Accuracy: 95.97%
-->

284
hfdocs/source/models/swsl-resnext.mdx
Unescape View File

@ -0,0 +1,284 @@
# SWSL ResNeXt
A **ResNeXt** repeats a [building block](https://paperswithcode.com/method/resnext-block) that aggregates a set of transformations with the same topology. Compared to a [ResNet](https://paperswithcode.com/method/resnet), it exposes a new dimension, *cardinality* (the size of the set of transformations) $C$, as an essential factor in addition to the dimensions of depth and width.
The models in this collection utilise semi-weakly supervised learning to improve the performance of the model. The approach brings important gains to standard architectures for image, video and fine-grained classification.
Please note the CC-BY-NC 4.0 license on theses weights, non-commercial use only.
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('swsl_resnext101_32x16d', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `swsl_resnext101_32x16d`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('swsl_resnext101_32x16d', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@article{DBLP:journals/corr/abs-1905-00546,
author = {I. Zeki Yalniz and
Herv{\'{e}} J{\'{e}}gou and
Kan Chen and
Manohar Paluri and
Dhruv Mahajan},
title = {Billion-scale semi-supervised learning for image classification},
journal = {CoRR},
volume = {abs/1905.00546},
year = {2019},
url = {http://arxiv.org/abs/1905.00546},
archivePrefix = {arXiv},
eprint = {1905.00546},
timestamp = {Mon, 28 Sep 2020 08:19:37 +0200},
biburl = {https://dblp.org/rec/journals/corr/abs-1905-00546.bib},
bibsource = {dblp computer science bibliography, https://dblp.org}
}
```
<!--
Type: model-index
Collections:
- Name: SWSL ResNext
Paper:
Title: Billion-scale semi-supervised learning for image classification
URL: https://paperswithcode.com/paper/billion-scale-semi-supervised-learning-for
Models:
- Name: swsl_resnext101_32x16d
In Collection: SWSL ResNext
Metadata:
FLOPs: 46623691776
Parameters: 194030000
File Size: 777518664
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Global Average Pooling
- Grouped Convolution
- Max Pooling
- ReLU
- ResNeXt Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- IG-1B-Targeted
- ImageNet
Training Resources: 64x GPUs
ID: swsl_resnext101_32x16d
LR: 0.0015
Epochs: 30
Layers: 101
Crop Pct: '0.875'
Batch Size: 1536
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/resnet.py#L1009
Weights: https://dl.fbaipublicfiles.com/semiweaksupervision/model_files/semi_weakly_supervised_resnext101_32x16-f3559a9c.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 83.34%
Top 5 Accuracy: 96.84%
- Name: swsl_resnext101_32x4d
In Collection: SWSL ResNext
Metadata:
FLOPs: 10298145792
Parameters: 44180000
File Size: 177341913
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Global Average Pooling
- Grouped Convolution
- Max Pooling
- ReLU
- ResNeXt Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- IG-1B-Targeted
- ImageNet
Training Resources: 64x GPUs
ID: swsl_resnext101_32x4d
LR: 0.0015
Epochs: 30
Layers: 101
Crop Pct: '0.875'
Batch Size: 1536
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/resnet.py#L987
Weights: https://dl.fbaipublicfiles.com/semiweaksupervision/model_files/semi_weakly_supervised_resnext101_32x4-3f87e46b.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 83.22%
Top 5 Accuracy: 96.77%
- Name: swsl_resnext101_32x8d
In Collection: SWSL ResNext
Metadata:
FLOPs: 21180417024
Parameters: 88790000
File Size: 356056638
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Global Average Pooling
- Grouped Convolution
- Max Pooling
- ReLU
- ResNeXt Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- IG-1B-Targeted
- ImageNet
Training Resources: 64x GPUs
ID: swsl_resnext101_32x8d
LR: 0.0015
Epochs: 30
Layers: 101
Crop Pct: '0.875'
Batch Size: 1536
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/resnet.py#L998
Weights: https://dl.fbaipublicfiles.com/semiweaksupervision/model_files/semi_weakly_supervised_resnext101_32x8-b4712904.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 84.27%
Top 5 Accuracy: 97.17%
- Name: swsl_resnext50_32x4d
In Collection: SWSL ResNext
Metadata:
FLOPs: 5472648192
Parameters: 25030000
File Size: 100428550
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Global Average Pooling
- Grouped Convolution
- Max Pooling
- ReLU
- ResNeXt Block
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Data:
- IG-1B-Targeted
- ImageNet
Training Resources: 64x GPUs
ID: swsl_resnext50_32x4d
LR: 0.0015
Epochs: 30
Layers: 50
Crop Pct: '0.875'
Batch Size: 1536
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/resnet.py#L976
Weights: https://dl.fbaipublicfiles.com/semiweaksupervision/model_files/semi_weakly_supervised_resnext50_32x4-72679e44.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 82.17%
Top 5 Accuracy: 96.23%
-->

256
hfdocs/source/models/tf-efficientnet-condconv.mdx
Unescape View File

@ -0,0 +1,256 @@
# (Tensorflow) EfficientNet CondConv
**EfficientNet** is a convolutional neural network architecture and scaling method that uniformly scales all dimensions of depth/width/resolution using a *compound coefficient*. Unlike conventional practice that arbitrary scales these factors, the EfficientNet scaling method uniformly scales network width, depth, and resolution with a set of fixed scaling coefficients. For example, if we want to use $2^N$ times more computational resources, then we can simply increase the network depth by $\alpha ^ N$, width by $\beta ^ N$, and image size by $\gamma ^ N$, where $\alpha, \beta, \gamma$ are constant coefficients determined by a small grid search on the original small model. EfficientNet uses a compound coefficient $\phi$ to uniformly scales network width, depth, and resolution in a principled way.
The compound scaling method is justified by the intuition that if the input image is bigger, then the network needs more layers to increase the receptive field and more channels to capture more fine-grained patterns on the bigger image.
The base EfficientNet-B0 network is based on the inverted bottleneck residual blocks of [MobileNetV2](https://paperswithcode.com/method/mobilenetv2), in addition to squeeze-and-excitation blocks.
This collection of models amends EfficientNet by adding [CondConv](https://paperswithcode.com/method/condconv) convolutions.
The weights from this model were ported from [Tensorflow/TPU](https://github.com/tensorflow/tpu).
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('tf_efficientnet_cc_b0_4e', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `tf_efficientnet_cc_b0_4e`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('tf_efficientnet_cc_b0_4e', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@article{DBLP:journals/corr/abs-1904-04971,
author = {Brandon Yang and
Gabriel Bender and
Quoc V. Le and
Jiquan Ngiam},
title = {Soft Conditional Computation},
journal = {CoRR},
volume = {abs/1904.04971},
year = {2019},
url = {http://arxiv.org/abs/1904.04971},
archivePrefix = {arXiv},
eprint = {1904.04971},
timestamp = {Thu, 25 Apr 2019 13:55:01 +0200},
biburl = {https://dblp.org/rec/journals/corr/abs-1904-04971.bib},
bibsource = {dblp computer science bibliography, https://dblp.org}
}
```
<!--
Type: model-index
Collections:
- Name: TF EfficientNet CondConv
Paper:
Title: 'CondConv: Conditionally Parameterized Convolutions for Efficient Inference'
URL: https://paperswithcode.com/paper/soft-conditional-computation
Models:
- Name: tf_efficientnet_cc_b0_4e
In Collection: TF EfficientNet CondConv
Metadata:
FLOPs: 224153788
Parameters: 13310000
File Size: 53490940
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- CondConv
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Techniques:
- AutoAugment
- Label Smoothing
- RMSProp
- Stochastic Depth
- Weight Decay
Training Data:
- ImageNet
ID: tf_efficientnet_cc_b0_4e
LR: 0.256
Epochs: 350
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 2048
Image Size: '224'
Weight Decay: 1.0e-05
Interpolation: bicubic
RMSProp Decay: 0.9
Label Smoothing: 0.1
BatchNorm Momentum: 0.99
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1561
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_cc_b0_4e-4362b6b2.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 77.32%
Top 5 Accuracy: 93.32%
- Name: tf_efficientnet_cc_b0_8e
In Collection: TF EfficientNet CondConv
Metadata:
FLOPs: 224158524
Parameters: 24010000
File Size: 96287616
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- CondConv
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Techniques:
- AutoAugment
- Label Smoothing
- RMSProp
- Stochastic Depth
- Weight Decay
Training Data:
- ImageNet
ID: tf_efficientnet_cc_b0_8e
LR: 0.256
Epochs: 350
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 2048
Image Size: '224'
Weight Decay: 1.0e-05
Interpolation: bicubic
RMSProp Decay: 0.9
Label Smoothing: 0.1
BatchNorm Momentum: 0.99
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1572
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_cc_b0_8e-66184a25.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 77.91%
Top 5 Accuracy: 93.65%
- Name: tf_efficientnet_cc_b1_8e
In Collection: TF EfficientNet CondConv
Metadata:
FLOPs: 370427824
Parameters: 39720000
File Size: 159206198
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- CondConv
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Techniques:
- AutoAugment
- Label Smoothing
- RMSProp
- Stochastic Depth
- Weight Decay
Training Data:
- ImageNet
ID: tf_efficientnet_cc_b1_8e
LR: 0.256
Epochs: 350
Crop Pct: '0.882'
Momentum: 0.9
Batch Size: 2048
Image Size: '240'
Weight Decay: 1.0e-05
Interpolation: bicubic
RMSProp Decay: 0.9
Label Smoothing: 0.1
BatchNorm Momentum: 0.99
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1584
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_cc_b1_8e-f7c79ae1.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.33%
Top 5 Accuracy: 94.37%
-->

262
hfdocs/source/models/tf-efficientnet-lite.mdx
Unescape View File

@ -0,0 +1,262 @@
# (Tensorflow) EfficientNet Lite
**EfficientNet** is a convolutional neural network architecture and scaling method that uniformly scales all dimensions of depth/width/resolution using a *compound coefficient*. Unlike conventional practice that arbitrary scales these factors, the EfficientNet scaling method uniformly scales network width, depth, and resolution with a set of fixed scaling coefficients. For example, if we want to use $2^N$ times more computational resources, then we can simply increase the network depth by $\alpha ^ N$, width by $\beta ^ N$, and image size by $\gamma ^ N$, where $\alpha, \beta, \gamma$ are constant coefficients determined by a small grid search on the original small model. EfficientNet uses a compound coefficient $\phi$ to uniformly scales network width, depth, and resolution in a principled way.
The compound scaling method is justified by the intuition that if the input image is bigger, then the network needs more layers to increase the receptive field and more channels to capture more fine-grained patterns on the bigger image.
The base EfficientNet-B0 network is based on the inverted bottleneck residual blocks of [MobileNetV2](https://paperswithcode.com/method/mobilenetv2).
EfficientNet-Lite makes EfficientNet more suitable for mobile devices by introducing [ReLU6](https://paperswithcode.com/method/relu6) activation functions and removing [squeeze-and-excitation blocks](https://paperswithcode.com/method/squeeze-and-excitation).
The weights from this model were ported from [Tensorflow/TPU](https://github.com/tensorflow/tpu).
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('tf_efficientnet_lite0', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `tf_efficientnet_lite0`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('tf_efficientnet_lite0', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{tan2020efficientnet,
title={EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks},
author={Mingxing Tan and Quoc V. Le},
year={2020},
eprint={1905.11946},
archivePrefix={arXiv},
primaryClass={cs.LG}
}
```
<!--
Type: model-index
Collections:
- Name: TF EfficientNet Lite
Paper:
Title: 'EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks'
URL: https://paperswithcode.com/paper/efficientnet-rethinking-model-scaling-for
Models:
- Name: tf_efficientnet_lite0
In Collection: TF EfficientNet Lite
Metadata:
FLOPs: 488052032
Parameters: 4650000
File Size: 18820223
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- RELU6
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: tf_efficientnet_lite0
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1596
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_lite0-0aa007d2.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 74.83%
Top 5 Accuracy: 92.17%
- Name: tf_efficientnet_lite1
In Collection: TF EfficientNet Lite
Metadata:
FLOPs: 773639520
Parameters: 5420000
File Size: 21939331
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- RELU6
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: tf_efficientnet_lite1
Crop Pct: '0.882'
Image Size: '240'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1607
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_lite1-bde8b488.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 76.67%
Top 5 Accuracy: 93.24%
- Name: tf_efficientnet_lite2
In Collection: TF EfficientNet Lite
Metadata:
FLOPs: 1068494432
Parameters: 6090000
File Size: 24658687
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- RELU6
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: tf_efficientnet_lite2
Crop Pct: '0.89'
Image Size: '260'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1618
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_lite2-dcccb7df.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 77.48%
Top 5 Accuracy: 93.75%
- Name: tf_efficientnet_lite3
In Collection: TF EfficientNet Lite
Metadata:
FLOPs: 2011534304
Parameters: 8199999
File Size: 33161413
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- RELU6
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: tf_efficientnet_lite3
Crop Pct: '0.904'
Image Size: '300'
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1629
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_lite3-b733e338.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.83%
Top 5 Accuracy: 94.91%
- Name: tf_efficientnet_lite4
In Collection: TF EfficientNet Lite
Metadata:
FLOPs: 5164802912
Parameters: 13010000
File Size: 52558819
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- RELU6
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: tf_efficientnet_lite4
Crop Pct: '0.92'
Image Size: '380'
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1640
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_lite4-741542c3.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 81.54%
Top 5 Accuracy: 95.66%
-->

669
hfdocs/source/models/tf-efficientnet.mdx
Unescape View File

@ -0,0 +1,669 @@
# (Tensorflow) EfficientNet
**EfficientNet** is a convolutional neural network architecture and scaling method that uniformly scales all dimensions of depth/width/resolution using a *compound coefficient*. Unlike conventional practice that arbitrary scales these factors, the EfficientNet scaling method uniformly scales network width, depth, and resolution with a set of fixed scaling coefficients. For example, if we want to use $2^N$ times more computational resources, then we can simply increase the network depth by $\alpha ^ N$, width by $\beta ^ N$, and image size by $\gamma ^ N$, where $\alpha, \beta, \gamma$ are constant coefficients determined by a small grid search on the original small model. EfficientNet uses a compound coefficient $\phi$ to uniformly scales network width, depth, and resolution in a principled way.
The compound scaling method is justified by the intuition that if the input image is bigger, then the network needs more layers to increase the receptive field and more channels to capture more fine-grained patterns on the bigger image.
The base EfficientNet-B0 network is based on the inverted bottleneck residual blocks of [MobileNetV2](https://paperswithcode.com/method/mobilenetv2), in addition to [squeeze-and-excitation blocks](https://paperswithcode.com/method/squeeze-and-excitation-block).
The weights from this model were ported from [Tensorflow/TPU](https://github.com/tensorflow/tpu).
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('tf_efficientnet_b0', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `tf_efficientnet_b0`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('tf_efficientnet_b0', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{tan2020efficientnet,
title={EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks},
author={Mingxing Tan and Quoc V. Le},
year={2020},
eprint={1905.11946},
archivePrefix={arXiv},
primaryClass={cs.LG}
}
```
<!--
Type: model-index
Collections:
- Name: TF EfficientNet
Paper:
Title: 'EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks'
URL: https://paperswithcode.com/paper/efficientnet-rethinking-model-scaling-for
Models:
- Name: tf_efficientnet_b0
In Collection: TF EfficientNet
Metadata:
FLOPs: 488688572
Parameters: 5290000
File Size: 21383997
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Techniques:
- AutoAugment
- Label Smoothing
- RMSProp
- Stochastic Depth
- Weight Decay
Training Data:
- ImageNet
Training Resources: TPUv3 Cloud TPU
ID: tf_efficientnet_b0
LR: 0.256
Epochs: 350
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 2048
Image Size: '224'
Weight Decay: 1.0e-05
Interpolation: bicubic
RMSProp Decay: 0.9
Label Smoothing: 0.1
BatchNorm Momentum: 0.99
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1241
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_b0_aa-827b6e33.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 76.85%
Top 5 Accuracy: 93.23%
- Name: tf_efficientnet_b1
In Collection: TF EfficientNet
Metadata:
FLOPs: 883633200
Parameters: 7790000
File Size: 31512534
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Techniques:
- AutoAugment
- Label Smoothing
- RMSProp
- Stochastic Depth
- Weight Decay
Training Data:
- ImageNet
ID: tf_efficientnet_b1
LR: 0.256
Epochs: 350
Crop Pct: '0.882'
Momentum: 0.9
Batch Size: 2048
Image Size: '240'
Weight Decay: 1.0e-05
Interpolation: bicubic
RMSProp Decay: 0.9
Label Smoothing: 0.1
BatchNorm Momentum: 0.99
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1251
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_b1_aa-ea7a6ee0.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 78.84%
Top 5 Accuracy: 94.2%
- Name: tf_efficientnet_b2
In Collection: TF EfficientNet
Metadata:
FLOPs: 1234321170
Parameters: 9110000
File Size: 36797929
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Techniques:
- AutoAugment
- Label Smoothing
- RMSProp
- Stochastic Depth
- Weight Decay
Training Data:
- ImageNet
ID: tf_efficientnet_b2
LR: 0.256
Epochs: 350
Crop Pct: '0.89'
Momentum: 0.9
Batch Size: 2048
Image Size: '260'
Weight Decay: 1.0e-05
Interpolation: bicubic
RMSProp Decay: 0.9
Label Smoothing: 0.1
BatchNorm Momentum: 0.99
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1261
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_b2_aa-60c94f97.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 80.07%
Top 5 Accuracy: 94.9%
- Name: tf_efficientnet_b3
In Collection: TF EfficientNet
Metadata:
FLOPs: 2275247568
Parameters: 12230000
File Size: 49381362
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Techniques:
- AutoAugment
- Label Smoothing
- RMSProp
- Stochastic Depth
- Weight Decay
Training Data:
- ImageNet
ID: tf_efficientnet_b3
LR: 0.256
Epochs: 350
Crop Pct: '0.904'
Momentum: 0.9
Batch Size: 2048
Image Size: '300'
Weight Decay: 1.0e-05
Interpolation: bicubic
RMSProp Decay: 0.9
Label Smoothing: 0.1
BatchNorm Momentum: 0.99
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1271
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_b3_aa-84b4657e.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 81.65%
Top 5 Accuracy: 95.72%
- Name: tf_efficientnet_b4
In Collection: TF EfficientNet
Metadata:
FLOPs: 5749638672
Parameters: 19340000
File Size: 77989689
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Techniques:
- AutoAugment
- Label Smoothing
- RMSProp
- Stochastic Depth
- Weight Decay
Training Data:
- ImageNet
Training Resources: TPUv3 Cloud TPU
ID: tf_efficientnet_b4
LR: 0.256
Epochs: 350
Crop Pct: '0.922'
Momentum: 0.9
Batch Size: 2048
Image Size: '380'
Weight Decay: 1.0e-05
Interpolation: bicubic
RMSProp Decay: 0.9
Label Smoothing: 0.1
BatchNorm Momentum: 0.99
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1281
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_b4_aa-818f208c.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 83.03%
Top 5 Accuracy: 96.3%
- Name: tf_efficientnet_b5
In Collection: TF EfficientNet
Metadata:
FLOPs: 13176501888
Parameters: 30390000
File Size: 122403150
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Techniques:
- AutoAugment
- Label Smoothing
- RMSProp
- Stochastic Depth
- Weight Decay
Training Data:
- ImageNet
ID: tf_efficientnet_b5
LR: 0.256
Epochs: 350
Crop Pct: '0.934'
Momentum: 0.9
Batch Size: 2048
Image Size: '456'
Weight Decay: 1.0e-05
Interpolation: bicubic
RMSProp Decay: 0.9
Label Smoothing: 0.1
BatchNorm Momentum: 0.99
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1291
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_b5_ra-9a3e5369.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 83.81%
Top 5 Accuracy: 96.75%
- Name: tf_efficientnet_b6
In Collection: TF EfficientNet
Metadata:
FLOPs: 24180518488
Parameters: 43040000
File Size: 173232007
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Techniques:
- AutoAugment
- Label Smoothing
- RMSProp
- Stochastic Depth
- Weight Decay
Training Data:
- ImageNet
ID: tf_efficientnet_b6
LR: 0.256
Epochs: 350
Crop Pct: '0.942'
Momentum: 0.9
Batch Size: 2048
Image Size: '528'
Weight Decay: 1.0e-05
Interpolation: bicubic
RMSProp Decay: 0.9
Label Smoothing: 0.1
BatchNorm Momentum: 0.99
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1301
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_b6_aa-80ba17e4.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 84.11%
Top 5 Accuracy: 96.89%
- Name: tf_efficientnet_b7
In Collection: TF EfficientNet
Metadata:
FLOPs: 48205304880
Parameters: 66349999
File Size: 266850607
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Techniques:
- AutoAugment
- Label Smoothing
- RMSProp
- Stochastic Depth
- Weight Decay
Training Data:
- ImageNet
ID: tf_efficientnet_b7
LR: 0.256
Epochs: 350
Crop Pct: '0.949'
Momentum: 0.9
Batch Size: 2048
Image Size: '600'
Weight Decay: 1.0e-05
Interpolation: bicubic
RMSProp Decay: 0.9
Label Smoothing: 0.1
BatchNorm Momentum: 0.99
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1312
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_b7_ra-6c08e654.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 84.93%
Top 5 Accuracy: 97.2%
- Name: tf_efficientnet_b8
In Collection: TF EfficientNet
Metadata:
FLOPs: 80962956270
Parameters: 87410000
File Size: 351379853
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Techniques:
- AutoAugment
- Label Smoothing
- RMSProp
- Stochastic Depth
- Weight Decay
Training Data:
- ImageNet
ID: tf_efficientnet_b8
LR: 0.256
Epochs: 350
Crop Pct: '0.954'
Momentum: 0.9
Batch Size: 2048
Image Size: '672'
Weight Decay: 1.0e-05
Interpolation: bicubic
RMSProp Decay: 0.9
Label Smoothing: 0.1
BatchNorm Momentum: 0.99
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1323
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_b8_ra-572d5dd9.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 85.35%
Top 5 Accuracy: 97.39%
- Name: tf_efficientnet_el
In Collection: TF EfficientNet
Metadata:
FLOPs: 9356616096
Parameters: 10590000
File Size: 42800271
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: tf_efficientnet_el
Crop Pct: '0.904'
Image Size: '300'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1551
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_el-5143854e.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 80.45%
Top 5 Accuracy: 95.17%
- Name: tf_efficientnet_em
In Collection: TF EfficientNet
Metadata:
FLOPs: 3636607040
Parameters: 6900000
File Size: 27933644
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: tf_efficientnet_em
Crop Pct: '0.882'
Image Size: '240'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1541
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_em-e78cfe58.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 78.71%
Top 5 Accuracy: 94.33%
- Name: tf_efficientnet_es
In Collection: TF EfficientNet
Metadata:
FLOPs: 2057577472
Parameters: 5440000
File Size: 22008479
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: tf_efficientnet_es
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1531
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_es-ca1afbfe.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 77.28%
Top 5 Accuracy: 93.6%
- Name: tf_efficientnet_l2_ns_475
In Collection: TF EfficientNet
Metadata:
FLOPs: 217795669644
Parameters: 480310000
File Size: 1925950424
Architecture:
- 1x1 Convolution
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inverted Residual Block
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Techniques:
- AutoAugment
- FixRes
- Label Smoothing
- Noisy Student
- RMSProp
- RandAugment
- Weight Decay
Training Data:
- ImageNet
- JFT-300M
Training Resources: TPUv3 Cloud TPU
ID: tf_efficientnet_l2_ns_475
LR: 0.128
Epochs: 350
Dropout: 0.5
Crop Pct: '0.936'
Momentum: 0.9
Batch Size: 2048
Image Size: '475'
Weight Decay: 1.0e-05
Interpolation: bicubic
RMSProp Decay: 0.9
Label Smoothing: 0.1
BatchNorm Momentum: 0.99
Stochastic Depth Survival: 0.8
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1509
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_l2_ns_475-bebbd00a.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 88.24%
Top 5 Accuracy: 98.55%
-->

154
hfdocs/source/models/tf-inception-v3.mdx
Unescape View File

@ -0,0 +1,154 @@
# (Tensorflow) Inception v3
**Inception v3** is a convolutional neural network architecture from the Inception family that makes several improvements including using [Label Smoothing](https://paperswithcode.com/method/label-smoothing), Factorized 7 x 7 convolutions, and the use of an [auxiliary classifer](https://paperswithcode.com/method/auxiliary-classifier) to propagate label information lower down the network (along with the use of batch normalization for layers in the sidehead). The key building block is an [Inception Module](https://paperswithcode.com/method/inception-v3-module).
The weights from this model were ported from [Tensorflow/Models](https://github.com/tensorflow/models).
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('tf_inception_v3', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `tf_inception_v3`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('tf_inception_v3', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@article{DBLP:journals/corr/SzegedyVISW15,
author = {Christian Szegedy and
Vincent Vanhoucke and
Sergey Ioffe and
Jonathon Shlens and
Zbigniew Wojna},
title = {Rethinking the Inception Architecture for Computer Vision},
journal = {CoRR},
volume = {abs/1512.00567},
year = {2015},
url = {http://arxiv.org/abs/1512.00567},
archivePrefix = {arXiv},
eprint = {1512.00567},
timestamp = {Mon, 13 Aug 2018 16:49:07 +0200},
biburl = {https://dblp.org/rec/journals/corr/SzegedyVISW15.bib},
bibsource = {dblp computer science bibliography, https://dblp.org}
}
```
<!--
Type: model-index
Collections:
- Name: TF Inception v3
Paper:
Title: Rethinking the Inception Architecture for Computer Vision
URL: https://paperswithcode.com/paper/rethinking-the-inception-architecture-for
Models:
- Name: tf_inception_v3
In Collection: TF Inception v3
Metadata:
FLOPs: 7352418880
Parameters: 23830000
File Size: 95549439
Architecture:
- 1x1 Convolution
- Auxiliary Classifier
- Average Pooling
- Average Pooling
- Batch Normalization
- Convolution
- Dense Connections
- Dropout
- Inception-v3 Module
- Max Pooling
- ReLU
- Softmax
Tasks:
- Image Classification
Training Techniques:
- Gradient Clipping
- Label Smoothing
- RMSProp
- Weight Decay
Training Data:
- ImageNet
Training Resources: 50x NVIDIA Kepler GPUs
ID: tf_inception_v3
LR: 0.045
Dropout: 0.2
Crop Pct: '0.875'
Momentum: 0.9
Image Size: '299'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/inception_v3.py#L449
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_inception_v3-e0069de4.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 77.87%
Top 5 Accuracy: 93.65%
-->

200
hfdocs/source/models/tf-mixnet.mdx
Unescape View File

@ -0,0 +1,200 @@
# (Tensorflow) MixNet
**MixNet** is a type of convolutional neural network discovered via AutoML that utilises [MixConvs](https://paperswithcode.com/method/mixconv) instead of regular [depthwise convolutions](https://paperswithcode.com/method/depthwise-convolution).
The weights from this model were ported from [Tensorflow/TPU](https://github.com/tensorflow/tpu).
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('tf_mixnet_l', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `tf_mixnet_l`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('tf_mixnet_l', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{tan2019mixconv,
title={MixConv: Mixed Depthwise Convolutional Kernels},
author={Mingxing Tan and Quoc V. Le},
year={2019},
eprint={1907.09595},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
<!--
Type: model-index
Collections:
- Name: TF MixNet
Paper:
Title: 'MixConv: Mixed Depthwise Convolutional Kernels'
URL: https://paperswithcode.com/paper/mixnet-mixed-depthwise-convolutional-kernels
Models:
- Name: tf_mixnet_l
In Collection: TF MixNet
Metadata:
FLOPs: 688674516
Parameters: 7330000
File Size: 29620756
Architecture:
- Batch Normalization
- Dense Connections
- Dropout
- Global Average Pooling
- Grouped Convolution
- MixConv
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Techniques:
- MNAS
Training Data:
- ImageNet
ID: tf_mixnet_l
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1720
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_mixnet_l-6c92e0c8.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 78.78%
Top 5 Accuracy: 94.0%
- Name: tf_mixnet_m
In Collection: TF MixNet
Metadata:
FLOPs: 416633502
Parameters: 5010000
File Size: 20310871
Architecture:
- Batch Normalization
- Dense Connections
- Dropout
- Global Average Pooling
- Grouped Convolution
- MixConv
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Techniques:
- MNAS
Training Data:
- ImageNet
ID: tf_mixnet_m
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1709
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_mixnet_m-0f4d8805.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 76.96%
Top 5 Accuracy: 93.16%
- Name: tf_mixnet_s
In Collection: TF MixNet
Metadata:
FLOPs: 302587678
Parameters: 4130000
File Size: 16738218
Architecture:
- Batch Normalization
- Dense Connections
- Dropout
- Global Average Pooling
- Grouped Convolution
- MixConv
- Squeeze-and-Excitation Block
- Swish
Tasks:
- Image Classification
Training Techniques:
- MNAS
Training Data:
- ImageNet
ID: tf_mixnet_s
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/efficientnet.py#L1698
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_mixnet_s-89d3354b.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 75.68%
Top 5 Accuracy: 92.64%
-->

387
hfdocs/source/models/tf-mobilenet-v3.mdx
Unescape View File

@ -0,0 +1,387 @@
# (Tensorflow) MobileNet v3
**MobileNetV3** is a convolutional neural network that is designed for mobile phone CPUs. The network design includes the use of a [hard swish activation](https://paperswithcode.com/method/hard-swish) and [squeeze-and-excitation](https://paperswithcode.com/method/squeeze-and-excitation-block) modules in the [MBConv blocks](https://paperswithcode.com/method/inverted-residual-block).
The weights from this model were ported from [Tensorflow/Models](https://github.com/tensorflow/models).
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('tf_mobilenetv3_large_075', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `tf_mobilenetv3_large_075`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('tf_mobilenetv3_large_075', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@article{DBLP:journals/corr/abs-1905-02244,
author = {Andrew Howard and
Mark Sandler and
Grace Chu and
Liang{-}Chieh Chen and
Bo Chen and
Mingxing Tan and
Weijun Wang and
Yukun Zhu and
Ruoming Pang and
Vijay Vasudevan and
Quoc V. Le and
Hartwig Adam},
title = {Searching for MobileNetV3},
journal = {CoRR},
volume = {abs/1905.02244},
year = {2019},
url = {http://arxiv.org/abs/1905.02244},
archivePrefix = {arXiv},
eprint = {1905.02244},
timestamp = {Tue, 12 Jan 2021 15:30:06 +0100},
biburl = {https://dblp.org/rec/journals/corr/abs-1905-02244.bib},
bibsource = {dblp computer science bibliography, https://dblp.org}
}
```
<!--
Type: model-index
Collections:
- Name: TF MobileNet V3
Paper:
Title: Searching for MobileNetV3
URL: https://paperswithcode.com/paper/searching-for-mobilenetv3
Models:
- Name: tf_mobilenetv3_large_075
In Collection: TF MobileNet V3
Metadata:
FLOPs: 194323712
Parameters: 3990000
File Size: 16097377
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Dense Connections
- Depthwise Separable Convolution
- Dropout
- Global Average Pooling
- Hard Swish
- Inverted Residual Block
- ReLU
- Residual Connection
- Softmax
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- RMSProp
- Weight Decay
Training Data:
- ImageNet
Training Resources: 4x4 TPU Pod
ID: tf_mobilenetv3_large_075
LR: 0.1
Dropout: 0.8
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 4096
Image Size: '224'
Weight Decay: 1.0e-05
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/mobilenetv3.py#L394
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_mobilenetv3_large_075-150ee8b0.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 73.45%
Top 5 Accuracy: 91.34%
- Name: tf_mobilenetv3_large_100
In Collection: TF MobileNet V3
Metadata:
FLOPs: 274535288
Parameters: 5480000
File Size: 22076649
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Dense Connections
- Depthwise Separable Convolution
- Dropout
- Global Average Pooling
- Hard Swish
- Inverted Residual Block
- ReLU
- Residual Connection
- Softmax
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- RMSProp
- Weight Decay
Training Data:
- ImageNet
Training Resources: 4x4 TPU Pod
ID: tf_mobilenetv3_large_100
LR: 0.1
Dropout: 0.8
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 4096
Image Size: '224'
Weight Decay: 1.0e-05
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/mobilenetv3.py#L403
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_mobilenetv3_large_100-427764d5.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 75.51%
Top 5 Accuracy: 92.61%
- Name: tf_mobilenetv3_large_minimal_100
In Collection: TF MobileNet V3
Metadata:
FLOPs: 267216928
Parameters: 3920000
File Size: 15836368
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Dense Connections
- Depthwise Separable Convolution
- Dropout
- Global Average Pooling
- Hard Swish
- Inverted Residual Block
- ReLU
- Residual Connection
- Softmax
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- RMSProp
- Weight Decay
Training Data:
- ImageNet
Training Resources: 4x4 TPU Pod
ID: tf_mobilenetv3_large_minimal_100
LR: 0.1
Dropout: 0.8
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 4096
Image Size: '224'
Weight Decay: 1.0e-05
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/mobilenetv3.py#L412
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_mobilenetv3_large_minimal_100-8596ae28.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 72.24%
Top 5 Accuracy: 90.64%
- Name: tf_mobilenetv3_small_075
In Collection: TF MobileNet V3
Metadata:
FLOPs: 48457664
Parameters: 2040000
File Size: 8242701
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Dense Connections
- Depthwise Separable Convolution
- Dropout
- Global Average Pooling
- Hard Swish
- Inverted Residual Block
- ReLU
- Residual Connection
- Softmax
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- RMSProp
- Weight Decay
Training Data:
- ImageNet
Training Resources: 16x GPUs
ID: tf_mobilenetv3_small_075
LR: 0.045
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 4096
Image Size: '224'
Weight Decay: 4.0e-05
Interpolation: bilinear
RMSProp Decay: 0.9
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/mobilenetv3.py#L421
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_mobilenetv3_small_075-da427f52.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 65.72%
Top 5 Accuracy: 86.13%
- Name: tf_mobilenetv3_small_100
In Collection: TF MobileNet V3
Metadata:
FLOPs: 65450600
Parameters: 2540000
File Size: 10256398
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Dense Connections
- Depthwise Separable Convolution
- Dropout
- Global Average Pooling
- Hard Swish
- Inverted Residual Block
- ReLU
- Residual Connection
- Softmax
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- RMSProp
- Weight Decay
Training Data:
- ImageNet
Training Resources: 16x GPUs
ID: tf_mobilenetv3_small_100
LR: 0.045
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 4096
Image Size: '224'
Weight Decay: 4.0e-05
Interpolation: bilinear
RMSProp Decay: 0.9
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/mobilenetv3.py#L430
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_mobilenetv3_small_100-37f49e2b.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 67.92%
Top 5 Accuracy: 87.68%
- Name: tf_mobilenetv3_small_minimal_100
In Collection: TF MobileNet V3
Metadata:
FLOPs: 60827936
Parameters: 2040000
File Size: 8258083
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Dense Connections
- Depthwise Separable Convolution
- Dropout
- Global Average Pooling
- Hard Swish
- Inverted Residual Block
- ReLU
- Residual Connection
- Softmax
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- RMSProp
- Weight Decay
Training Data:
- ImageNet
Training Resources: 16x GPUs
ID: tf_mobilenetv3_small_minimal_100
LR: 0.045
Crop Pct: '0.875'
Momentum: 0.9
Batch Size: 4096
Image Size: '224'
Weight Decay: 4.0e-05
Interpolation: bilinear
RMSProp Decay: 0.9
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/mobilenetv3.py#L439
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_mobilenetv3_small_minimal_100-922a7843.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 62.91%
Top 5 Accuracy: 84.24%
-->

358
hfdocs/source/models/tresnet.mdx
Unescape View File

@ -0,0 +1,358 @@
# TResNet
A **TResNet** is a variant on a [ResNet](https://paperswithcode.com/method/resnet) that aim to boost accuracy while maintaining GPU training and inference efficiency. They contain several design tricks including a SpaceToDepth stem, [Anti-Alias downsampling](https://paperswithcode.com/method/anti-alias-downsampling), In-Place Activated BatchNorm, Blocks selection and [squeeze-and-excitation layers](https://paperswithcode.com/method/squeeze-and-excitation-block).
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('tresnet_l', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `tresnet_l`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('tresnet_l', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@misc{ridnik2020tresnet,
title={TResNet: High Performance GPU-Dedicated Architecture},
author={Tal Ridnik and Hussam Lawen and Asaf Noy and Emanuel Ben Baruch and Gilad Sharir and Itamar Friedman},
year={2020},
eprint={2003.13630},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
<!--
Type: model-index
Collections:
- Name: TResNet
Paper:
Title: 'TResNet: High Performance GPU-Dedicated Architecture'
URL: https://paperswithcode.com/paper/tresnet-high-performance-gpu-dedicated
Models:
- Name: tresnet_l
In Collection: TResNet
Metadata:
FLOPs: 10873416792
Parameters: 53456696
File Size: 224440219
Architecture:
- 1x1 Convolution
- Anti-Alias Downsampling
- Convolution
- Global Average Pooling
- InPlace-ABN
- Leaky ReLU
- ReLU
- Residual Connection
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- AutoAugment
- Cutout
- Label Smoothing
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA 100 GPUs
ID: tresnet_l
LR: 0.01
Epochs: 300
Crop Pct: '0.875'
Momentum: 0.9
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/tresnet.py#L267
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-tresnet/tresnet_l_81_5-235b486c.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 81.49%
Top 5 Accuracy: 95.62%
- Name: tresnet_l_448
In Collection: TResNet
Metadata:
FLOPs: 43488238584
Parameters: 53456696
File Size: 224440219
Architecture:
- 1x1 Convolution
- Anti-Alias Downsampling
- Convolution
- Global Average Pooling
- InPlace-ABN
- Leaky ReLU
- ReLU
- Residual Connection
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- AutoAugment
- Cutout
- Label Smoothing
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA 100 GPUs
ID: tresnet_l_448
LR: 0.01
Epochs: 300
Crop Pct: '0.875'
Momentum: 0.9
Image Size: '448'
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/tresnet.py#L285
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-tresnet/tresnet_l_448-940d0cd1.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 82.26%
Top 5 Accuracy: 95.98%
- Name: tresnet_m
In Collection: TResNet
Metadata:
FLOPs: 5733048064
Parameters: 41282200
File Size: 125861314
Architecture:
- 1x1 Convolution
- Anti-Alias Downsampling
- Convolution
- Global Average Pooling
- InPlace-ABN
- Leaky ReLU
- ReLU
- Residual Connection
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- AutoAugment
- Cutout
- Label Smoothing
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA 100 GPUs
Training Time: < 24 hours
ID: tresnet_m
LR: 0.01
Epochs: 300
Crop Pct: '0.875'
Momentum: 0.9
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/tresnet.py#L261
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-tresnet/tresnet_m_80_8-dbc13962.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 80.8%
Top 5 Accuracy: 94.86%
- Name: tresnet_m_448
In Collection: TResNet
Metadata:
FLOPs: 22929743104
Parameters: 29278464
File Size: 125861314
Architecture:
- 1x1 Convolution
- Anti-Alias Downsampling
- Convolution
- Global Average Pooling
- InPlace-ABN
- Leaky ReLU
- ReLU
- Residual Connection
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- AutoAugment
- Cutout
- Label Smoothing
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA 100 GPUs
ID: tresnet_m_448
LR: 0.01
Epochs: 300
Crop Pct: '0.875'
Momentum: 0.9
Image Size: '448'
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/tresnet.py#L279
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-tresnet/tresnet_m_448-bc359d10.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 81.72%
Top 5 Accuracy: 95.57%
- Name: tresnet_xl
In Collection: TResNet
Metadata:
FLOPs: 15162534034
Parameters: 75646610
File Size: 314378965
Architecture:
- 1x1 Convolution
- Anti-Alias Downsampling
- Convolution
- Global Average Pooling
- InPlace-ABN
- Leaky ReLU
- ReLU
- Residual Connection
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- AutoAugment
- Cutout
- Label Smoothing
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA 100 GPUs
ID: tresnet_xl
LR: 0.01
Epochs: 300
Crop Pct: '0.875'
Momentum: 0.9
Image Size: '224'
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/tresnet.py#L273
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-tresnet/tresnet_xl_82_0-a2d51b00.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 82.05%
Top 5 Accuracy: 95.93%
- Name: tresnet_xl_448
In Collection: TResNet
Metadata:
FLOPs: 60641712730
Parameters: 75646610
File Size: 224440219
Architecture:
- 1x1 Convolution
- Anti-Alias Downsampling
- Convolution
- Global Average Pooling
- InPlace-ABN
- Leaky ReLU
- ReLU
- Residual Connection
- Squeeze-and-Excitation Block
Tasks:
- Image Classification
Training Techniques:
- AutoAugment
- Cutout
- Label Smoothing
- SGD with Momentum
- Weight Decay
Training Data:
- ImageNet
Training Resources: 8x NVIDIA 100 GPUs
ID: tresnet_xl_448
LR: 0.01
Epochs: 300
Crop Pct: '0.875'
Momentum: 0.9
Image Size: '448'
Weight Decay: 0.0001
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/9a25fdf3ad0414b4d66da443fe60ae0aa14edc84/timm/models/tresnet.py#L291
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-tresnet/tresnet_l_448-940d0cd1.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 83.06%
Top 5 Accuracy: 96.19%
-->

169
hfdocs/source/models/wide-resnet.mdx
Unescape View File

@ -0,0 +1,169 @@
# Wide ResNet
**Wide Residual Networks** are a variant on [ResNets](https://paperswithcode.com/method/resnet) where we decrease depth and increase the width of residual networks. This is achieved through the use of [wide residual blocks](https://paperswithcode.com/method/wide-residual-block).
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('wide_resnet101_2', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `wide_resnet101_2`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('wide_resnet101_2', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@article{DBLP:journals/corr/ZagoruykoK16,
author = {Sergey Zagoruyko and
Nikos Komodakis},
title = {Wide Residual Networks},
journal = {CoRR},
volume = {abs/1605.07146},
year = {2016},
url = {http://arxiv.org/abs/1605.07146},
archivePrefix = {arXiv},
eprint = {1605.07146},
timestamp = {Mon, 13 Aug 2018 16:46:42 +0200},
biburl = {https://dblp.org/rec/journals/corr/ZagoruykoK16.bib},
bibsource = {dblp computer science bibliography, https://dblp.org}
}
```
<!--
Type: model-index
Collections:
- Name: Wide ResNet
Paper:
Title: Wide Residual Networks
URL: https://paperswithcode.com/paper/wide-residual-networks
Models:
- Name: wide_resnet101_2
In Collection: Wide ResNet
Metadata:
FLOPs: 29304929280
Parameters: 126890000
File Size: 254695146
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Connection
- Softmax
- Wide Residual Block
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: wide_resnet101_2
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bilinear
Code: https://github.com/rwightman/pytorch-image-models/blob/5f9aff395c224492e9e44248b15f44b5cc095d9c/timm/models/resnet.py#L802
Weights: https://download.pytorch.org/models/wide_resnet101_2-32ee1156.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 78.85%
Top 5 Accuracy: 94.28%
- Name: wide_resnet50_2
In Collection: Wide ResNet
Metadata:
FLOPs: 14688058368
Parameters: 68880000
File Size: 275853271
Architecture:
- 1x1 Convolution
- Batch Normalization
- Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Connection
- Softmax
- Wide Residual Block
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: wide_resnet50_2
Crop Pct: '0.875'
Image Size: '224'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/5f9aff395c224492e9e44248b15f44b5cc095d9c/timm/models/resnet.py#L790
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/wide_resnet50_racm-8234f177.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 81.45%
Top 5 Accuracy: 95.52%
-->

230
hfdocs/source/models/xception.mdx
Unescape View File

@ -0,0 +1,230 @@
# Xception
**Xception** is a convolutional neural network architecture that relies solely on [depthwise separable convolution layers](https://paperswithcode.com/method/depthwise-separable-convolution).
The weights from this model were ported from [Tensorflow/Models](https://github.com/tensorflow/models).
## How do I use this model on an image?
To load a pretrained model:
```py
>>> import timm
>>> model = timm.create_model('xception', pretrained=True)
>>> model.eval()
```
To load and preprocess the image:
```py
>>> import urllib
>>> from PIL import Image
>>> from timm.data import resolve_data_config
>>> from timm.data.transforms_factory import create_transform
>>> config = resolve_data_config({}, model=model)
>>> transform = create_transform(**config)
>>> url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
>>> urllib.request.urlretrieve(url, filename)
>>> img = Image.open(filename).convert('RGB')
>>> tensor = transform(img).unsqueeze(0) # transform and add batch dimension
```
To get the model predictions:
```py
>>> import torch
>>> with torch.no_grad():
... out = model(tensor)
>>> probabilities = torch.nn.functional.softmax(out[0], dim=0)
>>> print(probabilities.shape)
>>> # prints: torch.Size([1000])
```
To get the top-5 predictions class names:
```py
>>> # Get imagenet class mappings
>>> url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
>>> urllib.request.urlretrieve(url, filename)
>>> with open("imagenet_classes.txt", "r") as f:
... categories = [s.strip() for s in f.readlines()]
>>> # Print top categories per image
>>> top5_prob, top5_catid = torch.topk(probabilities, 5)
>>> for i in range(top5_prob.size(0)):
... print(categories[top5_catid[i]], top5_prob[i].item())
>>> # prints class names and probabilities like:
>>> # [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]
```
Replace the model name with the variant you want to use, e.g. `xception`. You can find the IDs in the model summaries at the top of this page.
To extract image features with this model, follow the [timm feature extraction examples](https://rwightman.github.io/pytorch-image-models/feature_extraction/), just change the name of the model you want to use.
## How do I finetune this model?
You can finetune any of the pre-trained models just by changing the classifier (the last layer).
```py
>>> model = timm.create_model('xception', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)
```
To finetune on your own dataset, you have to write a training loop or adapt [timm's training
script](https://github.com/rwightman/pytorch-image-models/blob/master/train.py) to use your dataset.
## How do I train this model?
You can follow the [timm recipe scripts](https://rwightman.github.io/pytorch-image-models/scripts/) for training a new model afresh.
## Citation
```BibTeX
@article{DBLP:journals/corr/ZagoruykoK16,
@misc{chollet2017xception,
title={Xception: Deep Learning with Depthwise Separable Convolutions},
author={François Chollet},
year={2017},
eprint={1610.02357},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
<!--
Type: model-index
Collections:
- Name: Xception
Paper:
Title: 'Xception: Deep Learning with Depthwise Separable Convolutions'
URL: https://paperswithcode.com/paper/xception-deep-learning-with-depthwise
Models:
- Name: xception
In Collection: Xception
Metadata:
FLOPs: 10600506792
Parameters: 22860000
File Size: 91675053
Architecture:
- 1x1 Convolution
- Convolution
- Dense Connections
- Depthwise Separable Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: xception
Crop Pct: '0.897'
Image Size: '299'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/xception.py#L229
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-cadene/xception-43020ad28.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.05%
Top 5 Accuracy: 94.4%
- Name: xception41
In Collection: Xception
Metadata:
FLOPs: 11681983232
Parameters: 26970000
File Size: 108422028
Architecture:
- 1x1 Convolution
- Convolution
- Dense Connections
- Depthwise Separable Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: xception41
Crop Pct: '0.903'
Image Size: '299'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/xception_aligned.py#L181
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_xception_41-e6439c97.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 78.54%
Top 5 Accuracy: 94.28%
- Name: xception65
In Collection: Xception
Metadata:
FLOPs: 17585702144
Parameters: 39920000
File Size: 160536780
Architecture:
- 1x1 Convolution
- Convolution
- Dense Connections
- Depthwise Separable Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: xception65
Crop Pct: '0.903'
Image Size: '299'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/xception_aligned.py#L200
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_xception_65-c9ae96e8.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.55%
Top 5 Accuracy: 94.66%
- Name: xception71
In Collection: Xception
Metadata:
FLOPs: 22817346560
Parameters: 42340000
File Size: 170295556
Architecture:
- 1x1 Convolution
- Convolution
- Dense Connections
- Depthwise Separable Convolution
- Global Average Pooling
- Max Pooling
- ReLU
- Residual Connection
- Softmax
Tasks:
- Image Classification
Training Data:
- ImageNet
ID: xception71
Crop Pct: '0.903'
Image Size: '299'
Interpolation: bicubic
Code: https://github.com/rwightman/pytorch-image-models/blob/d8e69206be253892b2956341fea09fdebfaae4e3/timm/models/xception_aligned.py#L219
Weights: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_xception_71-8eec7df1.pth
Results:
- Task: Image Classification
Dataset: ImageNet
Metrics:
Top 1 Accuracy: 79.88%
Top 5 Accuracy: 94.93%
-->

67
hfdocs/source/results.mdx
Unescape View File

@ -0,0 +1,67 @@
# Results
CSV files containing an ImageNet-1K and out-of-distribution (OOD) test set validation results for all models with pretrained weights is located in the repository [results folder](https://github.com/rwightman/pytorch-image-models/tree/master/results).
## Self-trained Weights
The table below includes ImageNet-1k validation results of model weights that I've trained myself. It is not updated as frequently as the csv results outputs linked above.
|Model | Acc@1 (Err) | Acc@5 (Err) | Param # (M) | Interpolation | Image Size |
|---|---|---|---|---|---|
| efficientnet_b3a | 82.242 (17.758) | 96.114 (3.886) | 12.23 | bicubic | 320 (1.0 crop) |
| efficientnet_b3 | 82.076 (17.924) | 96.020 (3.980) | 12.23 | bicubic | 300 |
| regnet_32 | 82.002 (17.998) | 95.906 (4.094) | 19.44 | bicubic | 224 |
| skresnext50d_32x4d | 81.278 (18.722) | 95.366 (4.634) | 27.5 | bicubic | 288 (1.0 crop) |
| seresnext50d_32x4d | 81.266 (18.734) | 95.620 (4.380) | 27.6 | bicubic | 224 |
| efficientnet_b2a | 80.608 (19.392) | 95.310 (4.690) | 9.11 | bicubic | 288 (1.0 crop) |
| resnet50d | 80.530 (19.470) | 95.160 (4.840) | 25.6 | bicubic | 224 |
| mixnet_xl | 80.478 (19.522) | 94.932 (5.068) | 11.90 | bicubic | 224 |
| efficientnet_b2 | 80.402 (19.598) | 95.076 (4.924) | 9.11 | bicubic | 260 |
| seresnet50 | 80.274 (19.726) | 95.070 (4.930) | 28.1 | bicubic | 224 |
| skresnext50d_32x4d | 80.156 (19.844) | 94.642 (5.358) | 27.5 | bicubic | 224 |
| cspdarknet53 | 80.058 (19.942) | 95.084 (4.916) | 27.6 | bicubic | 256 |
| cspresnext50 | 80.040 (19.960) | 94.944 (5.056) | 20.6 | bicubic | 224 |
| resnext50_32x4d | 79.762 (20.238) | 94.600 (5.400) | 25 | bicubic | 224 |
| resnext50d_32x4d | 79.674 (20.326) | 94.868 (5.132) | 25.1 | bicubic | 224 |
| cspresnet50 | 79.574 (20.426) | 94.712 (5.288) | 21.6 | bicubic | 256 |
| ese_vovnet39b | 79.320 (20.680) | 94.710 (5.290) | 24.6 | bicubic | 224 |
| resnetblur50 | 79.290 (20.710) | 94.632 (5.368) | 25.6 | bicubic | 224 |
| dpn68b | 79.216 (20.784) | 94.414 (5.586) | 12.6 | bicubic | 224 |
| resnet50 | 79.038 (20.962) | 94.390 (5.610) | 25.6 | bicubic | 224 |
| mixnet_l | 78.976 (21.024 | 94.184 (5.816) | 7.33 | bicubic | 224 |
| efficientnet_b1 | 78.692 (21.308) | 94.086 (5.914) | 7.79 | bicubic | 240 |
| efficientnet_es | 78.066 (21.934) | 93.926 (6.074) | 5.44 | bicubic | 224 |
| seresnext26t_32x4d | 77.998 (22.002) | 93.708 (6.292) | 16.8 | bicubic | 224 |
| seresnext26tn_32x4d | 77.986 (22.014) | 93.746 (6.254) | 16.8 | bicubic | 224 |
| efficientnet_b0 | 77.698 (22.302) | 93.532 (6.468) | 5.29 | bicubic | 224 |
| seresnext26d_32x4d | 77.602 (22.398) | 93.608 (6.392) | 16.8 | bicubic | 224 |
| mobilenetv2_120d | 77.294 (22.706 | 93.502 (6.498) | 5.8 | bicubic | 224 |
| mixnet_m | 77.256 (22.744) | 93.418 (6.582) | 5.01 | bicubic | 224 |
| resnet34d | 77.116 (22.884) | 93.382 (6.618) | 21.8 | bicubic | 224 |
| seresnext26_32x4d | 77.104 (22.896) | 93.316 (6.684) | 16.8 | bicubic | 224 |
| skresnet34 | 76.912 (23.088) | 93.322 (6.678) | 22.2 | bicubic | 224 |
| ese_vovnet19b_dw | 76.798 (23.202) | 93.268 (6.732) | 6.5 | bicubic | 224 |
| resnet26d | 76.68 (23.32) | 93.166 (6.834) | 16 | bicubic | 224 |
| densenetblur121d | 76.576 (23.424) | 93.190 (6.810) | 8.0 | bicubic | 224 |
| mobilenetv2_140 | 76.524 (23.476) | 92.990 (7.010) | 6.1 | bicubic | 224 |
| mixnet_s | 75.988 (24.012) | 92.794 (7.206) | 4.13 | bicubic | 224 |
| mobilenetv3_large_100 | 75.766 (24.234) | 92.542 (7.458) | 5.5 | bicubic | 224 |
| mobilenetv3_rw | 75.634 (24.366) | 92.708 (7.292) | 5.5 | bicubic | 224 |
| mnasnet_a1 | 75.448 (24.552) | 92.604 (7.396) | 3.89 | bicubic | 224 |
| resnet26 | 75.292 (24.708) | 92.57 (7.43) | 16 | bicubic | 224 |
| fbnetc_100 | 75.124 (24.876) | 92.386 (7.614) | 5.6 | bilinear | 224 |
| resnet34 | 75.110 (24.890) | 92.284 (7.716) | 22 | bilinear | 224 |
| mobilenetv2_110d | 75.052 (24.948) | 92.180 (7.820) | 4.5 | bicubic | 224 |
| seresnet34 | 74.808 (25.192) | 92.124 (7.876) | 22 | bilinear | 224 |
| mnasnet_b1 | 74.658 (25.342) | 92.114 (7.886) | 4.38 | bicubic | 224 |
| spnasnet_100 | 74.084 (25.916) | 91.818 (8.182) | 4.42 | bilinear | 224 |
| skresnet18 | 73.038 (26.962) | 91.168 (8.832) | 11.9 | bicubic | 224 |
| mobilenetv2_100 | 72.978 (27.022) | 91.016 (8.984) | 3.5 | bicubic | 224 |
| resnet18d | 72.260 (27.740) | 90.696 (9.304) | 11.7 | bicubic | 224 |
| seresnet18 | 71.742 (28.258) | 90.334 (9.666) | 11.8 | bicubic | 224 |
## Ported and Other Weights
For weights ported from other deep learning frameworks (Tensorflow, MXNet GluonCV) or copied from other PyTorch sources, please see the full results tables for ImageNet and various OOD test sets at in the [results tables](https://github.com/rwightman/pytorch-image-models/tree/master/results).
Model code .py files contain links to original sources of models and weights.

35
hfdocs/source/scripts.mdx
Unescape View File

@ -0,0 +1,35 @@
# Scripts
A train, validation, inference, and checkpoint cleaning script included in the github root folder. Scripts are not currently packaged in the pip release.
The training and validation scripts evolved from early versions of the [PyTorch Imagenet Examples](https://github.com/pytorch/examples). I have added significant functionality over time, including CUDA specific performance enhancements based on
[NVIDIA's APEX Examples](https://github.com/NVIDIA/apex/tree/master/examples).
## Training Script
The variety of training args is large and not all combinations of options (or even options) have been fully tested. For the training dataset folder, specify the folder to the base that contains a `train` and `validation` folder.
To train an SE-ResNet34 on ImageNet, locally distributed, 4 GPUs, one process per GPU w/ cosine schedule, random-erasing prob of 50% and per-pixel random value:
```bash
./distributed_train.sh 4 /data/imagenet --model seresnet34 --sched cosine --epochs 150 --warmup-epochs 5 --lr 0.4 --reprob 0.5 --remode pixel --batch-size 256 --amp -j 4
```
<Tip>
It is recommended to use PyTorch 1.9+ w/ PyTorch native AMP and DDP instead of APEX AMP. --amp defaults to native AMP as of timm ver 0.4.3. --apex-amp will force use of APEX components if they are installed.
</Tip>
## Validation / Inference Scripts
Validation and inference scripts are similar in usage. One outputs metrics on a validation set and the other outputs topk class ids in a csv. Specify the folder containing validation images, not the base as in training script.
To validate with the model's pretrained weights (if they exist):
```bash
python validate.py /imagenet/validation/ --model seresnext26_32x4d --pretrained
```
To run inference from a checkpoint:
```bash
python inference.py /imagenet/validation/ --model mobilenetv3_large_100 --checkpoint ./output/train/model_best.pth.tar
```

63
hfdocs/source/training_hparam_examples.mdx
Unescape View File

@ -0,0 +1,63 @@
# Training Examples
## EfficientNet-B2 with RandAugment - 80.4 top-1, 95.1 top-5
These params are for dual Titan RTX cards with NVIDIA Apex installed:
```bash
./distributed_train.sh 2 /imagenet/ --model efficientnet_b2 -b 128 --sched step --epochs 450 --decay-epochs 2.4 --decay-rate .97 --opt rmsproptf --opt-eps .001 -j 8 --warmup-lr 1e-6 --weight-decay 1e-5 --drop 0.3 --drop-path 0.2 --model-ema --model-ema-decay 0.9999 --aa rand-m9-mstd0.5 --remode pixel --reprob 0.2 --amp --lr .016
```
## MixNet-XL with RandAugment - 80.5 top-1, 94.9 top-5
This params are for dual Titan RTX cards with NVIDIA Apex installed:
```bash
./distributed_train.sh 2 /imagenet/ --model mixnet_xl -b 128 --sched step --epochs 450 --decay-epochs 2.4 --decay-rate .969 --opt rmsproptf --opt-eps .001 -j 8 --warmup-lr 1e-6 --weight-decay 1e-5 --drop 0.3 --drop-path 0.2 --model-ema --model-ema-decay 0.9999 --aa rand-m9-mstd0.5 --remode pixel --reprob 0.3 --amp --lr .016 --dist-bn reduce
```
## SE-ResNeXt-26-D and SE-ResNeXt-26-T
These hparams (or similar) work well for a wide range of ResNet architecture, generally a good idea to increase the epoch # as the model size increases... ie approx 180-200 for ResNe(X)t50, and 220+ for larger. Increase batch size and LR proportionally for better GPUs or with AMP enabled. These params were for 2 1080Ti cards:
```bash
./distributed_train.sh 2 /imagenet/ --model seresnext26t_32x4d --lr 0.1 --warmup-epochs 5 --epochs 160 --weight-decay 1e-4 --sched cosine --reprob 0.4 --remode pixel -b 112
```
## EfficientNet-B3 with RandAugment - 81.5 top-1, 95.7 top-5
The training of this model started with the same command line as EfficientNet-B2 w/ RA above. After almost three weeks of training the process crashed. The results weren't looking amazing so I resumed the training several times with tweaks to a few params (increase RE prob, decrease rand-aug, increase ema-decay). Nothing looked great. I ended up averaging the best checkpoints from all restarts. The result is mediocre at default res/crop but oddly performs much better with a full image test crop of 1.0.
## EfficientNet-B0 with RandAugment - 77.7 top-1, 95.3 top-5
[Michael Klachko](https://github.com/michaelklachko) achieved these results with the command line for B2 adapted for larger batch size, with the recommended B0 dropout rate of 0.2.
```bash
./distributed_train.sh 2 /imagenet/ --model efficientnet_b0 -b 384 --sched step --epochs 450 --decay-epochs 2.4 --decay-rate .97 --opt rmsproptf --opt-eps .001 -j 8 --warmup-lr 1e-6 --weight-decay 1e-5 --drop 0.2 --drop-path 0.2 --model-ema --model-ema-decay 0.9999 --aa rand-m9-mstd0.5 --remode pixel --reprob 0.2 --amp --lr .048
```
## ResNet50 with JSD loss and RandAugment (clean + 2x RA augs) - 79.04 top-1, 94.39 top-5
Trained on two older 1080Ti cards, this took a while. Only slightly, non statistically better ImageNet validation result than my first good AugMix training of 78.99. However, these weights are more robust on tests with ImageNetV2, ImageNet-Sketch, etc. Unlike my first AugMix runs, I've enabled SplitBatchNorm, disabled random erasing on the clean split, and cranked up random erasing prob on the 2 augmented paths.
```bash
./distributed_train.sh 2 /imagenet -b 64 --model resnet50 --sched cosine --epochs 200 --lr 0.05 --amp --remode pixel --reprob 0.6 --aug-splits 3 --aa rand-m9-mstd0.5-inc1 --resplit --split-bn --jsd --dist-bn reduce
```
## EfficientNet-ES (EdgeTPU-Small) with RandAugment - 78.066 top-1, 93.926 top-5
Trained by [Andrew Lavin](https://github.com/andravin) with 8 V100 cards. Model EMA was not used, final checkpoint is the average of 8 best checkpoints during training.
```bash
./distributed_train.sh 8 /imagenet --model efficientnet_es -b 128 --sched step --epochs 450 --decay-epochs 2.4 --decay-rate .97 --opt rmsproptf --opt-eps .001 -j 8 --warmup-lr 1e-6 --weight-decay 1e-5 --drop 0.2 --drop-path 0.2 --aa rand-m9-mstd0.5 --remode pixel --reprob 0.2 --amp --lr .064
```
## MobileNetV3-Large-100 - 75.766 top-1, 92,542 top-5
```bash
./distributed_train.sh 2 /imagenet/ --model mobilenetv3_large_100 -b 512 --sched step --epochs 600 --decay-epochs 2.4 --decay-rate .973 --opt rmsproptf --opt-eps .001 -j 7 --warmup-lr 1e-6 --weight-decay 1e-5 --drop 0.2 --drop-path 0.2 --model-ema --model-ema-decay 0.9999 --aa rand-m9-mstd0.5 --remode pixel --reprob 0.2 --amp --lr .064 --lr-noise 0.42 0.9
```
## ResNeXt-50 32x4d w/ RandAugment - 79.762 top-1, 94.60 top-5
These params will also work well for SE-ResNeXt-50 and SK-ResNeXt-50 and likely 101. I used them for the SK-ResNeXt-50 32x4d that I trained with 2 GPU using a slightly higher LR per effective batch size (lr=0.18, b=192 per GPU). The cmd line below are tuned for 8 GPU training.
```bash
./distributed_train.sh 8 /imagenet --model resnext50_32x4d --lr 0.6 --warmup-epochs 5 --epochs 240 --weight-decay 1e-4 --sched cosine --reprob 0.4 --recount 3 --remode pixel --aa rand-m7-mstd0.5-inc1 -b 192 -j 6 --amp --dist-bn reduce
```
Write Preview
Loading…
Cancel
Save