* 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
* 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)
* 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)
* 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
For each competition, personal, or freelance project involving images + Convolution Neural Networks, I build on top of an evolving collection of code and models. This repo contains a (somewhat) cleaned up and paired down iteration of that code. Hopefully it'll be of use to others.
The work of many others is present here. I've tried to make sure all source material is acknowledged:
* Training/validation scripts evolved from early versions of the [PyTorch Imagenet Examples](https://github.com/pytorch/examples)
* CUDA specific performance enhancements have been pulled from [NVIDIA's APEX Examples](https://github.com/NVIDIA/apex/tree/master/examples)
* LR scheduler ideas from [AllenNLP](https://github.com/allenai/allennlp/tree/master/allennlp/training/learning_rate_schedulers), [FAIRseq](https://github.com/pytorch/fairseq/tree/master/fairseq/optim/lr_scheduler), and SGDR: Stochastic Gradient Descent with Warm Restarts (https://arxiv.org/abs/1608.03983)
* Random Erasing from [Zhun Zhong](https://github.com/zhunzhong07/Random-Erasing/blob/master/transforms.py) (https://arxiv.org/abs/1708.04896)
I've included a few of my favourite models, but this is not an exhaustive collection. You can't do better than [Cadene's](https://github.com/Cadene/pretrained-models.pytorch) collection in that regard. Most models do have pretrained weights from their respective sources or original authors.
* Instagram trained / ImageNet tuned ResNeXt101-32x8d to 32x48d from from [facebookresearch](https://pytorch.org/hub/facebookresearch_WSL-Images_resnext/)
* Squeeze-and-Excitation ResNet/ResNeXt (from [Cadene](https://github.com/Cadene/pretrained-models.pytorch) with some pretrained weight additions by myself)
* EfficientNet (from my standalone [GenEfficientNet](https://github.com/rwightman/gen-efficientnet-pytorch)) - A generic model that implements many of the efficient models that utilize similar DepthwiseSeparable and InvertedResidual blocks
Several (less common) features that I often utilize in my projects are included. Many of their additions are the reason why I maintain my own set of models, instead of using others' via PIP:
* All models have a common default configuration interface and API for
* accessing/changing the classifier - `get_classifier` and `reset_classifier`
* doing a forward pass on just the features - `forward_features`
* these makes it easy to write consistent network wrappers that work with any of the models
* All models have a consistent pretrained weight loader that adapts last linear if necessary, and from 3 to 1 channel input if desired
* The train script works in several process/GPU modes:
* NVIDIA DDP w/ a single GPU per process, multiple processes with APEX present (AMP mixed-precision optional)
* PyTorch DistributedDataParallel w/ multi-gpu, single process (AMP disabled as it crashes when enabled)
* PyTorch w/ single GPU single process (AMP optional)
* A dynamic global pool implementation that allows selecting from average pooling, max pooling, average + max, or concat([average, max]) at model creation. All global pooling is adaptive average by default and compatible with pretrained weights.
* A 'Test Time Pool' wrapper that can wrap any of the included models and usually provide improved performance doing inference with input images larger than the training size. Idea adapted from original DPN implementation when I ported (https://github.com/cypw/DPNs)
* AutoAugment (https://arxiv.org/abs/1805.09501) and RandAugment (https://arxiv.org/abs/1909.13719) ImageNet configurations modeled after impl for EfficientNet training (https://github.com/tensorflow/tpu/blob/master/models/official/efficientnet/autoaugment.py)
A CSV file containing an ImageNet-1K validation results summary for all included models with pretrained weights and default configurations is located [here](results/results-all.csv)
I've leveraged the training scripts in this repository to train a few of the models with missing weights to good levels of performance. These numbers are all for 224x224 training and validation image sizing with the usual 87.5% validation crop.
For the models below, the model code and weight porting from Tensorflow or MXNet Gluon to Pytorch was done by myself. There are weights/models ported by others included in this repository, they are not listed below.
The `tf_efficientnet`, `tf_mixnet` models require an equivalent for 'SAME' padding as their arch results in asymmetric padding. I've added this in the model creation wrapper, but it does come with a performance penalty.
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:
### 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.
[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.
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.
### 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.
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.
All development and testing has been done in Conda Python 3 environments on Linux x86-64 systems, specifically Python 3.6.x and 3.7.x. Little to no care has been taken to be Python 2.x friendly and I don't plan to 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.
This package can be installed via pip. Currently, the model factory (`timm.create_model`) is the most useful component to use via a pip install.
Install (after conda env/install):
```
pip install timm
```
Use:
```
>>> import timm
>>> m = timm.create_model('mobilenetv3_100', pretrained=True)
>>> m.eval()
```
### 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 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:
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):
* Complete feature map extraction across all model types and build obj detection/segmentation models and scripts (or integrate backbones with mmdetection, detectron2)
