From bc998cad9153316fe51891c4074997827549166e Mon Sep 17 00:00:00 2001 From: Ross Wightman Date: Wed, 18 Mar 2020 13:54:06 -0700 Subject: [PATCH 1/3] Experimenting with some MobileNetV2 variations to compare against EfficientNet-Lite --- timm/models/efficientnet.py | 29 +++++++++++++++++++++++++++-- 1 file changed, 27 insertions(+), 2 deletions(-) diff --git a/timm/models/efficientnet.py b/timm/models/efficientnet.py index cddb750e..52c5c81c 100644 --- a/timm/models/efficientnet.py +++ b/timm/models/efficientnet.py @@ -61,6 +61,10 @@ default_cfgs = { 'mnasnet_small': _cfg(url=''), 'mobilenetv2_100': _cfg(url=''), + 'mobilenetv2_100d': _cfg(url=''), + 'mobilenetv2_110d': _cfg(url=''), + 'mobilenetv2_140': _cfg(url=''), + 'fbnetc_100': _cfg( url='https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/fbnetc_100-c345b898.pth', interpolation='bilinear'), @@ -565,7 +569,7 @@ def _gen_mnasnet_small(variant, channel_multiplier=1.0, pretrained=False, **kwar return model -def _gen_mobilenet_v2(variant, channel_multiplier=1.0, pretrained=False, **kwargs): +def _gen_mobilenet_v2(variant, channel_multiplier=1.0, depth_multiplier=1.0, pretrained=False, **kwargs): """ Generate MobileNet-V2 network Ref impl: https://github.com/tensorflow/models/blob/master/research/slim/nets/mobilenet/mobilenet_v2.py Paper: https://arxiv.org/abs/1801.04381 @@ -580,7 +584,7 @@ def _gen_mobilenet_v2(variant, channel_multiplier=1.0, pretrained=False, **kwarg ['ir_r1_k3_s1_e6_c320'], ] model_kwargs = dict( - block_args=decode_arch_def(arch_def), + block_args=decode_arch_def(arch_def, depth_multiplier=depth_multiplier), stem_size=32, channel_multiplier=channel_multiplier, norm_kwargs=resolve_bn_args(kwargs), @@ -950,6 +954,27 @@ def mobilenetv2_100(pretrained=False, **kwargs): return model +@register_model +def mobilenetv2_100d(pretrained=False, **kwargs): + """ MobileNet V2 """ + model = _gen_mobilenet_v2('mobilenetv2_100d', 1.0, depth_multiplier=1.1, pretrained=pretrained, **kwargs) + return model + + +@register_model +def mobilenetv2_110d(pretrained=False, **kwargs): + """ MobileNet V2 """ + model = _gen_mobilenet_v2('mobilenetv2_110d', 1.1, depth_multiplier=1.2, pretrained=pretrained, **kwargs) + return model + + +@register_model +def mobilenetv2_140(pretrained=False, **kwargs): + """ MobileNet V2 """ + model = _gen_mobilenet_v2('mobilenetv2_140', 1.4, pretrained=pretrained, **kwargs) + return model + + @register_model def fbnetc_100(pretrained=False, **kwargs): """ FBNet-C """ From a6a5565de71b47969801294b594787926c1034a8 Mon Sep 17 00:00:00 2001 From: Ross Wightman Date: Thu, 19 Mar 2020 10:20:20 -0700 Subject: [PATCH 2/3] Fiddling... --- timm/models/efficientnet.py | 20 +++++++++++++------- 1 file changed, 13 insertions(+), 7 deletions(-) diff --git a/timm/models/efficientnet.py b/timm/models/efficientnet.py index 52c5c81c..42435b0d 100644 --- a/timm/models/efficientnet.py +++ b/timm/models/efficientnet.py @@ -322,6 +322,7 @@ class EfficientNet(nn.Module): # Stem if not fix_stem: stem_size = round_channels(stem_size, channel_multiplier, channel_divisor, channel_min) + print(stem_size) self.conv_stem = create_conv2d(self._in_chs, stem_size, 3, stride=2, padding=pad_type) self.bn1 = norm_layer(stem_size, **norm_kwargs) self.act1 = act_layer(inplace=True) @@ -569,7 +570,8 @@ def _gen_mnasnet_small(variant, channel_multiplier=1.0, pretrained=False, **kwar return model -def _gen_mobilenet_v2(variant, channel_multiplier=1.0, depth_multiplier=1.0, pretrained=False, **kwargs): +def _gen_mobilenet_v2( + variant, channel_multiplier=1.0, depth_multiplier=1.0, fix_stem_head=False, pretrained=False, **kwargs): """ Generate MobileNet-V2 network Ref impl: https://github.com/tensorflow/models/blob/master/research/slim/nets/mobilenet/mobilenet_v2.py Paper: https://arxiv.org/abs/1801.04381 @@ -584,8 +586,10 @@ def _gen_mobilenet_v2(variant, channel_multiplier=1.0, depth_multiplier=1.0, pre ['ir_r1_k3_s1_e6_c320'], ] model_kwargs = dict( - block_args=decode_arch_def(arch_def, depth_multiplier=depth_multiplier), + block_args=decode_arch_def(arch_def, depth_multiplier=depth_multiplier, fix_first_last=fix_stem_head), + num_features=1280 if fix_stem_head else round_channels(1280, channel_multiplier, 8, None), stem_size=32, + fix_stem=fix_stem_head, channel_multiplier=channel_multiplier, norm_kwargs=resolve_bn_args(kwargs), act_layer=nn.ReLU6, @@ -955,23 +959,25 @@ def mobilenetv2_100(pretrained=False, **kwargs): @register_model -def mobilenetv2_100d(pretrained=False, **kwargs): +def mobilenetv2_140(pretrained=False, **kwargs): """ MobileNet V2 """ - model = _gen_mobilenet_v2('mobilenetv2_100d', 1.0, depth_multiplier=1.1, pretrained=pretrained, **kwargs) + model = _gen_mobilenet_v2('mobilenetv2_140', 1.4, pretrained=pretrained, **kwargs) return model @register_model def mobilenetv2_110d(pretrained=False, **kwargs): """ MobileNet V2 """ - model = _gen_mobilenet_v2('mobilenetv2_110d', 1.1, depth_multiplier=1.2, pretrained=pretrained, **kwargs) + model = _gen_mobilenet_v2( + 'mobilenetv2_100d', 1.1, depth_multiplier=1.2, fix_stem_head=True, pretrained=pretrained, **kwargs) return model @register_model -def mobilenetv2_140(pretrained=False, **kwargs): +def mobilenetv2_120d(pretrained=False, **kwargs): """ MobileNet V2 """ - model = _gen_mobilenet_v2('mobilenetv2_140', 1.4, pretrained=pretrained, **kwargs) + model = _gen_mobilenet_v2( + 'mobilenetv2_110d', 1.2, depth_multiplier=1.4, fix_stem_head=True, pretrained=pretrained, **kwargs) return model From e34074b4dabab936fe0cf7997ff7601ac5b05839 Mon Sep 17 00:00:00 2001 From: Ross Wightman Date: Sun, 5 Apr 2020 17:31:02 -0700 Subject: [PATCH 3/3] Add final weights for MobileNet-V2 experiments --- README.md | 11 +++++++++++ timm/models/efficientnet.py | 24 ++++++++++++++---------- 2 files changed, 25 insertions(+), 10 deletions(-) diff --git a/README.md b/README.md index 56112e5c..6e6dd8bb 100644 --- a/README.md +++ b/README.md @@ -2,6 +2,13 @@ ## What's New +### 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) @@ -193,10 +200,12 @@ I've leveraged the training scripts in this repository to train a few of the mod | seresnext26tn_32x4d | 77.986 (22.014) | 93.746 (6.254) | 16.8M | bicubic | 224 | | efficientnet_b0 | 77.698 (22.302) | 93.532 (6.468) | 5.29M | bicubic | 224 | | seresnext26d_32x4d | 77.602 (22.398) | 93.608 (6.392) | 16.8M | bicubic | 224 | +| mobilenetv2_120d | 77.294 (22.706 | 93.502 (6.498) | 5.8M | bicubic | 224 | | mixnet_m | 77.256 (22.744) | 93.418 (6.582) | 5.01M | bicubic | 224 | | seresnext26_32x4d | 77.104 (22.896) | 93.316 (6.684) | 16.8M | bicubic | 224 | | skresnet34 | 76.912 (23.088) | 93.322 (6.678) | 22.2M | bicubic | 224 | | resnet26d | 76.68 (23.32) | 93.166 (6.834) | 16M | bicubic | 224 | +| mobilenetv2_140 | 76.524 (23.476) | 92.990 (7.010) | 6.1M | bicubic | 224 | | mixnet_s | 75.988 (24.012) | 92.794 (7.206) | 4.13M | bicubic | 224 | | mobilenetv3_large_100 | 75.766 (24.234) | 92.542 (7.458) | 5.5M | bicubic | 224 | | mobilenetv3_rw | 75.634 (24.366) | 92.708 (7.292) | 5.5M | bicubic | 224 | @@ -204,10 +213,12 @@ I've leveraged the training scripts in this repository to train a few of the mod | resnet26 | 75.292 (24.708) | 92.57 (7.43) | 16M | bicubic | 224 | | fbnetc_100 | 75.124 (24.876) | 92.386 (7.614) | 5.6M | bilinear | 224 | | resnet34 | 75.110 (24.890) | 92.284 (7.716) | 22M | bilinear | 224 | +| mobilenetv2_110d | 75.052 (24.948) | 92.180 (7.820) | 4.5M | bicubic | 224 | | seresnet34 | 74.808 (25.192) | 92.124 (7.876) | 22M | bilinear | 224 | | mnasnet_b1 | 74.658 (25.342) | 92.114 (7.886) | 4.38M | bicubic | 224 | | spnasnet_100 | 74.084 (25.916) | 91.818 (8.182) | 4.42M | bilinear | 224 | | skresnet18 | 73.038 (26.962) | 91.168 (8.832) | 11.9M | bicubic | 224 | +| mobilenetv2_100 | 72.978 (27.022) | 91.016 (8.984) | 3.5M | bicubic | 224 | | seresnet18 | 71.742 (28.258) | 90.334 (9.666) | 11.8M | bicubic | 224 | ### Ported Weights diff --git a/timm/models/efficientnet.py b/timm/models/efficientnet.py index 42435b0d..48678e67 100644 --- a/timm/models/efficientnet.py +++ b/timm/models/efficientnet.py @@ -60,10 +60,14 @@ default_cfgs = { 'semnasnet_140': _cfg(url=''), 'mnasnet_small': _cfg(url=''), - 'mobilenetv2_100': _cfg(url=''), - 'mobilenetv2_100d': _cfg(url=''), - 'mobilenetv2_110d': _cfg(url=''), - 'mobilenetv2_140': _cfg(url=''), + 'mobilenetv2_100': _cfg( + url='https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/mobilenetv2_100_ra-b33bc2c4.pth'), + 'mobilenetv2_110d': _cfg( + url='https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/mobilenetv2_110d_ra-77090ade.pth'), + 'mobilenetv2_120d': _cfg( + url='https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/mobilenetv2_120d_ra-5987e2ed.pth'), + 'mobilenetv2_140': _cfg( + url='https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/mobilenetv2_140_ra-21a4e913.pth'), 'fbnetc_100': _cfg( url='https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/fbnetc_100-c345b898.pth', @@ -953,31 +957,31 @@ def mnasnet_small(pretrained=False, **kwargs): @register_model def mobilenetv2_100(pretrained=False, **kwargs): - """ MobileNet V2 """ + """ MobileNet V2 w/ 1.0 channel multiplier """ model = _gen_mobilenet_v2('mobilenetv2_100', 1.0, pretrained=pretrained, **kwargs) return model @register_model def mobilenetv2_140(pretrained=False, **kwargs): - """ MobileNet V2 """ + """ MobileNet V2 w/ 1.4 channel multiplier """ model = _gen_mobilenet_v2('mobilenetv2_140', 1.4, pretrained=pretrained, **kwargs) return model @register_model def mobilenetv2_110d(pretrained=False, **kwargs): - """ MobileNet V2 """ + """ MobileNet V2 w/ 1.1 channel, 1.2 depth multipliers""" model = _gen_mobilenet_v2( - 'mobilenetv2_100d', 1.1, depth_multiplier=1.2, fix_stem_head=True, pretrained=pretrained, **kwargs) + 'mobilenetv2_110d', 1.1, depth_multiplier=1.2, fix_stem_head=True, pretrained=pretrained, **kwargs) return model @register_model def mobilenetv2_120d(pretrained=False, **kwargs): - """ MobileNet V2 """ + """ MobileNet V2 w/ 1.2 channel, 1.4 depth multipliers """ model = _gen_mobilenet_v2( - 'mobilenetv2_110d', 1.2, depth_multiplier=1.4, fix_stem_head=True, pretrained=pretrained, **kwargs) + 'mobilenetv2_120d', 1.2, depth_multiplier=1.4, fix_stem_head=True, pretrained=pretrained, **kwargs) return model