Merge pull request #88 from rwightman/attention

A lot of attention and much more

Author: rwightman

.gitignore

@@ -104,3 +104,5 @@ venv.bak/
 *.tar
 *.pth
 *.gz
+Untitled.ipynb
+Testing notebook.ipynb

README.md

@@ -2,6 +2,20 @@
 
 ## What's New
 
+### 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)
 
@@ -86,6 +100,7 @@ Included models:
     * 'Bag of Tricks' / Gluon C, D, E, S variations (https://arxiv.org/abs/1812.01187)
     * Instagram trained / ImageNet tuned ResNeXt101-32x8d to 32x48d from from [facebookresearch](https://pytorch.org/hub/facebookresearch_WSL-Images_resnext/)
     * Res2Net (https://github.com/gasvn/Res2Net, https://arxiv.org/abs/1904.01169)
+    * Selective Kernel (SK) Nets (https://arxiv.org/abs/1903.06586)
 * DLA
     * Original (https://github.com/ucbdrive/dla, https://arxiv.org/abs/1707.06484)
     * Res2Net (https://github.com/gasvn/Res2Net, https://arxiv.org/abs/1904.01169)
@@ -138,6 +153,8 @@ Several (less common) features that I often utilize in my projects are included.
 * 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)
 * AugMix w/ JSD loss (https://arxiv.org/abs/1912.02781), JSD w/ clean + augmented mixing support works with AutoAugment and RandAugment as well
 * SplitBachNorm - allows splitting batch norm layers between clean and augmented (auxiliary batch norm) data
+* DropBlock (https://arxiv.org/abs/1810.12890)
+* Efficient Channel Attention - ECA (https://arxiv.org/abs/1910.03151)
 
 ## Results
 
@@ -150,9 +167,11 @@ I've leveraged the training scripts in this repository to train a few of the mod
 |---|---|---|---|---|---|
 | efficientnet_b3a | 81.874 (18.126) | 95.840 (4.160) | 12.23M | bicubic | 320 (1.0 crop) |
 | efficientnet_b3 | 81.498 (18.502) | 95.718 (4.282) | 12.23M | bicubic | 300 |
+| skresnext50d_32x4d | 81.278 (18.722) | 95.366 (4.634) | 27.5M | bicubic | 288 (1.0 crop) |
 | efficientnet_b2a | 80.608 (19.392) | 95.310 (4.690) | 9.11M | bicubic | 288 (1.0 crop) |
 | mixnet_xl | 80.478 (19.522) | 94.932 (5.068) | 11.90M | bicubic | 224 |
 | efficientnet_b2 | 80.402 (19.598) | 95.076 (4.924) | 9.11M | bicubic | 260 |
+| skresnext50d_32x4d | 80.156 (19.844) | 94.642 (5.358) | 27.5M | bicubic | 224 |
 | resnext50d_32x4d | 79.674 (20.326) | 94.868 (5.132) | 25.1M | bicubic | 224 |
 | resnet50 | 79.038 (20.962) | 94.390 (5.610) | 25.6M | bicubic | 224 |
 | mixnet_l | 78.976 (21.024 | 94.184 (5.816) | 7.33M | bicubic | 224 |
@@ -165,6 +184,7 @@ I've leveraged the training scripts in this repository to train a few of the mod
 | seresnext26d_32x4d | 77.602 (22.398) | 93.608 (6.392) | 16.8M | 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 |
 | mixnet_s | 75.988 (24.012) | 92.794 (7.206) | 4.13M | bicubic | 224 |
 | mobilenetv3_100 | 75.634 (24.366) | 92.708 (7.292) | 5.5M | bicubic | 224 |
@@ -175,6 +195,7 @@ I've leveraged the training scripts in this repository to train a few of the mod
 | 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 |
 | seresnet18 | 71.742 (28.258) | 90.334 (9.666) | 11.8M | bicubic | 224 |
 
 ### Ported Weights

hubconf.py

@@ -0,0 +1,10 @@
+dependencies = ['torch']
+
+from timm.models import registry
+
+current_module = __import__(__name__)
+current_module.__dict__.update(registry._model_entrypoints)
+#for fn_name in registry.list_models():
+#    fn = registry.model_entrypoint(fn_name)
+#    setattr(current_module, fn_name, fn)
+

results/results-imagenet-a.csv

@@ -1,8 +1,14 @@
 model,top1,top1_err,top5,top5_err,param_count,img_size,cropt_pct,interpolation
+tf_efficientnet_l2_ns_475,62.3733,37.6267,87.0933,12.9067,480.31,475,0.936,bicubic
+tf_efficientnet_l2_ns,62.0267,37.9733,87.96,12.04,480.31,800,0.96,bicubic
+tf_efficientnet_b7_ns,45.76,54.24,74.2133,25.7867,66.35,600,0.949,bicubic
 ig_resnext101_32x48d,41.56,58.44,66.5467,33.4533,828.41,224,0.875,bilinear
+tf_efficientnet_b6_ns,40.4533,59.5467,68.8667,31.1333,43.04,528,0.942,bicubic
 ig_resnext101_32x32d,39.4267,60.5733,63.7867,36.2133,468.53,224,0.875,bilinear
-ig_resnext101_32x16d,36.0,64.0,59.0,41.0,194.03,224,0.875,bilinear
+tf_efficientnet_b5_ns,39.0133,60.9867,68.08,31.92,30.39,456,0.934,bicubic
+ig_resnext101_32x16d,36,64,59,41,194.03,224,0.875,bilinear
 swsl_resnext101_32x8d,32.0133,67.9867,59.44,40.56,88.79,224,0.875,bilinear
+tf_efficientnet_b4_ns,30.7867,69.2133,59.4667,40.5333,19.34,380,0.922,bicubic
 tf_efficientnet_b8_ap,29.5867,70.4133,56.9333,43.0667,87.41,672,0.954,bicubic
 tf_efficientnet_b8,29.3867,70.6133,57.0533,42.9467,87.41,672,0.954,bicubic
 ig_resnext101_32x8d,28.6667,71.3333,52.32,47.68,88.79,224,0.875,bilinear
@@ -13,6 +19,7 @@ tf_efficientnet_b7,25.28,74.72,51.6667,48.3333,66.35,600,0.949,bicubic
 tf_efficientnet_b6_ap,24.3467,75.6533,50.44,49.56,43.04,528,0.942,bicubic
 tf_efficientnet_b6,20.3733,79.6267,45.48,54.52,43.04,528,0.942,bicubic
 tf_efficientnet_b5_ap,19.4667,80.5333,44.7333,55.2667,30.39,456,0.934,bicubic
+tf_efficientnet_b3_ns,19.44,80.56,44.6533,55.3467,12.23,300,0.904,bicubic
 swsl_resnext50_32x4d,18.04,81.96,41.9733,58.0267,25.03,224,0.875,bilinear
 ssl_resnext101_32x16d,17.1867,82.8133,39.9333,60.0667,194.03,224,0.875,bilinear
 tf_efficientnet_b5,17.0533,82.9467,41.92,58.08,30.39,456,0.934,bicubic
@@ -23,13 +30,15 @@ tf_efficientnet_b4,13.32,86.68,35.5333,64.4667,19.34,380,0.922,bicubic
 pnasnet5large,13.0533,86.9467,32.2267,67.7733,86.06,331,0.875,bicubic
 nasnetalarge,12.56,87.44,33.4267,66.5733,88.75,331,0.875,bicubic
 ssl_resnext101_32x4d,12.1067,87.8933,31.8933,68.1067,44.18,224,0.875,bilinear
+tf_efficientnet_b2_ns,11.7333,88.2667,32.96,67.04,9.11,260,0.89,bicubic
 gluon_senet154,9.8933,90.1067,26.4267,73.5733,115.09,224,0.875,bicubic
 ssl_resnext50_32x4d,9.6533,90.3467,28.4667,71.5333,25.03,224,0.875,bilinear
 senet154,9.4667,90.5333,26.44,73.56,115.09,224,0.875,bilinear
-efficientnet_b3a,9.2533,90.7467,28.4267,71.5733,12.23,320,1.0,bicubic
+efficientnet_b3a,9.2533,90.7467,28.4267,71.5733,12.23,320,1,bicubic
 efficientnet_b3,8.9733,91.0267,28.2267,71.7733,12.23,300,0.904,bicubic
 inception_v4,8.8933,91.1067,24.68,75.32,42.68,299,0.875,bicubic
 gluon_seresnext101_64x4d,8.8667,91.1333,27.28,72.72,88.23,224,0.875,bicubic
+tf_efficientnet_b1_ns,8.6133,91.3867,27.2933,72.7067,7.79,240,0.882,bicubic
 gluon_xception65,8.44,91.56,25.12,74.88,39.92,299,0.875,bicubic
 gluon_resnet152_v1d,8.36,91.64,23.4267,76.5733,60.21,224,0.875,bicubic
 inception_resnet_v2,8.1733,91.8267,23.5733,76.4267,55.84,299,0.8975,bicubic
@@ -39,14 +48,15 @@ tf_efficientnet_b3,8.0133,91.9867,25.48,74.52,12.23,300,0.904,bicubic
 ens_adv_inception_resnet_v2,7.9733,92.0267,23.8667,76.1333,55.84,299,0.8975,bicubic
 gluon_resnet152_v1s,7.8533,92.1467,23.1867,76.8133,60.32,224,0.875,bicubic
 gluon_resnext101_64x4d,7.72,92.28,23.3067,76.6933,83.46,224,0.875,bicubic
+skresnext50_32x4d,7.08,92.92,23.0667,76.9333,27.48,224,0.875,bicubic
 ssl_resnet50,7.04,92.96,23.9067,76.0933,25.56,224,0.875,bilinear
-efficientnet_b2a,6.7467,93.2533,23.5067,76.4933,9.11,288,1.0,bicubic
+efficientnet_b2a,6.7467,93.2533,23.5067,76.4933,9.11,288,1,bicubic
 seresnext101_32x4d,6.4,93.6,21.4933,78.5067,48.96,224,0.875,bilinear
 efficientnet_b2,6.0933,93.9067,21.96,78.04,9.11,260,0.875,bicubic
 gluon_resnext101_32x4d,6.0133,93.9867,21.12,78.88,44.18,224,0.875,bicubic
 gluon_resnet101_v1d,5.92,94.08,19.9467,80.0533,44.57,224,0.875,bicubic
 gluon_seresnext50_32x4d,5.7867,94.2133,21.4533,78.5467,27.56,224,0.875,bicubic
-gluon_inception_v3,5.5067,94.4933,20.0,80.0,23.83,299,0.875,bicubic
+gluon_inception_v3,5.5067,94.4933,20,80,23.83,299,0.875,bicubic
 mixnet_xl,5.4667,94.5333,21.08,78.92,11.9,224,0.875,bicubic
 gluon_resnet101_v1s,5.28,94.72,19.56,80.44,44.67,224,0.875,bicubic
 hrnet_w64,5.16,94.84,19.4933,80.5067,128.06,224,0.875,bilinear
@@ -69,8 +79,9 @@ inception_v3,4.1867,95.8133,16.2933,83.7067,27.16,299,0.875,bicubic
 tf_efficientnet_b2_ap,4.16,95.84,18.3467,81.6533,9.11,260,0.89,bicubic
 seresnet152,4.1467,95.8533,15.9333,84.0667,66.82,224,0.875,bilinear
 resnext101_32x8d,4.1333,95.8667,16.92,83.08,88.79,224,0.875,bilinear
+tf_efficientnet_b0_ns,4.1333,95.8667,17.68,82.32,5.29,224,0.875,bicubic
 dpn98,4.08,95.92,15.96,84.04,61.57,224,0.875,bicubic
-res2net101_26w_4s,4.0,96.0,14.8667,85.1333,45.21,224,0.875,bilinear
+res2net101_26w_4s,4,96,14.8667,85.1333,45.21,224,0.875,bilinear
 efficientnet_b1,3.9733,96.0267,15.7733,84.2267,7.79,240,0.875,bicubic
 tf_efficientnet_b2,3.76,96.24,16.5867,83.4133,9.11,260,0.89,bicubic
 hrnet_w30,3.68,96.32,15.5733,84.4267,37.71,224,0.875,bilinear
@@ -102,6 +113,7 @@ dla60_res2net,2.64,97.36,14.1733,85.8267,21.15,224,0.875,bilinear
 gluon_resnet101_v1b,2.6133,97.3867,13.56,86.44,44.55,224,0.875,bicubic
 dla60x,2.6,97.4,13.3467,86.6533,17.65,224,0.875,bilinear
 mixnet_m,2.5467,97.4533,12.4133,87.5867,5.01,224,0.875,bicubic
+efficientnet_es,2.3733,97.6267,13.8267,86.1733,5.44,224,0.875,bicubic
 resnet152,2.36,97.64,12.2,87.8,60.19,224,0.875,bilinear
 swsl_resnet18,2.3467,97.6533,11.2267,88.7733,11.69,224,0.875,bilinear
 wide_resnet50_2,2.32,97.68,11.8267,88.1733,68.88,224,0.875,bilinear
@@ -133,7 +145,7 @@ ssl_resnet18,1.3867,98.6133,8.2,91.8,11.69,224,0.875,bilinear
 dla60,1.3333,98.6667,9.4667,90.5333,22.33,224,0.875,bilinear
 dpn68,1.32,98.68,8.8267,91.1733,12.61,224,0.875,bicubic
 res2net50_48w_2s,1.2933,98.7067,8.9333,91.0667,25.29,224,0.875,bilinear
-tf_mixnet_s,1.2667,98.7333,8.7467,91.2533,4.13,224,0.875,bicubic
+tf_mixnet_s,1.2667,98.7333,8.7333,91.2667,4.13,224,0.875,bicubic
 fbnetc_100,1.24,98.76,8.76,91.24,5.57,224,0.875,bilinear
 resnet26d,1.24,98.76,9.32,90.68,16.01,224,0.875,bicubic
 tf_mobilenetv3_large_100,1.1867,98.8133,7.9467,92.0533,5.48,224,0.875,bilinear
@@ -143,17 +155,18 @@ seresnet34,1.12,98.88,7.4267,92.5733,21.96,224,0.875,bilinear
 tf_efficientnet_es,1.12,98.88,8.5867,91.4133,5.44,224,0.875,bicubic
 spnasnet_100,1.1067,98.8933,8.2133,91.7867,4.42,224,0.875,bilinear
 dla34,1.08,98.92,7.68,92.32,15.78,224,0.875,bilinear
-resnet34,1.0,99.0,7.5333,92.4667,21.8,224,0.875,bilinear
+resnet34,1,99,7.5333,92.4667,21.8,224,0.875,bilinear
 gluon_resnet34_v1b,0.8933,99.1067,6.6,93.4,21.8,224,0.875,bicubic
 hrnet_w18_small_v2,0.8933,99.1067,7.3867,92.6133,15.6,224,0.875,bilinear
+skresnet18,0.88,99.12,7.3467,92.6533,11.96,224,0.875,bicubic
 tf_mobilenetv3_large_075,0.88,99.12,6.72,93.28,3.99,224,0.875,bilinear
 mnasnet_100,0.8667,99.1333,7.8267,92.1733,4.38,224,0.875,bicubic
 tf_mobilenetv3_small_100,0.7467,99.2533,4.6667,95.3333,2.54,224,0.875,bilinear
 seresnet18,0.7333,99.2667,6.0267,93.9733,11.78,224,0.875,bicubic
 densenet121,0.68,99.32,6.8933,93.1067,7.98,224,0.875,bicubic
 tf_mobilenetv3_small_075,0.6533,99.3467,4.1867,95.8133,2.04,224,0.875,bilinear
 tv_resnet34,0.6,99.4,5.5333,94.4667,21.8,224,0.875,bilinear
-resnet26,0.5867,99.4133,6.8933,93.1067,16.0,224,0.875,bicubic
+resnet26,0.5867,99.4133,6.8933,93.1067,16,224,0.875,bicubic
 dla46_c,0.52,99.48,4.1733,95.8267,1.31,224,0.875,bilinear
 dla60x_c,0.48,99.52,5.2133,94.7867,1.34,224,0.875,bilinear
 tf_mobilenetv3_large_minimal_100,0.48,99.52,4.88,95.12,3.92,224,0.875,bilinear
@@ -162,4 +175,4 @@ dla46x_c,0.4133,99.5867,4.44,95.56,1.08,224,0.875,bilinear
 gluon_resnet18_v1b,0.3867,99.6133,4.7867,95.2133,11.69,224,0.875,bicubic
 tf_mobilenetv3_small_minimal_100,0.36,99.64,2.8667,97.1333,2.04,224,0.875,bilinear
 resnet18,0.2933,99.7067,4.04,95.96,11.69,224,0.875,bilinear
-tv_resnet50,0.0,100.0,2.9067,97.0933,25.56,224,0.875,bilinear
+tv_resnet50,0,100,2.9067,97.0933,25.56,224,0.875,bilinear

results/results-imagenet.csv

@@ -1,7 +1,13 @@
 model,top1,top1_err,top5,top5_err,param_count,img_size,cropt_pct,interpolation
+tf_efficientnet_l2_ns,88.352,11.648,98.65,1.35,480.31,800,0.961,bicubic
+tf_efficientnet_l2_ns_475,88.234,11.766,98.546,1.454,480.31,475,0.936,bicubic
+tf_efficientnet_b7_ns,86.84,13.16,98.094,1.906,66.35,600,0.949,bicubic
+tf_efficientnet_b6_ns,86.452,13.548,97.882,2.118,43.04,528,0.942,bicubic
+tf_efficientnet_b5_ns,86.088,13.912,97.752,2.248,30.39,456,0.934,bicubic
 ig_resnext101_32x48d,85.428,14.572,97.572,2.428,828.41,224,0.875,bilinear
 tf_efficientnet_b8,85.37,14.63,97.39,2.61,87.41,672,0.954,bicubic
 tf_efficientnet_b8_ap,85.37,14.63,97.294,2.706,87.41,672,0.954,bicubic
+tf_efficientnet_b4_ns,85.162,14.838,97.47,2.53,19.34,380,0.922,bicubic
 tf_efficientnet_b7_ap,85.12,14.88,97.252,2.748,66.35,600,0.949,bicubic
 ig_resnext101_32x32d,85.094,14.906,97.438,2.562,468.53,224,0.875,bilinear
 tf_efficientnet_b7,84.936,15.064,97.204,2.796,66.35,600,0.949,bicubic
@@ -10,6 +16,7 @@ swsl_resnext101_32x8d,84.284,15.716,97.176,2.824,88.79,224,0.875,bilinear
 tf_efficientnet_b5_ap,84.252,15.748,96.974,3.026,30.39,456,0.934,bicubic
 ig_resnext101_32x16d,84.17,15.83,97.196,2.804,194.03,224,0.875,bilinear
 tf_efficientnet_b6,84.11,15.89,96.886,3.114,43.04,528,0.942,bicubic
+tf_efficientnet_b3_ns,84.048,15.952,96.91,3.09,12.23,300,0.904,bicubic
 tf_efficientnet_b5,83.812,16.188,96.748,3.252,30.39,456,0.934,bicubic
 swsl_resnext101_32x16d,83.346,16.654,96.846,3.154,194.03,224,0.875,bilinear
 tf_efficientnet_b4_ap,83.248,16.752,96.392,3.608,19.34,380,0.922,bicubic
@@ -18,13 +25,15 @@ tf_efficientnet_b4,83.022,16.978,96.3,3.7,19.34,380,0.922,bicubic
 pnasnet5large,82.736,17.264,96.046,3.954,86.06,331,0.875,bicubic
 ig_resnext101_32x8d,82.688,17.312,96.636,3.364,88.79,224,0.875,bilinear
 nasnetalarge,82.554,17.446,96.038,3.962,88.75,331,0.875,bicubic
+tf_efficientnet_b2_ns,82.38,17.62,96.248,3.752,9.11,260,0.89,bicubic
 swsl_resnext50_32x4d,82.182,17.818,96.23,3.77,25.03,224,0.875,bilinear
 efficientnet_b3a,81.866,18.134,95.836,4.164,12.23,320,1,bicubic
 ssl_resnext101_32x16d,81.844,18.156,96.096,3.904,194.03,224,0.875,bilinear
 tf_efficientnet_b3_ap,81.822,18.178,95.624,4.376,12.23,300,0.904,bicubic
 tf_efficientnet_b3,81.636,18.364,95.718,4.282,12.23,300,0.904,bicubic
 ssl_resnext101_32x8d,81.616,18.384,96.038,3.962,88.79,224,0.875,bilinear
 efficientnet_b3,81.494,18.506,95.716,4.284,12.23,300,0.904,bicubic
+tf_efficientnet_b1_ns,81.388,18.612,95.738,4.262,7.79,240,0.882,bicubic
 senet154,81.31,18.69,95.496,4.504,115.09,224,0.875,bilinear
 gluon_senet154,81.234,18.766,95.348,4.652,115.09,224,0.875,bicubic
 swsl_resnet50,81.166,18.834,95.972,4.028,25.56,224,0.875,bilinear
@@ -47,9 +56,10 @@ tf_efficientnet_b2_ap,80.3,19.7,95.028,4.972,9.11,260,0.89,bicubic
 seresnext101_32x4d,80.228,19.772,95.018,4.982,48.96,224,0.875,bilinear
 inception_v4,80.168,19.832,94.968,5.032,42.68,299,0.875,bicubic
 dpn107,80.156,19.844,94.91,5.09,86.92,224,0.875,bicubic
+skresnext50_32x4d,80.156,19.844,94.642,5.358,27.48,224,0.875,bicubic
 tf_efficientnet_b2,80.086,19.914,94.908,5.092,9.11,260,0.89,bicubic
 dpn92,80.008,19.992,94.836,5.164,37.67,224,0.875,bicubic
-ens_adv_inception_resnet_v2,79.982,20.018,94.938,5.062,55.84,299,0.8975,bicubic
+ens_adv_inception_resnet_v2,79.982,20.018,94.936,5.064,55.84,299,0.8975,bicubic
 gluon_seresnext50_32x4d,79.918,20.082,94.822,5.178,27.56,224,0.875,bicubic
 gluon_resnet152_v1c,79.91,20.09,94.84,5.16,60.21,224,0.875,bicubic
 dpn131,79.822,20.178,94.71,5.29,79.25,224,0.875,bicubic
@@ -85,6 +95,7 @@ tf_efficientnet_em,78.708,21.292,94.314,5.686,6.9,240,0.882,bicubic
 efficientnet_b1,78.698,21.302,94.144,5.856,7.79,240,0.875,bicubic
 dla169,78.688,21.312,94.336,5.664,53.99,224,0.875,bilinear
 seresnet152,78.66,21.34,94.37,5.63,66.82,224,0.875,bilinear
+tf_efficientnet_b0_ns,78.658,21.342,94.376,5.624,5.29,224,0.875,bicubic
 res2net50_26w_6s,78.57,21.43,94.124,5.876,37.05,224,0.875,bilinear
 resnext50_32x4d,78.512,21.488,94.042,5.958,25.03,224,0.875,bicubic
 dla102x,78.51,21.49,94.228,5.772,26.77,224,0.875,bilinear
@@ -99,6 +110,7 @@ dla60x,78.246,21.754,94.018,5.982,17.65,224,0.875,bilinear
 res2next50,78.246,21.754,93.892,6.108,24.67,224,0.875,bilinear
 hrnet_w30,78.206,21.794,94.222,5.778,37.71,224,0.875,bilinear
 res2net50_14w_8s,78.15,21.85,93.848,6.152,25.06,224,0.875,bilinear
+efficientnet_es,78.066,21.934,93.926,6.074,5.44,224,0.875,bicubic
 dla102,78.032,21.968,93.946,6.054,33.73,224,0.875,bilinear
 gluon_resnet50_v1c,78.012,21.988,93.988,6.012,25.58,224,0.875,bicubic
 seresnext26t_32x4d,77.998,22.002,93.708,6.292,16.82,224,0.875,bicubic
@@ -115,7 +127,7 @@ adv_inception_v3,77.582,22.418,93.736,6.264,23.83,299,0.875,bicubic
 gluon_resnet50_v1b,77.58,22.42,93.716,6.284,25.56,224,0.875,bicubic
 res2net50_48w_2s,77.522,22.478,93.554,6.446,25.29,224,0.875,bilinear
 dpn68b,77.512,22.488,93.822,6.178,12.61,224,0.875,bicubic
-inception_v3,77.438,22.562,93.474,6.526,27.16,299,0.875,bicubic
+inception_v3,77.44,22.56,93.474,6.526,27.16,299,0.875,bicubic
 resnet101,77.374,22.626,93.54,6.46,44.55,224,0.875,bilinear
 densenet161,77.358,22.642,93.638,6.362,28.68,224,0.875,bicubic
 tf_efficientnet_cc_b0_4e,77.306,22.694,93.334,6.666,13.31,224,0.875,bicubic
@@ -151,11 +163,12 @@ spnasnet_100,74.084,25.916,91.818,8.182,4.42,224,0.875,bilinear
 tf_mobilenetv3_large_075,73.438,26.562,91.35,8.65,3.99,224,0.875,bilinear
 tv_resnet34,73.312,26.688,91.426,8.574,21.8,224,0.875,bilinear
 swsl_resnet18,73.276,26.724,91.734,8.266,11.69,224,0.875,bilinear
+skresnet18,73.038,26.962,91.168,8.832,11.96,224,0.875,bicubic
 ssl_resnet18,72.61,27.39,91.416,8.584,11.69,224,0.875,bilinear
 hrnet_w18_small,72.342,27.658,90.678,9.322,13.19,224,0.875,bilinear
 tf_mobilenetv3_large_minimal_100,72.248,27.752,90.63,9.37,3.92,224,0.875,bilinear
 seresnet18,71.742,28.258,90.334,9.666,11.78,224,0.875,bicubic
-gluon_resnet18_v1b,70.836,29.164,89.76,10.24,11.69,224,0.875,bicubic
+gluon_resnet18_v1b,70.836,29.164,89.762,10.238,11.69,224,0.875,bicubic
 resnet18,69.748,30.252,89.078,10.922,11.69,224,0.875,bilinear
 tf_mobilenetv3_small_100,67.922,32.078,87.664,12.336,2.54,224,0.875,bilinear
 dla60x_c,67.892,32.108,88.426,11.574,1.34,224,0.875,bilinear