Pytorch implementation of Funnel Activation (FReLU): https://arxiv.org/pdf/2007.11824.pdf
Validation results are listed below:
Model | Activation | Err@1 | Err@5 |
---|---|---|---|
ResNet50 | FReLU | 22.40 | 6.164 |
Note that from the file resnet_frelu.py you can call ResNet18, ResNet34, ResNet50, ResNet101 and ResNet152 but the weights in this repo only available for ResNet50 and I never tried to train other models, so no guaranties there!
The code in this repo is based on pytorch imagenet example:
and original implementation of Funnel Activation in Megengine:
https://github.com/megvii-model/FunnelAct
Enjoy!