Hang Zhang

We used MXNet implementation to calculate FLOPs. https://github.com/Ldpe2G/DeepLearningForFun/tree/master/MXNet-Python/CalculateFlopsTool

Channel-wise multiplication is very efficient/cheap. The FLOPs is similar to SENet.

The computation of channel-wise multiplication is the same as SENet.

Hi, sorry about that. Wasabi cloud has a rule of limited downloading volume. I am looking for other solutions

Hi @DAVID-Hown could you try https://dl.dropboxusercontent.com/s/fi72lae1exon3o4/resnest50-528c19ca.pth?dl=0

I think your pytorch is out-of-date

Follow the instruction here, it should be working now https://github.com/zhanghang1989/ResNeSt#pytorch-models

Thanks for sharing!

This should be already included in the current code. Let me know if you have any difficulties in converting the model.

Sorry that I am not a TF user. It would be great if we can have community contributions on this.