Nearly Perfect & Easily Understandable PyTorch Implementation of SKNet(Selective Kernel Networks)

I reimplemented SKNET using PyTorch. Although there are several PyTorch implementations of SKNET, they are different from implementation described in the original paper, and I was hard to understand their implementations. So I did.

Refer to this part for the implementation of Selective Kernel Convolution

• Paper: Selective Kernel Networks
• Paper: Aggregated Residual Transformations for Deep Neural Networks
• Repository: pppLang/SKNet
• Repository: ResearchingDexter/SKNet_pytorch
• Repository: bearpaw/pytorch-classification

• Experiment on CIFAR100 with Resnet-18