PyTorch笔记 - A ConvNet for the 2020s (ConvNeXt) 网络

欢迎关注我的CSDN:https://blog.csdn.net/caroline_wendy
本文地址:https://blog.csdn.net/caroline_wendy/article/details/128236899

Paper:A ConvNet for the 2020s,FAIR,Berkeley
Code:https://github.com/facebookresearch/ConvNeXt

网络结构:

  • stem layer:骨干层,目的是降低计算复杂度。
  • Why does CNNs usually have a stem?

PyTorch笔记 - A ConvNet for the 2020s (ConvNeXt) 网络_第1张图片

实现ConvNeXt的Block,类似于ResNet,只不过是d7x7(depth-wise) -> LN -> 1x1(通道融合,提升通道) -> 1x1(通道融合,降低通道)

Block源码:继承于nn.Module

 
 

你可能感兴趣的:(深度学习,pytorch,深度学习,python)