1. Dynamic Convolution
paper的地址:https://link.zhihu.com/?target=https%3A//arxiv.org/pdf/1912.03458.pdf
pytorch代码:https://github.com/pinery-sls/Dynamic-convolution-Pytorch
💫可以替换常规卷积层
参考:https://zhuanlan.zhihu.com/p/208519425
2. CondConv
paper:https://arxiv.org/abs/1904.04971
pytorch代码: https://github.com/rwightman/gen-efficientnet-pytorch/blob/master/geffnet/conv2d_layers.py
tensorflow代码:https://github.com/tensorflow/tpu/blob/master/models/official/efficientnet/condconv/condconv_layers.py
3. Batch Convolutional Layers
受CondConv讨论启发,用于并行执行每个样本内核过滤器的分组卷积黑客:
https://github.com/pytorch/pytorch/issues/17983
4. MixConv
paper:https://arxiv.org/abs/1907.09595
pytorch代码:https://github.com/rwightman/gen-efficientnet-pytorch/blob/master/geffnet/conv2d_layers.py
tensorflow代码:https://github.com/tensorflow/tpu/blob/master/models/official/mnasnet/mixnet/custom_layers.py