深度學習之基礎模型-ShuffleNet

The new architecture utilizes two proposed operations, point wise group convolution and channel shuffle, to greatly reduce computation cost while maintaining accuracy.web 該論文利用了point wise group convol

>>阅读原文<<