xdt89460的博客

深度可分离卷积 https://yinguobing.com/separable-convolution/#fn2 https://blog.youkuaiyun.com/ly18846826264/article/details/105144998 MobileNetv1 https://www.jianshu.com/p/f601fa110eaf 在torch.nn.Conv2d中,group参数用来控制是否对输入的每个通道单独设置卷积: At groups= in_channels, each input cha