Conv_type
2020-03-15 本文已影响0人
yuanCruise
1.Convolutions
2.PointwiseConvolutions
来自Network IN Network
3.Spatial and Cross-Channel Convolutions
来自于Inception
4.Grouped Convolutions
来自AlexNet，后大量应用于ResNeXt（Inception-Resnet）
5.Separable Convolutions
DepthwiseSeparable Convolutions:深度可分离卷积，来自mobilenetv1
SpatiallySeparable Convolutions:空间可分离卷积
6.Flattened Convolutions
来自flattened convolutons
7.shuffle group Convoluions
来自shufflenet
8.Dilated Convolutions 空洞卷积
来自deeplab
9.Deformable Convolutions 可变形卷积
来自deformable
