Depthwise Separable Convolution

Introduced by Chollet in Xception: Deep Learning With Depthwise Separable Convolutions

While standard convolution performs the channelwise and spatial-wise computation in one step, Depthwise Separable Convolution splits the computation into two steps: depthwise convolution applies a single convolutional filter per each input channel and pointwise convolution is used to create a linear combination of the output of the depthwise convolution. The comparison of standard convolution and depthwise separable convolution is shown to the right.

Credit: Depthwise Convolution Is All You Need for Learning Multiple Visual Domains

Source: Xception: Deep Learning With Depthwise Separable Convolutions


Paper Code Results Date Stars


Task Papers Share
Image Classification 106 14.85%
Object Detection 73 10.22%
General Classification 51 7.14%
Semantic Segmentation 40 5.60%
Quantization 29 4.06%
Model Compression 17 2.38%
Instance Segmentation 14 1.96%
Knowledge Distillation 10 1.40%
Network Pruning 9 1.26%