This is a brief summary of paper for me to study and arrange for Xception: Deep Learning with Depthwise Separable Convolutions (Cholletl., CVPR 2017) I read and studied.

This paper is a research ralted to answer to the quextion which is how the convolutional network works.

They proposed depth-wise separable convolution neural networks.

They follow the following hypothesis: that the mapping of cross-channels correlation and spatial correlations in the featur maps of convolutional neural networks can be entirely decoupled.

It underlies the inception architecture, So they named their method Xception, which stands for “Extreme Inception”.

The following figure is their architectur of Xception:

Francois Cholletl., CVPR 2017

As you can see image above, they decoupled mapping of cross-channel correlation and spatial correlation.

Reference