Convolution effect layers * 1 1
Answer: FIG feature complete the polymerization or divergent channel
that is used to control the depth of the final output layer is characterized by convolving FIG 1 * 1, and thus indirectly affect the number of parameters associated therewith convolution layer
summary: to reduce the number of parameters
What is the role ResNet use identity mapping in the residual module
Summary: On the basis of no additional parameters and calculations on the amount, speed up training modules to enhance the training effect model
Bold Style
(excerpt from pytorch depth study of practical computer vision)