In 2014

Bright spot

  • Stack multiple 3*3 convolution kernels to replace large convolution kernels and reduce required parameters

In convolutional neural networks, the region size of the input layer corresponding to an element in the output results of a certain layer is called the receptive field. You can replace the 55 convolution kernel by stacking two 33 convolution kernels, and replace the 77 convolution kernel by stacking three 33 convolution kernels, and they have the same receptive field


F ( i ) = ( F ( i + 1 ) – 1 ) ∗ S t r i d e ∗ K s i z e F(i) = (F(i+1) – 1) * Stride *Ksize

F(I)F(I)F(I) is the receptive field of layer III

StrideStrideStride is the step of the third layer

KsizeKsizeKsize is the size of the convolution kernel or the sampling kernel

The network structure

  • The stride for conv is 1, and the padding is 1
  • The size of maxpool is 2, and the stride is 2