On the concepts of "input channel" and "output channel" in convolutional neural networks

The picture below is from ：https://segmentfault.com/q/1010000016667038

In a more concise word, it is ：

Input channel refers to the input of several two-dimensional information , That is very intuitive rgb The picture has r,g,b Three channels , This determines the number of channels of the convolution kernel , That is, the number of channels of the input image determines the number of convolution kernel channels ;（ In the picture , The first column has three matrices , That is, the input channel is 3, So behind , The second and third Columns , That is, two convolution kernels , They also have three matrices , That is, the number of convolution kernels is 3.）

Output channel refers to convolution （ relation ） Number of output channels after operation , It determines how many convolution kernels there are , That is, the number of output channels required , You need a few convolution kernels .（ In the figure , The second and third columns are convolution kernels of three channels respectively , After the convolution operation between the input image and these two convolution kernels , Two matrices of the fourth column are generated , That is, two channels are output ）

 

Finally, let's talk about the dynamic diagram demonstration , Enter a three channel image , And two three channel convolution kernels , Get two output channels .