When using the depth study of image processing, in order to fully understand each layer convolution operation and the extracted features of images, need to be visualized, so numerical convolution layer initialization is gaussian distribution, so after the convolution layer handles images inevitably have negative pixels, how these negative pixels visualization, entangled with the this problem for a long time, hope the great god could doubt dispels doubts

CodePudding user response:

Do you know the solution to this problem now