Divide an image into multiple patch, and then for each part of the convolution operation (sharing the same set of convolution kernels), and then combined the results of each part, in order to get and whole picture operation excuse me is there a way to deal with the result of the same pictures?