Hi,
I do basic CNN stuff with pytorch, but I always wondered one thing.
how the channel number that we set affects the image? I generally know that each channel represents one thing: R, G, B, vertical edges, horizontal edges, … colours, and other unknowns.
but is this the order? Say I want to observe G,B and Vertical edge information. How convnet will respond?
I have checked source code of convnet2d but didnt find quality answer. How the convnet changes/tracks the features by itself for each channel, while calculating in each forward pass?