In CNN model with keras API , we used an image as input then it works hidden layer where Convolution, Relu and Maxpool2D works and after then it works for Flatten, Dense layer and finally provide the desired output which is know as Classification. I just explained Sequentil model. Now I am confused about hidden layer. In hidden layer, is it works with an images 3 dimention?? I want to say, 3 dimensions as height, width and then in the last dimension the RGB colors. Actually what happening in the hidden layer.