I have been advised by my supervisor that if my U-Net segmentation network has RGB images at the input then I could use the channels for different images - median filter for R, normalization for G, canny-edge detection for B (example). I have no idea how to do that. Tried to find something similiar here but unfortunately without a success. Would be greateful if somebody could explain me this in detail, I'm new in DL. Thanks
An input to a U-Net are normally images with 3 (or 4) channels, as images have a red, green, and blue channel. It is however technically possibly to use other images (or arrays for that matter) as the input which a larger number of channels. If you wanted to use the three different methods that you mention, you would first have to apply all three methods to original image, which would then return an array with 1 or more channels. You then have to stack them in the channels direction, which would result in an array of shape
C is the total number of channels. The total number of channels in the input then depends on the image preprocessing methods you use and the number of channels of the original image.