I need to do a modulo transform on each value of an image that is stored as a tensor. For an image stored in PIL format, this can be done on the red channel using:
Cr = 50; maxV = 255
red, green, blue = img.split()
red = red.point(lambda p: (p+Cr) % max_V)
Not sure if there is a way to do the same in torch, without converting the image back to PIL and forth to torch tensor. The problem is that the tensor values are float and usually have -ve and +ve values. I can assume, for example, that maxV = 1 and Cr = 0.2.