Your method is very inspiring to me, I am also working on the corresponding GAN for field estimation, but I would like to ask that is there a part of your code that remaps the image back to the value, the color represents the corresponding value, but is it standard to calculate the value back this way?