I trained a small model for image super resolution using autoencoders. When I use the test data for making a prediction the output image is of different shades.
I use
