I am training a GAN model (DCGAN) to generate 128x128 images. Now, I wish to add a function which will take the generator output, perform some pre-defined operations on it, and return the modified output. The output of this function will be the "fake" images to be passed to the discriminator. Will such a model make sense and is there anything I should keep in mind while training this model?
PS: To explain the custom function further, the function takes an input, uses a transformation matrix to transform it into a different space. Adds some noise in the new space and transforms it back into the original form (but now with the added noise). I have multiple such pairs of transformation matrices. Also, I am using pyTorch.