[Python] What is the reason for my GAN to generate monochromatic images after jonly 1 epoch?...

text

On the GitHub above is a link to my code for a .ipynb file that should create and train a GAN based on the images in my images file. There are 5000 32x32 chest images in the file.

I have attempted to adjust the learning rates, yet every time after 1 epoch, the image end up either completely black except for the edges, or completely white for the edges. I have attempted to investigate a way to resolve it but I am unable to locate a fix, The losses for both the generator and discriminator start at around 0.7 during the initial epochs, but then fluctuate significantly, with one becoming very high and the other very low.

I have attempted adjusting the learning rates and validating the correct usage of images, but I have not discovered any outcomes.

Continue reading...