Closed sebastiangonsal closed 7 years ago
Did you write a GAN according to the original algorithm? These often suddenly explode and turn from good to complete garbage. With the DCGAN architecture that happens less often. WGAN has even better stability.
I did write it according to the DCGAN architecture. THanks
Hi, I have written a gan implementation but find that it starts generating random faces after a lot of iterations after giving good results until starting 5 -6 epochs.
Did you find your loss to increase in higher iterations or it just got better?