Hi, I am new to PyTorch and working with GANs. I looked around online already but I had a couple of basic beginner curiosities if anyone could help.
Not sure I understand three dimensional jpegs. HWC size? Is that similar to jpeg + frame size? I noticed a transformer resizes training images. Does the transformer crop images that aren’t square in a certain way?
Also any advice on batch size minimum to train? I created my own dataset but it’s generating static for fake images. I don’t think I have enough input. Is there a good amount of images to think about in terms of minimum training data? And anywhere I could learn more about GAN inputs?
Any ideas are welcomed!
Thanks