junyanz / pytorch-CycleGAN-and-pix2pix

Image-to-Image Translation in PyTorch
Other
22.8k stars 6.29k forks source link

How to set the number of training sets #1482

Closed yourbikun closed 1 year ago

yourbikun commented 2 years ago

I'm trying to adopt the idea of CycleGAN on solving a domain mapping problem. I have 100,000 images to do domain transformation, and about 10,000 images for domain B. The problem now is that I don't know how much to use as training set A and training set B. I see that the training set A and B in the data set you gave only have more than 1,000 pictures. I'm very confused right now and hope you guys can give me some advice. thanks

junyanz commented 2 years ago

More data often help. You can also run the same experiment with 10K vs. 1K images and do a comparison.

yourbikun commented 2 years ago

ok,I will try it thanks