Open rose-jinyang opened 1 year ago
Did you try to use the StyleGAN Generator? That should allow you to specify a latent vector for controlling the translated image. I am also very much interested in seeing CUT combined with a StyleGAN and with a style modulation layer. (The paper used a StyleGAN2, but without style modulation). Feel free to share.
Just found a paper that uses an approach similar to CUT and combines it with neural style transfer:
Domain Enhanced Arbitrary Image Style Transfer via Contrastive Learning, Zhang et al. 2022, at SIGGRAPH
Hello How are you? Thanks for contributing to this project. Is it possible to translate from one A to several possible Bs?