One-to-Many translation

[rose-jinyang]

Hello
How are you?
Thanks for contributing to this project.
Is it possible to translate from one A to several possible Bs?

[msseibel]

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.

[msseibel]

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