Talk: Unpaired Image-to-Image Translation (I2I) — CUT++

SDSC internal seminar on unpaired image-to-image translation following up on Contrastive Learning for Unpaired Image-to-Image Translation (CUT).

Image translation between domains while preserving source content benefits data augmentation, domain transfer, and artistic applications. I cover methods that propose non-cyclic mappings from source to target domain, enabling more efficient training than CycleGAN-style approaches.