Browse Teach Curate

Conditional Consistency Models for Multi-Domain Image Translation

Extending consistency models, which generate an image in a single denoising step, to conditional image-to-image translation across several domains. A task-specific conditional image is fed into the denoising network so the single-step output keeps the structure and context of the source domain, giving fast inference for jobs such as visible-to-infrared, medical stain, and low-light enhancement without the many sampling steps of diffusion or the training instability of conditional GANs.

Rosie Text

Conditional Consistency Guided Image Translation and Enhancement Amil Bhagat, Milind Jain, A. V.

A computer-vision paper that extends consistency models, a fast single-step alternative to diffusion models, to conditional multi-domain image-to-image translation. The authors propose Conditional Co…