UNIT-DDPM: UNpaired Image Translation with Denoising Diffusion Probabilistic Models
This addresses the problem of generating high-quality translated images for researchers and practitioners in computer vision, offering a stable alternative to adversarial methods.
The paper tackles unpaired image-to-image translation by proposing UNIT-DDPM, a method using denoising diffusion probabilistic models without adversarial training, achieving state-of-the-art Fréchet Inception Distance performance on multiple datasets.
We propose a novel unpaired image-to-image translation method that uses denoising diffusion probabilistic models without requiring adversarial training. Our method, UNpaired Image Translation with Denoising Diffusion Probabilistic Models (UNIT-DDPM), trains a generative model to infer the joint distribution of images over both domains as a Markov chain by minimising a denoising score matching objective conditioned on the other domain. In particular, we update both domain translation models simultaneously, and we generate target domain images by a denoising Markov Chain Monte Carlo approach that is conditioned on the input source domain images, based on Langevin dynamics. Our approach provides stable model training for image-to-image translation and generates high-quality image outputs. This enables state-of-the-art Fréchet Inception Distance (FID) performance on several public datasets, including both colour and multispectral imagery, significantly outperforming the contemporary adversarial image-to-image translation methods.