Regularized Distribution Matching Distillation for One-step Unpaired Image-to-Image Translation

TL;DR

This paper introduces a regularized version of Distribution Matching Distillation for unpaired image-to-image translation, achieving comparable or superior results to multi-step diffusion models in a one-step framework.

Contribution

It proposes a novel regularized distillation method tailored for unpaired I2I translation, extending the DMD framework beyond unconditional generation.

Findings

Performs on par or better than multi-step diffusion baselines

Effective in various 2D and dataset-to-dataset translation tasks

Maintains image quality with one-step generation

Abstract

Diffusion distillation methods aim to compress the diffusion models into efficient one-step generators while trying to preserve quality. Among them, Distribution Matching Distillation (DMD) offers a suitable framework for training general-form one-step generators, applicable beyond unconditional generation. In this work, we introduce its modification, called Regularized Distribution Matching Distillation, applicable to unpaired image-to-image (I2I) problems. We demonstrate its empirical performance in application to several translation tasks, including 2D examples and I2I between different image datasets, where it performs on par or better than multi-step diffusion baselines.