Pretraining is All You Need for Image-to-Image Translation

TL;DR

This paper introduces a pretraining-based framework for image-to-image translation that leverages pretrained diffusion models, significantly improving realism and fidelity across diverse tasks without extensive task-specific architecture design.

Contribution

It presents a simple, generic approach that adapts pretrained diffusion models for various image translation tasks, incorporating adversarial training and guidance sampling for enhanced quality.

Findings

Outperforms existing methods on benchmarks like ADE20K, COCO-Stuff, and DIODE.

Produces images with unprecedented realism and faithfulness.

Requires less task-specific architecture design.

Abstract

We propose to use pretraining to boost general image-to-image translation. Prior image-to-image translation methods usually need dedicated architectural design and train individual translation models from scratch, struggling for high-quality generation of complex scenes, especially when paired training data are not abundant. In this paper, we regard each image-to-image translation problem as a downstream task and introduce a simple and generic framework that adapts a pretrained diffusion model to accommodate various kinds of image-to-image translation. We also propose adversarial training to enhance the texture synthesis in the diffusion model training, in conjunction with normalized guidance sampling to improve the generation quality. We present extensive empirical comparison across various tasks on challenging benchmarks such as ADE20K, COCO-Stuff, and DIODE, showing the proposed pretraining-based image-to-image translation (PITI) is capable of synthesizing images of unprecedented realism and faithfulness.