Image is All You Need to Empower Large-scale Diffusion Models for In-Domain Generation

TL;DR

This paper introduces a guidance-decoupled prior preservation method that enables high-quality, controllable in-domain image generation using label-free data and pre-trained diffusion models, broadening their application scope.

Contribution

The authors propose a novel guidance-decoupled prior preservation technique and an efficient domain knowledge learning method to enhance in-domain diffusion model generation without labeled data.

Findings

Achieves superior in-domain image synthesis quality.

Maintains open-world control and unconditional generation capabilities.

Compatible with various diffusion-based control methods.

Abstract

In-domain generation aims to perform a variety of tasks within a specific domain, such as unconditional generation, text-to-image, image editing, 3D generation, and more. Early research typically required training specialized generators for each unique task and domain, often relying on fully-labeled data. Motivated by the powerful generative capabilities and broad applications of diffusion models, we are driven to explore leveraging label-free data to empower these models for in-domain generation. Fine-tuning a pre-trained generative model on domain data is an intuitive but challenging way and often requires complex manual hyper-parameter adjustments since the limited diversity of the training data can easily disrupt the model's original generative capabilities. To address this challenge, we propose a guidance-decoupled prior preservation mechanism to achieve high generative quality and controllability by image-only data, inspired by preserving the pre-trained model from a denoising guidance perspective. We decouple domain-related guidance from the conditional guidance used in classifier-free guidance mechanisms to preserve open-world control guidance and unconditional guidance from the pre-trained model. We further propose an efficient domain knowledge learning technique to train an additional text-free UNet copy to predict domain guidance. Besides, we theoretically illustrate a multi-guidance in-domain generation pipeline for a variety of generative tasks, leveraging multiple guidances from distinct diffusion models and conditions. Extensive experiments demonstrate the superiority of our method in domain-specific synthesis and its compatibility with various diffusion-based control methods and applications.