Mitigate Replication and Copying in Diffusion Models with Generalized Caption and Dual Fusion Enhancement

TL;DR

This paper proposes a method to reduce data replication in diffusion models by using generalized captions and dual fusion enhancement, significantly improving privacy without sacrificing image quality.

Contribution

It introduces a generality score for captions, employs large language models for caption generalization, and proposes a dual fusion approach to mitigate data replication in diffusion models.

Findings

Reduced replication by 43.5%

Maintained diversity and quality of generated images

Effective privacy enhancement in diffusion models

Abstract

While diffusion models demonstrate a remarkable capability for generating high-quality images, their tendency to `replicate' training data raises privacy concerns. Although recent research suggests that this replication may stem from the insufficient generalization of training data captions and duplication of training images, effective mitigation strategies remain elusive. To address this gap, our paper first introduces a generality score that measures the caption generality and employ large language model (LLM) to generalize training captions. Subsequently, we leverage generalized captions and propose a novel dual fusion enhancement approach to mitigate the replication of diffusion models. Our empirical results demonstrate that our proposed methods can significantly reduce replication by 43.5% compared to the original diffusion model while maintaining the diversity and quality of generations. Code is available at https://github.com/HowardLi0816/dual-fusion-diffusion.