Infusion: Preventing Customized Text-to-Image Diffusion from Overfitting

TL;DR

This paper introduces Infusion, a novel method for text-to-image customization that effectively prevents overfitting by balancing concept learning and knowledge preservation, requiring minimal parameters and outperforming existing methods.

Contribution

Infusion is a new approach that addresses overfitting in T2I customization by analyzing overfitting types, introducing metrics, and enabling efficient learning with only 11KB of parameters.

Findings

Infusion outperforms state-of-the-art methods in customized image generation.

It effectively prevents concept overfitting in limited modalities.

Requires only 11KB of trained parameters for effective customization.

Abstract

Text-to-image (T2I) customization aims to create images that embody specific visual concepts delineated in textual descriptions. However, existing works still face a main challenge, concept overfitting. To tackle this challenge, we first analyze overfitting, categorizing it into concept-agnostic overfitting, which undermines non-customized concept knowledge, and concept-specific overfitting, which is confined to customize on limited modalities, i.e, backgrounds, layouts, styles. To evaluate the overfitting degree, we further introduce two metrics, i.e, Latent Fisher divergence and Wasserstein metric to measure the distribution changes of non-customized and customized concept respectively. Drawing from the analysis, we propose Infusion, a T2I customization method that enables the learning of target concepts to avoid being constrained by limited training modalities, while preserving non-customized knowledge. Remarkably, Infusion achieves this feat with remarkable efficiency, requiring a mere 11KB of trained parameters. Extensive experiments also demonstrate that our approach outperforms state-of-the-art methods in both single and multi-concept customized generation.