LLM4GEN: Leveraging Semantic Representation of LLMs for Text-to-Image Generation

TL;DR

This paper introduces LLM4GEN, a framework that enhances text-to-image diffusion models by integrating large language model representations, improving semantic understanding, especially for complex prompts, and demonstrating significant performance gains.

Contribution

The paper presents a novel plug-and-play module, CAM, and an entity-guided regularization loss to improve semantic alignment and image quality in text-to-image generation models.

Findings

9.69% improvement in color accuracy for SD1.5

12.90% improvement in color accuracy for SDXL

Outperforms existing models in image-text alignment and human evaluations

Abstract

Diffusion models have exhibited substantial success in text-to-image generation. However, they often encounter challenges when dealing with complex and dense prompts involving multiple objects, attribute binding, and long descriptions. In this paper, we propose a novel framework called \textbf{LLM4GEN}, which enhances the semantic understanding of text-to-image diffusion models by leveraging the representation of Large Language Models (LLMs). It can be seamlessly incorporated into various diffusion models as a plug-and-play component. A specially designed Cross-Adapter Module (CAM) integrates the original text features of text-to-image models with LLM features, thereby enhancing text-to-image generation. Additionally, to facilitate and correct entity-attribute relationships in text prompts, we develop an entity-guided regularization loss to further improve generation performance. We also introduce DensePrompts, which contains $7,000$ dense prompts to provide a comprehensive evaluation for the text-to-image generation task. Experiments indicate that LLM4GEN significantly improves the semantic alignment of SD1.5 and SDXL, demonstrating increases of 9.69\% and 12.90\% in color on T2I-CompBench, respectively. Moreover, it surpasses existing models in terms of sample quality, image-text alignment, and human evaluation.