Subject-driven Text-to-Image Generation via Preference-based Reinforcement Learning

TL;DR

This paper introduces a preference-based reinforcement learning method for subject-driven text-to-image generation that improves training efficiency and model regularization without requiring complex text encoder training.

Contribution

It proposes the $\lambda$-Harmonic reward function and Reward Preference Optimization (RPO), enabling faster, more efficient subject-driven image synthesis with fewer negative samples and no need for text encoder optimization.

Findings

Achieves state-of-the-art CLIP-I score of 0.833 on DreamBench.

Requires only 3% of negative samples compared to DreamBooth.

Provides reliable model selection via $\lambda$-Harmonic reward function.

Abstract

Text-to-image generative models have recently attracted considerable interest, enabling the synthesis of high-quality images from textual prompts. However, these models often lack the capability to generate specific subjects from given reference images or to synthesize novel renditions under varying conditions. Methods like DreamBooth and Subject-driven Text-to-Image (SuTI) have made significant progress in this area. Yet, both approaches primarily focus on enhancing similarity to reference images and require expensive setups, often overlooking the need for efficient training and avoiding overfitting to the reference images. In this work, we present the $\lambda$-Harmonic reward function, which provides a reliable reward signal and enables early stopping for faster training and effective regularization. By combining the Bradley-Terry preference model, the $\lambda$-Harmonic reward function also provides preference labels for subject-driven generation tasks. We propose Reward Preference Optimization (RPO), which offers a simpler setup (requiring only $3\%$ of the negative samples used by DreamBooth) and fewer gradient steps for fine-tuning. Unlike most existing methods, our approach does not require training a text encoder or optimizing text embeddings and achieves text-image alignment by fine-tuning only the U-Net component. Empirically, $\lambda$-Harmonic proves to be a reliable approach for model selection in subject-driven generation tasks. Based on preference labels and early stopping validation from the $\lambda$-Harmonic reward function, our algorithm achieves a state-of-the-art CLIP-I score of 0.833 and a CLIP-T score of 0.314 on DreamBench.