Boost Your Human Image Generation Model via Direct Preference Optimization

2024-05-30

AI Summary

To address anatomical distortions and inaccurate pose modeling—key bottlenecks in human image generation that impair photorealism—this paper proposes HG-DPO, the first Direct Preference Optimization (DPO) framework leveraging high-fidelity real images as “preferred” samples. Unlike conventional DPO approaches that rely on generated images as positive examples, HG-DPO constructs fine-grained human-centric preference supervision using real images and introduces a curriculum learning strategy to mitigate training instability. The method integrates text-conditioned diffusion models with contrastive modeling between real and generated images, enabling both identity preservation and personalized text-to-image synthesis. Extensive evaluations across multiple benchmarks demonstrate significant improvements in anatomical plausibility, perceptual realism, and identity consistency, comprehensively outperforming existing state-of-the-art methods.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Humans and AI: Learning Human Values and Preferences
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Human image generation is a key focus in image synthesis due to its broad applications, but even slight inaccuracies in anatomy, pose, or details can compromise realism. To address these challenges, we explore Direct Preference Optimization (DPO), which trains models to generate preferred (winning) images while diverging from non-preferred (losing) ones. However, conventional DPO methods use generated images as winning images, limiting realism. To overcome this limitation, we propose an enhanced DPO approach that incorporates high-quality real images as winning images, encouraging outputs to resemble real images rather than generated ones. However, implementing this concept is not a trivial task. Therefore, our approach, HG-DPO (Human image Generation through DPO), employs a novel curriculum learning framework that gradually improves the output of the model toward greater realism, making training more feasible. Furthermore, HG-DPO effectively adapts to personalized text-to-image tasks, generating high-quality and identity-specific images, which highlights the practical value of our approach.

Problem

• Improving realism in human image generation models
• Enhancing DPO with real images for better outputs
• Adapting HG-DPO for personalized text-to-image tasks

Innovation

• Uses real images as winning samples in DPO
• Introduces curriculum learning for gradual realism
• Adapts DPO for personalized text-to-image tasks