Text-driven 3D Human Generation via Contrastive Preference Optimization
This addresses a specific challenge in 3D human generation for applications like virtual reality or gaming, representing an incremental advance over existing methods.
The paper tackles the problem of accurately aligning 3D human models with long and complex textual descriptions by proposing a framework that uses contrastive preferences with positive and negative prompts to assist Score Distillation Sampling, resulting in state-of-the-art improvements in texture realism and visual alignment.
Recent advances in Score Distillation Sampling (SDS) have improved 3D human generation from textual descriptions. However, existing methods still face challenges in accurately aligning 3D models with long and complex textual inputs. To address this challenge, we propose a novel framework that introduces contrastive preferences, where human-level preference models, guided by both positive and negative prompts, assist SDS for improved alignment. Specifically, we design a preference optimization module that integrates multiple models to comprehensively capture the full range of textual features. Furthermore, we introduce a negation preference module to mitigate over-optimization of irrelevant details by leveraging static-dynamic negation prompts, effectively preventing ``reward hacking". Extensive experiments demonstrate that our method achieves state-of-the-art results, significantly enhancing texture realism and visual alignment with textual descriptions, particularly for long and complex inputs.