VimoRAG: Video-based Retrieval-augmented 3D Motion Generation for Motion Language Models

This paper introduces VimoRAG, a novel video-based retrieval-augmented motion generation framework for motion large language models (LLMs). As motion LLMs face severe out-of-domain/out-of-vocabulary issues due to limited annotated data, VimoRAG leverages large-scale in-the-wild video databases to enhance 3D motion generation by retrieving relevant 2D human motion signals. While video-based motion RAG is nontrivial, we address two key bottlenecks: (1) developing an effective motion-centered video retrieval model that distinguishes human poses and actions, and (2) mitigating the issue of error propagation caused by suboptimal retrieval results. We design the Gemini Motion Video Retriever mechanism and the Motion-centric Dual-alignment DPO Trainer, enabling effective retrieval and generation processes. Experimental results show that VimoRAG significantly boosts the performance of motion LLMs constrained to text-only input.

Authors

• Haidong Xu
• Guangwei Xu Alibaba Cloud Computing
• Zhedong Zheng
• Xiatian Zhu University of Surrey
• Wei Ji School of Medicine, Yale University, New Haven, United States
• Xiangtai Li Skywork AI Nanyang Technological University
• Ruijie Guo Alibaba Cloud Computing
• Meishan Zhang Institute of Computing and Intelligence, Harbin Institute of Technology, Shenzhen, China
• Min Zhang College of Intelligence and Computing, Tianjin University
• Hao Fei National University of Singapore, Singapore

Keywords

No keywords are indexed for this paper.