CoMA: Compositional Human Motion Generation with Multi-modal Agents

CoMA: Compositional Human Motion Generation with Multi-modal Agents

Shanlin Sun*, Gabriel De Araujo*, Jiaqi Xu*, Shenghan Zhou*, Hanwen Zhang, Ziheng Huang, Chenyu You and Xiaohui Xie

(* Equal Contribution)

• Presented by University of California, Irvine; Southeast University; Chongqing University; Huazhong University of Science and Technology; Northeastern University; Stony Brook University
• :mailbox_with_mail: Primary contact: Shanlin Sun ( shanlins@uci.edu )

Highlights <a name="highlights"></a>

:star2: CoMA, a compositional human motion generation framework with multi-modal agents.

:star2: CoMA can generate high-quality motion sequences given long, complex and context-rich text prompts.

📰 News

📝 TODO List

• [ ] Release CoMA full implementation.
• [ ] Release MVC training code.
• [ ] Release SPAM training code.
• [ ] Release MVC inference code and checkpoints.
• [ ] Release SPAM inference code and checkpoints.