DragonMoCap: Monocular Video-based Motion Capture for Diverse Dragon Topologies
Published:
Language-grounded joint queries and non-parametric retrieval enable monocular motion capture across diverse dragon rigs, with zero-shot adaptation to unseen topologies.
Citation
Qing, Zhongfei, Xiaoxiang Zhao, Ling-Hao Chen, Zhongang Cai, Wenjia Wang, Yifan Wu, Qianying Liu, Wei Zeng, Bo Hu, and Songnan Li. (2026). "DragonMoCap: Monocular Video-based Motion Capture for Diverse Dragon Topologies." SIGGRAPH Asia.
BibTeX
@misc{qing2026dragonmocap,
title={DragonMoCap: Monocular Video-based Motion Capture for Diverse Dragon Topologies},
author={Zhongfei Qing and Xiaoxiang Zhao and Ling-Hao Chen and Zhongang Cai and Wenjia Wang and Yifan Wu and Qianying Liu and Wei Zeng and Bo Hu and Songnan Li},
year={2026},
eprint={2610.25437},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2610.25437}
}
DragonMoCap recovers dragon motion from monocular video using language-grounded joint queries that combine semantic and skeletal information. Non-parametric query-token retrieval enables zero-shot adaptation to unseen rigs without retraining, supporting diverse dragon topologies and production animation workflows.