Dragonmocap captures dragon motion from single videos for diverse designs
DragonMoCap: Monocular Video-based Motion Capture for Diverse Dragon Topologies
Computer Vision and Pattern Recognition
Summary
Motion capture usually requires complex setups, but capturing the movement of dragons with very different body shapes and joint structures is especially hard. The authors introduce DragonMoCap, a system that treats different dragon skeletons as flexible inputs, using language-based tokens to understand joints regardless of their names or counts. This lets it work well across many dragon designs, including ones it hasn’t seen before, and it can be used in real animation studios. They show it performs better than traditional methods that assume fixed body models.
What this means in practice
- •For animation studios: Capture dragon character motions from single videos for diverse dragon shapes without manual keyframing.$Commercial implications: Enables real-time dragon animation capture in studios, improving workflow speed and reducing manual work.
- •For game developers: Generate realistic dragon animations from monocular videos for use in games featuring various dragon designs.
Authors
Zhongfei Qing, Xiaoxiang Zhao, Ling-Hao Chen, Zhongang Cai, Wenjia Wang, Yifan Wu, Qianying Liu, Wei Zeng, Bo Hu, Songnan Li
Abstract
This work explores monocular motion capture for dragon-type characters, which currently relies on manual keyframing due to the lack of high-quality datasets and parametric models. The primary obstacle lies in the extreme topological diversity across dragon rigs, where varying joint counts and naming conventions render traditional fixed-template regressors inapplicable. To address this, we introduce DragonMoCap, a novel framework that treats the target rig as a dynamic input rather than a fixed prior. By deliberately scoping the framework to the most prevalent production topologies, we ensure high-fidelity results across diverse morphologies. The core of our approach is the design of language-grounded joint query tokens, which establish a dual-property representation: ensuring intra-rig discriminability while maintaining cross-rig similarity. We further propose a non-parametric query-token retrieval mechanism to enable zero-shot adaptation to unseen rigs without retraining. Extensive evaluations demonstrate that DragonMoCap consistently outperforms HMR-style baselines across diverse dragon topologies. The practical utility of our framework is further evidenced by its integration into professional animation workflows, highlighting its potential for real-time industrial applications.