GVHMR: Recovering World-Grounded Human Motion from Monocular Video, Now Open Source
rsasaki0109 · x · 2026-08-26
Zhejiang University's team open-sourced GVHMR (SIGGRAPH Asia 2024, accepted to TPAMI), which recovers world-grounded global human motion from monocular videos.
The key idea is a novel Gravity-View Coordinate system: using gravity and camera-view directions, the coordinates can be uniquely inferred from each image, decoupling pose prediction from the camera frame and solving drift and orientation ambiguity in global motion recovery.
It ships with Google Colab and HuggingFace demos. A March 2025 update replaced DPVO with a self-built SimpleVO for visual odometry — faster and more compatible — and added an option to specify full-frame focal length in millimeters. Use -s to skip visual odometry when the camera is static.
Related event: GVHMR: Open-Source Recovery of Global Human Motion from Monocular Video(2 posts)→
More from Research
- Yoav Goldberg Questions S1 Model's 'In-Context Learning' Claim — ivan_bezdomny · 2026-08-26
- Recuris: Recursive Experiential-Working Memory Evolution for Long-Horizon Agents — princetonu · 2026-08-26
- Meta^n: Recursive Self-Improvement through Emergent Depth — Zae Myung Kim · 2026-08-26
- Tencent's CAFE: Self-Improving Search Agents Need Co-Evolving Feedback — Tencent-Hunyuan · 2026-08-26
- Accelerated Understanding Launches AI Model Replacing Transformers with Neural Operators — badumtsssst · 2026-08-26
- Analyst replicates 300+ ICML papers finding errors using agents — _akhaliq · 2026-08-26