MotionDecode

🆕 Open-Source Release: Unitree G1 Retargeted Data !!We are releasing 1000 hours of robot-ready motion trajectories retargeted to the Unitree G1 humanoid. All data is provided in CSV format under the samples/ directory. Please indicate the source of the data when using it: from Chingmu.** ChingMu 1000-Hour Embodied Motion Dataset High-precision **optical motion capture** data for humanoid robots, dexterous hands, embodied AI, and virtual production. | | | |---|---:| | **Duration** | **1000+ hours** @ 120 Hz | | **Scenarios ** | 15+ real-world scenes | | **Tasks ** | 500+ standardized tasks | | **Objects** | 200+ tracked props (6D pose) | | **Modalities** | Skeleton · Finger · Object 6D · Video · Labels | | **Formats ** | BVH · Retargeted CSV · NPZ | ✅ **Access note:** This dataset is fully open and publicly accessible. Key Features **Optical ground truth** – sub-mm accuracy, 120 fps, no estimation errors. **Dexterous hands** – 20+ DoF per hand, synchronized with object 6DOP pose. **Robot-ready** – pre-retargeted to Unitree G1; custom retargeting available. **Real-world diversity** – 15+ scenarios, 500+ tasks, 200+ objects. **Multi-modal** – full-body skeleton, finger motion, object pose, multi-view video, semantic labels. **Quality assured** – every take passes automated cleaning + manual inspection; quality flags provided. Dataset Summary ChingMu 1000H is an optical motion capture dataset designed for training and validating embodied AI and humanoid robot controllers. It covers full-body skeleton, finger articulation, object 6D pose, multi-view video, and semantic labels across 15+ real-world scenarios (industrial, household, retail, healthcare, logistics, agriculture, performance). All data is cleaned, quality-assessed, and robot-retargeted. Data Format Specifications | Component | Format | Details | |---|---|---| | Raw motion | `.bvh` | Y-up, 120 fps, ZYX rotation, cm, 47–67 joints | | Retargeted trajectories | `.csv` | Root position (m), quaternion, joint angles