Alaaharoun's picture
Add dexterous hand sample: bulk_9_videos_933d5ad5 (9112 merged rows, wrist-relative fields + motion_intelligence)
acaf5f4 verified

Use cases

Export action label: Dexterous Hand movements

This dataset contains per-frame hand keypoints (MediaPipe Hands-style names when applicable) and, when enabled, dexterous_hand / motion_intelligence fields. See SCHEMA.md for the exact record shape. Typical applications:

1. Robotics manipulation & teleoperation

Train policies and estimators for reach-to-grasp, in-hand adjustment, tool use, and imitation learning from human demonstrations. Wrist trajectories and hand activity classes support retargeting to robot end-effectors; grip proxies help classify pinch vs power style grasps.

2. Gesture recognition & human-computer interaction

Build command gestures, static vs dynamic hand pose classifiers, and multimodal interfaces (vision + speech). Useful for accessibility (alternative input) and smart displays.

3. AR / VR & spatial computing

Drive hand-tracked UIs, avatar upper-body animation, and social presence in VR. Normalized keypoints integrate with Unity, Unreal, WebXR via your retargeting layer.

4. Fine motor control & rehabilitation

Track movement smoothness, range of motion proxies, and session-level trends from features.json and per-frame quality scores.

5. Animation & digital humans

Retarget performer hand motion to rigs; combine with artist rigs or additional mocap for film-quality fingers.

6. Industrial ergonomics & safety

Analyze repetitive hand/wrist paths and reach envelopes for workstation design.

7. Research & benchmarking

Use the export as a standardized JSONL stream for comparing filters and models on hand-centric labels.


Note: This document targets dexterous / hand exports. Full-body walking / gait use cases are not the focus here; see other Quality Vision walking samples if you need locomotion data.