← Use cases
Use case · Robotics
Egocentric data for robotics.
First-person footage of human hands performing real tasks — the demonstrations robots learn manipulation, grasping, and tool use from.
The need
Robots learn manipulation most efficiently from human demonstration. A head-mounted, first-person view captures the hand–eye coordination, contact moments, and task sequences that fixed third-person cameras miss — the raw material for imitation learning and manipulation policies.
What we provide
- Hand–object manipulation and grasping sequences
- Multi-step tool-use and assembly tasks
- Synced hand motion, IMU, and gaze alongside video
- Diverse objects, materials, and environments
How it's collected
- Scripted task scenarios plus natural, free-form activity
- Fully consented contributors
- Lightweight head-mounted multimodal capture rigs
- Varied settings — kitchen, workshop, lab, field
How it's delivered
- Frame-level action, object, and contact annotations
- Standard, model-ready formats and splits
- Documented provenance and licensing
More use cases
Need Robotics data?
Tell us the tasks and environments — we'll scope the capture.