Dataset · groundedsi
RoboCap-10k: robotics dataset with 3D poses
groundedsi/RoboCap-10k
10,000 hours of real human work video with head and hand pose annotations in video format for robotics.
What's inside
The corpus includes 10,000 hours of unscripted egocentric manipulation video tasks in English with metric 3D head and hand pose annotations, as well as two-level text annotations.
What you can do with it
- Training multimodal models for robotics and manipulator control
- Evaluating spatial perception and depth-estimation models
- Tasks
- Video classification, Depth estimation, Text → video, Unconditional image generation, Text → 3d
- Languages
- English
- Size
- 1M–10M rows
- Downloads this month
- 39.5K
- Likes
- 8
- Access
- on request on Hugging Face
How to load it
from datasets import load_dataset
ds = load_dataset("groundedsi/RoboCap-10k")