Yale-CMU-Berkeley (YCB) Object and Model Set
Yale University and UC Berkeley
3D models and images of over 80 everyday objects used to benchmark robot grasping.
A benchmarking set for robotics research, with mesh models, RGB, RGB-D, and point cloud images of over 80 physical objects, such as food boxes, tools, and kitchen items. The physical objects are also available from the YCB benchmarking project. Collected with two RGB-D systems.
Access
- Cost
- FreeNo charge to download; hosted in an AWS S3 bucket.
- License
- Open licenseCC BY 4.0
- Sign-up
- No sign-up
- How you get it
- API and Bulk downloadsee documentation
What's in it
- One row is
- One object, with models and multi-view images
- Covers
- Not geographic
- Years
- Not stated
- Updated
- Annual
- Size
- Over 80 objects
Quality
A long-standing community standard with shared physical objects. Small compared to modern vision datasets.
- Small object count
- Older sensors
Connects to
- OpenRoboCare Multi-Modal Expert Demonstration Dataset: Human demonstrations for another robot task
- RACECAR Dataset: Robotics sensor data in a very different setting
Questions it could answer
- Which objects are hardest to grasp with a given gripper?
- How closely do reconstructed models match physical dimensions?