davidwdw2026
Task01 v14: 109 successful dense Codegen trajectories
A validated subset of 109 successful episodes from BEHAVIOR-1K picking_up_trash task using Codegen with v14 teacher model, containing 61D state and 23D action representations across three RGB and three depth camera streams at 30 Hz.
Downloads113
Episodes109
Hours5.4505
Why This Matters for Physical AI
This densely annotated, validated trajectory dataset provides high-quality demonstrations for training robotic manipulation policies on trash-picking tasks, with precise action-state alignment critical for behavior cloning and imitation learning approaches.
Technical Profile
- Modalities
- rgbdepthproprioception
- Robot Embodiments
- R1Pro
- Action Space
- 23-DoF: action
- Environment
- simulation
- Task Types
- pick_and_placemanipulation
- Episodes
- 109
- Total Hours
- 5.4505
- Data Format
- LeRobot
- Annotation Types
- action_labelsstate_labels
Access
Need custom rgb data?
Claru builds purpose-built datasets for simulation applications with dense human annotations and quality assurance.
Request a Sample Pack