A dataset of simulated robot manipulation trajectories from the ManiSkill environment. The data appears to include RGB images and robot state information for tasks like cube picking. It was created by author 'surgingTu' and last updated on the Hugging Face platform in April 2026.
Use Cases
- Training visual imitation learning models based on the provided RGB trajectory data.
- Developing state estimation models for robotic arms using the combined image and state data.
- Benchmarking reinforcement learning algorithms on the PickCube task mentioned in the description.
- Analyzing robot manipulation performance across different viewpoints and robot configurations.
Strengths
- Includes multiple versions of data (e.g., pickcube_rgb_random_v2, pickcube_rgb_traj_v3) suggesting iterative refinement.
- Data appears to cover specific robotic tasks like 'PickCube' with different data collection strategies (trajectory vs. random).
Limitations
- Description metadata is limited; actual data quality requires manual inspection after download.
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count, file formats, and license information are unknown, which may limit suitability assessment.
Provenance
- Source
- Hugging Face, created by user 'surgingTu'.
- Collection Method
- Likely generated through simulation in the ManiSkill environment.
- Time Range
- Update logs in the description span from November to December of an unspecified year.
- Freshness
- Last updated 2026-04-03 15:32:57; freshness should be verified.
- Geography
- null