ember-lab-berkeley/robocasa365-pretrain-atomic
Robotics Dataset
A pretraining dataset of 7,356 human teleoperation episodes for 65 atomic single-skill tasks in RoboCasa environments, with 617 natural-language task descriptions, recorded at 20 fps with three RGB cameras on a PandaOmron robot, totaling about 21 hours of video.
Catalog metadata
- Repository owner
- ember-lab-berkeley
- Source
- Hugging Face / LeRobot
- Robot
- PandaOmron
- Episodes
- 7356
- Frames
- 1495313
- Cameras
- 3
- Frames per second
- 20
- License
- mit
Browse the catalog to compare robots, tasks, sensors, and licenses. Sign in to open protected download and model links.