Dataset · se3labs
Bimanual YAM Robot Teleoperation Across 10 Real Tasks
se3labs/real-robodojo-teleop-500
500 demonstrations (6.3 hours) of real bimanual manipulation using three stereo cameras and LeRobot format telemetry.
What's inside
Recorded via leader-follower teleoperation on a bimanual YAM station equipped with two 6-DOF arms and parallel grippers. Contains 500 successful episodes (6.3 hours total) across 10 real tasks from the RoboDojo benchmark (50 per task). Recorded at 30 fps in the LeRobot format, featuring video from three Intel RealSense D405 cameras along with 14-dimensional state and action vectors.
What you can do with it
- Training bimanual manipulation policies on real robotics data
- Fine-tuning imitation learning models within the LeRobot ecosystem
- Tasks
- Robotics
- Size
- 100K–1M rows
- Downloads this month
- 3.3K
- Likes
- 6
How to load it
from datasets import load_dataset
ds = load_dataset("se3labs/real-robodojo-teleop-500")