Skip to content

Dataset · EgoSteer

First-Person Bimanual Teleoperation Robotics Data

EgoSteer/EgoSteer-RealWorld

192 hours of video and telemetry of bimanual robot manipulation for training control systems.

What's inside

A real-world robotics dataset in English featuring 54,454 teleoperation episodes (192 hours or 20.75M frames) across 193 tasks. The data was recorded using a RealMan bimanual robot with Ruiyan anthropomorphic hands and two RGB-D cameras on the head and chest. It includes tabular data, time series, and video (volume ranging from 10M to 100M records).

What you can do with it

  • Training bimanual robot manipulator control models using egocentric video
  • Training fine-motor policies for multi-fingered robotic hands
  • Exploring the transfer of teleoperation demonstrations into autonomous robot skills
Tasks
Robotics
Languages
English
Size
10M–100M rows
Downloads this month
55.4K
Likes
8

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("EgoSteer/EgoSteer-RealWorld")