Skip to content

Dataset · Crobotia

Bimanual garment manipulation demonstrations for robots

Crobotia/iUMI

A large dataset (100K-1M records) with video, sensor readings, and clothing folding annotations in real homes.

What's inside

The Crobotia dataset for the IROS 2026 competition contains recordings of bimanual garment manipulations (unfolding, smoothing, and folding) in real home environments. Each episode includes left and right gripper videos, camera pose estimates, gripper opening angles, IMU data, audio, and multiple annotation layers; documentation is presented in English and Chinese.

What you can do with it

  • Training bimanual robot control policies for deformable objects (garments)
  • Multimodal training of robotic models on video, audio, and IMU data
Tasks
Robotics
Size
100K–1M rows
Downloads this month
29K
Likes
7

How to load it

from datasets import load_dataset

ds = load_dataset("Crobotia/iUMI")