Home · Open data · Manipulation · Robot video
PrimeBot Household Bimanual Manipulation Challenge Dataset
Roughly 532 hours and 32,518 episodes of wheeled bimanual robot teleoperation on laundry: washer, basket, folding; 720p, frame-level language, CC BY-SA.
Published by PrimeBot Research Institute (Swancor Advanced Materials), 2026-08
CC BY-SA 4.0
The publisher's licence permits commercial use and redistribution, with attribution. Hub card tag cc-by-sa-4.0; no licence sentence in the card body. Share-alike: adaptations must carry the same licence.
hf download challenge-2026/challenge_data --repo-type dataset --include official_data/robot_data/fold_cloth_batch2_train/* --local-dir ./challenge_dataPublic, no gate. About 5 TB in all; the command pulls the cloth-folding batch only. Teleoperation is under official_data/robot_data (four batches), UMI under official_data/umi_data, validation under validation_data.
Specification, as published
- Hours (derived)
- 532 h
- Episodes (stated)
- 32,518 episodes
- Scale
- All 500+ hours of real robot data released open source
- Task
- Laundry & folding, Pick & place, Other
- Environment
- Indoor – home
- Modality
- RGB, Proprioception, Actions, Language
- Capture
- Egocentric
- In frame
- Robot hands
- Captured in
- Asia
- Embodiment
- x2w wheeled bimanual robot (PrimeBot), UMI handheld grippers
- Frame rate
- 30 fps
- Resolution
- 1280x720 (teleoperation); 960x960 (UMI)
- Formats
- LeRobot v2.1, LeRobot v3.0, Parquet, MP4
- Size
- about 5 TB on the Hub
- Hosted on
- Hugging Face
Figures are the publisher's own. Kinetic Blocks has not measured this dataset. Derived from the meta/info.json of the four teleoperation training batches: 57,423,103 frames at 30 fps; the UMI data (about 205 hours by the changelog) is not included. Frames are timesteps, not summed across the three cameras.
About this dataset
The training data for PrimeBot's household bimanual manipulation challenge: teleoperation of the x2w wheeled dual-arm robot (URDF included) through a home laundry routine, opening and closing the washing machine, loading and unloading clothes, carrying the basket, and unfolding and folding garments. Four independent training batches carry frame-level or segment-level language instructions. Each episode has a head RealSense and two wrist ZED X One cameras at 1280x720 and 30 fps, with proprioception and an 89-dimensional action vector. The repo also holds about 205 hours of handheld UMI gripper data (960x960, LeRobot v3.0), a validation split with noised state and zeroed actions, and a base inference Docker image.
Publisher's own task names: open the washing machine door, close the washing machine door, put clothing into the washer, take clothing out of the washer, pick up the laundry basket, put the basket on the ground, put clothing on the sofa, unfold and fold clothing, place folded clothing in storage.
What Kinetic Blocks did, and did not do
Indexed and linked from the publisher. Not hosted, verified or graded by Kinetic Blocks. The publisher's terms govern. The licence shown is the one stated on the publisher's page, linked above. Nothing on this page is a claim by Kinetic Blocks.
Want the verified ones too?
Request access to the marketplace, where verified listings are checked against their files and graded A+ to F, and your money waits until the data holds up.