What’s Out There
← Back to swiping
Images & videoFreeCompany

FineART

Scale AI and Hugging Face LeRobot

40,543 robot episodes and 533,913 human-written subtask labels covering 1,718 hours.

A bimanual manipulation dataset on a Trossen Stationary AI robot, where every episode has dense human-written subtask annotations. One row is an episode with four camera streams, 14-dimensional joint and gripper state and action, and language. It spans 151 tasks, mostly pick-and-place plus contact-rich skills.

Access

Cost
FreeNo charge.
License
Open licenseCC BY 4.0
Sign-up
No sign-up
How you get it
API and Bulk downloadLeRobot v3.0 (Parquet and MP4)

What's in it

One row is
One teleoperated robot episode (about 2.5 minutes)
Covers
Not geographic
Years
2026
Updated
One-time release
Size
40,543 episodes, 185.5 million frames, 1,718 hours, 151 tasks
4 camera video streamsjoint and gripper stateactionsepisode summarysubtask segments with text

Quality

4 of 5

Detailed documentation, an audited human annotation process, and standard format. A single robot type limits generality.

  • Single robot embodiment
  • About 77 percent of hours are everyday pick-and-place
  • Annotation language style is specific to one vendor

Connects to

  • ABC-130k: Larger bimanual teleoperation set with separate subtask annotations
  • AgiBot World 2026: Real-world dual-arm set on another platform

Questions it could answer

  • Which subtasks most often precede failures?
  • How does subtask count relate to episode length?

Similar data sets

  • ABC-130k

    Billed as the largest open-source robot teleoperation dataset: bimanual trajectories on two-arm YAM stations.

    Images & videoXDOFFree

  • AgiBot World 2026

    Real-world robot demonstrations from homes and commercial spaces, collected on the AgiBot G2 platform.

    Images & videoAgiBotFree

  • OpenRoboCare

    315 sessions of occupational therapists doing caregiving tasks, with video, touch, and gaze.

    Images & videoEmPRISE Lab at Cornell UniversityFree

  • Kinetics-700

    650,000 ten-second YouTube clips of people doing 700 things, from yoga to changing a tire.

    Images & videoGoogle DeepMind (hosted by the Common Visual Data Foundation)Free

  • GDPval

    220 real work tasks across 44 occupations, with reference files, to test AI on economically valuable output.

    LaborOpenAIFree

  • Ego4D

    3,670 hours of daily life filmed from head-mounted cameras by 923 people in 9 countries.

    Images & videoEgo4D consortium (Meta AI and university partners)By application