CC BY 4.0
The publisher's licence permits commercial use and redistribution, with attribution.
Full set is a 36 GB tar.gz on Google Drive; a ~100 MB sampler is at gdown https://drive.google.com/uc?id=1YX2TgT8IKSn9V4wGCwdzbRnS53yicV2P&export=download (robonet_sampler.tar.gz). Google Drive quota limits can block large gdown pulls.
Specification, as published
- Episodes (stated)
- 162,000 episodes
- Scale
- 15 million video frames, 7 different robot platforms, 113 unique camera viewpoints
- Task
- Pick & place, Other
- Environment
- Indoor – lab
- Modality
- RGB, Actions, Proprioception
- Capture
- Exocentric
- In frame
- Robot hands
- Captured in
- United States of America
- Embodiment
- Sawyer, Franka Panda, Baxter, Fetch +3 more
- Formats
- HDF5
- Size
- 36 GB
- Hosted on
- Google Drive
Figures are the publisher's own. Kinetic Blocks has not measured this dataset.
About this dataset
RoboNet pools random and scripted object-interaction videos from seven robot arms (Sawyer, Franka Panda, Baxter, Fetch, Google R3, Kuka LBR iiwa, WidowX) recorded at four institutions. Each trajectory pairs RGB frames from one or more of 113 fixed viewpoints with the commanded actions and robot state. The corpus was built for video prediction and inverse-model pretraining rather than task-labelled demonstrations. It ships as a single 36 GB HDF5 archive on Google Drive.
Publisher's own task names: random object interaction, scripted object interaction, visual foresight, inverse model learning.
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.