Select an episode to see its details.
Select an episode to see its details.
DataHive helps labs collect data for HiveBoard. Choose a section to get started; you can switch at any time from the tabs above.
Six simple steps from your robot to HiveBoard. DataHive keeps track of where you are.
DataHive helps labs collect data for HiveBoard: run the trials, record and annotate each episode, check it against HiveBoard's rules, and upload it to your lab's dataset. Data stays in samples/ until you upload.
Runner plans, times and records the trials. Annotate is where you review, annotate, validate and upload episodes.
Manual: you time the trial here and upload the HDF5 and one video per camera. Automatic: your robot script records the episode and you annotate it right after. Switch any time from the trial screen.
Yes. If you built your own collection pipeline instead of using the Runner's manual or automatic mode, you can copy your recorded episode files directly into the samples/ directory (under samples/{session_id}/episodes/). DataHive will automatically detect them so you can review, annotate, validate, and upload them in Annotate.
The HDF5 file and one video for every camera in your robot profile. The trial counts as valid only when all are uploaded and the episode passes validation.
Yes. Fill in the required fields under Robot Profile first. Editing it later only affects new episodes.
A complete profile, 100 Hz or faster, 1 to 600 seconds, matching videos on every camera, and an annotation that agrees with the recording.
Five on each of the 13 tasks, 65 in total. The Runner plans exactly that by default.
On your machine, until you press Upload or Sync. Uploads go to the Hugging Face dataset set in your configuration.