OhhO Data
Collect, label and ship robot demonstration data. An end-to-end pipeline in LeRobot-compatible format, with teleop recording, an episode viewer and CLI tools — the fuel your policies learn from.
Collect. Label. Ship.
Robot intelligence is only as good as the demonstrations it learns from. OhhO Data is the pipeline that turns teleoperation sessions into clean, training-ready datasets.
Record episodes from a leader arm and mobile base, synchronized with multi-camera video, and store them in the standard LeRobot dataset format — Parquet plus MP4. Review and curate with the episode viewer, then ship straight to training.
Data is designed for the whole loop: the same schema your robot records is the schema your model trains on and the policy runs in production — so what you collect is exactly what you deploy.
- Synchronized recording. Leader arm, base velocity and multiple camera streams aligned to a tight sync tolerance, frame by frame.
- Standard format. LeRobot-compatible Hugging Face datasets (Parquet + MP4) for imitation learning, plus MCAP — the ROS 2-native bag format — for raw ROS 2 topic recording. No bespoke converters, no lock-in.
- Episode viewer. Scrub, inspect and keep-or-discard episodes before they pollute a training run.
- One schema, end to end. A 9-DOF mobile-manipulation state/action that matches the recorder, the trainer and the policy.
- CLI-first. Scriptable record / inspect / push commands that fit into a data-ops workflow.
Related: OhhO Train, OhhO Serve, OhhO Pilot