What a robotics episode contains
- Sensor streams: RGB, depth, proprioception, sometimes force/torque.
- Action stream: joint or end-effector commands at a fixed rate.
- Language annotation: the instruction or a post-hoc description of the task.
- Outcome label: success, failure, or partial, often human-judged.
Cross-embodiment pooling
The Open X-Embodiment collaboration pooled datasets from many robot types into a shared format (Open X-Embodiment Collaboration, arXiv:2310.08864). For owners, this signals that format conversion and per-robot metadata (kinematics, control frequency, camera placement) are part of what makes data reusable.
QA specific to robotics
| Check | Why |
|---|---|
| Instruction matches observed behaviour | Post-hoc captions drift from what happened |
| Success label criteria written down | 'Success' varies by operator |
| Timestamp alignment across sensors | Misalignment silently corrupts training |
Rights notes
Teleoperation data recorded in homes or workplaces may capture people and private spaces. Treat it like video for consent purposes, and confirm that operators' contracts cover their demonstrations.
Have an annotated dataset or specialist expertise?
Tell us what you own, how the labels were reviewed and which rights you can license. A short overview of your dataset or annotation business is a useful place to start.
Email us about your dataset
Prepare for a licensing conversation Prefer to write directly? hello@callwavemedia.com