Introduction
RoboX is a crowdsourced data platform for robotics imitation learning. Contributors use their smartphones to record first-person video across data campaigns covering object manipulation, social interaction, and scene understanding.
All data is anonymized on-device before upload and processed through a layered annotation pipeline that extracts new data dimensions over time from a single recording.
Instead of relying on expensive fleet vehicles or controlled lab setups, RoboX distributes collection across real people in real environments. Every phone that joins a campaign becomes a first-person data sensor.
How RoboX Works
Select a Campaign: Choose from active data missions.
Choose a Campaign: Browse active campaigns and pick a recording task.
Record a Clip: Follow the instructions and capture first-person video of the task.
On-Device Anonymization: PII is stripped locally before anything leaves your phone.
Layered Annotation: Each clip is processed server-side through an annotation pipeline designed to add new data layers over time.
Dataset Packaging: Annotated datasets are structured for use by robotics labs and research teams.
Privacy and Data Control
All personal identifiers are stripped away before data ever leaves the phone. Everything is processed locally and securely.
No exact locations are stored or shared
Faces, plates, and identifying details are removed on-device
Data is encrypted before transmission
Delete your data anytime through your account settings.
Data is collected only during active campaigns you choose to join, never in the background
All collection is opt-in and contributor-initiated
On-chain data provenance ensures every clip is traceable and verifiable
No data is sold to third parties without contributor consent
RoboX is GDPR-compliant
User decides what to record, when to record, and whether to keep or delete it.
Last updated