Robotics data formats and delivery
How recorded and annotated data is packaged, reviewed, and handed over to robotics, embodied AI, and computer vision teams.
What we deliver today
MP4 video
Recorded footage is delivered as MP4 video files.
File naming & upload
Files follow a naming convention and upload procedure agreed at the start of each project.
Project-specific metadata
Metadata fields such as task, environment, and timestamps are defined with the client for each program.
Annotation & processing
Annotation and processing workflows — filtering, timestamping, and labeling — are applied per project guidelines.
Quality review & acceptance
Deliveries pass multi-stage review against the agreed acceptance criteria before handover.
Secure delivery
The delivery method and access controls are arranged with each client.
Specialized formats we can evaluate
Some robotics teams work with formats such as LeRobot datasets, RLDS, HDF5, or synchronized multi-sensor data. These can be discussed and evaluated for your project; they are not listed here as currently supported formats.
Tell us your target format and we will confirm feasibility before scoping.
Conversion and pipeline support we can scope
LeRobot dataset conversion
If your team trains with LeRobot, we can review your target structure and confirm whether we can package recordings to match it.
RLDS for VLA models
For vision-language-action pipelines using RLDS, we can discuss episode structure, metadata, and task labels for your project.
HDF5 demonstration data
Where HDF5 is required, we can evaluate how video, labels, and metadata would be organised before any commitment.
These are discussed per project and confirmed in scoping; they are not standard deliverables today.
Have a Data Collection or Annotation Project?
Whether you need real-world data collected, an existing dataset annotated, or an end-to-end data pipeline, Veda Solutions can build and scale the workflow around your requirements.