Everything a training team needs in order to read a release
Dataset structure, data formats, annotation schema, versioning, licensing and security — with the loader code and manifest excerpts a technical reviewer asks for.
Start here
Getting started
A release is not a folder of videos. It is video plus structure, annotations, provenance, licence and documentation, versioned and delivered as one immutable unit.
Dataset structure
Every release uses the same directory layout and the same file names, so a loader written against one dataset works against the next one.
Data formats
Formats are chosen so a robotics team can read a release with tools it already has, without a proprietary runtime.
Reference
Python SDK
The SDK is a thin layer over the release layout: it resolves versions, verifies checksums and hands you episodes and frames. Everything it does can also be done with pyarrow and a video reader.
API reference
A read-only REST API over the datasets and versions you hold a licence for. It returns metadata and short-lived download URLs; it never streams video itself.
Annotation schema
The schema describes one attempt at one task in enough structure to train against and to debug against. It is versioned independently of the datasets that use it.
Operations
Versioning
A training run should always be attributable to an exact dataset state. Releases are immutable and identified by a version you can pin.
Licensing
Licence scope is agreed before delivery and shipped inside the release, so what you may do with a dataset is answerable from the files themselves.
Security
Raw footage is the sensitive asset in this pipeline. These are the controls that apply to it from upload to delivery.
Start with a scoped pilot
Bring a task. On the call we scope the capture protocol, the episode volume, the annotation schema and the delivery format.
Pilots typically start from approximately €25,000. This is an indicative figure only—final scope and pricing are set after a technical qualification call.