Solutions · Data collection

Captured where the work happens.

Egocentric, exocentric and multi-sensor capture in real environments, on your site or in our lab. Every episode arrives synchronized, reviewed twice and rights-cleared.

Scope a collection batch Watch a sample episode

Modalities & rigs

What we capture

Egocentric & handheld

Head-mounted cameras and handheld UMI grippers for first-person demonstrations of real tasks.
Multi-sensor rigs

Multi-camera, stereo, depth, LiDAR and IMU, time-aligned per episode. The rig used is recorded in every episode’s metadata.
Scripted and in-the-wild

Scripted protocols when you need controlled variations. In-the-wild sessions when you need the long tail of real work.

Sites

Where we collect

On your site: Your facility, our capture team; We bring rigs and a capture team to your facility and work within your safety and access rules.
In our lab: Incheon, South Korea; Controlled, repeatable sessions at our research lab.
Environments: Eight environments; Food & hospitality, logistics, manufacturing, textile, retail, workshop, agriculture and home. See Coverage

Process

How a collection batch runs

01 · PROTOCOL

Agreed before capture

Tasks, variations, success criteria and episode counts, agreed before capture starts.
02 · CAPTURE

Recorded to protocol

Demonstrators record to the protocol. Scene, task, rig and consent are logged for every episode.
03 · REVIEW

Reviewed twice

Two independent reviewers check every episode. Rejections come back with reason codes.
04 · DELIVERY

Your format, your bucket

Synchronized files, metadata and QC verdicts, in your format, to your bucket.

Deliverables

What you receive

Consent on every file

Every demonstrator and every site gives written consent before capture. Each episode carries a consent reference and its provenance, so any file can be traced back to where and how it was recorded.