Annotation
Temporal action segmentation (verb + object taxonomy), object and interaction labels, task/environment metadata, edge-case and recovery tagging. Delivered in your schema — JSON, COCO-style, or custom — with documentation.
The data
What lands in your pipeline.
Every program is built to spec, but this is our standard delivery baseline.
Capture specification
| Parameter | Baseline |
|---|---|
| Viewpoint | Head-mounted, first-person (hands-centered framing) |
| Resolution / rate | 1080p–4K, 30–60 fps, linear FOV |
| Session structure | Full-task continuity, first touch to final check |
| Environments | Lived-in EU households and real workplaces |
| Audio | Optional task narration track |
| Extended modalities | On request: synced depth, hand pose, IMU — scoped per program |
Beyond the pixels
Annotated, checked, and cleared for training.
Quality assurance
Two-stage review (capture QA before annotation, label QA before delivery), inter-annotator agreement tracking, rejection of sub-spec footage at our cost.
Rights & compliance
Full consent chain per session (worker + premises), GDPR-compliant processing, EU data residency, face/bystander anonymization where required. Licensing is flexible: exclusive or non-exclusive, per-program or per-dataset — priced accordingly.
Gallery
Annotated frames, as delivered.
[Media] Frame 01
Kitchen — wiping a surface, labels on.
[Media] Frame 02
Bathroom — scrubbing tiles, different home.
[Media] Frame 03
Laundry — folding, object labels.
[Media] Frame 04
Living room — tidying clutter.
Request the full spec sheet and a sample batch.
Up to 2 hours of annotated footage, captured to the spec we agree — free, within 48 hours.