Vision & video
Visual data captured by people, for models that must see.
Human-generated image and video data for visual understanding, grounding and multimodal systems — collected against explicit capture specifications.
What we collect
Signals, structured.
Images
Object, scene and interaction imagery captured to defined briefs.
Video
Continuous recordings of tasks, movement and events over time.
Environment data
Indoor and outdoor settings, lighting conditions and viewpoints.
Visual annotation
Boxes, segmentation, attributes, captions and temporal labels.
Potential applications
- Detection, segmentation and grounding
- Video understanding and world models
- Multimodal vision-language models
- Domain-specific evaluation datasets
Collection workflow
- 01Capture specification covering devices, angles, lighting and consent.
- 02Contributor briefing with reference examples.
- 03Collection through controlled capture workflows.
- 04Annotation passes with reviewer calibration.
- 05Structured dataset with metadata prepared for delivery.
Building an AI system that needs better data?
Tell us what your model needs to learn. We'll explore how a purpose-built human data program could support it.