Physical-world data / FIUND
First-person video for understanding human actions.
First-person video that follows a person through real tasks, with hands, objects and the surrounding environment in view.
Training & evaluation · Scope confirmed for your project
Head-mounted recordings
Perspective → sequence → interpretation
SOURCE CONTEXT
The task, from the wearer’s perspective.
A useful sequence shows the action and the context around it: the tools, the objects, the pauses and the outcome.
Head-mounted recordings
Narrated everyday tasks
Hands interacting with tools
Embodied AI evaluation · Human-action understanding
TECHNICAL BRIEF
Define the capture and coverage.
These are decisions to agree for your collection, rather than specifications assumed across every source.
Camera position
Describe what needs to stay visible and from which perspective. State whether changing viewpoints, camera motion or multiple views are useful or unwanted.
Task boundaries
Describe the beginning, intermediate steps and outcome a useful example should contain. Include interruptions or failed attempts if they matter to your task.
Frame rate and audio
Set the source quality and file representation your system needs. Original recordings, exports and additional preparation are scoped separately.
BEFORE DELIVERY
A defined scope.
A considered handoff.
Evaluate the fit.
Agree what a useful sample must demonstrate, including the source conditions and required relationships.
Confirm the permissions.
Review the intended AI uses, relevant exclusions and documentation for the selected material.
Agree the package.
Specify files, metadata, preparation and acceptance criteria before proceeding with delivery.
START A CONVERSATION
Let’s scope your
data request.
Tell us what your model needs from egocentric video. We’ll assess the sourcing options and discuss a suitable next step.
Availability and collection terms are confirmed after review.
Frequently asked questions
Can I request continuous sequences rather than edited clips?
Specify the sequence length, camera perspective and surrounding context your task needs. We confirm whether suitable source recordings support that brief and identify any additional capture or preparation.
What should I include in a egocentric video brief?
Start with your model task and the source material you need. Include camera position, task boundaries, frame rate and audio so we can assess fit and propose a useful evaluation sample.
When are availability, pricing and permissions confirmed?
After we assess your brief and identify suitable material or a capture scope. Samples, permitted AI uses, preparation and delivery terms are agreed for the proposed collection before a purchase.