Egocentric data

Video from the worker's point of view

The short answer

Egocentric data is video shot from where your eyes are. The phone sits on your forehead, so the video shows your hands, your rag and the sink in front of you, which a camera in the corner of the room can't.

A cleaner wiping a restroom sink, seen from the corner of the room. His hands are small and partly blocked. — Left: a camera in the corner. Right: a phone on the cleaner's forehead.

The same kind of sink seen from the cleaner's forehead: both hands, the rag and the spray bottle fill the frame. — Left: a camera in the corner. Right: a phone on the cleaner's forehead.

A cleaner wiping a restroom counter with a phone held just above her eyes by a head strap, tilted down. — The phone sits just above the eyes, tilted down at the hands.

Outside viewA camera across the room. The hands are tiny and often blocked.

Headcam viewA phone on the forehead. Hands, tools and the surface fill the frame.

Two cameras, one restroom

Picture two cameras on the same restroom job: one up in the corner by the door, and the other on your cleaner's forehead, angled down.

The corner camera sees a person bent over a sink, while the headcam sees the spray bottle, the rag folding in the palm, and the ring of soap scum around the drain. That close view is what "egocentric" means, and the word itself just means "centered on me."

Why the hands matter so much

A machine that wants to clean has to learn what hands do, like how hard you press on a mirror or where your fingers go when you pull a full trash liner out of a can.

From the corner, the body blocks most of that, but from the forehead the full palm and all five fingers stay in the shot. That is why we ask your crew to keep their hands in view when they grab things.

People are already collecting it

Robot and AI teams now gather this kind of video on purpose. In 2025, Figure, a humanoid robot company, reported that its robot learned to find its way through cluttered homes using only first-person video of people, with no robot demos used for that skill.

Ropedia, a research group, has posted a set called Xperience-10M that holds 10,000 hours of first-person video, along with depth, hand motion and written labels. Both are examples of the field, and neither one is a Colby customer.

Why this matters to your crew

Your crew already does the work these teams want to see. When a cleaner wipes down a real restroom with the phone on their forehead and both hands in view, that hour shows a machine something a corner camera can't. Clear hands are also a big part of what gets an hour accepted.

What other owners ask

Does egocentric video show the cleaner's face?

The phone points down and out from the forehead, so it shows hands, tools and the surface. It does not point back at the person wearing it.

Why not just use the building's security cameras?

Those sit high and far away, so they show where a person walked but miss what the hands did with the rag and the spray bottle.

Where to read more

These are public examples from other companies. They are not Colby partners or customers.