A camera at a door
A camera at a door, and a ledger of who crossed it
Recognition on our own servers, consent recorded in the database, and every threshold a setting rather than a constant.
What it is
A camera pointed at a place, a list of people who have agreed to be recognised, and a record of who came in and who went out. For a building, a workshop, a site entrance — anywhere attendance is physical.
The recognition happens on our machines. No face is sent to a third-party service.
Consent is a column, not a policy document
Enrolment is recorded per person. Somebody who has not agreed is not enrolled, and an unrecognised face stays an unrecognised face.
What this does not do
The recognition thresholds have never been calibrated against real photographs of real people. They are a starting point taken from the model family’s published behaviour, and the settings page says so in as many words. Calibrating them needs real faces at a real door, and until somebody does that this is a feature to pilot rather than to rely on.
