Use cases
Face Matcher covers two deployment shapes. They share one server and one watchlist database and differ only in where the input comes from: live video from cameras, or single images sent to the API. A single installation can combine them, for example cameras at a checkpoint and an identification API used by a registration desk.
Real-time face identification on cameras
Face Matcher detects and identifies faces across many camera streams at once, with a delay measured in fractions of a second, so an operator or a downstream system can react while the person is still in front of the camera. Its detection handles faces covered by a face mask, low indoor light and infrared night-vision streams. Typical sites are airport security zones, border crossing points and other secured perimeters where the people passing through must be checked against a watchlist.
Inputs and outputs
The inputs are RTSP streams from IP cameras, processed on the server, and edge streams from smart cameras or AI boxes that run the Embedded Stream Processor and publish face crops and templates over MQTT instead of video. Every stream produces face detections with tracking, a passive liveness result, estimated age and gender, and a match or no-match result against the watchlists. See Cameras for the two source types.
Results reach your systems as real-time notifications over GraphQL subscriptions and RabbitMQ, and are stored for later search through the REST and GraphQL APIs and Station's event history. See Integration for the API surface.
Typical roles
Security integrators, providers of video management and security systems, and operators of airports, border sites and government premises.
Face matching API service
Face Matcher also works without any camera. A client sends a base64-encoded image to the REST API and receives an identification (1:N search against watchlists), a verification (1:1 comparison of two faces) or a liveness check; the same API creates, updates and deletes watchlists and their members. Station's Identify a face page uses the same calls from a browser, with an uploaded photo or the computer's webcam.
This is the mode for kiosks, registration desks, back-office deduplication and any application that already has the image and only needs the answer. The REST API can be restricted to this feature set (Watchlist mode) when the deployment should not process video at all; see REST API.
The same deployment runs on premises or in a private cloud, and the sizing is driven by request rate and watchlist size rather than by cameras; see Hardware requirements.