Face Matcher
Face Matcher is a real-time face identification server. It takes video from IP cameras and from smart cameras with on-device processing, detects and tracks faces, extracts biometric templates and matches them against watchlists you manage. Every detection and match is published as it happens over REST, GraphQL (including subscriptions), RabbitMQ and S3, so your own systems can act on the result within a fraction of a second.
Face Matcher is the base module of the Smart Corridors & e-Gates solution, which builds the corridor and e-gate logic on top of it. It is also delivered on its own for projects that need face identification without the corridor layer, for example an identification feed for an existing security system or an identification API behind your own application.
What you get
The package is a set of Docker images and a Docker Compose deployment for Linux, published in the face-matcher repository. It contains:
| Part | What it does |
|---|---|
| Engine services | Camera processing, face detection, template extraction, watchlist matching, passive liveness, face search over stored detections, pedestrian and object detection, edge stream ingestion, Leader and Follower watchlist synchronization. |
| Station | The web UI: watchlists, cameras, live camera previews, 1:N identification from an uploaded image and the event history. |
| Dependencies | PostgreSQL, RabbitMQ and SeaweedFS (S3-compatible storage), started by the same scripts. |
| Scripts and configuration | start.sh, stop.sh, factory-reset.sh, .env, .env.station, branding files and the license folder. See The release package. |
The outputs are events: a face was detected, a face matched a watchlist member (with score, liveness result and, if enabled, the face template), or no match was found. Face crops and full frames are stored in S3 and referenced from the events, and the event history is searchable in Station and through the APIs.
Scope
Face Matcher is delivered as-is with a fixed feature set. There is no custom feature work or one-off extension; integrations are built on the documented APIs and configuration. Some capabilities of the earlier video processing platform, such as offline video processing, access-control logic and non-face biometrics, are not part of Face Matcher and are not described in this documentation.
Who this documentation is for
- Decision makers and product owners evaluating whether Face Matcher fits a site: start with the overview.
- Engineers installing a first instance and running a first identification: getting started.
- Integrators consuming events and managing watchlists from their own software: platform and integration.
- Operators working with cameras, watchlists and events day to day: Station.
How this documentation is organized
- Overview: use cases, accuracy and thresholds, data and privacy. The shared biometrics glossary explains the terms.
- Getting started: hardware, license, installation, release package, first identification.
- [Guides](/docs/border-control/smart-biometric-corridors/face-matcher- Guides: camera placement, scaling, GPU, multi-server and multi-site setups, backup, monitoring, HTTPS, authentication, upgrades.
- Integration: REST and GraphQL APIs, RabbitMQ notifications, S3 storage, enrollment and identification samples.
- Manuals: the complete reference for the Face Matcher Server (cameras, processing pipeline, settings, liveness, watchlists, face search, events, topologies, ports, data retention), Station and the Stream Processor for smart cameras and AI boxes.