Research · Face recognition
Face Finder
Face recognition that groups every person across thousands of photos on a private computer, and recognises people live on CCTV for attendance and security.
- Client
- Product
- Type
- Face recognition
- Industry
- Consumer and industrial
The challenge
Finding a person should not mean uploading every photo to a cloud service, or watching hours of CCTV. We built recognition that runs where the images already are.
What we did
- 01InsightFace detection and 512-dimension embeddings stored in PostgreSQL with pgvector.
- 02HDBSCAN clustering that copes with different lighting, angles and expressions.
- 03The same engine runs on CCTV streams at the edge, with encrypted face data under your control.
What we delivered
- Desktop photo app
- Live recognition for CCTV
- Attendance and path reports
- Zone alerts
Where it runs
One technology, 3 places it matters.
- 01Consumer
Private photo library
Families keep years of photos across drives, and cloud “People” features mean uploading every one of them.
- Every face found and grouped by person, entirely on the user’s own computer
- Search by photo across the whole archive
- Merge, move and undo for groups
- 02Industrial sites
Real-time face recognition on CCTV
Registers and punch cards are slow and easy to cheat, and on a large site nobody can say where a person was last seen.
- Automatic attendance as people walk past existing CCTV cameras
- People path finder: where a person went, camera by camera, with times
- Runs at the edge, with encrypted face data under your control
- 03Industrial sites
Intrusion detection
Restricted areas are watched by people scanning many screens at once, so crossings get missed.
- Restricted zones drawn on any camera view
- An instant alert to the control room when someone enters a zone
- A clip and timestamp saved for every event
Have a project in mind? Let’s talk about it.
Tell us what you want to build. A senior team member replies within one working day.