Face recognition, new-person detection and cross-camera movement history, running on a PC in your own network. No cloud, no images leaving the building.
ZoneX-Vision reads the video streams of ordinary surveillance cameras — IP, or analog behind an encoder — recognises people against your own database of faces and identities, flags anyone it has never seen before, and follows each person from camera to camera.
Ask for anyone's day and you get a timeline: which cameras, which zones, from when to when. The server runs on one Windows PC and is used from any browser on the LAN.
KEY FEATURES
All cameras in a grid or one at a time with colour-coded labels: known by face, probable by clothing, new, watchlist. An activity badge shows the AI is really working.
Every identity with photos, category (employee, visitor, watchlist, banned, unknown), first/last seen and reference faces that improve with each clear sighting.
A clear, unknown face seen for a few seconds becomes a new identity and raises an alert. Name it, classify it or merge it later.
By camera or by zone, zoomable, with time spent in each place and day-by-day history. Wrong attributions are fixed with "Not him".
When the face is hidden, the person is still followed by body tracking and, optionally, by a same-day clothing signature.
New person, watchlist or banned person seen, possible duplicate identity, camera offline, AI engine problem.
RTSP / ONVIF streams, connection test, zones, and per-camera recognition settings for dense crowds, soft or distant images.
CPU, memory, GPU load and temperature, per-camera frame rate and AI state, stream health and a filterable live log.
Admin, operator and technician roles; technicians see positions only, never names. Configurable retention and one-click deletion of a person.
HOW IT WORKS
REQUIREMENTS & DELIVERY
| Server | Windows 10/11 64-bit, 16 GB RAM, SSD, wired LAN |
|---|---|
| GPU | NVIDIA GeForce RTX / Quadro (6 GB+) recommended; CPU-only for 1–2 cameras |
| Capacity | 8–16 cameras per server (RTX 3060 class, 1080p) |
| Cameras | RTSP / ONVIF, 1080p recommended, main stream ≥ 4 Mbit/s |
| Clients | Any modern browser on the LAN — PC, tablet or phone |
| Delivery | Installation, camera onboarding and tuning, staff enrolment, operator training and a maintenance plan. Typically one day for up to 16 cameras. |
Facial recognition is subject to local data-protection law. The operator is responsible for informing the people concerned and for the lawful use of the system.
FREQUENTLY ASKED QUESTIONS
No. ZoneX-Vision runs on a PC in your local network, with a local database and no cloud service or subscription.
Yes. It reads RTSP / ONVIF streams from IP cameras and analog cameras behind an encoder.
It depends on local data-protection law. The operator must inform the people concerned, have a lawful basis and set the retention period configured in the platform.