AI Edge-Vision Platform

Vision that knows every vehicle on the road.

Sentinel detects and tracks every vehicle, reads its number plate and tags its type, colour and make — all in real time on the edge. A cloud dashboard ties the fleet together, and a review-and-retrain loop keeps accuracy climbing.

Runs on the edge Indian ANPR Gets sharper as it runs
truck#118car#1422-wheeler#156Live 12 trackedKA 01 AB 123498%
Overview

One platform for every plate and every face

Sentinel reads number plates and faces at the edge, classifies the vehicle, and checks every read against your watch-lists in real time. Matches raise an instant alert and every event stays searchable for later. It also links to SoftLogix Radar Enforcement, tying recognition to plate-level enforcement end to end.

Edge vision nodes

AI cameras and on-site edge compute sit at every location, capturing each vehicle and face the moment it passes — no footage shipped off-site to be read.

Recognition engine

ANPR, 1:N facial recognition and vehicle-attribute classification all run on-device, so a plate or a face is resolved in real time without a cloud round-trip.

Command console

Every read rolls up to one console with watch-lists, rule-based alerts and a searchable history spanning all of your sites.

Key Features

Recognition, alerting and search in one stack

From on-device ANPR and facial recognition to watch-list alerts and a searchable, audit-ready history — everything intelligent vehicle and person monitoring needs.

Number-plate recognition

Real-time ANPR tuned for Indian plates — single and double-row, regional and commercial formats — reading reliably by day and at night with IR illumination.

Facial recognition

1:N face matching against enrolled watch-lists flags a person of interest the instant they appear in frame.

Vehicle attribute classification

Beyond the plate, each vehicle is classified by make, model, colour and type — car, truck, two-wheeler, auto or bus.

Watch-lists & rule-based alerts

Blacklist, BOLO, stolen and VIP lists drive rule-based alerts that fire the instant a plate or face matches.

Live operations dashboard

A real-time WebSocket event feed streams every read — plate, snapshot, site, time and direction of travel — as it happens.

Searchable history & audit trail

Every read is logged and queryable, with full snapshots and an export-ready trail for investigations and audits.

Multi-site, multi-camera

A distributed-edge architecture spans many camera sites and locations, all managed and monitored from a single console.

Direction & zone context

Direction-of-travel, lane and zone context, plus entry, exit and dwell tracking add meaning to every event — not just a raw plate.

Privacy & role-based access

Privacy controls, role-based access and audit-ready logging keep sensitive recognition data governed and accountable.

How it works

Capture. Recognise. Alert.

The same edge-first loop runs at every site — turning what each camera sees into a recognised plate, a matched face and an instant alert, with the console reserved for monitoring and search.

01

Capture

Edge cameras at every site capture each vehicle and face as it passes — full-frame context plus a tight crop of the plate and the person.

02

Recognise

On-device ANPR, 1:N face matching and attribute classification read the plate, match the face and classify the vehicle in real time — no cloud round-trip.

03

Alert

A match against your watch-lists raises an instant alert, while every read rolls up to the command console for live monitoring and later search.

Specifications

Built for recognition at the edge

ANPR and IR cameras paired with NVIDIA Jetson edge compute, on-device ANPR + AFR engines, and a role-based command console.

ANPR / IR cameras

Day-and-night capture cameras with IR illumination for clean plate reads after dark.

Edge appliance

NVIDIA Jetson edge appliance runs all recognition on-device, one node per site.

Plate OCR

Indian-format OCR — single and double-row plates, regional and commercial series.

Face-match engine

1:N facial recognition against enrolled watch-lists with configurable match thresholds.

Recognition latency

Sub-second on-device recognition — plate and face resolved without a cloud round-trip.

Connectivity

Ethernet with WiFi / 4G fallback; secure link from each site to the command console.

Data retention & storage

Configurable retention with local edge buffering and central archival of reads and snapshots.

Live dashboard

Real-time WebSocket event feed with role-based access for operators and supervisors.

Watch-list capacity

Large enrolled watch-lists — blacklist, BOLO, stolen and VIP — synced to every edge node.

Audit logging

Tamper-evident, audit-ready logging of every read, match and operator action.

Ready to see Sentinel in action?

See ANPR and facial recognition running on live sites — book a demo, or talk to our team about your locations.