Product · Computer vision · industrial security

Eagle AI Camera Surveillance

Eagle AI turns a site's existing IP cameras into an operations platform. The AI engine detects and tracks people and vehicles, names enrolled employees by face, reads license plates, and understands audio. Enterprise modules turn those sightings into automatic attendance, zone access enforcement, visitor control, gate and weighbridge compliance, and a tamper-evident audit trail — in a command center where guards work a triaged alert queue instead of video walls. Every event is processed on site; nothing leaves the plant.

Pilot-ready · fully on-premises · Built and operated by XISLABS

Deployment
100% on-premises (Docker)
Designed for
Hundreds of cameras · multiple zones
Alert taxonomy
17 alert types
Measured accuracy
Face 92–94% · LPR 92–97%
Placeholder: surveillance cameras
Command center — KPIs, triaged alert console, camera fleet, and zone occupancy
§ 01

Why it exists

The problem Eagle AI solves

The problem

Large industrial sites run hundreds or thousands of cameras that nobody can watch. Attendance, visitor logs, zone access, vehicle inspections, and weighbridge checks live in separate systems — or on paper — and security teams find out about incidents after the fact, if at all.

The product

Eagle AI unifies it in one codebase, one login, one database, and one audit trail. Built as a unified fork of the open-source Frigate NVR, it adds face recognition with 12-slot enrolment, license plate reading, semantic footage search, and a set of enterprise modules for workforce, visitors, zone policy, gates and vehicles, and audit — designed for a large cement plant with hundreds of cameras, a workforce in the thousands, and multiple security zones.

§ 02

Capabilities

What Eagle AI does

Every capability below ships in the product today — nothing on a roadmap slide.

Face recognition

12-slot enrolment per employee — seven face angles drive recognition, with body and top shots stored for future re-identification. Measured live at 92–94% confidence.

License plate recognition

Plate reads with a dedicated gate-camera mode; verified reads at 92–97%.

Semantic footage search

Find footage in plain language — “person in orange vest”.

17-type alert console

Critical, warning, and info alerts with severity triage, acknowledge/resolve, and storm-guard so the same subject in the same zone isn't re-raised while an open alert is fresh.

Automatic attendance

First sighting is check-in, last is check-out — with search, HR override, CSV export, and time analytics of hours in own, other, and unauthorised zones.

Zone access policy

A department × zone policy matrix with maximum-stay durations, live occupancy, overstays, and a violations feed.

Gates, vehicles & weighbridge

Inspection compliance (stop + trunk open), weighbridge entry/exit weight against approved dispatch with auto-hold, loading-bay counts, and badge-plus-face identity fusion.

Tamper-evident audit

A hash-chained event log with one-click integrity verification and per-audience CSV report packs — every export itself audited.

§ 03

Enterprise modules

From sightings to operations

The AI engine watches; these modules turn what it sees into attendance, access control, compliance, and evidence.

AI Dashboard

Overview KPIs, alert console, camera fleet wall, zone occupancy, people tracking with movement paths, and AI shift briefings.

Workforce

Employee directory with HRMS import and webhook, departments, face enrolment, attendance, and time analytics.

Visitors

Gate registration with host, temporary card, allowed zones, and face capture; a live on-site board with risk flags.

Zone Access

Zone directory with risk levels and camera mapping, policy matrix, overstays, and violations.

Gates & Vehicles

Inspection compliance, weighbridge, loading bays, metal-detector screening, and badge-reader access control.

Audit & Reports

Hash-chained tamper-evident log, searchable structured events, and report packs for management, security, HR, and audit.

§ 04

Inside the product

Eagle AI in use

The main surfaces of the product — what operators and end users actually work in.

Placeholder: a surveillance camera lens close-up
Workforce — 12-slot face enrolment and automatic attendance
Placeholder: a wall of monitoring dashboard screens
Gates & vehicles — inspection compliance, weighbridge variance, badge-plus-face fusion
Placeholder: an operator at multiple monitors
Audit & reports — hash-chained event log with one-click integrity verification
§ 05

How it works

From setup to daily use

The path from first day to routine operation with Eagle AI.

  1. 01

    Cameras stream in

    Existing IP cameras (RTSP) are decoded on site; a motion gate keeps AI detection efficient.

  2. 02

    The AI engine identifies

    Object detection and tracking, then enrichments: face recognition, plate reading, audio, and semantic embeddings.

  3. 03

    Policies turn sightings into events

    Zone rules, attendance, presence trails, and behaviour analytics — loitering, crowd density, gate bypass, badge missing.

  4. 04

    Guards work a triaged queue

    Alerts arrive with severity in a tab-based command center; every action lands in the audit chain.

Under the hood

  • Frigate NVR (MIT)
  • Python / FastAPI
  • React + TypeScript
  • Docker
  • MQTT
  • REST API (40+ endpoints)
  • Coral · OpenVINO · TensorRT · ROCm · Hailo · RKNN
§ 06

Who it's for

Where Eagle AI fits

Built for

  • Cement, steel, and process plants
  • Sites with gates, weighbridges, and loading bays
  • Security teams that need auditable evidence
  • Operators who cannot send footage off-site

Plant security operations

Unauthorised-area, after-hours, and perimeter-breach alerts across zones, with people tracking and movement paths.

Attendance without badges

Camera-based check-in and check-out for large workforces, with HR export and override.

Gate & dispatch compliance

Vehicle inspections, weighbridge variance holds, and loading-bay anomalies caught as they happen.

Auditable incident evidence

A tamper-evident trail and structured reports for management, security, HR, and audit.

§ 07

Answers

Eagle AI: common questions

Does any footage leave the site?

No. Every event is processed on-premises inside a Docker deployment on the plant's own hardware. Nothing is sent to a cloud service.

Does it work with our existing cameras?

Eagle AI ingests standard IP camera streams (RTSP). It supports hardware acceleration for Coral, OpenVINO, TensorRT, ROCm, Hailo, and RKNN.

How accurate is the recognition?

Measured on the development rig rather than claimed: face recognition at 92–94% confidence after 12-slot enrolment, and verified license plate reads at 92–97%. Enrolment photos should include frames from the deployment cameras themselves.

Can it connect to our HR and access-control systems?

Yes. The employee directory supports HRMS CSV import and a live webhook, and badge readers feed an access-control ingestion API that fuses badge scans with face identity to flag mismatches.

§ 10 / Contact

See Eagle AI for your business

Book a walkthrough with the team that built it — we'll map it to your workflow and show what it looks like on your data.

WhatsApp