Axix Visual Intelligence
- +Works with your existing Dahua, Hikvision, or any RTSP/HTTP camera
- +$0 additional camera hardware cost
- +CPU-first inference with OpenVINO — less GPU required
- +On-premise — your video never leaves your network
📹 Axix Technologies · AI Video Intelligence Platform
Your Cameras See
Everything.
Now Make Them Understand It.
Axix Visual Intelligence turns your IP camera network into a site safety and analytics layer — people and access analytics, cross-camera movement analytics, zone enforcement, footfall counting, and real-time guard alerts. All on standard CPUs. All on your own premises.
CPU-only inference — no GPU requiredOn-premise — your data never leaves your networkReal-time alerts — mobile app + web dashboardSupports 10+ simultaneous IP camerasOpenVINO optimized for Intel/AMD CPUsRSA-signed license enforcement per installation
8
AI modules for people analytics
5
Real-time alert types
10+
Cameras — no GPU
0
Cloud dependencies
Camera Feed · Live
On-prem AI · no GPU · no cloud
Identity matched · Employee ID verified
Zone: Authorized · Cross-camera link active
Footfall count today: 142
Zone: Authorized · Cross-camera link active
Footfall count today: 142
✓ No-go zone clear · Perimeter A
✓ Guard alert sent · Flutter app
✓ 8 cameras processing · CPU only
The Problem
Traditional CCTV Is Reactive.
You Only Watch Recordings After Something Goes Wrong.
You Only Watch Recordings After Something Goes Wrong.
Five problems that every security manager with a standard CCTV system faces — and that Axix Visual Intelligence eliminates.
Hours of footage, no actionable intelligence.
Standard CCTV systems record video and nothing else. Finding a specific person or incident means manually scrubbing through hours of footage. Axix Visual Intelligence gives you answers in real time — who is on premises, where they are, and whether they are authorized.
Unauthorized access goes undetected until it is too late.
Restricted areas are only as secure as the guard watching the monitor. When no one is watching, unauthorized persons enter unchallenged. Axix Visual Intelligence detects and alerts on intrusions and zone violations the moment they happen — automatically.
Following movement across multiple cameras is slow and manual.
When an incident occurs, piecing together a person's path across five cameras takes hours and multiple people. Axix Visual Intelligence links movement across your entire camera network in real time using AI — connecting appearances even when faces are not clearly visible.
You have no idea how many unique visitors enter your facility daily.
Raw motion detection and headcount sensors count every detection, including duplicates. Axix Visual Intelligence uses ReID-based identity matching to count only unique individuals — giving you accurate daily footfall, hourly trends, and occupancy insights.
Guards are stuck at a desk to monitor alerts.
Security events happen in the field. Guards need alerts on their phones, not just on a control room monitor. Axix Visual Intelligence includes a Flutter mobile app that delivers real-time alerts directly to guard devices — scoped to their assigned area and alert types.
Axix Visual Intelligence solves all five.
Real-time intelligence. Automated enforcement. Mobile alerts. Accurate analytics. All on-premise.
See it live →How It Works
From Camera Stream to
Actionable Intelligence in Milliseconds.
Actionable Intelligence in Milliseconds.
Four stages — connection, AI processing, alert delivery, and audit. Every stage runs on your premises, on your hardware.
1
Connect Your IP Cameras
Add any IP camera (RTSP, HTTP, Dahua, Hikvision, or others) from the web dashboard. Test the connection, configure per-camera AI settings, assign it to a zone, and go live. The system starts processing frames immediately.
2
AI Processes Every Frame in Real Time
Each feed goes through: YOLO person detection → ByteTrack tracking → ReID appearance matching → face recognition against enrolled identities → zone evaluation → alert generation. All running on your CPU via OpenVINO-optimized models.
3
Alerts Reach Your Team Instantly
Zone violations, blocked person detections, intrusions, and camera health failures are delivered via Socket.IO to the web dashboard and to guard devices via the Flutter mobile app — within seconds of detection, with images and timestamps.
4
Review, Audit, and Report
Every detection, check-in, violation, and alert is stored with images and timestamps. Search by person, camera, date, or alert type. Export compliance reports as CSV or PDF. Review the full cross-camera timeline for any individual.
Platform Capabilities
Every Capability Your Security
Operation Needs — in One Platform.
Operation Needs — in One Platform.
15 built-in modules covering identity, access control, analytics, operations, and mobile delivery — nothing to integrate, nothing to build.
Cluster A — Face Intelligence
Face Enrollment with Quality Gate
Enrollment validates image quality before indexing so weak or blurry profiles never pollute recognition results. This improves confidence and reduces false matches.
FAISS Similarity Search
Fast nearest-neighbor matching over face embeddings provides production-grade identity retrieval with tuned thresholds for real-world camera conditions.
Allow, Restrict, and Alert Lists
Each person can be categorized by operational intent. The system applies policy-specific alert logic for authorized staff, unknown visitors, and restricted individuals.
Spoof Detection Ready
Anti-spoof hooks are built into the pipeline for future activation. Deployment teams can enable additional liveness logic per site policy when needed.
Cluster B — Spatial Alerting & Policy Enforcement
No-Go Area Intrusion Detection
Absolute restricted areas trigger immediate alarms for any detected person. Designed for server rooms, perimeter fences, substations, and high-risk control zones.
Access-Controlled Zone Enforcement
Zones support role-based or person-based authorization. Only unauthorized entries generate violations while allowed personnel movement stays silent.
Alert Prioritization & Escalation
Intelligent scoring ranks each event using event severity, identity confidence, and zone criticality so SOC teams focus on the incidents that require immediate action.
Evidence Snapshot Capture
Each alert includes pre/post evidence frames and metadata. Teams can investigate without scrubbing full footage, reducing response and audit overhead.
Cluster C — Person ReID Movement Engine
Cross-Camera ReID Tracking
OSNet embeddings maintain person continuity across non-overlapping cameras, even when faces are occluded, masked, or facing away.
Color + Appearance Fusion
The tracker combines clothing-color histograms, ReID vectors, and optional face identity to improve matching robustness in crowded scenes.
Movement Timeline Reconstruction
Operators get complete path visualization by timestamp and camera sequence, making incident forensics and audit reporting significantly faster.
M1
Detector
YOLOv8 person/face candidate generation.
M2
Tracker
Frame-to-frame identity continuity.
M3
ReID Encoder
OSNet embeddings for re-identification.
M4
Color Profiler
HSV clothing signature extraction.
M5
Face Matcher
FAISS-based known identity lookup.
M6
Fusion Engine
Weighted scoring and confidence merge.
M7
Path Graph
Cross-camera movement history map.
M8
Alert Router
Rule application and delivery routing.
Event Flows
Operational Event Chains
Every alert type is routed through deterministic flow chains so teams can understand exactly how the system classified, scored, and escalated each event.
Flow 1 — Alert List Face Match
Camera Frame → Face Embedding → FAISS Match → Alert List Hit → Mobile Alert
Flow 2 — Unauthorized Zone Entry
Person Detected → Zone Check → Authorization Check → Violation Event → Supervisor Escalation
Flow 3 — Cross-Camera Movement Linking
Unknown Person → ReID Vector → Camera Transition → Path Confidence → Live Path View
Flow 4 — Evidence & Audit Trail
Event Trigger → Snapshot Capture → Metadata Index → Dashboard Timeline → Post-Incident Report
AI Model Stack
Production Models by Function
Model selection prioritizes CPU inference speed, operational reliability, and deployment simplicity for on-prem industrial workloads.
| Function | Model | Purpose |
|---|---|---|
| Person Detection | YOLOv8n | Fast CPU-optimized person detection on edge frames. |
| Face Detection | YuNet / OpenCV DNN | Low-latency face region extraction for enrollment and live events. |
| Face Embedding | ArcFace-Compatible Encoder | Dense identity vectors for FAISS similarity matching. |
| Person ReID | OSNet-x0.5 | Appearance embedding for cross-camera tracking beyond face visibility. |
| Similarity Search | FAISS Index | Sub-second nearest-neighbor retrieval over identity vectors. |
| Inference Runtime | OpenVINO | Intel/AMD CPU acceleration without dedicated GPU hardware. |
No GPU required. Entire inference path runs on optimized CPU infrastructure.
Architecture
Composable, On-Prem Ready Stack
The architecture is modular, auditable, and designed for secure deployments across single-site, multi-site, and isolated network environments.
Camera Ingestion
RTSP/HTTP stream ingest with per-channel buffering and reconnection handling
AI Runtime
Python inference workers + OpenVINO optimized models
API Layer
FastAPI services for event, identity, and alert orchestration
Messaging
Redis/Kafka compatible event bus for scalable deployments
Data Layer
PostgreSQL for metadata + Redis for low-latency caching
Dashboard
Next.js control interface for operators and supervisors
Mobile App
Flutter app for field guards with scoped alert feeds
Security
RSA-signed licensing and role-based access control
Deployment Patterns
Single-Site Edge Node
docker compose up -d
Ideal for one facility. Camera ingestion, AI workers, API, and dashboard run on one on-prem server.
Multi-Plant Cluster
k8s + event bus scale
Regional sites push events into a central cluster while retaining local camera and inference privacy controls.
Air-Gapped Deployment
offline package install
For regulated environments requiring isolated networks. No cloud dependencies for daily video analytics operations.
System Requirements
Hardware Profiles
Choose a baseline profile based on camera count, retention policy, and response-time requirements.
Minimum Deployment
CPU
Intel i5 / Ryzen 5 (4+ cores)
RAM
16 GB
Storage
256 GB SSD
Camera Capacity
Up to 5 streams (15-20 FPS)
Recommended Deployment
CPU
Intel i7/i9 or Ryzen 7/9 (8+ cores)
RAM
32 GB
Storage
500 GB+ NVMe SSD
Camera Capacity
10+ streams (25-30 FPS)
Use Cases
Built for Real Operational Scenarios
From perimeter defense to internal compliance zones, Axix Visual Intelligence supports high-stakes environments where speed and reliability are critical.
Industrial Perimeter Security
Detect unauthorized entries around plant boundaries and alert guards before physical escalation.
Corporate Campus Access Monitoring
Monitor known and unknown visitors across lobbies, gates, and restricted office zones.
Control Room & No-Go Enforcement
Apply absolute exclusion to high-risk rooms where any human presence is a critical event.
Warehouse Theft Prevention
Follow suspicious movement between loading, storage, and dispatch zones with rapid evidence retrieval.
Oil & Gas Facility Monitoring
Support remote-site safety with on-prem AI analytics and low-bandwidth alert routing.
Hospital & Campus Safety
Monitor sensitive zones, improve response readiness, and maintain searchable incident history.
Pricing
Cloud SaaS Subscription Pricing.
Deploy on Your Infrastructure.
Deploy on Your Infrastructure.
Monthly subscription per site. Per-camera billing. Cancel anytime.
Use the IP cameras you already own. No proprietary hardware, no forced camera replacement — save $500–$3,000+ per camera vs. platforms that require you to buy their cameras.
Free Trial
7 days · Full platform · No credit card
$0 USD/month
7 days · Up to 3 cameras
- ✓Up to 3 IP cameras
- ✓Face recognition & zone enforcement
- ✓No-go area intrusion detection
- ✓Web dashboard
- ✓Email support
Standard
Per camera · Face recognition + 1 zone
$69 USD/camera/month
e.g. 5 cameras = $345 USD/month
per month, billed monthly
7-day free trial included
Cancel anytime
- ✓Face recognition & attendance
- ✓1 access zone per site
- ✓No-go area intrusion detection
- ✓Blocked person alerts
- ✓Web dashboard
- ✓Flutter mobile app for guards
- ✓Email support
Cancel anytime · Cancellation policy
VisaMastercardAmexPayPal
Most Popular
Premium
Per camera · Full Axix Visual Intelligence suite
$99 USD/camera/month
e.g. 5 cameras = $495 USD/month
per month, billed monthly
7-day free trial included
Cancel anytime
- ✓Everything in Standard
- ✓Zone management & access control
- ✓8-module cross-camera movement analytics
- ✓Footfall analytics & occupancy trends
- ✓Kafka-based scalable processing
- ✓Zone Violation Gallery
- ✓Audit reports (CSV + PDF)
- ✓Priority support
Cancel anytime · Cancellation policy
VisaMastercardAmexPayPal
You Already Own the Hardware. Why Pay for More?
Verkada and Rhombus require their proprietary cameras before their AI platform works at all. Axix Visual Intelligence runs on the IP cameras you already have installed.
Proprietary Camera Platforms (Verkada, Rhombus)
- –Requires purchasing their specific camera hardware — existing cameras cannot be used
- –$200–$3,000+ per camera in mandatory hardware cost before the platform works
- –Cloud-tethered — video and metadata processed off-premises
- –Camera + license bundled — canceling the license can disable the camera
A 10-camera facility switching to Verkada: ~$6,000–$12,000 in mandatory new camera hardware before any software cost. The same facility on Axix Visual Intelligence: $0 hardware — just connect your existing cameras.
How the Numbers Actually Compare
Software licensing and all 15 built-in modules side-by-side — Premium saves ~50% vs typical Rhombus Enterprise license (~$199/camera/mo).
| Metric | Axix Visual Intelligence (Premium) | Verkada | Rhombus (Enterprise/AI tier) |
|---|---|---|---|
| Software/license price | $99/camera/mo (Premium) · $69 Standard | ~$150/camera/mo (license only) | ~$199/camera/mo (Enterprise) |
| Typical software savings vs Rhombus (~$199/cam) | ~50% lower on Premium ($99 vs ~$199) | ~34% lower vs ~$150/cam (Premium) | Baseline reference |
| Mandatory camera hardware | None — use existing cameras | $500–$3,000+/camera, required | $200–$1,200+/camera, required |
| GPU required | Less GPU required (CPU/OpenVINO-first) | N/A (proprietary hardware handles it) | N/A (proprietary hardware handles it) |
| Data stays on-premise | Yes | No | No |
| Deployment model | Your infrastructure | Verkada camera + cloud | Rhombus camera + cloud |
| Face recognition & 8-module movement analytics (suite) | All 8 modules included (Premium) | Add-on / higher tier | Included in Enterprise |
| M1 — Person/face detector (YOLOv8) | Yes | Partial / tiered | Included in Enterprise |
| M2 — Frame-to-frame tracker | Yes | Add-on / higher tier | Included in Enterprise |
| M3 — ReID encoder (OSNet) | Yes | Add-on / higher tier | Partial / tiered |
| M4 — Clothing color profiler | Yes | No | No |
| M5 — Face matcher (FAISS) | Yes | Partial / tiered | Included in Enterprise |
| M6 — Fusion engine (ReID + color + face) | Yes | No | Partial / tiered |
| M7 — Cross-camera path graph | Yes | Add-on / higher tier | Partial / tiered |
| M8 — Alert router & escalation | Yes | Partial / tiered | Included in Enterprise |
| Face enrollment with quality gate | Yes | Partial / tiered | Included in Enterprise |
| Allow / restrict / alert lists | Yes | Partial / tiered | Included in Enterprise |
| No-go zone intrusion detection | Yes | Add-on / higher tier | Partial / tiered |
| Access-controlled zone enforcement | Yes | Add-on / higher tier | Partial / tiered |
| Alert prioritization & escalation | Yes | Partial / tiered | Included in Enterprise |
| Evidence snapshot capture | Yes | Yes | Included in Enterprise |
| Cross-camera movement timeline | Yes | Add-on / higher tier | Partial / tiered |
*Competitor pricing based on publicly available vendor pricing pages and third-party review platforms (G2, Vendr, TrustRadius) as of July 2026. Verkada and Rhombus pricing is quote-based and varies by camera model, volume, and contract length; figures reflect typical published ranges.
FAQ
Common Questions.
Everything teams ask before deploying Axix Visual Intelligence in production environments.
No. Axix Visual Intelligence is fully on-premise. All AI processing, storage, and alert delivery happen on your local network. The only outbound connection is optional license validation. Camera streams, recognition data, and alert images never leave your network.
No. The system is specifically designed for CPU-only inference using Intel OpenVINO optimization. It runs on standard Intel i5/i7 or AMD Ryzen 5/7 hardware. No NVIDIA GPU, no cloud AI API, no additional hardware cost beyond your existing server.
On minimum hardware (4-core CPU, 16 GB RAM), approximately 5 cameras at 15–20 FPS. On recommended hardware (8+ core CPU, 32 GB RAM), 10+ cameras at 25–30 FPS. Kafka-based deployment scales horizontally by adding more processing nodes for larger installations.
Any camera that provides an RTSP or HTTP stream is supported. Tested and optimized configurations exist for Dahua and Hikvision. Any other brand with standard RTSP output works as well — the system is brand-agnostic at the stream level.
The system uses a FAISS-based similarity search with a match threshold tuned for production environments (~0.32 cosine distance). Accuracy depends on enrollment photo quality, lighting conditions, and camera angle. The system includes a photo quality check during enrollment. Anti-spoofing and ANPR are available as configuration flags for future activation.
Axix Visual Intelligence processes identity matching data entirely on your own premises — no biometric data is transmitted to Axix or any third party. However, biometric data collection is regulated in some jurisdictions (e.g., Illinois BIPA in the US, GDPR in the EU), and you are responsible for obtaining appropriate employee/visitor consent and posting required notices at your facility. We recommend consulting your legal counsel to confirm compliance requirements for your specific location before enrolling individuals.
Yes. The Flutter mobile app runs on Android and iOS. Guards download the app, configure the server IP on first launch, and log in with credentials created by the admin. Alerts are scoped to each guard's assigned area and alert types — guards only see what is relevant to their post.
No-go areas are absolute exclusion zones — no one is ever allowed, ever. Any person detected triggers an immediate alarm. Zones are access-controlled areas with an allow list — specific employees, visitors, or card colors are allowed. Only unauthorized persons trigger a zone violation. Both systems can run simultaneously on the same camera.
Yes. The 8-module movement analytics system uses ReID appearance embeddings (body shape, clothing color, gait) — not just face recognition. OSNet-x0.5 generates 512-dimensional appearance vectors that are effective even when the face is not visible. The system fuses ReID (60% weight), clothing color (30%), and face identity (10%) for robust cross-camera matching.
All records are stored in PostgreSQL on your local machine. Storage duration is limited only by your disk capacity. The system includes auto-cleanup scripts. For 10 cameras at 25 FPS with violation and tracking images, plan for 500 GB+ SSD and review image cleanup policies quarterly.
Axix Visual Intelligence optionally integrates with 360 FaceVision HR for payroll and attendance workflows. For other integrations, the system exposes REST APIs and webhook callbacks for all alert types, check-in events, zone violations, and AI service events — allowing connection to any external system.
Ready to Make Your
Cameras Intelligent?
See Axix Visual Intelligence in action on your own camera feed. Request a demo and we will walk you through a live deployment on your hardware — no cloud, no GPU, no obligation.
📹
Visual Intelligence Assistant
● Online · Axix Technologies
Hi! I'm the Axix Visual Intelligence assistant. What can I help you with?