Live AI detection on existing cameras

Real-time AI on the cameras
you already own.

Real-time detection on the IP cameras you already own. Edge-processed, ERP-connected, live in twelve weeks.

Start the pilot → See it in action ↓
aqivon.live / node-01 / 16 cameras
142 ms avg / v 2.4.1 / on-prem
PERSON 0.97 PERSON 0.92
CAM 03 · LOBBY LIVE
2 detections · 1 tracked
RESTRICTED PERSON 0.95
CAM 07 · DOCK ALERT
Restricted-zone breach
ID 4012 · 0.99
CAM 11 · ENTRY LIVE
Recognized · ID 4012
FORKLIFT 0.94 PERSON 0.91 PALLET 0.88
CAM 14 · FLOOR LIVE
3 objects · 0 flagged
~150ms
Detection latency, edge to dashboard
90%
Reduction in manual monitoring effort
Hours → 15min
Incident detection time
12wks
From contract to live deployment
Live with a multi-site electronics manufacturer across 18+ countries. Delivered in twelve weeks.
Customer reference available on request.
Why now

Your cameras are running. Your operations are still reacting to what someone happens to see.

Most facilities monitor what cameras show. They don't act on what cameras see. The gap costs more than people think.

Manual zone monitoring
Safety incidents surface in the next morning's report.
Manual gate counts
Capacity decisions made on guesses.
Manual after-hours patrols
Breaches detected at shift change, not in real time.
Manual audit prep
10 to 15 hours of someone's quarter, every quarter.
Common deployments

Four solutions most facilities start with.

Most facilities start with one of four. Each combines a few of the capabilities below and integrates into the system of record your team already uses.

Attendance & access control

Contactless attendance through facial recognition as employees enter or exit, validated against geo-fenced camera zones so check-ins only register where they should. Anti-spoofing prevents buddy punching. Events sync to NetSuite, SAP, or your HRMS for shift, leave, and overtime calculation.

Capabilities used
Facial recognition · Restricted-zone monitoring · ERP & HRMS integration · Audit logging
Where it fits
Multi-site employers, shift-based operations, regulated workplaces that need an audit trail on who was where when.
SOP & safety compliance

Continuous monitoring of standard operating procedures across the floor. The system flags missing PPE, unsafe machine handling, restricted-zone entries while equipment is active, and skipped process steps. Per-operator compliance metrics surface in the dashboard. Events log into your QMS or SAP for audit and corrective action.

Capabilities used
PPE compliance · Pose & movement · Restricted-zone monitoring · Facial recognition for operator ID
Where it fits
Regulated manufacturing, processing plants, food production, anywhere ISO or safety audits drive operations.
Quality inspection on the line

Pass and fail flagging on production lines. Flagged items route to operator review. The system tracks variance across runs so process drift surfaces before a batch is lost. Quality events log to SAP quality records or NetSuite.

Capabilities used
Object detection · Counting · Quality & defect signaling · ERP integration
Where it fits
Manufacturing lines, packing operations, electronics assembly, any production environment where defect cost compounds.
Security & intrusion operations

Real-time detection of after-hours entry, tailgating, loitering, restricted-area access, and unauthorized movement near high-value assets. Alerts route to your security team in seconds. Events auto-create records in NetSuite security or SAP plant maintenance.

Capabilities used
Person detection · Restricted-zone monitoring · Facial recognition · Real-time alerts · ERP integration
Where it fits
Multi-site facilities, after-hours operations, restricted research or production zones, high-value asset storage.
Capabilities

Eight detection models. Tuned per facility.

The building blocks behind every deployment above. Most facilities enable four to six. Configuration is per camera-zone, not per facility. The lobby and the dock typically need very different rule sets.

PERSON DETECTED PERSON 0.96 PERSON 0.94 2 tracked · 0.95 avg

Person detection & tracking

Count occupants, track movement across cameras, re-identify across non-overlapping fields of view.

ZONE BREACH RESTRICTED ZONE PERSON 0.93 Polygon · 04:22:08

Restricted-zone monitoring

Polygon geofencing per camera. Auto-alert on entry. Auto-ticket creation in ERPs.

OBJECTS 3 FOUND FORKLIFT 0.94 PERSON 0.91 PALLET 0.88 forklift · person · pallet

Object & vehicle detection

Forklifts, pallets, packages, vehicles. Labeled, counted, tracked across zones.

POSE FLAGGED PERSON-DOWN 0.89 Stillness · 12s

Pose, movement & fall detection

Skeleton tracking for fall detection, ergonomics, and behavioral flags. Tuned for industrial environments.

IDENTITY MATCHED ID 4012 · 0.99 Audit-logged · 30d retention

Facial recognition

Employee or visitor ID matching. Configurable retention. Disabled by default where regulated.

SAFETY 2 FLAGGED NO HELMET 0.91 NO VEST 0.87 PPE policy · Zone B

PPE & safety compliance

Hard-hat, hi-vis, gloves. Per-zone safety rules with configurable enforcement.

QUALITY FAIL PASS 0.96 PASS 0.94 FAIL 0.91 Variance · batch 4421

Quality & defect signaling

Pass and fail flagging on production lines. Run-level variance detection. Routes flagged items to operator review.

SPATIAL · 7 DAYS Floor 2 · heatmap Dwell + floor-plan view

Spatial & heatmap analytics

Heatmaps of dwell time and movement, plus floor-plan visualization so security and ops teams see events in spatial context.

Privacy posture

Facial recognition, configured by policy.

The most-regulated CV capability we ship. Disabled by default in jurisdictions where it's restricted (Illinois BIPA, Texas, Washington). Templates are deletable on request. Every face viewed is audit-logged with role and timestamp. Available on request. Never on by default.

RETENTION
30d default · 0 supported
DATA LOCATION
Your network, never our cloud
ACCESS
Role-based · audit-logged
OPT-IN
Per facility, per zone
Architecture

Software on the hardware you already own.

No new cameras. No cloud subscription. No minimum camera count, from a single feed to sixty-plus. A single 1U edge device in your existing rack, an afternoon of network configuration, and a quarter to roll out across the facility. Video never leaves the premises.

01 / SOURCE

Existing IP cameras

RTSP, ONVIF, PoE+, H.264/H.265. If your cameras output any of these, they speak Aqivon.

02 / EDGE

1U edge device

Scales from a single camera to 64+ per node. Add more nodes for multi-site deployments. Edge inference runs on NVIDIA Jetson-class GPU hardware, with lighter silicon available for low-camera-count sites.

03 / DETECT

Real-time inference

~150ms detection latency. Eight detection models. Per-zone configuration via dashboard.

04 / ROUTE

ERP & webhooks

Detections stream to NetSuite, SAP, or any system that accepts a JSON POST.

CAMERAS CAM 03 CAM 07 CAM 11 CAM 14 EDGE NODE Aqivon Edge on-prem · 1U ~150ms inference DESTINATIONS Oracle NetSuite SAP Webhooks (REST) VIDEO NEVER LEAVES YOUR NETWORK
What you get

What's in the box at week twelve.

A pilot proves the product works on your cameras. A deployment puts the system into the operations workflow. Here's what that looks like at the end of twelve weeks.

  1. An Aqivon edge node racked, configured, and tuned for your camera count.
  2. Eight detection models active, with per-camera-zone rules your team can change.
  3. Native NetSuite or SAP connector, or a webhook into whatever you run.
  4. Dashboard with role-based access, audit logging, and configurable retention.
  5. Documented zone policies and event-to-record mappings, version-controlled.
  6. Quarterly model tuning included for the first year. Drift handled before it costs us.
How we're different

Three things other vendors don't ship.

Unlike vendors that require new cameras, Aqivon runs on the IP cameras you already own.

Unlike cloud CV platforms, video never leaves your network.

Unlike generic CV, we ship with native NetSuite and SAP connectors. Detections become records in your ERP, not dashboards no one checks.

Integrations

Detections flow into the systems you already run.

A camera event is only useful if it lands somewhere a human or system will act on it. Aqivon ships with native connectors for the two ERPs most customers run; everything else via webhook.

Live in NetSuite

Every event becomes a row a human can act on.

Quarterly audit prep that used to take 10 to 15 hours now takes about an hour.

  • Auto-ticket creation for restricted-zone breaches
  • Audit log sync for compliance review
  • Configurable event-to-record mappings per facility
Security Records
Audit Log
Occupancy
14:22:08 Restricted-zone breach · CAM 07 · DOCK ALERT
14:21:30 After-hours entry · CAM 03 · LOBBY LOGGED
11:08:42 Identity match · ID 4012 · CAM 11 LOGGED
09:14:55 Occupancy threshold · Floor 2 · 24/25 NOTED
06:42:11 Person detected · CAM 03 · LOBBY LOGGED

Oracle NetSuite

● Native connector · GA
  • Auto-record creation for security events
  • Configurable detection→record mapping
  • Audit log sync for quarterly review

SAP

● Native connector · GA
  • Auto-ticket on restricted-zone breach
  • Occupancy data → operational dashboards
  • Custom event-to-ticket workflows

Webhooks & REST

● JSON POST · Everything else
  • Salesforce, Slack, PagerDuty, your homegrown ops platform
  • Every detection published as a webhook
  • If it can receive JSON, it can receive Aqivon
Pilot offer

Two cameras. Two weeks. No commitment.

There's no minimum camera count to start. The pilot uses two of your existing IP cameras pointed at the two zones that matter most in your operation, running the detection models we'd use in production. At the end of two weeks we share what we caught, what we missed, and what a full rollout would look like.

If the pilot doesn't catch what we said it would, we walk away. No invoice, no equipment to return, no awkward email chain.
  • 01

    Share your camera count and where they're pointed

    A rough number is fine.

  • 02

    We confirm camera-brand and codec compatibility on the call

    Almost always yes.

  • 03

    We scope the pilot and a path to full deployment

    Pilots usually start within two weeks of the call.

Start the pilot

Average response under 4 hours, US business hours.

Not ready for a call? Get the 2-page brief. No form, no follow-up.
FAQ

Common questions.

Do we need to replace any of our cameras?
Almost never. If your cameras output RTSP, ONVIF, H.264, or H.265, they work with Aqivon. We confirm camera-brand compatibility on the first call before any commitment.
Is there a minimum number of cameras?
No. We've deployed on a single camera and on facilities with sixty-plus. The edge node handles up to 64+ feeds per device, and multi-site rollouts add more nodes.
Where does the video go?
Nowhere. Inference runs on the edge node sitting in your rack. Video stays on your network. We never see it and neither does anyone else outside your facility.
How accurate is the detection?
Depends on the model and the environment. For person and object detection in typical industrial lighting, well above 95% after tuning. For facial recognition, above 99% on enrolled IDs. We benchmark on your footage during the pilot, not on someone else's lab data.
Can the models be tuned for our environment?
Every deployment includes training on your footage. We don't ship a generic model and call it tuning. We train on the objects, layout, and workflows that matter at your facility. The pilot validates the training before you commit to a full rollout.
How fast can we see results?
Pilots usually run within two weeks of the first call. Full deployments are live in twelve weeks from contract. The pilot is what most buyers use to make the internal case.
How is Aqivon different from a cloud CV platform?
Three things. Cloud platforms require sending video off-premise. We don't. Cloud platforms scale by selling more cloud compute. We scale by adding edge nodes you own. Cloud platforms hand you a dashboard. We hand you ERP-integrated detections that become records, not screens.
What about compliance?
Aqivon is built to operate inside regulated environments. Video stays on your network, which removes most data-residency questions before they're asked. Facial recognition is off by default in jurisdictions that restrict it (Illinois BIPA, Texas, Washington) and is fully audit-logged where enabled. For healthcare deployments we configure for HIPAA. For European operators we configure for GDPR. Your compliance team reviews the configuration during scoping.
Ready to start the pilot? Start the pilot →