Intelligent vision. Real world impact.

From SurveillanceCameras toIntelligent Eyes

Cameras record what happened. Trinocular AI understands what is happening, in real time.

See the Demo See Live Detection

Schools · Factories · Warehouses · Retail

01 Camera Any ONVIF or RTSP feed

02 Live Video Continuous capture

03 AI Models People, objects, zones

04 Detection Matched to your rules

05 Alert To named staff

Latency: Sub-second; Detection to alert on a configured rule
Cameras: ONVIF / RTSP; Works with compatible existing CCTV
Deployment: Edge or Cloud; Deploy where the footage has to stay
Tuning: Per site; Models, thresholds and routing configured

Trinocular AI · Live Detection

Detect the Incident While It Is Happening

Trinocular AI identifies predefined anomalies while they are happening.

Live Trinocular AI · Live Monitor

Cameras: 128
Streaming: 4
Flagged: 1

Pause Play

CAM-014 Logistics Flagged 10:42:07

CAM-102 Perimeter Flagged 10:42:07

CAM-221 Access 10:42:07

CAM-310 Residential 10:42:07

Detection Ledger 6 events

14:27:16 Weapon detected CAM-03 · Concourse 0.92
14:18:32 Fire and smoke detected CAM-07 · Warehouse, Zone 4 0.98
21:14:02 Crowding detected CAM-07 · Plaza, Zone 3 48 people
10:42:07 Unusual group forming CAM-014 · School, Walkway 0.88
10:40:31 Person in machine zone CAM-102 · Factory, Cell 3 0.95
10:39:14 Restricted zone entry CAM-310 · Warehouse, Aisle 12 0.91

Accuracy depends on the models deployed, camera quality and placement, lighting and configured rules.

Trinocular AI · Detections

What Trinocular AI Flags

Each detection is a configured model and rule set, running on the cameras you already have.

A transit concourse camera view. Trinocular AI has drawn a detection box around a handgun carried at a walking man's side, labelled Weapon Detected, Gun, 92 per cent. — CAM 03 · Concourse · 14:27:16 Weapon 0.92

Trinocular AI · Threat Detection

Weapon Detection

A visible firearm in a public space is flagged the moment it enters frame, not after someone reviews the footage.

Detect Handguns and long guns in open carry
Locate Camera, zone and direction of travel
Alert Named staff and on-site security, with the clip
Escalate Lockdown and emergency routing per site

See It on Your Cameras

Detection covers visibly carried weapons. Accuracy depends on camera placement, lighting, distance and the models deployed. It is a force multiplier for your team, not a replacement for one.

A boundary camera looking up into bare tree branches against a blue sky. Trinocular AI has drawn a small detection box around a single branch alight, roughly a foot of flame, labelled Fire 0.88. — CAM-B1 · Boundary Trees North · 11:38:20 Fire 0.88

Trinocular AI · Safety Detection

Fire and Smoke Detection

A ceiling sensor needs the smoke to reach it. A camera needs line of sight, which is how this was caught at one branch rather than at a treeline.

Detect Flame at the source, before it spreads
Locate Zone, camera and distance to the boundary
Alert Site safety and the fire marshal, with the clip
Escalate Evacuation and emergency services routing

See It on Your Cameras

Works on the cameras already covering the site, with no additional sensors. How early a flame is caught depends on line of sight, lighting, and how much of it the camera can resolve at that distance.

A ceiling camera looking down a building corridor. Trinocular AI has drawn a counting zone across the walkway and reports a crowd count of six or more as people move through it. Faces are obscured, because the count does not need them. — CAM-07 · Main Corridor · 11:24:10 Crowd 6+

Trinocular AI · Crowd Intelligence

Crowd and Density Detection

Occupancy is counted per zone and compared against the threshold you set, so a crush is flagged while it is still forming.

Count Live occupancy per zone, not per camera
Measure Density, dwell time and direction of flow
Alert Stewards and control room at your threshold
Report Peak load and flow history per entrance

See It on Your Cameras

Counts people rather than identifying them, so the zone stays measurable without holding identity; the faces above are obscured for that reason. Accuracy depends on camera angle, height and how much the crowd overlaps in frame.

A factory conveyor camera view. Trinocular AI has drawn a detection box around each panel passing the inspection station, labelled Tile OK at 98 per cent, while a worker in a hi-vis vest and helmet looks on. — CAM 12 · Line 2 · 09:06:41 Tile OK 0.98

Trinocular AI · Production Intelligence

Production Monitoring

Every panel on the line is checked as it passes, so a defect is caught at the station that made it rather than at dispatch.

Inspect Surface defects, edges and finish per item
Count Throughput and reject rate by line and shift
Flag Station and timestamp for every reject
Trend Drift caught before it becomes a batch

See It on Your Cameras

Runs on the line camera you already have. What can be judged depends on resolution, lighting and how visible the defect is from the mounted angle.

A construction site camera view. Trinocular AI has drawn a detection box around each of five workers, every one labelled PPE Compliant at 98 per cent, with a side panel confirming hard hat, glasses, vest and gloves active and five of five people compliant. — CAM 05 · North Gate · 07:48:12 5 of 5 compliant

Trinocular AI · Safety Compliance

PPE Compliance

Hard hat, vest, glasses and gloves are checked on every person entering the area, continuously, without a supervisor standing at the gate.

Verify Helmet, vest, eye protection and gloves
Count People on site against people compliant
Flag Missing item, with the frame that shows it
Record Compliance rate by gate, shift and contractor

See It on Your Cameras

Checks what the camera can see, so an item hidden by posture or angle may not be confirmed. Rules and required items are set per site and per zone.

A monitor mounted beside a CCTV camera at an industrial yard. The screen shows a live feed with a cyan boundary drawn around a fenced restricted area, labelled Area Detected, Zone 01, Status Monitoring, beside a Restricted Area, Authorised Personnel Only sign. — CCTV 04 · Zone 01 · 14:18:32 Zone armed

Trinocular AI · Zone Intelligence

No Go Area Monitoring

You draw the boundary once. Anyone crossing it is flagged with the time, the camera and the direction they entered from.

Define Any shape, any camera, as many zones as needed
Watch Entry, loitering and crossing against the rule
Alert Named staff the moment the line is crossed
Evidence Clip of the crossing attached to the event

See It on Your Cameras

Zones are configured per camera and can be scheduled, so an area that is restricted overnight stays open during a shift.

A motorway gantry camera view. Trinocular AI has drawn a box around a grey sedan's number plate reading AB12 CDE, with a side panel listing the plate, vehicle type, colour, a speed of 56 miles per hour and 99.4 per cent confidence. — CAM 07 · Gantry · 14:35:12 AB12 CDE · 0.994

Trinocular AI · Vehicle Intelligence

Vehicle Monitoring

Plate, type, colour and speed are read as the vehicle passes, so a gate log writes itself and a watchlist hit is immediate.

Read Number plate, vehicle type and colour
Measure Speed and direction through the lane
Match Watchlists, permit lists and expected arrivals
Log Every pass, searchable by plate or window

See It on Your Cameras

Plate reading depends on angle, speed, plate condition and lighting. Partial reads are flagged as partial rather than guessed.

An airport terminal camera view. Trinocular AI has highlighted one walking man with a detection box labelled Restricted Person, confidence 98 per cent, while other travellers around him are not marked. — CAM 22 · Terminal B · 11:02:48 Match 0.98

Trinocular AI · Access Intelligence

Restricted Person Detection

A person on your restricted list is flagged as they move through a public space, while everyone around them stays unidentified.

Match Against the list your organisation maintains
Locate Camera, terminal and direction of travel
Alert Security desk, with the frame that matched
Audit Every match and every list change recorded

See It on Your Cameras

Operates only against a list you control, under your own policy and retention rules. Lawful basis, notice and oversight are a deployment decision, and we will not configure it without them.

A warehouse aisle camera view. Trinocular AI has drawn a detection box around each carton on the racking, every one labelled with its SKU, detected quantity and match status, with a floor monitor reporting 419 of 421 matches reconciled. — CAM 18 · Zone B5 · 16:40:09 419 / 421 matched

Trinocular AI · Warehouse Intelligence

Inventory Reconciliation

Stock on the rack is read against stock in the system continuously, so a discrepancy surfaces on the shift it happened rather than at the next count.

Read SKU, label and carton count per bay
Match Against expected quantity in your system
Flag Shortfalls, overstock and misplaced pallets
Reconcile Running match rate by zone and aisle

See It on Your Cameras

Reads what faces the camera, so labels turned to the back or stock buried behind a pallet will not be counted. It narrows a stock take, it does not remove one.

A port terminal inspection gantry camera view. Trinocular AI has drawn a detection box around a dented and gouged panel on a red container, labelled Damage Detected, type dent and gouge, severity high, confidence 96 per cent, beside the container number MSCU 123456. — GATE 02 · Inspection · 08:14:55 MSCU 123456 · 0.96

Trinocular AI · Terminal Intelligence

Container Damage Detection

Damage is logged against the container number as it passes the gate, so liability is settled on arrival rather than argued later.

Read Container number, ISO type and seal
Inspect Dents, gouges, holes and door damage
Grade Damage type and severity against your scale
Evidence Timestamped images filed to the move record

See It on Your Cameras

Covers the faces the gantry can see. Roof and underside need their own camera positions. Severity grading follows the scale your terminal already uses.

Tailored

What do you need caught?

Objects, states or behaviours specific to your site
Counts, conditions and compliance rules
Anything your team currently watches for by eye

Describe Your Use Case

Trinocular AI · Tailored Models

Something Else on Your Site

These are the detections we are asked for most. They are not the limit. If you can point a camera at it and describe what good and bad look like, it can usually be modelled.

Describe What you need caught, and what to ignore
Assess Whether your cameras and angles can see it
Configure A model and rule set for your site
Review Tuned against your footage before it goes live

We will tell you plainly when a camera cannot support what you are asking for. A model that reports confidently and wrongly is worse than no model at all.

CAM 03 · Concourse · 14:27:16 Weapon 0.92

Trinocular AI · Threat Detection