Turn any camera into an AI sensor.

Lenstat runs vision AI on edge gateways at your site. Each gateway serves many cameras and understands what they see as it happens. The video never leaves your site.

100×+more efficient than central GPU systems
>99%fewer false alerts with drawn zones
0frames stored
AInspector brings Lenstat to drones, robots, vehicle and body-worn cameras.
Cameras turned into sensors at a city corner Entrance, 1,283 todayEntrance, 1,284 todayPlaza, 212 peopleCritical. Person down in the plazaSend first aid

Built to run where video is born.

Lenstat runs on NPU-powered edge gateways, each connected to many cameras. The heavy work happens on site, so there are no GPUs to buy, far less cloud to pay for, and the video stays where it was recorded.

  • Runs at the edgeOne gateway serves many cameras.
  • No GPU fleetLow power and low cost, on site.
  • Private by defaultVideo never leaves your site.
One gateway, many cameras video stays on site Findings11 cameras12 camerasPlaza, 212 peoplePlaza, 236 peopleGate 3, 42 per minGate 3, 44 per minQueue 5, wait 3:40Queue 5, wait 2:55Camera 12 added
99.9%

of the video processing and AI runs on the gateway.

90%+

less cloud bandwidth, storage and compute to pay for.

0

GPUs needed for video processing or AI at the edge.

Compared with video analytics on central GPU servers.

Three kinds of intelligence, each where it belongs.

The camera sees. The gateway reads what is happening. A reasoning layer ties it together across your sites, so anyone can ask a question in plain words and get the answer.

Perceive PersonCar
01

Perceive

Detects, tracks and labels people, vehicles and objects in every frame.

On the gateway
Understand WalkingWaiting0:000:060:120:180:240:300:360:420:48Zone: lobby
02

Understand

Reads posture, behaviour and context: who is queuing, who has fallen, who is lingering.

On the gateway
Reason Busiest entrance now?Reading 3 sitesEast entrance, 312
03

Reason

Connects events across every camera and site, and answers your questions in plain language.

Across your estate

Every site on its own gateway. One live picture of the estate.

There is no central bottleneck. Each gateway reads its own cameras in real time, and only small, anonymised findings travel on, where they join into one picture of your operation. Grow from one camera to thousands without changing your network.

  • Works with the cameras you haveRuns on its own or beside your video management system. Nothing to rip out.
  • Footfall and heatmapsCounts, flow and dwell from every camera, joined into one heatmap of the estate.
  • Sensor fusionWireless air-quality and gas sensors add what a camera cannot see.
One live picture of the estate Stand 12, 3 of 5 doneStand 12, 4 of 5 doneMall east, 1,284 inMall east, 1,291 inPlot 7, 11 on sitePlot 7, 12 on site3 sites livecounts and events

Private by design.

No video or images are stored and no personal data is kept. Faces are blurred in memory and dropped at once. The findings travel across your sites. Identity never does.

  • Nothing storedFrames are read on the gateway and dropped. Only the findings you choose are kept.
  • Re-identification without biometricsFollows the same person from camera to camera, without faces or identities.
  • On premises or air-gappedFor sensitive and sovereign sites, with no connection out.
  • GDPR, CCPA and APADesigned for the privacy laws of the places we work.
What leaves the gatewayFindings only
A live list of what leaves the gateway: counts, times and events, one after another, and never a frame or a face 14:02:10North mall entrance, 1,284 in today 14:02:16Stand 12, fuelling complete 14:02:31Level 3 site, PPE on, 11 of 11 14:02:44Arena gate 3, 42 people a minute 14:03:02Store 4 checkout, 6 waiting 14:03:15Car park B, a car going the wrong way 14:03:40Washroom 2, vape detected 14:04:05Yard 1, robot stopped short of a person 14:02:10North mall entrance, 1,284 in today 14:02:16Stand 12, fuelling complete 14:02:31Level 3 site, PPE on, 11 of 11 14:02:44Arena gate 3, 42 people a minute 14:03:02Store 4 checkout, 6 waiting 14:03:15Car park B, a car going the wrong way 14:03:40Washroom 2, vape detected 14:04:05Yard 1, robot stopped short of a person

No frames. No faces. No names.

One platform. Many things to watch.

The same intelligence counts people, times a queue, catches a fall and spots a car going the wrong way. Pick one to see it.

What it sees In 1,284, out 1,1904, out 1,1914, out 1,1925, out 1,192 Queue 5, wait 3:40Dwell 12:012345 Critical. Person down, plazaSend first aid High. Wrong way, aisle BTell car park staff PalletPalletPalletForkliftPallets in bay, 14 Air sensor, washroom 2Vape detected, no camera inside

Counts and flow from drawn lines and zones: footfall, occupancy and direction, per entrance and per hour.

Ask in plain words. Get the answer and the chart.

LensChat sits on top of everything Lenstat sees. Ask in English or Arabic, and get answers, charts and comparisons in seconds, each one traceable to the moment it came from.

  • Charts on requestFootfall, dwell, queues, vehicles and anomalies, for any camera or site.
  • TraceableEvery answer links back to the moments it counted.
LensChat
Sample data
Hello. I can see all 1,124 cameras across your sites. Ask me about footfall, dwell, queues, vehicles or anomalies.

Set up for the place it watches.

The same edge intelligence, configured for the realities of each sector. The video is read on the gateway and never leaves the site.

Retail

Turn footfall into a weekly operating rhythm.

For store and mall operators. See how shoppers move, where they linger and when queues build, store by store, with nobody watching a feed.

  • Footfall and unique visitsOccupancy by hour, so staffing matches demand.
  • QueuesLive wait times call for another till before shoppers walk away.
Lenstat for retail
Retail store, checkout queue Entrance, 1,283 todayEntrance, 1,284 todayDwell 2:40, promotion tableHigh. 6 waiting at checkoutOpen till 3
Aviation and airports

Every turnaround, watched to the second.

For airport operations and ground handlers. Turnaround milestones, ground-service activity and apron safety, live, so a delay shows as it starts.

  • Turnaround timelineEvery milestone from on-stand to doors closed, against the schedule.
  • Apron safetyHi-vis, restricted zones and vehicle proximity on every stand.
Aircraft turnaround Stand 12, turnaroundOn stand10:02Bridge on10:03Fuelfuelling10:10Bagsloading10:11Cateringpendinglate 6 minStand 12, T-12:30High. Catering late, stand 12Call the catering team
Construction

A safer site that reports on itself.

For contractors and site safety managers. PPE, danger zones, plant movement and progress, watched all day and flagged the moment something is off.

  • PPE checksHelmet and vest on every worker, with a flag when one is missing.
  • Danger zonesAn alert when someone walks under a crane load or into an excavation.
Construction site safety Lift zonePPE on, 11 of 11Critical. Person under the loadStop the lift
Venues

Crowd safety and access, before it becomes a problem.

For stadiums, arenas, halls and transport hubs. Crowd density, gate flow and entry queues, live, to keep every space within safe limits.

  • Crowd densityLive occupancy against the limit of each zone.
  • Gates and exitsThroughput per gate, and a blocked exit flagged at once.
Arena entry, crowd and gates Zone AZone BZone C Gate 3, 42 per minZone A, 180 peopleHigh. Zone C near its limitOpen gate 6, send stewards
UAV and robotics

Eyes on every drone and robot you run.

For warehouses, ports and industrial yards. Drones, ground robots and the people around them, in one live picture.

  • Routes and geofencesDraw the corridors and no-fly zones, and know the moment anything leaves them.
  • People firstA safe distance between people and robots, with an alert before contact.
Robot and drone yard Robot routes, 4 activeDrone corridorCritical. Robot 1 m from a personStop the robot
AInspector

Proactive security from any camera you fly, drive or wear.

AInspector is Lenstat's visual GenAI for drones, ground robots, vehicle and body-worn cameras. Three AI agents check every scene on an edge AI gateway at your site, and an alert needs two of them to agree.

3AI agents on every scene
2must agree before an alert
0faces or plates stored
Rail corridor inspection Person approaching Track bed, drawn areaCritical. Person on tracksStop trains on this line

See Lenstat on your own cameras.

Tell us what you need to watch for, and we will show you what Lenstat sees on your own streams, usually within a week.

GDPR, CCPA and APA

We reply within one business day. Your details are sent securely.