Vision AI

Make every camera more intelligent.

Turn the cameras you already have into counts, readings and alerts. Detect and count people, vehicles and objects, read plates and labels, and send your team only what needs a person.

  • People and vehicles
  • Counting
  • Plates and text
  • Alerts
Vision AI detecting people and vehicles across a factory, a construction site, a warehouse and a road.

What it does

Detect, count and read, from cameras you already own.

Detect

People, vehicles and objects, and whether something is where it shouldn’t be.

Count

Footfall, vehicles through a gate, items on a line, people in a queue.

Read

Number plates, labels, container codes and meter readings, straight from the frame.

Track and time

How long something waits, where it goes and when it leaves.

Alert

An event goes to the right person with the clip attached, so they can decide.

Where it helps

Wherever someone is watching a screen today.

  • Illustration: traffic and cities scene with detection boxes

    Traffic and cities

    • Vehicle counts by lane and by type
    • Number-plate reading at gates and car parks
    • Congestion, wrong-way driving and blocked junctions
  • Illustration: industrial and manufacturing scene with detection boxes

    Industrial and manufacturing

    • Hard hats, vests and other safety gear
    • People entering restricted or machine zones
    • Counts and defects on the line
  • Illustration: hospitals and healthcare scene with detection boxes

    Hospitals and healthcare

    • Waiting-room occupancy and queue times
    • Restricted areas and after-hours access
    • Beds, trolleys and equipment where they’re needed
  • Illustration: retail scene with detection boxes

    Retail

    • Footfall and busy hours, store by store
    • Queue length at the tills
    • Empty shelves before customers notice
  • Illustration: warehousing and logistics scene with detection boxes

    Warehousing and logistics

    • Trucks in and out, with plates read
    • Dock and bay occupancy
    • Pallets and containers counted and identified
  • Illustration: construction and security scene with detection boxes

    Construction and security

    • Perimeter and after-hours intrusion
    • Safety gear on site
    • Machinery and people near danger zones

How it works

One question at a time.

  1. 01

    Pick one question

    A single thing worth counting, reading or catching, and how your team handles it today.

  2. 02

    Test on your footage

    We compare its counts and readings with a person’s on your own recordings, and show you the result.

  3. 03

    Go live

    Events reach your team with the clip attached. The system flags; people decide.

  4. 04

    Add the next one

    Once the first question holds up, we add cameras or new things to look for.

Works with your setup

Plugs into what you already run.

Cameras

  • Most IP cameras that stream over RTSP or ONVIF
  • Existing NVR and DVR recorders
  • New cameras only where a view is missing

Where it runs

  • A small server on your site
  • Your own data centre or private cloud
  • Our cloud, if footage is allowed to leave

Where alerts go

  • WhatsApp, Slack or email
  • A dashboard for counts and history
  • Your own systems, through an API or webhook

Private by default

Your video doesn’t have to leave the building.

Camera footage is some of the most sensitive data a business holds. We can run the models on your own hardware, next to your recorders.

  • Footage stays on your network
  • Nothing is sent out unless you decide it should be
  • You choose who sees which cameras and alerts

Questions

What people ask before they start.

  • It depends on the camera angle, light and distance, so we don’t quote a number before we’ve seen your footage. We measure accuracy on your own cameras against a person’s count before anything goes live, and you see the result.

Have cameras running and nobody watching them?