Computer vision for safer traffic
In our Innovation Lab we build our own computer vision and machine learning solutions. SmartBox watches railway crossings, other cameras tune intersections, alternating traffic and the movement of people in spaces.
Watching the crossing.
The camera sees the road and the rails.
What machine learning actually is
An ordinary program does exactly what a developer writes. Machine learning works the other way round: a model learns from examples. We show it thousands of shots of cars, people or trains and it learns to recognise them on footage it has never seen. Computer vision is the part of machine learning that works with the image from a camera.
Examples
The model learns from footage, not from rules written by hand.
Recognition
On a new shot it finds cars, people, vehicles and how they move.
Data
Video turns into numbers and events, not more hours of recordings.
We research and test it in real conditions, from traffic and railway crossings to guarding premises.
Cameras that understand what they see
Railways
SmartBox
A box with a camera at a railway crossing. It recognises a car stuck between the barriers and sends a warning before the train arrives.
Traffic
Smart intersections
Cameras count cars, cyclists and pedestrians and suggest signal timing that matches the real traffic, not a fixed plan.
Traffic
Alternating traffic
Where roadworks leave one lane open, the camera watches both queues and gives longer green to the side where more cars wait.
Spaces
People in spaces
Anonymous movement detection in halls and buildings. It shows where people queue, which areas are crowded and how the flow changes during the day.
Safety
Pedestrian crossings
Spotting risky situations at crossings in time, such as a pedestrian stepping out in front of a fast car.
What a camera can recognise
The same approach works wherever you need to know what is happening in front of a camera. These are the areas we work on.
Security
Guarding premises and fences
Entry into a restricted zone
Movement outside working hours
Traffic
Counting cars and cyclists
Queues and delays
Crossings and railway crossings
People and spaces
Number of people in a space
Queues and crowded places
Anonymous, nobody is identified
Vehicles
Types of vehicles
Entries into a site
Parking occupancy
From image to decision in a moment
The camera does more than record. A model on the device itself evaluates the scene and acts right away.
- 01
Camera
Captures the scene day and night, in rain or fog.
- 02
Edge model
A computer vision model runs right in the box, without sending video anywhere.
- 03
Real‑time decision
It recognises cars, people and their movement and decides whether something is wrong.
- 04
Action
It sends a warning, adjusts the signals or adds a record to the statistics.
Machine learning beyond cameras
Images are where we started. The same methods help with numbers and text, and these are the directions we research next.
We estimate demand, load or stock from the history you already have.
Reading invoices and forms into data, with a person checking the result.
Spotting unusual values in data, for example a device about to fail.
Routing requests and messages to the right person automatically.
Built for the street, not the lab
Every prototype is tested in real conditions before we call it ready.
Talk to the labPrivacy first
We count and recognise movement, we do not identify people.
Edge processing
Video stays in the device, only events and numbers leave it.
Any weather
Trained on day, night, rain and snow, not only on sunny footage.
Connects to existing systems
Works with existing cameras, traffic controllers and dispatch software.
Where to next
What we do and who we do it for.
- AI & automationThe routine runs by itself, people approve what matters.
- IntegrationsWe connect what you already run, even on days when one system is down.
- Cloud & DevOpsHosting, CI/CD, backups and monitoring, in the cloud or on your servers.
- ConsultingA conversation with a developer about an offer, code or an idea.
- AnalysisA clear scope, priorities and a price you can plan with.
Tell us about your project or problem
A crossing, an intersection or a space where you need to know what is happening. We reply with questions and, once the scope is clear, a proposal.