
- Client
- Confidential
- Sector
- Traffic management
- Scope
- Detection, classification & counting
- Type
- AI video analysis
This was a private project. Client details are kept confidential.
01 — The challenge
What needed to change.
Traffic analysis depends on knowing how many vehicles of each type pass a location and when. Counting manually or from recorded footage is slow, expensive and hard to repeat consistently over long periods.
The client needed an automated way to turn camera footage into reliable vehicle counts per type.
02 — What we built
The solution, piece by piece.
01 / Component
Detection & Classification
- An AI model that finds every vehicle in the video.
- Each vehicle is sorted by type, such as car, bus or truck.
- Fast enough to work on live video.
02 / Component
Counting & Output
- Each vehicle is followed through the scene so it is counted only once.
- Counts are recorded per vehicle type and time period.
- Results are delivered as data the client can use in their traffic analysis.
03 — How we delivered it
Step by step.
- 01 / STEP
Data
We prepared footage from the client’s camera positions to train and test the model in the real conditions it would run in.
- 02 / STEP
Model
We trained the AI to tell apart the vehicle types the client needed.
- 03 / STEP
System
We built the full system around it, from video in to counts out.
- 04 / STEP
Validation
We compared the system’s counts with manually checked footage before handing it over.
Need something similar?
