About
Our client, a German-based company, uses cameras on highways and city crossroads to monitor traffic and generate data. These video streams train their ML system to recognize vehicle types. However, the system struggled in poor lighting and adverse weather conditions like rain and snow. It also had limited data with varying angles and setups. Additionally, the system could only identify broad vehicle categories, unable to distinguish sub-types.
Challenge, approach, and impact
Needs fixing
Our Client monitors traffic using cameras on highways and city crossroads, generating data for their ML system to recognize vehicle types. However, performance dropped in low light and bad weather, and the system lacked varied angles and setups. It also failed to distinguish sub-types of vehicles, grouping all trucks together and missing specific types like tow trucks and emergency vehicles. How do we improve the system?
Training begins
Since real-world traffic conditions can be unpredictable, we have decided to move to a simulated environment to save time and quickly cover many scenarios. The idea was to generate around 15.000 images to train the computer vision model, using different weather conditions, camera angles, vehicle types, etc.
How we built
Testimonials
Christoph
ITS United
“Throughout the entire project, communication was fluent and based on need. Direct communication through messaging apps made the process very smooth. Due to this, the project was delivered on time and in scope.“
Team structure
Client team
Christoph
CEO
Project stakeholder
The client stakeholders were working closely with the team at Marble IT
Agency team
1 x ML Engineer
Production
