How Generali Employee Benefits built a future-ready data foundation
Reviewing parking violations shouldn’t be manual guesswork. We built a computer vision POC on Google Cloud to help Parking Brussels assess images faster and make more accurate decisions.
The challenge
Parking Brussels relies on image-based checks to decide whether a fine should be issued. Many of these images are low quality, making it difficult to see exactly what is on the image.
As a result, operators spend a lot of time reviewing cases manually. This leads to human errors and complaints. With AI assisting in image analysis, this process can be optimized.
Our approach
We developed a POC on Google Cloud to test how computer vision could support this process. We started with Cloud Vision API, but quickly saw it was too general for these specific use cases.
We then explored Gemini and later trained custom models using AutoML. After testing different setups, we concluded that the custom trained AutoML models performed best on the images we got. In some cases, combining multiple models delivered even better results.
Dolor enim eu tortor urna sed duis nulla. Aliquam vestibulum, nulla odio nisl vitae. In aliquet pellentesque aenean hac vestibulum turpis mi bibendum diam. Tempor integer aliquam in vitae malesuada
The outcome
ScanCar helps operators review cases faster and with more clarity. By supporting image analysis with AI, Parking Brussels can reduce manual effort and improve consistency in decision-making.
Next to that, the license plate detection model increased accuracy by 3 percent compared to the existing setup.
Have a talk
with us
