Insight into actual passenger numbers using computer vision

To understand the total number of passengers in its vehicles, GVB carries out manual counts. This work is labour-intensive, prone to errors and provides only a snapshot. We developed software for GVB that can take over this task. Using computer vision and images from a vehicle’s security cameras, it can count the passengers. The images are not viewed or shared; they are only analysed by the model. This protects passengers’ privacy. The solution gives GVB insight into actual passenger numbers over a longer period, so it can act on what it learns. For example, it can identify which routes have more passengers who do not pay.  

It has been a challenging project that has taught us a great deal about innovating with data while keeping things practical. Good collaboration between different experts shows what is possible.