Cyclists are exposed to great risks in inner cities. Vehicles often overtake without consideration of distances. This increases cyclists' feelings of insecurity and leads to reduced bicycle use. Car drivers are unaware of their fault and have difficulty estimating their distance to vulnerable road users. A driving assistance system could inform car users of critical situations while driving, what indirectly contributes to better cycling conditions. The BikeDetect project wants to improve the detection of cyclists in traffic scenarios. For this usecase, AI approaches are generally suitable. The AI model Two Wheels consists of implementations with Faster R-CNN, SSD, and YOLOv8, with YOLO demonstrating the best performance for camera based cyclist detection. Since cameras are associated with high costs, cyclists can also be detected using other technologies. For example, LiDAR offers the possibility of classifying objects based on geometric features. Thermal sensors can distinguish living from non-living objects based on temperature. In this specific application, the distance to the objects (the cyclist) is of interest. Ultrasonic or radar are suitable for measuring distances. The project faces the challenge of identifying approaches for fusion of classification and distance data streams. BikeDetect is currently in the preliminary study phase. Initial data is being generated in a parking lot test scenario. Based on the results of the study, the sensors that could form the basis of a driving assistance system are selected for a prototype that will be tested on a measuring vehicle in the inner city traffic of Osnabrück (Lower Saxony, Germany) during two days in September/October 2025. The goal is to determine overtaking distances in real road traffic scenarios through data fusion. The lecture or workshop will present the applied hardware and software technologies and discuss with participants how a future driving assistance system could be designed to significantly improve cyclist safety.