Deep Learning Based Vehicle Detection on Cross-Roads
murat GENÇER Nesrin Aydın Atasoy
Abstract- Although the number of lanes of roads in residential cities is constant, the number of vehicle in traffic is increasing every year. This causes traffic congestion at certain times, such as before and after work hours. Thus, instead of traditional methods, intelligent systems have become a necessity to control traffic lights. For this problem, there are traffic signaling applications developed with using image processing and artificial intelligence techniques in literature. In this study, an application was developed to provide more detailed data for traffic signaling applications. Used Faster R-CNN model was trained and tested in Karabük and trained model detected 76 of 79 vehicles in 23 test frames and achieved 96% success.