An Automated System for Wrong-Way Vehicle Detection using YOLO and DeepSORT

Wrong-way driving is one of the main causes of wrong-way crashes. This also contributes to increased traffic flow and accidents. As a consequence, every year thousands of people die in road accidents, and valuable working hours are wasted. Bangladesh is a highly populated country where the traffic police are still handling traffic monitoring manually. Consequently, they often overlook or ignore the wrong-way vehicle. To address these problems, the paper introduces an automated wrong-way vehicle detection system that can be used for monitoring wrong-way vehicles. The proposed system has three phases: Detect objects from the video clips, track the detected objects using DeepSORT, and lastly, identify the vehicles driving in the wrong direction. Additionally, we count the number of vehicles traveling in each direction, which can be used for further statistical analysis. Our proposed wrong-way detection model achieved an accuracy of 97.53%. The model is also converted to TFLite and Core ML model so that the model can be deployed on Android and macOS/iOS devices. The proposed solution will be a fundamental tool for vehicle monitoring, lowering road accidents, traffic law violations, criminal activities, and traffic congestion.