AN OPTIMIZED REAL-TIME VEHICLE DETECTION FRAMEWORK FOR VIDEO STREAMS USING YOLOv4

With the number of vehicles growing at an accelerated rate, efficient traffic monitoring and vehicle detection systems are needed. The current methods of traffic monitoring are largely manual, thus requiring lots of time and manpower. In this paper, a YOLOv4 object detection algorithm is used to develop an optimized real-time vehicle detection framework for video streams. The proposed system receives image and video inputs, and outputs vehicle detection results in the form of a bounding box, class label and confidence score. The system offers individual modules for detection of images and live stream vehicles. The object detection is performed by YOLOv4 for fast and efficient detection. The developed framework can be adopted for traffic surveillance, vehicle monitoring and intelligent transportation applications. The experimental results show the vehicle detection capabilities of the system from traffic images and live video streams.