Implementing YOLOv9 and Faster R-CNN for Multi-Class Vehicle Recognition

Computer vision has become one of the rapidly developing fields of knowledge. One of the implementations of computer vision is object detection and recognition. Two prominent object detection algorithms that have gained significant attention are YOLOv9 and Faster R-CNN. The purpose of this study is to apply the YOLOV9 and Faster R-CNN algorithms to detect objects in the form of vehicles on the road such as motorcycles, cars, buses and trucks. Data are taken from CCTV recordings belonging to the department of transportation. 5963 image data have been collected from May to June 2024. Data were divided to 70% as training data, 20% as validation data and 10% as testing data. Further, all data was preprocessed by converting to YOLO and faster R-CNN format then resizing the image to 640x640. The analysis results show that YOLOv9 algorithm is superior to faster R-CNN with F1-score value of 0.938 and mAP@50:95 of 0.937. The results of testing datasets also show that YOLOv9 is good to detect vehicle on the road both during the day and night.