Comparative Analysis and Evaluation of Pothole Detection Algorithms

Abstract

Potholes are one of the biggest reasons of menace on the streets. They fill up after rains with water and become hard to detect and anyone do not know how deep they go. Lots of people get into accidents because of these potholes. It is important that they are reported to the government and even keep a record of them to help people avoid them. The proposed system makes use of Yolov7 object detection algorithm. The highest mAP score is obtained by YOLOv7 with 84.7%, which is followed by YOLOv3 which got 66.3%, YOLOv3 Tiny obtained 59.5%, and MobileNet SSD with 51.9%. The mAP calculates a score by comparing the detected box to the ground-truth bounding box. The better the score, the more reliable the model's detections are. It is one of the fastest available around the world today. This model came from the makers of YOLO after the YOLOv4 version. The technology may relate to current security cameras and assist transportation agencies in promptly locating and filling potholes, making road networks safer and more effective. Pothole damage is a common reason for car insurance claims. By using YOLOv7 to detect and record the location of potholes, insurance companies can better manage claims and reduce fraud.