Pothole Detection: A Performance Comparison Between YOLOv7 and YOLOv8
Zineb Haimer, Khalid Mateur, Youssef Farhan, Abdessalam Aït Madi
Abstract
This study focuses on the comparison of two state-of-the-art object detection models, YOLOv7 and YOLOv8, for the task of pothole detection in road images. The detection of potholes is crucial for maintaining the safety of roads and highways, and the use of computer vision algorithms has shown promising results in automating this process. Our aim is to determine which of these two models offers the best trade-off between speed and accuracy in pothole detection. We conducted experiments on a dataset of road images containing potholes and evaluated the performance of each model using object detection metrics such as precision, recall, mean average precision (mAP) and F1-score accuracy. Our findings reveal that YOLOv8 slightly outperforms YOLOv7 in terms of object detection metrics, achieving a mAP of 78.6%. These results demonstrate the potential of YOLOv8 for real-time pothole detection in road safety applications.