Automated Road Damage Detection Using UAV Images and Deep Learning Techniques

Abstract

Automated road damage identification utilizing UAV photos and sophisticated deep learning algorithms is presented in this research as a novel methodology. While keeping roads in good repair is essential for travel safety, gathering data by hand can be a dangerous and time-consuming ordeal. Our solution is to use UAVs together with AI to make road damage identification far more efficient and accurate. To identify objects in UAV photos, our approach makes use of three cutting-edge algorithms: YOLOv5, YOLOv7, and YOLOv5. Extensive testing and training using Chinese and Spanish datasets show that YOLOv7 produces the best accuracy. In addition, we expand our study by presenting YOLOv8, an algorithm that surpasses existing algorithms and shows significantly better prediction accuracy when trained on road damage data. These results demonstrate the promise of unmanned aerial vehicles (UAVs) and deep learning for detecting road damage, which should lead to further developments in this area. Unmanned Aerial Vehicle (UAV), road damage detection, object detection, YOLOV5, YOLOV7, and YOLOV8 are index words.