INTEGRATED UAV IMAGERY AND DEEP LEARNING TECHNIQUES AUTOMATED ROAD DAMAGE DETECTION
Keywords:
UAV, road damage detection, deep learning, object-detection, YOLOV5, YOLOV7, YOLOV8.Abstract
This paper introduces an innovative approach for automated road damage detection using Unmanned
Aerial Vehicle (UAV) images and advanced deep learning techniques. Road infrastructure maintenance is crucial for
safe transportation, but manual data collection is often labor-intensive and risky. In response, we employ UAVs and
Artificial Intelligence (AI) to significantly enhance the efficiency and accuracy of road damage detection. Our
method leverages three state-of-the-art algorithms, YOLOv5, and YOLOv7, for object detection in UAV images.
Extensive training and testing with datasets from China and Spain reveal that YOLOv7 yields the highest precision.
Furthermore, we extend our research by introducing YOLOv8, which, when trained on road damage data,
outperforms other algorithms, demonstrating even greater prediction accuracy. These findings underscore the
potential of UAVs and deep learning in road damage detection, paving the way for future advancements in this field.
