Urban Applications of Deep Learning: Potholes Detection Using YOLOv10 Model in Al-Bayda, Libya

The city of Al Bayda is a regional center with many agricultural villages that lie in the northeast of Libya. Poor road maintenance, reaching the end of their service life, and high traffic loads have led to a marked deterioration in the condition of the city's roads. This research used open-source data from Kartaview and OSM (OpenStreetMap) integrated with YOLO (You Only Look Once) object detection model to detect potholes in the city's road network. Therefore, this study aims on improving road safety and efficiency related to maintenance, using volunteer-generated geospatial data along with the latest YOLOv10 model, to create a scalable and reliable method of pothole detection. Results Confirm image diversity and accurate annotation are crucial in model training to achieve mAP (Mean Average Precision) above 50%. This identifies an effective tool for pothole detection that may be scalable in the context of urban infrastructure management and is a cost-effective approach that can ensure traffic safety and early maintenance actions while reducing the environmental impact of vehicle downtime. Keywords: YOLO, Deep Learning, Pothole Detection, Infrastructure Management, Kartaview, OpenStreetMap (OSM).