PotholeNet11:A YOLO11-Powered Real-Time Framework for Comprehensive Pothole Detection and Road Safety

Roads are the nation's leading transportation system, but the global road network is only 39% paved. Moreover, potholes on pavements pose a significant safety hazard, leading to high maintenance costs and road accidents. Manual detection of potholes proves inefficient and costly, requiring extensive human labor before repair activities begin. Therefore, there is a need to create automated detection systems for potholes. Merging computer vision with deep learning approaches, offering a better alternative for real-time pothole detection adeptly. The study utilizes 9,240 annotated pothole images, divided into 66% for training, 23% for validation, and 11% for testing, to train and evaluate YOLO11n, YOLO11s, YOLO11m, YOLO11l, and YOLO11x models. YOLO11s delivered the best mAP50 score of 0.848, with a precision of 0.887 and recall of 0.778, surpassing other models, making YOLO11s the best match for real-time pothole detection. This research is one of the first to evaluate YOLO11’s detection capability, which assures the possibility of improving pothole detection and maintenance strategies, reducing costs and accident probabilities.