Road Surface Anomaly Detection: YOLOv8 Nano for Accurate and Fast Pothole Segmentation

Road Surface Anomaly Detection: YOLOv8 Nano for Accurate and Fast Pothole Segmentation

Ananya Rajankar
Department of CSE(AIML),

Pimpri Chinchwad College of Engineering,
Pune, India
ananya.rajankar23@pccoepune.org

Shreyas Bile
Department of CSE(AIML),

Pimpri Chinchwad College of Engineering,
Pune, India
shreyas.bile23@pccoepune.org

Mrs. Suvarna Bhatsangave

Department of CSE(AIML),

Pimpri Chinchwad College of Engineering,
Pune, India
suvarna.bhatsangave@pccoepune.org

Abstract—Road potholes present significant challenges for vehicle maintenance and driver safety, causing substantial economic losses and traffic hazards worldwide. Traditional detection methods are often labor-intensive and inefficient, emphasizing the need for automated, accurate pothole identification systems. In this study, we present a deep learning approach utilizing the YOLOv8 nano segmentation model to detect and segment potholes with high precision. Our custom dataset, consisting of annotated road images, facilitates robust model training and validation. The proposed model is evaluated on metrics such as accuracy, precision, and Intersection over Union (IoU), achieving promising results that validate its effectiveness for real-time road monitoring. This method offers a scalable, automated solution with potential for integration into intelligent transportation and road maintenance systems, aiding authorities in proactive maintenance to enhance road safety.

Keywords— Pothole detection, YOLOv8, deep learning, road maintenance, segmentation, computer vision, intelligent transportation systems, automated inspection, real-time monitoring, road safety