<![CDATA[Road infrastructure damage, particularly potholes, is a critical issue that contributes to an increased risk of traffic accidents in Indonesia. This condition highlights the need for a road monitoring system that is capable of operating quickly, accurately, and continuously. This study aims to develop and evaluate an automated pothole detection system based on computer vision using the YOLOv8 (You Only Look Once version 8) method, a deep learning algorithm designed for real-time object detection. Dataset collection was conducted through two approaches, namely acquiring data from the Kaggle platform and capturing real-world road images directly using a smartphone camera to enhance data diversity and represent actual road conditions. All collected images were annotated using the Roboflow platform to generate labeled data suitable for model training. The YOLOv8 model was trained using a total of 345 images, consisting of 241 training images, 69 validation images, and 35 testing images. The training results indicate that the model achieved a mean Average Precision (mAP) of 93%, with a precision of 88% and a recall of 82%, demonstrating strong detection and localization performance. Furthermore, real-time testing using a smartphone camera showed that the system achieved an accuracy of over 85% under real-world conditions. These findings demonstrate that YOLOv8 has strong potential as an efficient and reliable automated pothole detection system, supporting road infrastructure maintenance and enhancing road user safety]]>
