Road Condition Recognition Based on Object Classification Using YOLOv8
The quality of road infrastructure is one of the important factors in supporting the safety and comfort of road users as well as the smooth distribution of transportation. Road maintenance requires periodic monitoring by authorized institutions or agencies. Manual road condition monitoring tends to require considerable time, cost, and manpower, and is also prone to subjectivity. Therefore, a computational system capable of performing this task is needed. Based on this background, this study aims to develop a computer vision-based application for recognizing road conditions. Data consisting of road images with proper annotations (damaged or good) were used to train the YOLOv8 vision model. Our test found that the system accuracy, precision, recall, and F1-score is 0.96, 0.93, 1.00, and 0.96 respectively. The developed application allows users to input road images through a live camera and obtain real-time road condition classification results.