JAILCD2025-009: Enhancing Road Maintenance Strategies in Bandung City Through YOLO and CNN-Based Damage Detection

Abstract

Maintaining road infrastructure is critical to urban mobility, economic growth, and public safety. In Bandung City, Indonesia, over 30% of roads are in poor condition, posing significant challenges for efficient maintenance. Traditional manual road damage detection methods are labor-intensive, prone to errors, and inefficient. To address this, an AI-powered hybrid system combining YOLOv8 and Convolutional Neural Networks (CNN) was developed for real-time road damage detection and classification. High-resolution images of Bandung’s roads were processed to create a diverse dataset encompassing cracks, potholes, and other defects. YOLOv8 identified damaged regions rapidly, while CNN provided precise classification. The hybrid model achieved an average accuracy of 93.06% with consistent performance across various datasets, outperforming traditional and standalone AI methods. By streamlining road damage detection, this system optimizes resource allocation, enhances maintenance prioritization, and supports sustainable urban mobility. The proposed approach demonstrates scalability and effectiveness, making it a viable solution for road infrastructure management in Bandung and similar urban areas.