Road Damage Detection and Reporting System Using YOLOv7: Enhancing Public Participation in Infrastructure Maintenance

Abstract

Road damage in Lampung Province poses a serious threat to public safety and transportation, necessitating immediate government assistance. Despite the necessity for fast repair, there has been a lack of an appropriate platform for public reporting, making road inspections inefficient and error prone. This study makes two significant contributions: (1) the creation of a comprehensive dataset documenting road damage in Lampung, using samples collected in Pesawaran Regency, and (2) the development of an automated road damage identification and reporting system based on the YOLOv7 object detection model. The technology detects damage types such as alligator cracking, corrugation, depression, and potholes and provides real-time geolocation to improve reporting accuracy. Furthermore, it allows the public to report road damage. The model was trained on 1,200 pictures over 100 epochs, yielding a mean average precision (mAP) of 0.6. System performance was evaluated using precision and recall metrics, and compatibility testing confirmed that the system works reliably across various devices and browsers. This research enhances road maintenance by making inspections faster and more precise, while also promoting greater public involvement. Future work could integrate additional data sources, such as satellite imagery or IoT technology. to further increase the system’s scalability and accuracy.