Deep Learning-Based Road Damage Detection and Classification Using Computer Vision
Jurnal Rekayasa Sipil dan Arsitektur · 2026
DOI: 10.51903/bxvt9389 ↗Kurz gefasst
Fachartikel in der Zeitschrift Jurnal Rekayasa Sipil dan Arsitektur, 2026. Verfasst von Calvin Arsento und 1 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-SA 4.0).
Abstract (Originalsprache)
Road damage detection is essential for supporting efficient road infrastructure inspection, yet conventional visual assessment is time-consuming and may produce inconsistent results when dealing with diverse pavement distress patterns. This study aims to comparatively evaluate YOLOv8s, YOLOv10s, YOLOv11s, and an optimized YOLOv11s configuration for detecting and classifying longitudinal cracks, transverse cracks, alligator cracks, and potholes using the RDD2022 dataset. A quantitative experimental approach was employed using 55,007 annotated damage instances, with all models trained under consistent experimental conditions and evaluated using precision, recall, F1-score, mAP@50, mAP@50–95, and inference time. The results show a progressive improvement across the evaluated configurations, with optimized YOLOv11s achieving the highest performance at 83.4% precision, 80.9% recall, 82.1% F1-score, 88.4% mAP@50, and 59.6% mAP@50–95, with an inference time of 8.7 ms. Class-specific analysis showed that longitudinal cracks achieved the highest F1-score of 85.5%, whereas alligator cracks recorded the lowest at 79.7%, with the largest classification confusion occurring between potholes and alligator cracks. The novelty of this study lies in the systematic cross-generation comparison of YOLOv8s, YOLOv10s, YOLOv11s, and optimized YOLOv11s under consistent experimental conditions while jointly examining overall performance, class-specific behavior, classification errors, and inference efficiency. The study contributes a detailed characterization of model performance across heterogeneous road-damage categories and provides quantitative evidence for selecting an effective YOLO configuration for automated road infrastructure inspection.
Zitieren
APA
Arsento, C., & Weningrum, N. (2026). Deep Learning-Based Road Damage Detection and Classification Using Computer Vision. Jurnal Rekayasa Sipil dan Arsitektur. https://doi.org/10.51903/bxvt9389
Vancouver
Arsento C, Weningrum N. Deep Learning-Based Road Damage Detection and Classification Using Computer Vision. Jurnal Rekayasa Sipil dan Arsitektur. 2026. doi:10.51903/bxvt9389
BibTeX
@article{Arsento2026,
author = {Arsento, Calvin and Weningrum, Nadia},
title = {Deep Learning-Based Road Damage Detection and Classification Using Computer Vision},
journal = {Jurnal Rekayasa Sipil dan Arsitektur},
year = {2026},
doi = {10.51903/bxvt9389}
}
EUVIMED