Concrete surface deterioration detection and classification using YOLO26n: an AI-driven approach
- 1Department of Civil Engineering, Chittagong University of Engineering and Technology
Abstract
An efficient and reliable system for the classification and detection of concrete surface deterioration is essential for timely maintenance and extending the service life of concrete structures. Conventional manual inspection methods are labor-intensive, time consuming, and prone to human error, which can lead to delayed remediation and increased repair costs. This study presents an artificial intelligence–based approach using the You Only Look Once 26 Nano (YOLO26n) model to automatically detect and classify various types of concrete surface degradation, including exposed reinforcement, rust stains, scaling, spalling, cracks, and efflorescence. A dataset comprising 6,806 images, collected and manually annotated from publicly available sources, was partitioned into training, validation, and test sets using an 80%, 10%, 10% split. To enhance model robustness and generalization, multiple data augmentation techniques such as scaling, translation, and color jittering were applied. The proposed model was implemented using the Ultralytics YOLO framework, and its performance was evaluated using precision, recall, F1 score, and mean Average Precision (mAP). Experimental results demonstrate that the YOLO26n model achieves a precision of 95.8%, recall of 97.7%, F1-score of 0.967, mAP@50 of 98.9%, and mAP@50–95 of 85.5%. These results indicate strong detection accuracy and localization performance, outperforming traditional manual inspection methods as well as existing deep learning models such as YOLOv5n, YOLOv8n, YOLO11n, and YOLO12n. The proposed automated system significantly reduces inspection time, improves detection consistency, and enables efficient and cost-effective infrastructure maintenance.
Related
- Deterioration behavior and model of alkali-activated fly ash-slag concrete under freeze–thaw cycles
- A deep learning-based silhouette gait detection framework for fall risk classification in Parkinson’s disease
- Exploratory Analysis of Circulating miR‐486‐3p and miR‐21‐5p in Early Detection and Subtype Classification of Luminal A and B Breast Cancer in Peruvian Women
- Customizable bead-based platform for EV surface protein immunocapture and detection in plasma
- Deep learning based edge segmentation and classification for plasterboard defect detection
BibTeX
@article{akter2026concrete,
title = {{Concrete surface deterioration detection and classification using YOLO26n: an AI-driven approach}},
author = {Syeda Tamzida Akter and Sayed Shahriar Lamun and Md. Basir Zisan and Md. Asifur Rahman},
journal = {Scientific Reports},
year = {2026},
month = sep,
doi = {10.1038/s41598-026-68706-0},
url = {https://doi.org/10.1038/s41598-026-68706-0},
note = {CC BY},
}
RIS
TY - JOUR
TI - Concrete surface deterioration detection and classification using YOLO26n: an AI-driven approach
AU - Syeda Tamzida Akter
AU - Sayed Shahriar Lamun
AU - Md. Basir Zisan
AU - Md. Asifur Rahman
PY - 2026
DA - 2026/09/01
JO - Scientific Reports
DO - 10.1038/s41598-026-68706-0
UR - https://doi.org/10.1038/s41598-026-68706-0
AB - An efficient and reliable system for the classification and detection of concrete surface deterioration is essential for timely maintenance and extending the service life of concrete structures. Conventional manual inspection methods are labor-intensive, time consuming, and prone to human error, which can lead to delayed remediation and increased repair costs. This study presents an artificial intelligence–based approach using the You Only Look Once 26 Nano (YOLO26n) model to automatically detect and classify various types of concrete surface degradation, including exposed reinforcement, rust stains, scaling, spalling, cracks, and efflorescence. A dataset comprising 6,806 images, collected and manually annotated from publicly available sources, was partitioned into training, validation, and test sets using an 80%, 10%, 10% split. To enhance model robustness and generalization, multiple data augmentation techniques such as scaling, translation, and color jittering were applied. The proposed model was implemented using the Ultralytics YOLO framework, and its performance was evaluated using precision, recall, F1 score, and mean Average Precision (mAP). Experimental results demonstrate that the YOLO26n model achieves a precision of 95.8%, recall of 97.7%, F1-score of 0.967, mAP@50 of 98.9%, and mAP@50–95 of 85.5%. These results indicate strong detection accuracy and localization performance, outperforming traditional manual inspection methods as well as existing deep learning models such as YOLOv5n, YOLOv8n, YOLO11n, and YOLO12n. The proposed automated system significantly reduces inspection time, improves detection consistency, and enables efficient and cost-effective infrastructure maintenance.
SN - 2045-2322
ER -