ISSN No:2250-3676 ----- Crossref DOI Prefix: 10.64771 ----- Impact Factor: 9.625
   Email: editor@ijesat.com , ijesatj@gmail.com,   

(Peer Reviewed, Referred & Indexed Journal)


    REAL-TIME ROADWAY DEFECT LOCALIZATION WITH YOLO26: EFFICIENT VISION FOR SMART INFRASTRUCTURE

    Mosapoina Rajesh, M. Sreedhar

    Author

    ID: 4035

    DOI: Https://doi.org/10.64771/ijesat.2026.v26.i11.4035

    Abstract :

    Despite The Long And Complicated History Of The Problem, Pothole Detection Is Still A Significant Challenge In Intelligent Transportation And Road Asset Management Due To The Fact That Irregular Degradation Of The Road Surface Is Directly Related To Safety, The Integrity Of Vehicles, And Maintenance Prioritization. This Paper Uses The Lightweight Real-time Object Detector YOLO26n As A Standard Model For The Pothole Localization Problem, And Evaluates It On The Two Available Datasets On Kaggle That Include Various Road Surface Textures, Lighting Conditions, And Pothole Defect Forms. The Model Is Trained For 100 Epochs Using An Image Resolution Of 640 Images And A Batch Size Of 16, On Dual NVIDIA Tesla T4 GPUs, So It Can Be Evaluated Under Real Training Conditions To Understand How Well The Model Performs In Different Scenarios. Experimental Results Show A Precision Of 0.7567, Recall Of 0.6549, MAP50 Of 0.7477, And MAP50-95 Of 0.4786. The Values Show That YOLO26n Is A Reliable Object Detection Method To Define Pothole Regions At Certain Overlap Values, Whereas At Smaller Overlap Values, It Is Challenging To Achieve Tighter Localization Because The Geometry Of The Pothole Is Not Regular, The Contrast Is Low And The Visibility Of The Pothole Is Unclear. Further Validation Results, Precision-recall Results, Confidence Curves And Confusion Matrices Also Validate The Fact That The Detector Performs Well On Simple Pothole Cases And Still Fails On Some Cases Due To Missed Detections And Imprecise Boundaries In Complex Cases. The Overall Success Of This Work Suggests That YOLO26n Is A Promising Lightweight Baseline Model For Pothole Detection And Can Be Applied To Reproducible Benchmarking On Public Datasets. The Study Also Points To The Need For Further Research In Precise Defect Localization, Which Will Be Encouraged By Segmentation, Refinement Strategies Or Improved Postprocessing For Infrastructure Monitoring Applications.

    Published:

    12-11-2026

    Issue:

    Vol. 26 No. 11 (2026)


    Page Nos:

    100-104


    Section:

    Articles

    License:

    This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

    How to Cite

    Mosapoina Rajesh, M. Sreedhar, REAL-TIME ROADWAY DEFECT LOCALIZATION WITH YOLO26: EFFICIENT VISION FOR SMART INFRASTRUCTURE , 2026, International Journal of Engineering Sciences and Advanced Technology, 26(11), Page 100-104, ISSN No: 2250-3676.

    DOI: https://doi.org/10.64771/ijesat.2026.v26.i11.4035