2020/10/20 by Vishal Mandal, Mandal, Vishal, Abdul Rashid Mussah +3
Engineering · #Asphalt Pavement Performance Evaluation #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Infrastructure Maintenance and Monitoring #Vehicle License Plate Recognition
paper · pdf · doi:10.48550/arxiv.2010.10681
openalex publication_date 2020/10/20 · openalex created_date 2022/07/25 · openalex updated_date 2026/07/28
Automatic detection and classification of pavement distresses is critical in\ntimely maintaining and rehabilitating pavement surfaces. With the evolution of\ndeep learning and high performance computing, the feasibility of vision-based\npavement defect assessments has significantly improved. In this study, the\nauthors deploy state-of-the-art deep learning algorithms based on different\nnetwork backbones to detect and characterize pavement distresses. The influence\nof different backbone models such as CSPDarknet53, Hourglass-104 and\nEfficientNet were studied to evaluate their classification performance. The\nmodels were trained using 21,041 images captured across urban and rural streets\nof Japan, Czech Republic and India. Finally, the models were assessed based on\ntheir ability to predict and classify distresses, and tested using F1 score\nobtained from the statistical precision and recall values. The best performing\nmodel achieved an F1 score of 0.58 and 0.57 on two test datasets released by\nthe IEEE Global Road Damage Detection Challenge. The source code including the\ntrained models are made available at [1].\n