A Comparative Analysis of YOLOv8 and Vision Transformers for Road Damage Detection and Classification Using the RDD2022 Dataset

Shaikh, Abdul Ahad (2025) A Comparative Analysis of YOLOv8 and Vision Transformers for Road Damage Detection and Classification Using the RDD2022 Dataset. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (9MB) | Preview
Preview	PDF (Configuration Manual) Download (5MB) | Preview

Abstract

The problem of road damage detection is a long-standing issue of road authorities, because manual inspection is time-consuming, labour-intensive, and has the tendency of inconsistency. This study meets the requirement of the automation, precision, and scalability of detection strategies that can detect different types of road defects in various settings. To address this requirement, this paper compares two recent deep-learning models, including YOLOv8 that is a object-level detector and Vision Transformers (ViT), a model that is an image-level classification network. What this work contributes is the adoption of an end-to-end pipeline, i.e. the conversion of the dataset into a trainable format, evaluation and the comparative analysis of the various models, and also illustrating how each model performs on the RDD2022 dataset. The experiments demonstrate that YOLOv8 can reach an mAP at 0.50 of 0.5627, i.e. more than 56 percent of predicted bounding boxes overlap ground truth at 50 or above, and ViT can gain higher accuracy 71.48 at validation, which means that the model is capable of high performance in single-label classification. The findings indicate that YOLOv8 is more compatible with the existing state-of-the-art localisation strategies, and ViT has more useful applications in larger categorisation problems. Nevertheless, there are still shortcomings when it comes to the way ViT recognizes numerous small cracks, and YOLO perceives repair types as visual uncertainties.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Sahni, Vikas UNSPECIFIED
Subjects:	T Technology > TE Highway engineering. Roads and pavements Q Science > QH Natural history > QH301 Biology > Methods of research. Technique. Experimental biology > Data processing. Bioinformatics > Artificial intelligence Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Artificial intelligence Q Science > QH Natural history > QH301 Biology > Methods of research. Technique. Experimental biology > Data processing. Bioinformatics > Artificial intelligence > Computer vision Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Artificial intelligence > Computer vision
Divisions:	School of Computing > Master of Science in Artificial Intelligence
Depositing User:	Ciara O'Brien
Date Deposited:	02 Sep 2026 10:46
Last Modified:	02 Sep 2026 10:46
URI:	https://norma.ncirl.ie/id/eprint/9768

Actions (login required)

View Item