Answers are grounded in this paper and cite their sources.
The problem of potholes has been inexhaustible. to ensure road safety, transportation reliability, and timely infrastructure maintenance. This is because manual inspection methods are not scalable. and uniformity, which drives the desire to create automated AI-driven. detection systems. This paper gives a detailed Comparative analysis of two modern object detection architectures. Two additional models may be used to detect potholes in real-time: the YOLO family: YOLOv8 and YOLOv11, for real-time pothole estimation of severity and detection. A fully reproducible pipeline is constructed, including dataset preparation and annotation. parsing, augmentation techniques, model training, and performance analysis based on the major metrics, i.e., Precision, Recall, etc. mAP50, and mAP50–95. Estimates are additionally determined by the system. pothole severity using bounding-box area analysis. Therefore, enabling prioritisation of important road defects. In an experimental study on a pothole dataset, it can be seen that YOLOv11 results are shown below. Higher convergence speed, higher detection accuracy and improved. performs better than weak visual conditions. To YOLOv8 because of architectural upgrades, like better. optimized, improved feature aggregation, and backbone design. Training dynamics. In this study, I have given specific information on. discusses the practical differences between the two architectures and their trade-offs. Deployment thoughts and denotes ways of integrating. transformer-based models.
No takes yet. Share an insight, caveat, or question.
Sama et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: