Pothole Detection: Machine Learning and Computer Vision Approach

Potholes have become a major threat to road safety in recent times, accidents and fatalities have been associated with them in large numbers, especially when drivers are unable to notice the potholes and avoid them in due time. Hence, Pothole detection and mitigation are critical aspects of maintaining road infrastructure and ensuring commuter safety. Traditional methods of identification of potholes using manual inspections are both time-consuming and labour-intensive. So, in the recent years many machine learning models and algorithms like You Only Look Once (YOLO), Convolutional Neural Networks (CNNs), etc have emerged to accurately detect and help mitigate potholes. In this paper, a comparative analysis of these methodologies is presented based on accuracy, real-time performance and cost-efficiency.