ISpot: An intelligent real-time pothole spot identification model using a CNN algorithm

Abstract

Maintaining roads is a very intricate and significant global concern. Detecting road abnormalities, including potholes, is crucial in road monitoring and management. Identifying potholes is essential to minimize road accidents and car damage and improve travel comfort. Authorities have long seen road maintenance as a significant concern. However, the absence of accurate identification and connecting of road potholes exacerbates the problem. An end-to-end system named Intelligent Spotting (iSpot) of Pathole has been developed to address this issue by providing real-time identification, tracking, and geographical mapping of potholes around the city. A Convolutional Neural Network (CNN) framework is suggested and assessed using a real-world dataset for detecting potholes. Real-time maps displaying pothole locations are created using the Google Maps Application Programming Interface (API). Both pothole identification and mapping are combined into an Android application to offer a comprehensive service via this technology. The suggested model outperforms the baseline techniques regarding accuracy, precision, and F score.