An Analytical Study of Various Pothole Detection and Prevention Techniques

With today's fast population growth, a significant rise can be seen in road traffic, resulting in improper road conditions, like potholes. These cause a lot of problems, like road accidents and heavy damage to vehicles. Several solutions have been used to address this issue, including manual reporting to local government agencies via sensors and other hardware-based technologies. However, many of the options listed above have drawbacks such as high implementation costs, danger, and inaccuracy. As a result, the aim of this paper is to examine the accuracy in identifying potholes on the road in order to offer drivers safety precautions before meeting one. We collected a substantial amount of data consisting of pothole images taken in diverse conditions, built multiple pothole detection techniques, and conducted a comparative analysis on accuracy and loss functions. The Convolutional Neural Network fared the best in detecting potholes, according to the data. The results of this survey will aid in guiding academics to focus on this critical application of CNN in the field of pothole detection.