Classification of Road Cracks using Deep Neural Networks

Roads play an important role in the economic development of a country. Along with economic development, they even provide social benefits such as traveling with ease and reduce the accidents caused due to poor road maintenance. With the above-said importance of the road, a means of transport, poorly maintained roads may reduce mobility and even lead to a significant increase in road accidents and vehicle operating cost. These cons of poor road maintenance emphasize the need for frequent and quick road damage detection. Road damage detection is an important part of road maintenance, which requires a huge manual effort to get the task done. This work can be done within a minimum time by developing a model that will be able to detect the type of cracks present in the image. In this project, the user is provided with an option to upload the damaged road image which is analyzed against the yolov3 model which can detect whether the road crack is among pothole, alligator crack, longitudinal crack, and lateral crack. This result will be stored in an excel sheet along with the location details of the road crack. Next to this, this research study defines the condition of the road b y using the same image as an input to a CNN model. In this way, a supervised deep neural network has been trained to detect road damage and save the data attained.