Classification of Potholes and Plain Roads For Self-Driving Cars

Autonomous vehicles are crucial in applying traffic intelligence. The safety of self-driving automobiles is thus impacted by the smoothness of the road. Road potholes may cause a variety of problems, including auto damage and the likelihood of crashes. So, real-time pothole identification on the road should allow self-driving cars to adjust their behavior. This issue is dealt with using a variety of techniques, such as 3D laser aging and using vibration-based sensors. These techniques were, however, constrained by factors like high costs and the danger of being discovered. Therefore, it is vital to quickly and accurately automate the process of identifying potholes. In this paper, we have implemented transfer learning and CNN-based solutions to solve image classification problems using data from potholes and plain road images. The Kaggle dataset is used in this work. The goal of the work is to build models that identify if the image has potholes or is a plain road without potholes. Accordingly, it is a binary classification problem. The convolutional neural network deep learning model is stored and used for the prediction of potholes and plain roads which resulted in a final accuracy of 97.14%.