Enabling Safe Driving through AI - Based Speedbump and Pothole Detection for Indian Roads

The objective of this study is to use Computer Vision and Artificial Intelligence techniques for detecting Potholes and Speed bumps on Indian roads to improve driver safety Supervised machine learning AI technique was utilized by using the Convolutional Neural Network (CNN) to create an image detection model which divided the Indian road images into 3 categories - normal, Speed Bump and Pothole. Our code was able to accurately classify the images into one of the three categories. We trained a Convolutional Neural Network using over 800 unique Indian road images on Speed bumps and Potholes. We got an accuracy of 91% using this model, indicating that a reasonable sized CNN model can very accurately detect these Indian road conditions. We compare the complexity of this model versus a model that is not trained with India road images, to showcase the added complexity of building a model for Indian conditions.