Lane Detection for Self-Driving Car using Deep Neural Networks and Raspberry Pi

Abstract

The use of Computer Vision has indeed proven to be an integral part of the advancements in the recent technologies. Among the many applications of Computer Vision, the Self-Driving car is one of the most important incorporations in the automotive industry. In this paper we propose an implementation of Lane Detection using neural networks, and how it serves to be an important component in vision controlled self-driving car. Lane detection helps in the autonomous working of the car and can be achieved using Convolutional Neural Networks (CNN). The CNN model enables the car to learn from different types of roads, thereby allowing it to predict the direction the car has to take on any given road with minimal error. We evaluate the model to obtain the accuracy and loss during the training and validation of the data. We achieved a maximum accuracy of 96.68% and 96.18% during training and validation respectively. The model is then used to determine the direction in which the car has to move and this output is used for the motor control of the vehicle prototype designed accordingly.