Self Driving Car Simulation Using Deep Learning

Abstract

Every year, traffic accidents account for 2.2% of worldwide deaths. That accumulates to a few millions a year. On prime of this, some individuals are seriously lacerate in auto-related accidents annually in large numbers mostly due to human error. From distracted driving to drunk driving to reckless driving to careless driving, one poor or inattentive call may be the distinction between a typical drive and a critical state of affairs. However what if we tend to might neutralize human error from the equation?The proposed system of ‘Self Driving Autonomous Car’ is designed using Deep Learning and CNN and Computer Vision. Imagine stepping into your car and however speaking a location into your vehicle’s interface, then lease it drive you to your destination whereas you read a book, surf the net, or nap without fear concerning something. Computer Vision techniques via Open-CV are applied to establish lane lines for a self-driving car. CNN is trained to spot numerous traffic signs using Keras (open-source platform).To generalize the behavior of cars on different tracks, it is not possible to collect and process a huge amount of data, that is why augmentation is done which generalizes data on new tracks. Augmentation can be done using various techniques like crop, zooming, Flipping, and changing brightness