Data preprocessing and balancing to enhance end to end learning in self-driving vehicle
Syed Zishan Ali, Rishabh Sharda, Abhishek Dewangan, Sourabh Chawda, Rahul Sharma · International journal of advance research, ideas and innovations in technology · 2018
Driving a vehicle has always been a demanding task be it any vehicle since robotics and artificial intelligence has progressed multi-folds in the last decade which gave us the technological grounds to automate many processes which include driving. Developing autonomous vehicle is the current research area for many corporate like are Google, Tesla, Nvidia, and Uber. Several proposed methodology for them is Nvidia’s Behavioural Cloning, CommaAI’s OPENPILOT, Tesla’s AUTOPILOT all of which uses the camera to process surrounding of the vehicle. In this paper, we discuss Nvidia’s recent work (behavioral cloning) and incorporate their work with few techniques of our own like filtering the repeating data and augment the input data to reduce the amount of data collection required.