Analysis of training time optimization for self-driving car using alternate max pooling layers

Abstract

In the modern era, the vehicles are focused to be automated to give human driver relaxed driving. In the field of automobile various aspects have been considered which makes a vehicle automated. As Udacity, in 2016, with the Google Self-Driving Car founder Sebastian Thrun open sourced their self driving car simulation environment, a new door opened up in self driving vehicle research. In this paper we have focused on comparing between a popular neural network model introduced by NVIDIA and a model made with max pooling optimized neural network. Max pooling is a method of making Convolutional Neural Network simpler. We have also worked vigorously on creating a custom urban environment based on Dhaka, where we have run the car autonomously, gathered the training time data, and the instances it fails to drive along the trained roadway. The idea described in this paper is related to deep learning algorithm analysis and comparison, which is the core part of the self driving car. The analysis hence gives us a great realization of machine learning techniques and their effectiveness in practical situations. All being said, this paper, however approaches to solve one big problem, to find out practicality of a popular model compared to an unconventional model, put in a real scenario.