In this tutorial we are going to ultimately build an autonomous driving Neural Network model, trained from manual driving simulator, using the behavioral cloning technique, in Python.
The tutorial was part of the "Complete Self-Driving Car Course - Applied Deep Learning" course of Udemy.
This repository spans across 9 sections-tutorials to teach the basic aspects of Neural Networks and how they advance autonomous driving, before actually building our model.
In this section we learn to use the opencv library (cv2) to get the basics of computer vision, in a hands-on application of lines detection in autonomous driving.
In this section we learn the basics of Artificial Neural Networks (NN) and how to manually build a linear classification NN, also called Perceptron.
In this section we upgrade our previous knowledge by building a Neural Network model, using Keras library, which is perhaps the most known python library for NN.
In this section we extend the previous knowledge by building a Deep Neural Network, which essentially allows non-linear classification between 2 classes.
In this section, we extend the Deep Neural Network, by implementing non-linear classification for a multiple-class problem, which is usually the case in real-world.
In this section, we implement the knowledge we got for Neural Networks, in a famous handwritten digits dataset called MNIST.
In this section we learn the basics of Convolutional Neural Networks, and how these outweigh images classification problems. We apply it to the MNIST dataset to identify the differences.
In this section we extend our knowledge regarding Convolutional Neural Network, by applying these in a 43-class dataset of road signs. We also learn techniques to optimize the network in order to handle such complex datasets.
The last section of this course regards how to use all our previous knowledge about Deep and Convolutional Neural Networks as well as Continuous (polynomial) Regression, in order to apply it to the last model we will build, ultimately building an End-To-End Autonomous Driving model, to emulate conventional (manual) driving, using the Behavioral Cloning technique.