Guide to Convolutional Neural Networks

Abstract

This book tries to adequately explain fundamentals of neural network and show how to implement and assess them in practice. Specifically, Chap. 2 covers basic concepts related to classification and it derives the idea of feature learning using neural network starting from linear classifiers. Then, Chap. 3 shows how to derive convolutional neural networks from fully connected neural networks. It also reviews classical network architectures and mentions different techniques for evaluating neural networks. Next, Chap. 4 thoroughly talks about a practical library for implementing convolutional neural networks. It also explains how to use Python interface of this library in order to create and evaluate neural networks. The next two chapters explain practical examples about detection and classification of traffic signs using convolutional neural networks. Finally, the last chapter introduces a few techniques for visualizing neural networks using Python interface. Graduate/undergraduate students as well as machine vision practitioners can use the book to gain a hand-on knowledge in the field of convolutional neural networks. Exercises have been designed such that they will help readers to acquire deeper knowledge in the field. Last but not least, Python scripts have been provided so reader will be able to reproduce the results and practice the topics of this book easily.