Deep Learning Models: A Practical Approach for Hands-On Professionals

✍ Scribed by Jonah Gamba

• Publisher: Springer
• Year: 2024
• Tongue: English
• Leaves: 211
• Category: Library

✦ Synopsis

This book focuses on and prioritizes a practical approach, minimizing theoretical concepts to deliver algorithms effectively. With deep learning emerging as a vibrant field of research and development in numerous industrial applications, there is a pressing need for accessible resources that provide comprehensive examples and quick guidance. Unfortunately, many existing books on the market tend to emphasize theoretical aspects, leaving newcomers scrambling for practical guidance. This book takes a different approach by focusing on practicality while keeping theoretical concepts to a necessary minimum. The book begins by laying a foundation of basic information on deep learning, gradually delving into the subject matter to explain and illustrate the limitations of existing algorithms. A dedicated chapter is allocated to evaluating the performance of multiple algorithms on specific datasets, highlighting techniques and strategies that can address real-world challenges when deep learning is employed. By consolidating all necessary information into a single resource, readers can bypass the hassle of scouring scattered online sources, gaining a one-stop solution to dive into deep learning for object detection and classification. To facilitate understanding, the book employs a rich array of illustrations, figures, tables, and code snippets. Comprehensive code examples are provided, empowering readers to grasp concepts quickly and develop practical solutions. The book covers essential methods and tools, ensuring a complete and comprehensive coverage that enables professionals to implement deep learning algorithms swiftly and effectively.
This book is designed to equip professionals with the necessary skills to thrive in the active field of deep learning, where it has the potential to revolutionize traditional problem-solving approaches. This book serves as a practical companion, enabling readers to grasp concepts swiftly and embark on building practical solutions.

✦ Table of Contents

Preface
Acknowledgements
Contents
1 Basic Approaches in Object Detection and Classification by Deep Learning
1.1 Introduction
1.2 Conventional Methods of Object Detection and Machine Learning
1.2.1 K-Nearest Neighbors (KNN)
1.2.2 Linear Discriminant Analysis (LDA) and Quadratic Discriminant Analysis (QDA)
1.2.3 Support Vector Machine (SVM)
1.2.4 Random Forest
1.2.5 Gradient Boosting Machines
1.3 Deep Learning as Part of Artificial Intelligence
1.4 Frameworks for Deep Learning
1.5 Selection of Target Areas for This Book
1.6 Concluding Remarks
1.7 Self-evaluation Exercises
References
2 Requirements for Hands-On Approach to Deep Learning
2.1 Introduction
2.2 Basic Python Arrays for Deep Learning
2.3 Setting Up Environment
2.3.1 OS Support for Offline Environments
2.3.2 Windows Environment Creation Example
2.3.3 Options to Consider for Online Environments
2.4 Concluding Remarks
2.5 Self-evaluation Exercises
References
3 Building Deep Learning Models
3.1 Introduction: Neural Networks Basics
3.1.1 Shallow Networks
3.1.2 Convolutional Neural Networks (CNNs)
3.1.3 Recurrent Neural Networks (RNNs)
3.1.4 Long Short-Term Memory (LSTM)/Gated Recurring Units (GRUs)
3.2 Using Keras for as Deep Learning Framework
3.2.1 Overview of Library
3.2.2 Usability
3.3 Concluding Remarks
3.4 Self-evaluation Exercises
References
4 The Building Blocks of Machine Learning and Deep Learning
4.1 Introduction
4.2 Categorization of Machine Learning
4.3 Methods of Evaluating Machine Learning Models
4.3.1 Data Preprocessing for Deep Learning
4.3.2 Problem of Overfitting
4.4 The Machine Learning Workflow
4.5 Concluding Remarks
4.6 Self-evaluation Exercises
References
5 Remote Sensing Example for Deep Learning
5.1 Introduction
5.2 Background of the Remote Sensing Example
5.3 Remote Sensing: Land Cover Classification
5.4 Background of Experimental Comparison of Keras Applications Deep Learning Models Performance on EuroSAT Dataset
5.4.1 Information Input Data Requirements
5.4.2 Input Restrictions (from Keras Application Page)
5.4.3 Training and Test Results
5.5 Application of EuroSAT Results to Uncorrelated Dataset
5.5.1 Evaluation of 10-Classes with Best EuroSAT Weights
5.5.2 Training Results with 6 Classes—Unbalanced/Balanced Case
5.5.3 Training Results with 5 Classes
5.6 Concluding Remarks
References