Guide to Convolutional Neural Networks (eBook)
XXIII, 282 Seiten
Springer International Publishing (Verlag)
978-3-319-57550-6 (ISBN)
This must-read text/reference introduces the fundamental concepts of convolutional neural networks (ConvNets), offering practical guidance on using libraries to implement ConvNets in applications of traffic sign detection and classification. The work presents techniques for optimizing the computational efficiency of ConvNets, as well as visualization techniques to better understand the underlying processes. The proposed models are also thoroughly evaluated from different perspectives, using exploratory and quantitative analysis.
Topics and features: explains the fundamental concepts behind training linear classifiers and feature learning; discusses the wide range of loss functions for training binary and multi-class classifiers; illustrates how to derive ConvNets from fully connected neural networks, and reviews different techniques for evaluating neural networks; presents a practical library for implementing ConvNets, explaining how to use a Python interface for the library to create and assess neural networks; describes two real-world examples of the detection and classification of traffic signs using deep learning methods; examines a range of varied techniques for visualizing neural networks, using a Python interface; provides self-study exercises at the end of each chapter, in addition to a helpful glossary, with relevant Python scripts supplied at an associated website.This self-contained guide will benefit those who seek to both understand the theory behind deep learning, and to gain hands-on experience in implementing ConvNets in practice. As no prior background knowledge in the field is required to follow the material, the book is ideal for all students of computer vision and machine learning, and will also be of great interest to practitioners working on autonomous cars and advanced driver assistance systems.Preface 6
Books Website 7
Contents 8
Acronyms 12
List of Figures 14
1 Traffic Sign Detection and Recognition 23
1.1 Introduction 23
1.2 Challenges 24
1.3 Previous Work 27
1.3.1 Template Matching 27
1.3.2 Hand-Crafted Features 27
1.3.3 Feature Learning 29
1.3.4 ConvNets 32
1.4 Summary 34
2 Pattern Classification 37
2.1 Formulation 38
2.1.1 K-Nearest Neighbor 39
2.2 Linear Classifier 42
2.2.1 Training a Linear Classifier 44
2.2.2 Hinge Loss 52
2.2.3 Logistic Regression 56
2.2.4 Comparing Loss Function 59
2.3 Multiclass Classification 63
2.3.1 One Versus One 63
2.3.2 One Versus Rest 66
2.3.3 Multiclass Hinge Loss 68
2.3.4 Multinomial Logistic Function 70
2.4 Feature Extraction 73
2.5 Learning ?(x) 80
2.6 Artificial Neural Networks 83
2.6.1 Backpropagation 87
2.6.2 Activation Functions 93
2.6.3 Role of Bias 100
2.6.4 Initialization 101
2.6.5 How to Apply on Images 101
2.7 Summary 103
2.8 Exercises 104
3 Convolutional Neural Networks 106
3.1 Deriving Convolution from a Fully Connected Layer 106
3.1.1 Role of Convolution 111
3.1.2 Backpropagation of Convolution Layers 113
3.1.3 Stride in Convolution 115
3.2 Pooling 116
3.2.1 Backpropagation in Pooling Layer 118
3.3 LeNet 119
3.4 AlexNet 121
3.5 Designing a ConvNet 122
3.5.1 ConvNet Architecture 123
3.5.2 Software Libraries 124
3.5.3 Evaluating a ConvNet 126
3.6 Training a ConvNet 132
3.6.1 Loss Function 133
3.6.2 Initialization 134
3.6.3 Regularization 136
3.6.4 Learning Rate Annealing 142
3.7 Analyzing Quantitative Results 145
3.8 Other Types of Layers 147
3.8.1 Local Response Normalization 147
3.8.2 Spatial Pyramid Pooling 148
3.8.3 Mixed Pooling 148
3.8.4 Batch Normalization 148
3.9 Summary 149
3.10 Exercises 149
4 Caffe Library 152
4.1 Introduction 152
4.2 Installing Caffe 153
4.3 Designing Using Text Files 153
4.3.1 Providing Data 158
4.3.2 Convolution Layers 160
4.3.3 Initializing Parameters 162
4.3.4 Activation Layer 163
4.3.5 Pooling Layer 165
4.3.6 Fully Connected Layer 166
4.3.7 Dropout Layer 167
4.3.8 Classification and Loss Layers 167
4.4 Training a Network 173
4.5 Designing in Python 175
4.6 Drawing Architecture of Network 178
4.7 Training Using Python 178
4.8 Evaluating Using Python 178
4.9 Save and Restore Networks 182
4.10 Python Layer in Caffe 183
4.11 Summary 185
4.12 Exercises 185
5 Classification of Traffic Signs 188
5.1 Introduction 188
5.2 Related Work 190
5.2.1 Template Matching 191
5.2.2 Hand-Crafted Features 191
5.2.3 Sparse Coding 192
5.2.4 Discussion 192
5.2.5 ConvNets 193
5.3 Preparing Dataset 194
5.3.1 Splitting Data 195
5.3.2 Augmenting Dataset 198
5.3.3 Static Versus One-the-Fly Augmenting 206
5.3.4 Imbalanced Dataset 206
5.3.5 Preparing the GTSRB Dataset 208
5.4 Analyzing Training/Validation Curves 209
5.5 ConvNets for Classification of Traffic Signs 210
5.6 Ensemble of ConvNets 220
5.6.1 Combining Models 221
5.6.2 Training Different Models 222
5.6.3 Creating Ensemble 223
5.7 Evaluating Networks 224
5.7.1 Misclassified Images 229
5.7.2 Cross-Dataset Analysis and Transfer Learning 230
5.7.3 Stability of ConvNet 235
5.7.4 Analyzing by Visualization 238
5.8 Analyzing by Visualizing 238
5.8.1 Visualizing Sensitivity 239
5.8.2 Visualizing the Minimum Perception 240
5.8.3 Visualizing Activations 241
5.9 More Accurate ConvNet 243
5.9.1 Evaluation 245
5.9.2 Stability Against Noise 247
5.9.3 Visualization 250
5.10 Summary 251
5.11 Exercises 252
6 Detecting Traffic Signs 256
6.1 Introduction 256
6.2 ConvNet for Detecting Traffic Signs 257
6.3 Implementing Sliding Window Within the ConvNet 260
6.4 Evaluation 264
6.5 Summary 267
6.6 Exercises 267
7 Visualizing Neural Networks 268
7.1 Introduction 268
7.2 Data-Oriented Techniques 269
7.2.1 Tracking Activation 269
7.2.2 Covering Mask 269
7.2.3 Embedding 270
7.3 Gradient-Based Techniques 270
7.3.1 Activation Maximization 271
7.3.2 Activation Saliency 274
7.4 Inverting Representation 275
7.5 Summary 278
7.6 Exercises 278
A Appendix Gradient Descend 280
A.1 Momentum Gradient Descend 288
A.2 Nesterov Accelerated Gradients 291
A.3 Adaptive Gradients (Adagrad) 292
A.4 Root Mean Square Propagation (RMSProp) 293
A.5 Shuffling 294
Glossary 296
Index 300
| Erscheint lt. Verlag | 17.5.2017 |
|---|---|
| Zusatzinfo | XXIII, 282 p. 150 illus., 111 illus. in color. |
| Verlagsort | Cham |
| Sprache | englisch |
| Themenwelt | Mathematik / Informatik ► Informatik |
| Technik ► Elektrotechnik / Energietechnik | |
| Technik ► Fahrzeugbau / Schiffbau | |
| Schlagworte | Caffe library • convolutional neural network • Deep learning • pattern recognition • Traffic sign detection • Visualization |
| ISBN-10 | 3-319-57550-3 / 3319575503 |
| ISBN-13 | 978-3-319-57550-6 / 9783319575506 |
| Haben Sie eine Frage zum Produkt? |
PDF (Wasserzeichen)Größe: 14,1 MB
DRM: Digitales Wasserzeichen
Dieses eBook enthält ein digitales Wasserzeichen und ist damit für Sie personalisiert. Bei einer missbräuchlichen Weitergabe des eBooks an Dritte ist eine Rückverfolgung an die Quelle möglich.
Dateiformat: PDF (Portable Document Format)
Mit einem festen Seitenlayout eignet sich die PDF besonders für Fachbücher mit Spalten, Tabellen und Abbildungen. Eine PDF kann auf fast allen Geräten angezeigt werden, ist aber für kleine Displays (Smartphone, eReader) nur eingeschränkt geeignet.
Systemvoraussetzungen:
PC/Mac: Mit einem PC oder Mac können Sie dieses eBook lesen. Sie benötigen dafür einen PDF-Viewer - z.B. den Adobe Reader oder Adobe Digital Editions.
eReader: Dieses eBook kann mit (fast) allen eBook-Readern gelesen werden. Mit dem amazon-Kindle ist es aber nicht kompatibel.
Smartphone/Tablet: Egal ob Apple oder Android, dieses eBook können Sie lesen. Sie benötigen dafür einen PDF-Viewer - z.B. die kostenlose Adobe Digital Editions-App.
Zusätzliches Feature: Online Lesen
Dieses eBook können Sie zusätzlich zum Download auch online im Webbrowser lesen.
Buying eBooks from abroad
For tax law reasons we can sell eBooks just within Germany and Switzerland. Regrettably we cannot fulfill eBook-orders from other countries.
aus dem Bereich
