دانلود کتاب Concepts and Real-Time Applications of Deep Learning

فهرست مطالب :

Foreword
Preface
Contents
Part I: Concepts of Deep Learning: Recognition Systems
Emotion Recognition from Speech Using Deep Neural Network
1 Introduction
2 Background
2.1 Challenges
3 Methodology
3.1 Types of Dataset
3.1.1 Free License
Danish Emotional Database (DES)
3.1.2 Commercially Available
Chinese Language Database (CASIA)
3.1.3 Public and Free
Berlin Emotional Database
RAVDESS
3.2 Datasets
3.2.1 CREMA-D
3.2.2 The MSP-IMPROV
3.3 Feature Extraction and Feature Selection
3.3.1 Prosodic Features
Linear Predictive Coding (LPC)
Mel-Frequency Cepstral Coefficient (MFCC)
Short-Term Fourier Transform (STFT)
Principal Component Analysis (PCA)
3.4 Classification
3.4.1 Convolutional Neural Network (CNN)
3.4.2 Support Vector Machine (SVM)
3.4.3 Deep Neural Network (DNN)
3.4.4 Long Short-Term Memory (LSTM)
3.4.5 Gated Recurrent Unit (GRU)
3.4.6 Bag of Visual Words (BovW)
3.4.7 Ensemble Learning
3.5 Emotion Recognition
3.5.1 Evolution
3.5.2 Significance
3.6 Experiments and Results
3.6.1 Performance of Feature Extraction
3.6.2 Performance of Different Classification Models
3.6.3 Activation and Dropout
4 Conclusion and Future Recommendations
References
Text-Independent Speaker Recognition Using Deep Learning
1 Introduction
2 Dataset and Setup
3 Related Works
4 Our Approach
5 Mel-Frequency Cepstral Coefficients (MFCCs)
6 Feature Extraction
7 Convolutional Neural Networks
8 Model Architecture
9 Hyperparameter Tuning of CNN
10 Results
10.1 Training and Testing of Data
10.2 Comparison with Other Models
11 Conclusion and Discussion
References
A Qualitative and Quantitative Research of Machine Learning Algorithms for Gait Analysis and Recognition
1 Introduction
2 Existing Research
3 Dataset
4 Gait Energy Image
5 CNNs for Gait Recognition
6 Results
7 Conclusion
References
Emotion Recognition from Speech Signals Using Machine Learning and Deep Learning Techniques
1 Introduction
2 Related Work
3 Methodology
3.1 Dataset
3.2 Pre-processing and Feature Extraction
3.3 SVM
3.4 Multilayer Perceptron Neural Network
3.5 CNN
4 Result
5 Conclusion and Future Work
References
Micro-expression Detection Using Main Directional Maximal Differential Analysis (MDMD) Method
1 Introduction
1.1 Challenges
2 Motivation
3 Datasets Involved
3.1 SAMM
3.1.1 Experimental Setup
3.2 CAS(ME)2
3.3 SMIC
4 Related Work
4.1 Apex-Time Network (ATNet)
4.2 Micro-expression STCNN
4.3 Method Using a Recurrent Neural Network on Optical Flow Features
5 Methodology
5.1 Preprocessing
5.2 Main Directional Maximal Difference Analysis (MDMD)
5.3 Evaluation
5.3.1 Evaluation for a Single Video
5.3.2 Evaluation for the Entire Dataset
6 Results and Discussion
7 Conclusion
References
Part II: Concepts of Deep Learning: Healthcare Systems
Survival Prediction of Cancer Patient Using Machine Learning
1 Introduction: Background and Driving Forces
1.1 Introduction
1.2 Literature Survey
2 Methodology and Dataset
2.1 Dataset
2.2 Algorithms
2.2.1 Neural Network
2.2.2 Bagging
2.2.3 Boosting
3 Result and Discussion
3.1 Description of Parameters in Algorithm
3.2 Confusion Matrix
3.3 Receiver Operating Characteristic Curve
3.4 Comparative Analysis of Algorithms Used
4 Conclusion
5 Future Scope
References
Skin Lesion Segmentation Using Deep Convolutional Networks
1 Introduction
2 Previous Work
3 Convolutional Neural Networks
4 Wide Residual Networks
5 Proposed Model
5.1 Double Wide Residual Networks
5.2 HRB Block
6 Experimental Setup
6.1 Dataset
6.2 Training, Validation, and Test Sets
6.3 Data Pre-Processing
7 Results and Discussion
7.1 Results
7.2 Discussion
8 Conclusion
References
Bone Cancer Survivability Prognosis with KNN and Genetic Algorithms
1 Introduction
2 Survivability Analysis with Machine Learning for Cancer
2.1 Breast Cancer Prediction with KNN
2.2 Lung Cancer Prediction with KNN
2.3 Skin Cancer Prediction with KNN
3 Methodology
3.1 Research Parameters
4 Findings
4.1 Low Risk
4.2 Medium Risk
4.3 High Risk
5 Conclusion
References
BeamAtt: Generating Medical Diagnosis from Chest X-Rays Using Sampling-Based Intelligence
1 Introduction
2 Related Work
3 Proposed Model: BeamAtt
3.1 Overview
3.2 Image Feature Encoder
3.3 GRU-Based Decoder with Visual Attention
3.4 Generating Inference
4 Experiments and Results
4.1 Data
4.2 Implementation Details
4.3 Evaluation
5 Conclusion and Future Work
References
Part III: Real-Time Applications of Deep Learning
CNN-Based Driver Drowsiness Detection System
1 Introduction
1.1 Accurately Identifying the Time the Alarm Beeps
1.2 Quality of the Image Being Fed as the Input
1.3 Significant Change or Drop of Accuracy with or without Spectacles
1.4 Frame Drop during the Image Feed
2 Motivation
3 Dataset
4 Methodology
4.1 Predictive Task
4.2 Using Convolutional Neural Network
5 The Model Architecture
6 Machine Learning Approaches
6.1 Support Vector Machine Classifier
6.2 Random Forest
7 Comparison of Techniques
8 Conclusion
References
Forecasting Using Deep Learning Approaches
1 Introduction
1.1 Main Objectives of the Chapter
1.2 Forecasting and its Need
1.3 Forecasting Applications
2 Time Series Forecasting
2.1 Forecasting Models
2.1.1 Naïve Forecasting Model
2.1.2 Linear Regression Model
2.1.3 Exponential Smoothing Model
2.1.4 ARIMA
2.2 Problems in the Classical Models
3 Deep Learning
3.1 Deep Learning Models for Forecasting
3.2 Fully Connected Neural Network (FCNN)
3.3 Convolutional Neural Network (CNN)
3.4 Recurrent Neural Network (RNN)
3.5 Long Short-Term Memory (LSTM)
3.6 Gated Recurrent Unit (GRU)
3.7 Generative Adversarial Network (GAN)
3.8 Deep Reinforcement Learning
4 Experimental Results
5 Conclusion
References
A Low-Cost IOT and Deep Learning Enabled Precision Agriculture Support System for Indian Diverse Environment
1 Introduction
2 Related Works
3 Convolutional Neural Network
3.1 Convolutional Layers
3.2 Pooling Layers
3.3 Fully Connected Layers
3.4 Weights
4 ResNet50 and SE-Resnet50
4.1 ResNet50
4.2 SE-ResNet50
5 Methodology
6 Experimental Results
6.1 Dataset Description
6.2 Experimentation
6.3 Quantitative Results
7 Conclusion and Future Work
References
Index