Handbook Of Pattern Recognition And Computer Vision

✍ Scribed by C H Chen

• Publisher: World Scientific Publishing Company
• Year: 2020
• Tongue: English
• Leaves: 403
• Edition: 6
• Category: Library

✦ Synopsis

Written by world-renowned authors, this unique compendium presents the most updated progress in pattern recognition and computer vision (PRCV), fully reflecting the strong international research interests in the artificial intelligence arena.

Machine learning has been the key to current developments in PRCV. This comprehensive volume complements the previous five editions of the book. It places great emphasis on the use of deep learning in many aspects of PRCV applications, not readily available in other reference text.

✦ Table of Contents

CONTENTS
Dedication
Preface
PART 1: THEORY, TECHNOLOGY AND SYSTEMS
A Brief Introduction to Part 1 (by C.H. Chen)
Chapter 1.1 Optimal Statistical Classification
1 Introduction
2 Optimal Bayesian Classifier
2.1 OBC Design
3 OBC for the Discrete Model
4 OBC for the Gaussian Model
5 Multi-class Classification
5.1 Optimal Bayesian Risk Classification
6 Prior Construction
7 Optimal Bayesian Transfer Learning
8 Conclusion
References
Chapter 1.2 Deep Discriminative Feature Learning Method for Object Recognition
1. Introduction
2. Entropy-Orthogonality Loss Based Deep Discriminative Feature Learning Method
2.1. Framework
2.2. Entropy-Orthogonality Loss (EOL)
2.3. Optimization
3. Min-Max Loss Based Deep Discriminative Feature Learning Method
3.1. Framework
3.2. Min-Max Loss
3.2.1. Min-Max Loss Based on Intrinsic and Penalty Graphs
3.2.2. Min-Max Loss Based on Within-Manifold and Between-Manifold Distances
3.3. Optimization
3.3.1. Optimization for Min-Max Loss Based on Intrinsic and Penalty Graphs
3.3.2. Optimization for Min-Max Loss Based on Within-Manifold and Between-Manifold Distances
4. Experiments with Image Classification Task
4.1. Experimental Setups
4.2. Datasets
4.3. Experiments using QCNN Model
4.4. Experiments using NIN Model
4.5. Feature Visualization
5. Discussions
References
Chapter 1.3 Deep Learning Based Background Subtraction: A Systematic Survey
1. Introduction
2. Background Subtraction
2.1. Convolutional Neural Networks
2.2. Multi-scale and Cascaded CNNs
2.3. Fully CNNs
2.4. Deep CNNs
2.5. Structured CNNs
2.6. 3D CNNs
2.7. Generative Adversarial Networks (GANs)
3. Experimental Results
4. Conclusion
References
Chapter 1.4 Similarity Domains Network for Modeling Shapes and Extracting Skeletons without Large Datasets
1. Introduction
2. RelatedWork
3. Similarity Domains
4. Similarity Domains Network
5. Parametric Shape Modeling with SDN
6. Extracting the Skeleton from SDs
7. Experiments
7.1. Parametric Shape Modeling with SDs
7.2. Skeleton Extraction From the SDs
8. Conclusion
Acknowledgement
References
Chapter 1.5 On Curvelet-Based Texture Features for Pattern Classification (Reprinted from Chapter 1.7 of 5th HBPRCV)
1. Introduction
2. The Method of Curvelet Transform
3. Curvelet-based Texture Features
4. A Sample Application Problem
5. Summary and Discussion
Appendix
References
Chapter 1.6 An Overview of Efficient Deep Learning on Embedded Systems
1. Introduction
2. Overview of Deep Neural Networks
2.1. Convolutional Neural Networks
2.2. Computational Costs of Networks
3. Hardware for DNN Processing
3.1. Microprocessors
3.2. DSPs
3.3. Embedded GPUs
3.4. FPGAs
3.5. ASICs
4. The Methods for Efficient DNN Inference
4.1. Reduce Number of Operations and Model Size
4.1.1. Quantization
4.1.2. Network Pruning
4.1.3. Compact Network Architectures
4.2. Optimize Network Structure
4.3. Winograd Transform and Fast Fourier Transform
5. Summary
References
Chapter 1.7 Random Forest for Dissimilarity-Based Multi-View Learning
1. Introduction
2. Random Forest Dissimilarity
2.1. Random Forest
2.2. Using Random Forest for measuring dissimilarities
3. The Dissimilarity Representation for Multi-view Learning
3.1. The dissimilarity space
3.2. Using dissimilarity spaces for multi-view learning
4. Combining Views with Weighted Combinations
4.1. Static combination
4.2. Dynamic combination
4.2.1. Generation of the pool of classifiers
4.2.2. Evaluation and selection of the best classifier
5. Experiments
5.1. Experimental protocol
5.2. Results and discussion
6. Conclusion
Acknowledgement
References
Chapter 1.8 A Review of Image Colourisation
1. Introduction
2. Colourisation by Reference Image
3. Colourisation by Scribbles
4. Colourisation by Deep Learning
5. Other Related Work
6. Conclusion
References
Chapter 1.9 Recent Progress of Deep learning for Speech Recognition
1. Introduction
2. End-to-End Models
2.1. Connectionist Temporal Classification
2.2. RNN Transducer
2.3. Attention-based Encoder-Decoder
2.4. Practical Issues
2.4.1. Tokenization
2.4.2. Language Model Integration
2.4.3. Context Modeling
3. Robustness
3.1. Knowledge Transfer with Teacher-Student Learning
3.2. Adversarial Learning
3.3. Speech Separation
4. Summary and Future Directions
5. Acknowledgement
References
PART 2: APPLICATIONS
A Brief Introduction to Part 2 (by C.H. Chen)
Chapter 2.1 Machine Learning in Remote Sensing
1. Introduction
2. The Traditional Processing Chain of PolSAR Image Analysis
3. Holistic Feature Extraction and Model Training
3.1. Random Forests
3.1.1. RF-based Feature Learning
3.1.2. Batch Processing for Random Forests
3.1.3. Stacking of Random Forests
3.2. Deep Convolutional Networks
4. Conclusion
References
Chapter 2.2 Hyperspectral and Spatially Adaptive Unmixing for Analytical Reconstruction of Fraction Surfaces from Data with Corrupt Pixels
1. Introduction
2. Spatially adaptive hyperspectral unmixing based on analytical 2D surfaces
2.1. Sum of anisotropic 2-D Gaussians for analytical reconstruction of fraction surfaces
2.2. Reconstruction of fraction surfaces from noisy data with a high percentage of corrupted pixels
3. Evaluation and results
4. Conclusions
References
Chapter 2.3 Image Processing for Sea Ice Parameter Identification from Visual Images
1. Introduction
2. Ice Pixel Detection
3. Ice Floe Identification
3.1. Ice Boundary based Segmentation
3.2. Ice Shape Enhancement
4. A Case Study and Its Application
4.1. MIZ Image Processing
4.1.1. Marginal Ice Pixel Extraction
4.1.2. Marginal Ice Boundary Detection
4.1.3. Marginal Ice Shape Enhancement and Final Image Processing Result
4.2. Digital Ice Field Generation
5. Discussions and Further Work
5.1. Ice Pixel Detection
5.2. Ice Boundary Detection
5.3. Ice Field Generation
References
Chapter 2.4 Applications of Deep Learning to Brain Segmentation and Labeling of MRI Brain Structures
1. Introduction
2. Methods
2.1. Hardware and software
2.1.1. Hardware
2.1.2. Software
2.2. Brain segmentation
2.2.1. Neural network architecture
2.2.2. Ground truth data for training and testing
2.2.3. Training the CNN
2.2.4. Image pre- and post-processing
2.2.5. Measures for evaluating the brain mask predictions
2.3. Structural edge detection
2.3.1. Neural network architecture
2.3.2. Ground truth data for training and testing
2.3.3. Training the CNN
2.3.4. Edge-enhanced longitudinal same-subject scan registration
2.3.5. Tests of the CNN edge-predictions used in volume change
2.3.6. Voxel-based analysis in template space
2.3.7. Statistical evaluations
3. Results
3.1. Brain segmentation
3.1.1. Model generalization
3.1.2. Model consistency
3.1.3. Resource efficiency
3.2. Edge labeling
3.2.1. Characteristics of CNN edge labels
3.2.2. Characteristics of longitudinal registration incorporating CNN labels
3.2.3. Statistical power computations
4. Discussion
4.1. Brain segmentation
4.1.1. Limitations of our brain segmentation
4.1.2. Future work
4.2. Brain edge labeling in longitudinal registration
5. Conclusion
References
Chapter 2.5 Automatic Segmentation of IVUS Images Based on Temporal Texture Analysis
1. Introduction
2. The Data Base
3. The Proposed Method
4. Implementation and Results
5. Concluding Remarks
References
Chapter 2.6 Deep Learning for Historical Document Analysis
1. State of the Art
1.1. Automated Document Image Analysis
1.2. Deep Learning
1.3. Synthetic Image Generation
1.4. Digital Humanities
2. Cross-Depicted Motif Categorization
2.1. Classification System
2.2. Problem Definition
2.3. Experiments
3. Towards large Historical Document Datasets with Historical Image Synthesis
3.1. Task
3.2. Data Pre-processing
3.3. Model Architecture and Training
3.4. Results
3.5. Future Work
References
Chapter 2.7 Signature Verification via Graph-Based Methods
1. Introduction
2. From Signature Images to Graphs to Verifications
3. Graph Edit Distance and its Approximations
3.1. Graph Edit Distance
3.2. Bipartite Graph Edit Distance
3.3. Hausdorff Edit Distance
3.4. Normalization of Graph Edit Distance
4. Experimental Evaluation
4.1. Experimental Setup
4.1.1. Datasets
4.1.2. Evaluation Metric
4.2. Empirical Results
5. Conclusions and Recent Work
References
Chapter 2.8 Cellular Neural Network for Seismic Pattern Recognition
1. Introduction
1. Introduction
2. Cellular Neural Network
2.1. Structure of Cellular Neural Network
2.2. Discrete-time Cellular Neural Network
2.3. Design of Feedback Template with Linear Neighboring
2.4. Stability
2.5. Design of DT-CNN for Associative Memories
3. Pattern Recognition Using DT-CNN Associative Memory
4. Experiments
4.1. Preprocessing on Seismic Data
4.2. Experiment 1: Experiment on Simulated Seismic Patterns
4.2.1. Experiment on Simulated Seismic Patterns
4.2.2. Comparison with Hopfield Associative Memory
4.3. Experiment 2: Experiment on Simulated Seismic Images
5. Conclusions
Acknowledgements
References
Chapter 2.9 Incorporating Facial Attributes in Cross-modal Face Verification and Synthesis
1. Introduction
2. Attribute-Guided Face Verification
2.1. Center loss
2.2. Proposed loss function
2.2.1. Network structure
2.2.2. Attribute-centered loss
2.2.3. A special case and connection to the data fusion
2.3. Implementation details
2.3.1. Network structure
2.3.2. Data description
2.3.3. Network training
2.4. Evaluation
2.4.1. Experiment setup
2.4.2. Experimental results
3. Attribute-guided sketch-to-photo synthesis
3.1. Conditional generative adversarial networks (cGANs)
3.1.1. Training procedure
3.2. CycleGAN
3.2.1. Architecture
3.3. Conditional CycleGAN (cCycleGAN)
3.4. Architecture
3.5. Training procedure
3.6. Experimental results
3.6.1. Datasets
3.6.2. Results on FERET and WVU multi-modal
3.6.3. Results on CelebA and synthesized sketches
3.6.4. Evaluation of synthesized photos with a face verifier
4. Discussion
References
Chapter 2.10 Connected and Autonomous Vehicles in the Deep Learning Era: A Case Study on Computer-Guided Steering
1. Introduction
2. Related Work: AI Applications in CAVs
3. Relevant Issues
4. A Case study: Our Proposed Approach
5. Experiment Setup
5.1. Dataset
5.2. Data Preprocessing
5.3. Vehicle-assisted Image Sharing
5.4. Evaluation Metrics
5.5. Baseline Networks
5.6. Implementation and Hyperparameter Tuning
6. Analysis and Results
7. Concluding Remarks
Acknowledgement
References
Index