Binary Representation Learning on Visual Images: Learning to Hash for Similarity Search

✍ Scribed by Zheng Zhang

• Publisher: Springer
• Year: 2024
• Tongue: English
• Leaves: 210
• Edition: 2024
• Category: Library

✦ Synopsis

This book introduces pioneering developments in binary representation learning on visual images, a state-of-the-art data transformation methodology within the fields of machine learning and multimedia. Binary representation learning, often known as learning to hash or hashing, excels in converting high-dimensional data into compact binary codes meanwhile preserving the semantic attributes and maintaining the similarity measurements.

The book provides a comprehensive introduction to the latest research in hashing-based visual image retrieval, with a focus on binary representations. These representations are crucial in enabling fast and reliable feature extraction and similarity assessments on large-scale data. This book offers an insightful analysis of various research methodologies in binary representation learning for visual images, ranging from basis shallow hashing, advanced high-order similarity-preserving hashing, deep hashing, as well as adversarial and robust deep hashing techniques. These approaches can empower readers to proficiently grasp the fundamental principles of the traditional and state-of-the-art methods in binary representations, modeling, and learning. The theories and methodologies of binary representation learning expounded in this book will be beneficial to readers from diverse domains such as machine learning, multimedia, social network analysis, web search, information retrieval, data mining, and others.

✦ Table of Contents

Preface
Contents
Acronyms
1 Introduction
1.1 What Is Binary Representation Learning
1.2 Binary Representation Learning on Visual Images
1.2.1 Asymmetric Discrete Hashing
1.2.2 Ordinal-Preserving Hashing
1.2.3 Deep Collaborative Hashing
1.2.4 Trustworthy Deep Hashing
1.3 Evaluation Datasets and Protocols
1.3.1 Publicly Available Benchmark Datasets
1.3.2 Widely Used Evaluation Protocols
1.4 Book Structure
References
2 Scalable Supervised Asymmetric Hashing
2.1 Introduction
2.2 Scalable Supervised Asymmetric Hashing
2.2.1 Problem Definition
2.2.2 The Proposed Scalable Supervised Asymmetric Hashing
2.2.3 Optimization of SSAH
2.2.4 Out-of-Sample Extension
2.2.5 Convergence Analysis
2.3 Experiment Evaluation
2.3.1 Experimental Settings
2.3.2 Experimental Results on CIFAR-10 Tiny Images
2.3.2.1 Retrieval Performance
2.3.2.2 Parameter Sensitivity
2.3.2.3 Convergence Analysis
2.3.3 Retrieval Results on Caltech-256 for Object Retrieval
2.3.4 Retrieval Results on SUN397 for Scene Retrieval
2.3.5 Retrieval Results on ImageNet Large Dataset
2.3.6 Retrieval Results on Multi-label NUS-WIDE Dataset
2.3.7 Discussion
2.3.7.1 Comparison with Deep Hashing Methods
2.3.7.2 Asymmetric Hashing Discussion
2.4 Conclusion
References
3 Inductive Structure Consistent Hashing
3.1 Introduction
3.2 Inductive Structure Consistent Hashing
3.2.1 Notation and Problem Formulation
3.2.2 Inductive Semantic Space Construction
3.2.3 Visual to Semantic Bridging
3.2.4 Prototype Binary Code Learning
3.2.5 Objective Function
3.2.6 Optimization
3.2.7 Out-of-Sample Extension
3.3 Experiments
3.3.1 Datasets
3.3.2 Experiment Settings
3.3.2.1 Comparison Methods
3.3.2.2 Evaluation Protocols
3.3.3 Experimental Results
3.3.3.1 Accuracy Comparison with the State of the Arts
3.3.3.2 Comparison with Deep Hashing Methods
3.3.4 Further Evaluation
3.3.4.1 Ablation Study
3.3.4.2 Efficiency Comparison
3.3.4.3 Convergence and Sensitivity Analysis
3.3.4.4 Visualization
3.4 Conclusion
References
4 Probability Ordinal-Preserving Semantic Hashing
4.1 Introduction
4.2 Probability Ordinal-Preserving Hashing
4.2.1 Notations and Definitions
4.2.2 Ordinal-Preserving Hashing
4.2.3 Ordinal-Preserving Hashing: A Bayesian Perspective
4.2.3.1 Probabilistic Ordinal Similarity Preservation
4.2.3.2 Probabilistic Quantization Function
4.2.3.3 Probabilistic Semantic-Preserving Function
4.2.3.4 The Overall Objective Function
4.3 Optimization of POSH
4.4 Algorithm Extension and Theoretic Analysis
4.4.1 Nonlinear Anchor Feature Embedding
4.4.2 Out-of-Sample Extension
4.4.3 Convergence Analysis
4.4.4 Computational Complexity
4.5 Experimental Evaluation
4.5.1 Datasets
4.5.2 Baseline Methods and Implementation Details
4.5.3 Evaluation Protocols
4.5.4 Experimental Results on CIFAR-10
4.5.5 Experimental Results on MNIST
4.5.6 Experimental Results on NUS-WIDE
4.5.7 Comparison with State-of-the-Art Deep Hashing Models
4.5.8 Experimental Analysis
4.5.8.1 Parameter Sensitivity Analysis
4.5.8.2 Convergence Study
4.5.8.3 Visualization Analysis
4.6 Conclusion
References
5 Ordinal-Preserving Latent Graph Hashing
5.1 Introduction
5.2 Original-Preserving Latent Graph Hashing
5.2.1 Notations and Problem Definition
5.2.2 The General Framework of Graph-Based Discriminative Hashing
5.2.3 Discriminative Latent Hashing Space Construction
5.2.4 Feature-Level Ordinal-Correlation Preservation
5.2.5 Final Objective Function
5.3 Learning Algorithm
5.4 Out-of-Example Extension and Analysis
5.4.1 Out-of-Example Extension
5.4.2 Convergence Analysis
5.4.3 Computational Analysis
5.5 Experimental Results
5.5.1 Experimental Configuration
5.5.2 Implementation Details
5.5.3 Evaluation Protocols
5.5.4 Comparison Results
5.5.4.1 Experimental Comparison Results on CIFAR-10
5.5.4.2 Experimental Comparison Results on Caltech-256
5.5.4.3 Experimental Comparison Results on ESP-GAME
5.5.4.4 Experimental Comparison Results on MNIST
5.5.4.5 Experimental Comparison Results on ImageNet
5.5.5 Training Efficiency Study
5.6 Conclusion
References
6 Deep Collaborative Graph Hashing
6.1 Introduction
6.2 Deep Collaborative Graph Hashing
6.2.1 Notation and Problem Definition
6.2.2 Visual Image Representation Learning
6.2.3 Semantic Feature Encoder Network
6.2.4 Collective Latent Graph Embedding
6.2.5 The Final Objective Function
6.2.6 Training Strategy
6.2.6.1 Network Parameter Updating
6.2.6.2 Binary Codes Learning
6.2.7 Rationality of Collaborative Learning
6.3 Experimental Results
6.3.1 Dataset
6.3.2 Implementation Details
6.3.3 Evaluation Metrics
6.3.4 Comparison Results
6.3.4.1 Experimental Results on CIFAR-10
6.3.4.2 Experimental Results on MIRFLICKR
6.3.4.3 Experimental Results on NUS-WIDE
6.3.5 Ablation Study
6.3.6 Parameter Sensitivity
6.3.7 Visualization
6.4 Conclusion
References
7 Semantic-Aware Adversarial Training
7.1 Introduction
7.2 Semantic-Aware Adversarial Training
7.2.1 Preliminaries
7.2.2 An Overall Illustration
7.2.3 Generation of Mainstay Code
7.2.4 Semantic-Aware Adversarial Attack
7.2.5 Semantic-Aware Adversarial Training
7.2.6 Discussion on the Difference from the Related Works
7.3 Experiments
7.3.1 Experimental Setup
7.3.1.1 Datasets
7.3.1.2 Baselines
7.3.1.3 Implementation Details
7.3.1.4 Protocols
7.3.2 Adversarial Attack Results
7.3.3 Adversarial Defense Results
7.3.4 Analysis and Discussions
7.3.4.1 Attack Results in Theory
7.3.4.2 Effect of T in PGD
7.3.4.3 Analysis on Hyper-parameters
7.3.4.4 Perceptibility
7.3.4.5 Universality on Different Hashing Models
7.4 Conclusion
References
Index