Pattern Recognition and Computer Vision 5th Chinese Conference, PRCV 2022 Shenzhen, China, November 4–7, 2022 Proceedings, Part III

Preface
Organization
Contents – Part III
3D Computer Vision and Reconstruction, Robots and Autonomous Driving
Locally Geometry-Aware Improvements of LOP for Efficient Skeleton Extraction
	1 Introduction
	2 Related Work
	3 The Improved LOP
		3.1 Overview
		3.2 Bilateral Filter Based Weighting
		3.3 Adaptive Radius
	4 Experimental Results
	5 Conclusions
	References
Spherical Transformer: Adapting Spherical Signal to Convolutional Networks
	1 Introduction
	2 Related Work
	3 The Proposed Approach
		3.1 Spherical Sampling
		3.2 Spherical Transformer Module
		3.3 Network Architecture
	4 Experiments
		4.1 Spherical MNIST
		4.2 3D Object Classification
		4.3 Spherical Image Semantic Segmentation
	5 Conclusion
	References
Waterfall-Net: Waterfall Feature Aggregation for Point Cloud Semantic Segmentation
	1 Introduction
	2 Related Work
	3 Waterfall-Net
		3.1 Cascaded Sub-Networks Encoder
		3.2 Learn to Upsample
	4 Experiments
		4.1 Analysis of Waterfall-Net Architecture
		4.2 Results and Visualization
	5 Conclusion
	References
Sparse LiDAR and Binocular Stereo Fusion Network for 3D Object Detection
	1 Introduction
	2 Related Work
		2.1 LiDAR-Based 3D Object Detection
		2.2 Monocular-Based 3D Object Detection
		2.3 Stereo-Based 3D Object Detection
		2.4 Multi-modal 3D Object Detection
	3 Proposed Method
		3.1 Feature Extraction
		3.2 Attention Fusion
		3.3 3D Object Information Regression Prediction
		3.4 Implementation Details
	4 Experiments
		4.1 KITTI Dataset
		4.2 Evaluation Metrics
		4.3 Main Results
		4.4 Ablation Study
	5 Conclusion
	References
Full Head Performance Capture Using Multi-scale Mesh Propagation
	1 Introduction
	2 Related Work
	3 Template Fitting Based Dynamic Full Head Performance Capture
		3.1 Per-frame Multi-view Scan Reconstruction
		3.2 Template Warping
		3.3 Multi-scale Mesh Propagation
		3.4 Ear Reconstruction
	4 Experimental Results
	5 Conclusion
	References
Learning Cross-Domain Features for Domain Generalization on Point Clouds
	1 Introduction
	2 Related Work
		2.1 Deep Learning on Point Cloud
		2.2 Unsupervised Domain Adaptation
		2.3 Domain Generalization
	3 Network Architecture
		3.1 Point Set Mask
		3.2 Cross-Domain Mixup
		3.3 Hierarchical Feature Alignment
		3.4 Overall Loss
	4 Experiments and Results
		4.1 Dataset
		4.2 Comparative Methods
		4.3 Implementation Details
		4.4 Results
		4.5 Ablation Study
	5 Conclusion
	References
Unsupervised Pre-training for 3D Object Detection with Transformer
	1 Introduction
	2 Related Work
		2.1 Object Detection on Point Clouds
		2.2 Unsupervised Representation Learning on Point Clouds
	3 UP3DETR
		3.1 Pre-training
		3.2 Fine-tuning
	4 Experiments
		4.1 ScanNetV2 Object Detection
		4.2 SUN RGB-D Object Detection
		4.3 Ablations
	5 Conclusion
	References
Global Patch Cross-Attention for Point Cloud Analysis
	1 Introduction
	2 Related Work
		2.1 Multi-view Based and Voxelized Methods
		2.2 Point Based Method
		2.3 Attention Based Method
	3 Method
		3.1 Overview
		3.2 Global Patch Construction
		3.3 Local-Global Feature Aggregation
	4 Experiment
		4.1 Point Cloud Analysis
		4.2 Analysis of GPCAN
	5 Conclusion
	References
EEP-Net: Enhancing Local Neighborhood Features and Efficient Semantic Segmentation of Scale Point Clouds
	1 Introduction
	2 Related Work
		2.1 Projection-Based Methods
		2.2 Discretization-Based Methods
		2.3 Point-Based Methods
	3 EEP-Net
		3.1 Architecture of EEP Module
		3.2 Global Feature (GF)
		3.3 Architecture of EEP-Net
	4 Experiments
		4.1 Evalution on S3DIS Dataset
		4.2 Ablation Study
	5 Conclusion
	References
CARR-Net: Leveraging on Subtle Variance of Neighbors for Point Cloud Semantic Segmentation
	1 Introduction
	2 Related Work
	3 Method
		3.1 CARR Module
		3.2 CARRs
		3.3 Overall Architecture
	4 Experiments
		4.1 Set up
		4.2 Evaluation on S3DIS Dataset
		4.3 Evaluation on SemanticKITTI Dataset
		4.4 Ablation Study
	5 Conclusion
	References
3D Meteorological Radar Data Visualization with Point Cloud Completion and Poisson Surface Reconstruction
	1 Introduction
	2 Method Description
		2.1 Data Format Conversion
		2.2 Echo Models and Completion Algorithms
		2.3 Drawing 3D Surface Using Bilateral-Filtering Poisson Surface Reconstruction Algorithm (BPSR)
	3 Experiments and Analysis
		3.1 Comparison of Meteorological Radar Data Completion
		3.2 Comparison of Point Cloud Completion Experiments
	4 Conclusion
	References
JVLDLoc: A Joint Optimization of Visual-LiDAR Constraints and Direction Priors for Localization in Driving Scenario
	1 Introduction
	2 Related Work
		2.1 Visual-LiDAR SLAM
		2.2 Using Vanishing Points as Direction Constraint
	3 Notation
	4 Method
		4.1 Tracking
		4.2 Local Mapping
		4.3 Global Mapping Using Direction Priors
	5 Experiments
		5.1 Improvements over Prior Map
		5.2 Effects of Direction Priors
		5.3 Comparison to Other Methods on KITTI Odometry Dataset
		5.4 Ablation Study
	6 Conclusion
	References
A Single-Pathway Biomimetic Model for Potential Collision Prediction
	1 Introduction
	2 Related work
	3 Proposed Method
		3.1 Problem Formulation
		3.2 The Single-Pathway LGMD2
		3.3 Collision Prediction Criteria
	4 Experiments and Analyses
		4.1 Datasets and Competing Methods
		4.2 Parameter Setting
		4.3 Evaluation Metrics
		4.4 Experimental Results
	5 Conclusions
	References
PilotAttnNet: Multi-modal Attention Network for End-to-End Steering Control
	1 Introduction
	2 Related Work
		2.1 Driving Model
		2.2 Driving Dataset
	3 Method
		3.1 Spatial Information Encoding
		3.2 The End-to-End Attentional Driving Model
	4 Experiments
		4.1 Dataset Description
		4.2 Evaluation Metrics
		4.3 Results
	5 Conclusion
	References
Stochastic Navigation Command Matching for Imitation Learning of a Driving Policy
	1 Introduction
	2 Related Works
	3 Method
		3.1 Problem Formulation
		3.2 Backbone Network
		3.3 Navigation Command
		3.4 Multi-branch Architecture
		3.5 Stochastic Navigation Command Matching
		3.6 Training
	4 Experiments
		4.1 Experiment Setting
		4.2 Quantitative Comparison
		4.3 Qualitative Comparison
		4.4 Visualization Results
	5 Conclusions
	References
Recognition, Remote Sensing
Group Activity Representation Learning with Self-supervised Predictive Coding
	1 Introduction
	2 Related Work
	3 Approach
		3.1 Spatial Graph Transformer Encoder
		3.2 Temporal Causal Transformer Decoder
		3.3 Joint Learning Measure
	4 Experiments
		4.1 Datasets
		4.2 Implementation Details
		4.3 Ablations on Volleyball
		4.4 Comparison with State-of-the-Art
	5 Conclusion
	References
Skeleton-Based Action Quality Assessment via Partially Connected LSTM with Triplet Losses
	1 Introduction
	2 Related Work
		2.1 Action Quality Assessment
		2.2 Graph-Based Methods
	3 Methods
		3.1 Joints Graph and Activation Matrix
		3.2 Partially Connected Layer
		3.3 Partially Connected LSTM
		3.4 Triplet Loss
	4 Experiments
		4.1 Evaluation Datasets and Settings
		4.2 Data Preprocess
		4.3 Experimental Results and Analysis
		4.4 Complexity Analysis
	5 Conclusion
	References
Hierarchical Long-Short Transformer for Group Activity Recognition
	1 Introduction
	2 Related Work
		2.1 Group Activity Recognition
		2.2 Transformer
	3 Methodology
		3.1 Overview of HLSTrans
		3.2 Long-Short Transformer Block
		3.3 Hierarchical Structure
		3.4 Position Bias
	4 Experiments
		4.1 Datasets
		4.2 Implementation
		4.3 Comparison to Others
	5 Conclusion
	References
GNN-Based Structural Dynamics Simulation for Modular Buildings
	1 Introduction
	2 Methodology
		2.1 Graph Representation Method
		2.2 GNN Model
	3 Numerical Studies
		3.1 Numerical Examples of Three Spring-Mass Systems
		3.2 Training and Prediction Results
	4 Conclusion
	References
Semantic-Augmented Local Decision Aggregation Network for Action Recognition
	1 Introduction
	2 Proposed Approach
		2.1 LDNet
		2.2 Semantic Information Module
		2.3 Combining Semantic Information Module with LDNet
	3 Experiments
		3.1 Datasets and Implementation Details
		3.2 Ablation Study
	4 Conclusions
	References
Consensus-Guided Keyword Targeting for Video Captioning
	1 Introduction
	2 Related Work
		2.1 Video Captioning
		2.2 Video Captioning Datasets
	3 Method
		3.1 Encoder-Decoder Framework
		3.2 Consensus-Guided Loss
		3.3 Keyword Targeting Loss
		3.4 Consensus-Guided Keyword Targeting Captioning Model
	4 Experiments
		4.1 Datasets and Metrics.
		4.2 Implementation Details
		4.3 Quantitative Results
		4.4 Qualitative Results
	5 Conclusion
	References
Handwritten Mathematical Expression Recognition via GCAttention-Based Encoder and Bidirectional Mutual Learning Transformer
	1 Introduction
	2 Related Work
		2.1 Image-to-Markup
		2.2 CNN
		2.3 Global Contextual Attention
		2.4 Transformer
		2.5 Mutual Learning
	3 Methodology
		3.1 Encoder
		3.2 Decoder
		3.3 Positional Encoding
	4 Experiments
		4.1 Datasets
		4.2 Comparison with Prior Works
		4.3 Ablation Study
		4.4 The Program with GUI
	5 Conclusion
	References
Semi- and Self-supervised Learning for Scene Text Recognition with Fewer Labels
	1 Introduction
	2 Background and Related Work
		2.1 Scene Text Recognition
		2.2 Datasets
		2.3 Self-supervised Learning
	3 Method
		3.1 Architecture
		3.2 Data Augmentation
		3.3 Loss Function
		3.4 Pseudo-labeling
	4 Experiments
		4.1 Performance on Real Scene Datasets
		4.2 Performance on Synthetic Datasets
		4.3 Comparison with State-of-the-Art Models
	5 Conclusion
	References
TMCR: A Twin Matching Networks for Chinese Scene Text Retrieval
	1 Introduction
	2 Related Work
	3 Methods
		3.1 Detection Module
		3.2 Recognition Module
		3.3 Similarity Module
		3.4 Loss and Training
	4 Experiments
		4.1 Dataset and Implementation Details
		4.2 Comparisons with State-of-the-Art
		4.3 Ablation Study
		4.4 Model Generalization
	5 Conclusion
	References
Thai Scene Text Recognition with Character Combination
	1 Introduction
	2 Methodology
		2.1 Recognition Architecture
		2.2 Thai Character Combination
	3 Experimental Setting
		3.1 Thai STR Datasets
		3.2 Data Preparing
		3.3 Model Configurations and Training
		3.4 Evaluation Metric
	4 Experimental Results and Analyses
		4.1 Experimental Results
		4.2 The Effectiveness of TCC
		4.3 Failure Cases Analysis
	5 Conclusion
	References
Automatic Examination Paper Scores Calculation and Grades Analysis Based on OpenCV
	1 Introduction
	2 Methodology
		2.1 Image Acquisition
		2.2 Image Processing
		2.3 Data Processing
	3 Experimental Results and Discussions
	4 Conclusions
	References
Efficient License Plate Recognition via Parallel Position-Aware Attention
	1 Introduction
	2 Related Work
		2.1 License Plate Datasets
		2.2 License Plate Recognition
	3 Methods
		3.1 Feature Encoder
		3.2 Parallel Position-Aware Attention
		3.3 Character Decoder
		3.4 Loss Function
	4 Data Synthesis
	5 Experiments
		5.1 DataSets
		5.2 Experiment Settings
		5.3 Experimental Results
	6 Conclusions
	References
Semantic-Aware Non-local Network for Handwritten Mathematical Expression Recognition
	1 Introduction
	2 Related Works
		2.1 Grammar-Based HMER
		2.2 Encoder-Decoder Based HMER
	3 Methodology
		3.1 Non-local Neural Networks
		3.2 FastText Language Model
	4 Experiments
		4.1 Datasets
		4.2 Metrics
		4.3 Results
	5 Conclusion
	References
Math Word Problem Generation with Memory Retrieval
	1 Introduction
	2 Related Work
		2.1 Math Word Problem Generation
		2.2 Memory Retrieval for Text Generation
	3 Problem Setup
		3.1 MWPG
		3.2 Low-Resource MWPG
	4 Proposed Approach
		4.1 Overview
		4.2 Retrieval Module
		4.3 Generation Module
		4.4 Training
	5 Experiments
		5.1 Datasets
		5.2 Metrics
		5.3 Implementation Details
		5.4 Baselines
		5.5 Quantitative Results
		5.6 Qualitative Results
	6 Conclusions
	References
Traditional Mongolian Script Standard Compliance Testing Based on Deep Residual Network and Spatial Pyramid Pooling
	1 Introduction
	2 Related Work
	3 Model Architecture
		3.1 Convolutional Layers with Residual Learning
		3.2 The Spatial Pyramid Pooling Layer
	4 Experiment and Analysis
		4.1 Data and Experimental Environment
		4.2 Evaluation Metrics
		4.3 Results
	5 Conclusion
	References
FOV Recognizer: Telling the Field of View of Movie Shots
	1 Introduction
	2 Related Works
		2.1 Human Detection
		2.2 Field of View Recognition Method
	3 Movie Field of View Dataset(MFOVD)
	4 Field of View Recognition Method
	5 Experiments
		5.1 Recognition on Movie Field of View Dataset(MFOVD)
		5.2 Recognition on a Full Movie
	6 Conclusion
	References
Multi-level Temporal Relation Graph for Continuous Sign Language Recognition
	1 Introduction
	2 Related Work
		2.1 Sign Language Recognition
		2.2 Video Contexts Modeling
		2.3 Graph Convolutional Network
	3 Our Approach
		3.1 Visual Model
		3.2 Multi-level Temporal Relation Graph
		3.3 Alignment Model
	4 Results and Discussion
		4.1 Dataset
		4.2 Experimental Setup
		4.3 Comparison with SOTA Methods
		4.4 Model Validity Experiment
	5 Conclusions
	References
Beyond Vision: A Semantic Reasoning Enhanced Model for Gesture Recognition with Improved Spatiotemporal Capacity
	1 Introduction
	2 Related Work
		2.1 Temporal Information Model
		2.2 Attention Mechanism
		2.3 Semantic Information Model
	3 Method
		3.1 The Overview of the Network
		3.2 Long and Short-term Temporal Shift Module (LS-TSM)
		3.3 Spatial Attention Module
		3.4 Label Relation Module
	4 Experiment
		4.1 Datasets
		4.2 Implementation Details
		4.3 Comparision with the State of the Art
		4.4 Ablation Study
	5 Conclusion
	References
SemanticGAN: Facial Image Editing with Semantic to Realize Consistency
	1 Introduction
	2 Related Works
		2.1 Generative Adversarial Networks
		2.2 Facial Image Editing
	3 Proposed Method
		3.1 Preliminary
		3.2 Attribute-Related Fine Editing
		3.3 Attribute-Independent Optimization
	4 Experiments
		4.1 Implementation Details
		4.2 Attribute Face Editing
		4.3 Editing with SemanticGAN
		4.4 Ablation Studies
	5 Conclusion
	References
Least-Squares Estimation of Keypoint Coordinate for Human Pose Estimation
	1 Introduction
	2 Proposed Method
		2.1 Encoding
		2.2 Decoding
	3 Experiments
		3.1 Datasets and Evaluation
		3.2 Comparison with Other Methods
		3.3 Ablation Study
	4 Conclusion
	References
Joint Pixel-Level and Feature-Level Unsupervised Domain Adaptation for Surveillance Face Recognition
	1 Introduction
	2 Related Work
		2.1 Deep Face Recognition
		2.2 Unsupervised Domain Adaptation
	3 Methodology
		3.1 Training of Feature Extractor
		3.2 Training of Domain Classifier
		3.3 Training of Style Transformer
	4 Experiment
		4.1 Datasets
		4.2 Details of Training
		4.3 Ablation Experiment
		4.4 Quantity Comparison
		4.5 Comparison
	5 Conclusion
	References
Category-Oriented Adversarial Data Augmentation via Statistic Similarity for Satellite Images
	1 Introduction
	2 Related Works
		2.1 Data Augmentation
		2.2 Appearance Properties
	3 Proposed Method
		3.1 Problem Definition and Basic Solutions
		3.2 Statistic Similarity Evaluation
		3.3 Adversarial Generation Between Similar Categories
		3.4 Task of Object Detection
	4 Experimental Results
	5 Conclusion
	References
A Multi-scale Convolutional Neural Network Based on Multilevel Wavelet Decomposition for Hyperspectral Image Classification
	1 Introduction
	2 Related Works
		2.1 2D Discrete Wavelet Transform
		2.2 DenseNet
	3 Proposed Framework
	4 Experimental Results and Discussion
		4.1 HIS Datasets
		4.2 Experimental Setting
		4.3 Results and Discussion
	5 Conclusion
	References
High Spatial Resolution Remote Sensing Imagery Classification Based on Markov Random Field Model Integrating Granularity and Semantic Features
	1 Introduction
	2 Background on MRF-Based Methods
		2.1 MRF Model with Different Granularities
		2.2 MRF Model with Multilayer
	3 Proposed Method
		3.1 MRF Model
		3.2 Proposed MRF-MM Model
	4 Experimental Results
		4.1 Data
		4.2 Classification Experiment
		4.3 Test of the MRF-MM Model Parameters
	5 Conclusion
	References
Feature Difference Enhancement Fusion for Remote Sensing Image Change Detection
	1 Introduction
	2 Related Work
		2.1 Traditional Change Detection Methods
		2.2 Deep Learning Based Change Detection Methods
	3 Method
		3.1 Overall Structure of Proposed CD Architecture
		3.2 Difference Enhancement Fusion Module (DEFM)
	4 Experiments
		4.1 Experimental Setup
		4.2 Experimental Results
		4.3 Ablation Studies
	5 Conclusion
	References
WAFormer: Ship Detection in SAR Images Based on Window-Aware Swin-Transformer
	1 Introduction
	2 Related Works
		2.1 SAR Target Detection Based on Deep Learning
		2.2 Vision Transformer
	3 Method
		3.1 Motivation
		3.2 Overview
		3.3 Variable Size Window Self-attention
		3.4 WAFormer Block
	4 Experiments
		4.1 Dataset and Evaluation Metrics
		4.2 Implementation Details
		4.3 Comparison Results
		4.4 Related Configuration Adjustment
	5 Conclusion
	References
EllipseIoU: A General Metric for Aerial Object Detection
	1 Introduction
	2 Related Work
		2.1 Aerial Object Detection
		2.2 IoU-Based Metrics
	3 EllipseIoU Loss
		3.1 EllipseIoU
		3.2 EllipseIoU Loss
		3.3 Discussion on Several IoU-Based Metrics
	4 Experimental Results
		4.1 Datasets
		4.2 Results on DOTA and HRSC2016 Dataset
	5 Conclusion
	References
Transmission Tower Detection Algorithm Based on Feature-Enhanced Convolutional Network in Remote Sensing Image
	1 Introduction
	2 Dataset Production
		2.1 Data Collection
		2.2 Dataset Production
		2.3 Dataset Expansion
		2.4 YOLOv3 Algorithm Principle
		2.5 Algorithms in This Paper
	3 Experiments and Results Analysis
		3.1 Experimental Setup
		3.2 Analysis of Results
	4 Conclusion
	References
Vision Analysis and Understanding
Mining Diverse Clues with Transformers for Person Re-identification
	1 Introduction
	2 Related Work
	3 Proposed Method
		3.1 Vision Transformer as Feature Extractor
		3.2 Person ReID with Transformers
		3.3 MDCTNet Architecture
	4 Experiments
		4.1 Datasets
		4.2 Implementation Details
		4.3 Ablation Study
		4.4 Comparison with State-of-the-Arts
	5 Conclusion
	References
Mutual Learning Inspired Prediction Network for Video Anomaly Detection
	1 Introduction
	2 Mutual Learning Inspired Prediction Network
		2.1 Framework
		2.2 Boundary Perception-Based Mimicry Loss
		2.3 Self-supervised Weighted Loss
		2.4 Objective Function
		2.5 Anomaly Detection on Testing Data
	3 Experiment
		3.1 Dataset
		3.2 Evaluation Metrics
		3.3 Comparison with Existing Methods
		3.4 Running time
	4 Conclusion
	References
Weakly Supervised Video Anomaly Detection with Temporal and Abnormal Information
	1 Introduction
	2 Related Work
		2.1 Unsupervised Video Anomaly Detection
		2.2 Weakly-Supervised Video Anomaly Detection
		2.3 Multiple Instance Learning
		2.4 Pair-Based Loss in Deep Metric Learning
	3 Approach
		3.1 Temporal Strengthen Network
		3.2 Multi-positive Sample MIL
		3.3 Anomaly Samples' N-Pair Loss
	4 Experiment
		4.1 Datasets and Metrics
		4.2 Implementation Details
		4.3 Results on ShanghaiTech
		4.4 Results on UCF-Crime
		4.5 Ablation Studies
		4.6 Qualitative Analyse
	5 Conclusion
	References
Towards Class Interpretable Vision Transformer with Multi-Class-Tokens
	1 Introduction
	2 Related Work
		2.1 Vision Transformer
		2.2 Heatmap-Based Visual Interpretability
	3 Proposed Method
		3.1 Overview of Proposed Approach
		3.2 Multi-Class-Tokens and Cross Attention
		3.3 Non-parametric Scoring Function
		3.4 Heatmap Based Per-class Interpretability
	4 Experiments
		4.1 Datasets
		4.2 Implementation Details
		4.3 Comparison Results
		4.4 Ablation Study
	5 Conclusion
	References
Multimodal Violent Video Recognition Based on Mutual Distillation
	1 Introduction
	2 Related Work
		2.1 Violent Video Recognition
		2.2 Self-supervised Learning
		2.3 Knowledge Distillation
	3 Methods
		3.1 Mutual Distillation for Violent RGB Feature
		3.2 MAF-Net
	4 Experiments
		4.1 Datasets and Metrics
		4.2 Experiments on Mutual Distillation for Violent RGB Feature
		4.3 Experiments on Multimodal Feature Fusion
		4.4 Comparison with Others
	References
YFormer: A New Transformer Architecture for Video-Query Based Video Moment Retrieval
	1 Introduction
	2 Related Work
		2.1 Transformer in Computer Vision
		2.2 Video Moment Retrieval
	3 YFormer for Video Moment Retireval
		3.1 Spatio-Temporal Feature Extractor
		3.2 Semantic Relevance Matcher
		3.3 Prediction Heads
		3.4 Losses
	4 Experiments
		4.1 Experiment Setup
		4.2 Quantitative Results
		4.3 Qualitative Results
		4.4 Ablation Study
	5 Conclusion
	References
Hightlight Video Detection in Figure Skating
	1 Introduction
	2 Related Works
		2.1 Action Quality Assessment
		2.2 Figure Skating
		2.3 Temporal Action Segmentation
	3 Approach
		3.1 Overview
		3.2 Video Segmentation
		3.3 Tube Self-Attention
		3.4 Frame Scoring
	4 Experiments
		4.1 Dataset
		4.2 Implementation Details
		4.3 Results During Training
		4.4 Ablation Study
		4.5 Results in the Singles Figure Skating Competition
	5 Conclusion
	References
Memory Enhanced Spatial-Temporal Graph Convolutional Autoencoder for Human-Related Video Anomaly Detection
	1 Introduction
	2 Related Work
		2.1 Video Anomaly Detection
		2.2 Graph Convolutional Networks
		2.3 Memory Networks
	3 Method
		3.1 Preprocessing
		3.2 Network Architecture
		3.3 Loss Function
		3.4 Anomaly Detection
	4 Experiments
		4.1 Datasets
		4.2 Implementation Details
		4.3 Evaluation
		4.4 Ablation Studies
	5 Conclusions
	References
Background Suppressed and Motion Enhanced Network for Weakly Supervised Video Anomaly Detection
	1 Introduction
	2 Related Work
		2.1 Unsupervised Video Anomaly Detection
		2.2 Weakly Supervised Video Anomaly Detection
	3 Background Suppressed and Motion Enhanced Network (BSMEN)
		3.1 Motion Discrimination Sequence Extraction (MDSE)
		3.2 Background Suppressed and Motion Enhanced Module (BSMEM)
		3.3 Loss Function
	4 Experiments
		4.1 Datasets
		4.2 Evaluation Metric
		4.3 Implementation Details
		4.4 Comparison with State-of-the-Art Methods
		4.5 Ablation Studies
	5 Conclusions
	References
Dirt Detection and Segmentation Network for Autonomous Washing Robots
	1 Introduction
	2 Relate Works
	3 Method
		3.1 SVDD (Support Vector Data Description)
		3.2 Deep SVDD
		3.3 DDSN (Dirt Detection and Segmentation Network)
	4 Evaluation
		4.1 Experiment Setup
		4.2 Experimental Result on MVTecAD Dataset
		4.3 Experimental Result on Dirt Dataset
	5 Conclusion
	References
Finding Beautiful and Happy Images for Mental Health and Well-Being Applications
	1 Introduction
	2 Related Work
	3 A Beautiful Natural Image Database (BNID)
		3.1 Collecting the Images
		3.2 Collecting Beatutifulness and Happiness Scores
		3.3 Analysis of Beautifulness and Happiness Scores
	4 Beautifulness and Happiness Assessment
		4.1 Image Beautifulness Assessment
		4.2 Loss Functions
		4.3 Final Score
		4.4 Extension to Image Happiness Prediction
	5 Experiments
		5.1 Image Beautifulness Assessment Results
		5.2 Image Happiness Assessment Results
	6 Concluding Remarks
	References
Query-UAP: Query-Efficient Universal Adversarial Perturbation for Large-Scale Person Re-Identification Attack
	1 Introduction
	2 Related Work
	3 Methodology
		3.1 Problem Definition
		3.2 Loss Function
		3.3 Query-UAP Attack
	4 Experiment
		4.1 Experimental Settings
		4.2 Comparison with State of the Arts
		4.3 Ablation Study
	5 Conclusion
	References
Robust Person Re-identification with Adversarial Examples Detection and Perturbation Extraction
	1 Introduction
	2 Related Work
		2.1 Adversarial Attack
		2.2 Adversarial Defense
	3 Proposed Method
		3.1 Networks Architecture
		3.2 Adversarial Examples Generation
		3.3 Perturbation Extractor and Purification
		3.4 Adversarial Example Detector
	4 Experiments
		4.1 Experimental Settings
		4.2 Robustness of Purification
		4.3 Adversarial Detection
		4.4 Ablation Experiments
	5 Conclusion
	References
Self-supervised and Template-Enhanced Unknown-Defect Detection
	1 Introduction
	2 Related Work
	3 Approach
		3.1 Framework
		3.2 Feature Fusion Module
		3.3 Loss Function
		3.4 Defect Detection
	4 Experiments
		4.1 Dataset
		4.2 Training Parameters
		4.3 Results
	5 Conclusion
	References
JoinTW: A Joint Image-to-Image Translation and Watermarking Method
	1 Introduction
	2 Related Work
	3 Proposed Work
		3.1 Problem Statement
		3.2 Method Overview
		3.3 The Watermark Extractor
		3.4 Watermark Embedding Generator
		3.5 The Adversary Net
		3.6 Training Details
		3.7 Loss Functions
	4 Experimental Results
		4.1 Datasets
		4.2 Training Details
		4.3 Visual Quality Study of the Generated Images
		4.4 Qualities of the Extracted Watermarks
		4.5 Watermark Robustness Under Agnostic Distortions
	5 Conclusions
	References
Author Index