Pattern Recognition and Computer Vision: 5th Chinese Conference, PRCV 2022, Shenzhen, China, November 4–7, 2022, Proceedings, Part I (Lecture Notes in Computer Science, 13534)

The 4-volume set LNCS 13534, 13535, 13536 and 13537 constitutes the refereed proceedings of the 5th Chinese Conference on Pattern Recognition and Computer Vision, PRCV 2022, held in Shenzhen, China, in November 2022. The 233 full papers presented were carefully reviewed and selected from 564 submissions. The papers have been organized in the following topical sections: Theories and Feature Extraction; Machine learning, Multimedia and Multimodal; Optimization and Neural Network and Deep Learning; Biomedical Image Processing and Analysis; Pattern Classification and Clustering; 3D Computer Vision and Reconstruction, Robots and Autonomous Driving; Recognition, Remote Sensing; Vision Analysis and Understanding; Image Processing and Low-level Vision; Object Detection, Segmentation and Tracking. Preface Organization Contents – Part I Theories and Feature Extraction Architecture Colorization via Self-supervised Learning and Instance Segmentation 1 Introduction 2 Related Work 2.1 Architecture Colorization Method 2.2 Natural Image Colorizing Methods 2.3 Nature Image Generation Methods 3 Methods 3.1 Datasets 3.2 Instance Segmentation Predictive Networks 3.3 Colorizing Module 3.4 Fusion Model 4 Experiments 4.1 Datasets 4.2 Instance Semantic Segmentation 4.3 Model Analysis 4.4 Fusion Function 5 Limitations and Discussion 6 Conclusion and Future Work References Dual-Rank Attention Module for Fine-Grained Vehicle Model Recognition 1 Introduction 2 Related Works 2.1 Weakly-Supervised Method 2.2 Researches Related to Attention Mechanism 2.3 Researches Related to STN 3 Methodology 3.1 The LTRA Module 3.2 The HFCA Module 4 Experimental Results and Analysis 4.1 Datasets and Experimental Environments 4.2 Implementation Details 4.3 Experimental Results and Analysis 5 Conclusion References Multi-view Geometry Distillation for Cloth-Changing Person ReID 1 Introduction 2 Related Work 3 Method 3.1 3D Grouping Geometry Graph Convolution Network 3.2 3D-Guided Appearance Learning 3.3 Multi-view Interactive Learning 3.4 Uncertainty-Aware Loss Function 4 Experiment 4.1 Comparison with State-of-the-Art 4.2 Comparison with Existing 3D Cloth-Changing ReID Method 4.3 Ablation Study 5 Conclusions References Triplet Ratio Loss for Robust Person Re-identification 1 Introduction 2 Related Work 3 Method 3.1 Triplet Loss 3.2 Triplet Ratio Loss 4 Experiment 4.1 Implementation Details 4.2 Impact of the Hyper-Parameter 4.3 Triplet Ratio Loss vs. Triplet Loss 4.4 Comparison with State-of-the-Art 5 Conclusion References .26em plus .1em minus .1emTFAtrack: Temporal Feature Aggregation for UAV Tracking and a Unified Benchmark 1 Introduction 2 Related Work 2.1 Tracking by Detection 2.2 Temporal-Based Tracking 2.3 Attention Mechanism 3 Overview 3.1 Baseline Tracker 3.2 Backbone 3.3 Convolutional Block Attention Module 3.4 Temporal Feature Aggregation for Tracking 4 T2UAV Benchmark 4.1 Large-Scale Data Collection 4.2 High-Quality Annotation 5 Experiments 5.1 Evaluated Algorithms 5.2 Evaluation Metrics 5.3 Overall Performance 6 Conclusion References Correlated Matching and Structure Learning for Unsupervised Domain Adaptation 1 Introduction 2 Method 2.1 Formulation 2.2 Optimization 3 Experimental Results 3.1 Datasets and Parameter Setting 3.2 Results 3.3 Parameters Sensitivity and Convergence Analysis 4 Conclusion References Rider Re-identification Based on Pyramid Attention 1 Introduction 2 Related Work 2.1 Rider Re-identification 2.2 Attention Model 3 Proposed Method 3.1 Pyramid Attention Network 3.2 Attention Model 3.3 Loss Function 4 Experiments 4.1 Datasets and Implementation Details 4.2 Ablation Study 4.3 Comparison with the State-of-the-Art 4.4 Attention Visualization 5 Conclusion References Temporal Correlation-Diversity Representations for Video-Based Person Re-Identification 1 Introduction 2 Related Work 3 The Proposed Method 3.1 Overall Architecture 3.2 Temporal-Guided Frame Feature Enhancement 3.3 Temporal Feature Integration 3.4 Loss Function and Optimization 4 Experiments 4.1 Datasets and Evaluation Protocol 4.2 Implementation Details 4.3 Ablation Study 4.4 Comparison with State-of-the-Art Methods 4.5 Visualization Analysis 5 Conclusions References FIMF Score-CAM: Score-CAM Based Visual Explanations via Fast Integrating Multiple Features of Local Space for Deep Networks 1 Introduction 2 Related Works 3 Our Method 4 Experiments 5 Completeness of Class Saliency Maps and Model Performance Evaluation 6 Perturbation Analysis 7 Localization Evaluation 8 Comparison of Model Calculation Efficiency 9 Conclusion References Learning Adaptive Progressive Representation for Group Re-identification 1 Introduction 2 Related Work 2.1 Person Re-identification 2.2 Transformer 2.3 Group Re-identification 3 Method 3.1 APLN 3.2 Loss Functions 4 Experiment 4.1 Datasets 4.2 Implementation Detail 4.3 Group Re-identification 4.4 Ablation Study 5 Conclusion References General High-Pass Convolution: A Novel Convolutional Layer for Image Manipulation Detection 1 Introduction 2 Related Works 2.1 Research History 2.2 Bayar Convolution 3 Proposed Method 3.1 High-Pass Filter 3.2 General High-Pass Convolution 4 Experiments and Evaluation 4.1 Automatic Dataset Generation 4.2 JPEG Compression 4.3 Experimental Results 5 Conclusions References Machine Learning, Multimedia and Multimodal Thangka Mural Line Drawing Based on Dense and Dual-Residual Architecture 1 Introduction 2 Related Work 2.1 Traditional Non-learning Edge Detectors 2.2 Classic Learning-Based Edge Detection Methods 2.3 Deep Learning Methods 3 Methodology 3.1 Dense and Dual-Residual Architecture (DDR) 3.2 2-Phase Loss Function Strategy 4 Experimental Results and Analysis 4.1 Dataset and Experiment Setup 4.2 Performance on BIPED Testset 4.3 Performance on Thangka Testset 5 Conclusion References Self-supervised Adaptive Kernel Nonnegative Matrix Factorization 1 Introduction 2 RBF Kernel Nonnegative Matrix Factorization 3 The Proposed Method 3.1 Self-supervised Adaptive Kernel Nonnegative Matrix Factorization 3.2 Image Recognition 4 Experimental Results and Analyses 4.1 Datasets and Parameters Setting 4.2 Experiment on Images Classification 4.3 Experiment of Kernel Width Robustness 5 Conclusion References Driver Behavior Decision Making Based on Multi-Action Deep Q Network in Dynamic Traffic Scenes 1 Introduction 2 Related Work 3 A Multi-Action Deep Q-Network Based Decision Framework for Driving Behavior 3.1 Multi-Action Deep Q-Network 3.2 Improved Training Methods for Autonomous Vehicles 3.3 Curiosity-Driven Exploration Strategies 4 Experiments 4.1 OpenAI GYM 4.2 CARLA Simulation Platform 5 Conclusions References Federated Twin Support Vector Machine 1 Introduction 2 Related Work 2.1 Twin Support Vector Machines 2.2 Stochastic Gradient Twin Support Vector Machines 2.3 Federated Learning 3 Federated Twin Support Vector Machines 3.1 Model Frame 3.2 Participants 3.3 Coordinator 3.4 Federated Support Vector Machine 4 Experiments 4.1 Dataset and Experiment Setup 4.2 Accuracy Experiment 4.3 Runtime Experiments 4.4 Ablation Experiment 4.5 Performance of FTSVM and FSVM 5 Conclusion References Adversarial VAE with Normalizing Flows for Multi-Dimensional Classification 1 Introduction 2 Background 2.1 Variational Autoencoders 2.2 Normalizing Flows 2.3 Adversarial Learning 3 ADVAE-Flow 3.1 Flow-Based Encoder 3.2 Decoder 3.3 Discriminator 3.4 Loss Function 4 Experiments 4.1 Experimental Setup 4.2 Analysis on Results 4.3 Sensitivity Analysis and Ablation Study 5 Conclusion References Fuzzy Twin Bounded Large Margin Distribution Machines 1 Introduction 2 Background 3 FTBLDM 3.1 Fuzzy Membership Assignment 3.2 Model Construction of FTBLDM 3.3 The Optimization and Dual Problem 4 Experiments 4.1 Experiments on Artificial Datasets 4.2 Experiments on UCI Datasets 4.3 Experiments on Noisy Datasets 5 Conclusion References Harnessing Multi-Semantic Hypergraph for Few-Shot Learning 1 Introduction 2 Related Work 3 Method 3.1 Preliminary: Graph-Based FSL 3.2 Few-Shot Multi-Semantic Hypergraph 3.3 Multi-Semantic Hypergraph Neural Network 3.4 Objective 4 Experiments 4.1 Datasets 4.2 Experimental Setup 4.3 Comparison Results 4.4 Ablation Studies 5 Conclusion References Deep Relevant Feature Focusing for Out-of-Distribution Generalization 1 Introduction 2 Related Work 2.1 Domain Generalization (DG) 2.2 Feature Decorrelation 3 Neural Activation Network (NacNet) 3.1 Original Network 3.2 Threshold Activation (TA) Network 3.3 Training Details 4 Experiments 4.1 Datasets 4.2 Comparisons and Setting 4.3 Multi-source Setting 4.4 Ablation 4.5 Single-Source Setting 5 Conclusion References Attributes Based Visible-Infrared Person Re-identification 1 Introduction 2 Related Work 3 Method 3.1 Architecture Overview 3.2 Attributes Re-Weighting Module (RW) 3.3 Attributes-Guided Attention Module (AA) 3.4 Identity-Guided Attention Module (IA) 3.5 Attention-Align Mechanism (ALG) 3.6 Optimization 4 Experiments 4.1 Datasets and Evaluation Metrics 4.2 Implementation Details 4.3 Comparison with State-of-the-Art Methods 4.4 Ablation Study 4.5 Other Analysis 5 Conclusions References A Real-Time Polyp Detection Framework for Colonoscopy Video 1 Introduction 2 Methods 2.1 Feature Aggregation Network Based on CNN and Transformer 2.2 Temporal Information Fusion Module 3 Experiments 3.1 Experimental Datasets 3.2 Evaluation Metrics 3.3 Comparative Experiment of One-Stage Object Detection Algorithms 3.4 Ablation Experiment 3.5 Comparative Experiment with Recent Research 3.6 Comparative Experiments Under Different IOU Thresholds 4 Conclusion References Dunhuang Mural Line Drawing Based on Bi-Dexined Network and Adaptive Weight Learning 1 Introduction 2 Related Works 3 Bi-Dexined Network for Edge Detection 3.1 Network Architecture 3.2 Adaptive Weight Processing Module 3.3 Loss Functions 4 Experimental Results and Analysis 4.1 Dataset Descriptions and Experiment Setup 4.2 Performance on Public TestSet 4.3 Performance on Mural TestSet 5 Conclusion References Attention-Based Fusion of Directed Rotation Graphs for Skeleton-Based Dynamic Hand Gesture Recognition 1 Introduction 2 Method 2.1 Double-Stream Directed Rotation Graph Feature 2.2 Attention-Based Double-Stream Fusion Framework 3 Experiments 3.1 Experimental Setup 3.2 Ablation Studies 3.3 Comparison with State-of-the-Arts Methods 4 Conclusion References SteelyGAN: Semantic Unsupervised Symbolic Music Genre Transfer 1 Introduction 2 Related Work 3 SteelyGAN 3.1 The Architecture of SteelyGAN 3.2 Loss Functions 4 Experiments and Analysis 4.1 Free MIDI Library Dataset 4.2 Genre Classification 4.3 Genre Transfer 4.4 Objective Evaluation of Generated Music 4.5 Subjective Evaluation of Generated Music 5 Conclusion and Discussion References Self-supervised Learning for Sketch-Based 3D Shape Retrieval 1 Introduction 2 Related Work 2.1 Sketch Based 3D Shape Retrieval 2.2 Self-supervised Learning 3 The Proposed Approach 3.1 Multi-views Generation 3.2 Architecture 4 Experiments 4.1 Datasets 4.2 Implementation Details 4.3 Evaluation Metric 4.4 Results 5 Conclusion References Preference-Aware Modality Representation and Fusion for Micro-video Recommendation 1 Introduction 2 Related Work 3 Methodology 3.1 Preference-Aware Modality Representation Layer 3.2 Dynamic Modality Fusion Layer 4 Experiments 4.1 Datasets 4.2 Experimental Settings 4.3 Performance Comparison 4.4 Ablation Study 5 Conclusion References Multi-intent Compatible Transformer Network for Recommendation 1 Introduction 2 Related Work 3 Multi-intent Compatible Transformer Network 3.1 Input and Initialization 3.2 Intent Disentangling 3.3 Multi-intent Compatible Aggregation 3.4 Matching and Model Optimization 4 Experiment 4.1 Experimental Setting 4.2 Performance Comparison 5 Conclusion References OpenMedIA: Open-Source Medical Image Analysis Toolbox and Benchmark Under Heterogeneous AI Computing Platforms 1 Introduction 2 Algorithms 2.1 Medical Image Classification 2.2 Medical Image Segmentation 2.3 Weakly Supervised Image Localisation 2.4 Medical Image Detection 3 Benchmarks and Results 3.1 Datasets 3.2 Metrics 3.3 Evaluation on Image Classification 3.4 Evaluation on Image Segmentation 3.5 Evaluation on Weakly Supervised Image Localisation 3.6 Evaluation on Image Detection 3.7 Evaluation on Time Efficiency 4 Conclusions and Future Works References CLIP Meets Video Captioning: Concept-Aware Representation Learning Does Matter 1 Introduction 2 Related Work 3 On INP vs. CLIP for Video Captioning 3.1 Overview of Video Captioning 3.2 A Transformer Baseline for Video Captioning 3.3 Experimental Setup 3.4 Analysis 4 Approach 4.1 Formulation of Concept Detection 4.2 Dual Concept Detection (DCD) 4.3 Experimental Setup 4.4 Comparison with State-of-the-Art Methods 4.5 Ablation Study 5 Conclusions References Attention-Guided Multi-modal and Multi-scale Fusion for Multispectral Pedestrian Detection 1 Introduction 2 Related Work 2.1 General Object Detection 2.2 Multi-modal Fusion 2.3 Multi-scale Fusion 3 Methodology 3.1 Multi-modal and Multi-scale Sparse Sampling 3.2 Attention-Guided Fine-Grained Fusion 4 Experiments 4.1 Datasets 4.2 Parameter Setting 4.3 Evaluation Metrics 4.4 Results Analysis 5 Conclusion References XPNet: Cross-Domain Prototypical Network for Zero-Shot Sketch-Based Image Retrieval 1 Introduction 2 Related Work 2.1 Zero-Shot SBIR 2.2 Prototype and Cross-Domain Retrieval 2.3 Metric Learning for SBIR and Adversarial Generation 3 Methodology 3.1 Problem Formulation 3.2 Cross-Domain Prototype Network 3.3 Hard Triplet Generator 3.4 Optimization Objectives 3.5 Episode Training 3.6 Zero-shot SBIR and Few-shot Sketch Classification 4 Experiments 4.1 Datasets and Experimental Settings 4.2 Results 4.3 Ablation Study 5 Conclusion References A High-Order Tensor Completion Algorithm Based on Fully-Connected Tensor Network Weighted Optimization 1 Introduction 1.1 Nuclear Norm Approximation 1.2 Low-rank Tensor Decomposition 2 Preliminaries 2.1 Notations 2.2 Fully-Connected Tensor Network Decomposition 3 Fully Connected Tensor Network Weighted Optization 4 Experimental 4.1 Synthetic Data Experiments 4.2 Real Data Experiments 5 Conclusions References Momentum Distillation Improves Multimodal Sentiment Analysis 1 Introduction 2 Related Work 2.1 Multimodal Sentiment Analysis 2.2 Multimodal Sarcasm Detection 3 Method 3.1 Encoders Module 3.2 Multimodal Interaction Module 3.3 Multimodal Fusion Module and Classification Layer 3.4 Momentum Distillation 3.5 Model Variants 4 Experiments 4.1 Datasets and Implementation Details 4.2 Experimental Results and Analysis 4.3 Ablation Study 5 Conclusions References Synthesizing Counterfactual Samples for Overcoming Moment Biases in Temporal Video Grounding 1 Introduction 2 Related Work 2.1 Temporal Video Grounding 2.2 Moment Biases 2.3 Counterfactual Sample Synthesis 3 Methodology 3.1 Problem Formulation 3.2 Preliminaries 3.3 Synthesizing Counterfactual Samples 3.4 Training Process 4 Experiments 4.1 Datasets and Metric 4.2 Implementation Details 4.3 Comparisons 4.4 Ablation Studies and Analyze 4.5 Visualizations 5 Conclusion References Multi-grained Cascade Interaction Network for Temporal Activity Localization via Language 1 Introduction 2 Related Work 3 Proposed Method 3.1 Video Encoder 3.2 Query Encoder 3.3 Multi-grained Cascade Interaction Network 3.4 Attention Based Temporal Location Regression 3.5 Training 4 Experiments 4.1 Datasets 4.2 Implementation Details 4.3 Ablation Study 4.4 Performance Comparison with Other Representative Methods 5 Conclusion References Part-Based Multi-Scale Attention Network for Text-Based Person Search 1 Introduction 2 Related Works 2.1 Person Re-identification 2.2 Text-Based Person Search 3 Our Approach 3.1 Multi-scale Visual Representation Learning 3.2 Multi-scale Textual Representation Learning 3.3 Multi-scale Feature Matching 4 Experiments 4.1 Experimental Setup 4.2 Comparison with State-of-the-Art Methods on CUHK-PEDES 4.3 Ablation Studies 5 Conclusion References Deliberate Multi-Attention Network for Image Captioning 1 Introduction 2 Multi-modal Deliberate Multi-Attention Network 2.1 Language Module 2.2 Attention Module 2.3 Sentinel and Generation Module 2.4 Training Strategy 3 Experiments 3.1 Dataset and Implementation Details 3.2 Quantitative Analysis 3.3 Qualitative Analysis 4 Conclusions References CTFusion: Convolutions Integrate with Transformers for Multi-modal Image Fusion 1 Introduction 2 Related Work 2.1 Convolutional and Attention Mechanisms 2.2 Vision Transformer 3 Method 3.1 Early Convolutions Block 3.2 Multi-path Transformer Fusion Block 3.3 Loss Function 4 Experiment 4.1 Experimental Settings 4.2 Infrared and Visible Image Fusion 4.3 MRI and PET Image Fusion 4.4 Ablation Study 5 Conclusion References Heterogeneous Graph-Based Finger Trimodal Fusion 1 Introduction 2 Finger Trimodal Bonding Fusion Based on Weighted Graph 2.1 Inter-modal Connection 2.2 Intra-modal Connection 3 Experiment 3.1 Dataset and Experimental Settings 3.2 The Selection of the K Value and Intra-modal Connection Method 3.3 Contrastive Experiment of Different Graph Fusion Methods 4 Conclusion References Disentangled OCR: A More Granular Information for ``Text''-to-Image Retrieval 1 Introduction 2 Related Work 2.1 OCR Related Cross-Modal Tasks 2.2 Cross-Modal Retrieval 3 Methods 3.1 Text OCR Detection 3.2 Methods to Perform Retrieval with Disentangled OCR 4 The TextCaps-OCR Dataset 4.1 Origin and Storage Structure 4.2 Comparison with Other Datasets 5 Experiment 5.1 Performance of Text OCR Detection 5.2 Ablations 5.3 Comparison to the State of the Art 6 Conclusion References III Optimization and Neural Network and Deep Learning Cloth-Aware Center Cluster Loss for Cloth-Changing Person Re-identification 1 Introduction 2 Related Work 3 Method 3.1 Framework Details 3.2 Cloth-Aware Center Cluster Loss 3.3 Optimization 4 Experiments 4.1 Datasets and Evaluation Protocol 4.2 Implementation Details 4.3 Comparison with State-of-the-Art Methods 4.4 Ablation Study 4.5 Visualization and Analysis 5 Conclusion References Efficient Channel Pruning via Architecture-Guided Search Space Shrinking 1 Introduction 2 Motivation 3 Channel Pruning Method 4 Experiments 4.1 Implemented Details 4.2 Results on ImageNet 4.3 Results on COCO 4.4 Ablation 5 Conclusion References EFG-Net: A Unified Framework for Estimating Eye Gaze and Face Gaze Simultaneously 1 Introduction 2 Related Work 2.1 Appearance-Based Gaze Estimation Framework 2.2 Gaze Features Communication 3 Method 3.1 Framework 3.2 Feature Extraction 3.3 Feature Communication 3.4 Predicting Heads and Loss Function 4 Experiments 4.1 Datasets 4.2 Data Pre-processing 4.3 Implementation Details 4.4 Comparison with Appearance-based Methods 4.5 Ablation Study 4.6 Robustness Analysis 4.7 Qualitative Results 4.8 Discussion 5 Conclusion References Local Point Matching Network for Stabilized Crowd Counting and Localization 1 Introduction 2 Related Work 2.1 Crowd Counting 2.2 Crowd Localization 2.3 Set-Prediction-Based Detector 3 Methods 3.1 The End-to-End Network 3.2 Local Matching Strategy 3.3 Design Details of LPMN 4 Experiments 4.1 Implementation Details and Metrics 4.2 Crowd Counting 4.3 Crowd Localization 4.4 Ablation Study 5 Conclusion A Supplementary A.1 Counting Metrics A.2 Analysis for the Distribution of Positive Samples A.3 Analysis for Hyperparameters A.4 Discussion on Label Assignment References Discriminative Distillation to Reduce Class Confusion in Continual Learning 1 Introduction 2 Related Work 3 Method 3.1 Overview of the Proposed Framework 3.2 Expert Classifier 3.3 Knowledge Distillation 4 Experiments 4.1 Experimental Settings 4.2 Effectiveness Evaluation 4.3 Ablation Study 5 Conclusions References Enhancing Transferability of Adversarial Examples with Spatial Momentum 1 Introduction 2 Related Work 3 Methodology 3.1 Spatial Momentum Iterative Attack 3.2 The Difference with Existing Attacks 4 Experiments 4.1 Experimental Settings 4.2 Impact of Hyper-Parameter n 4.3 Ablation Study 4.4 Comparisons with State-of-the-Art Attacks 4.5 Performances Combined with Other Methods 4.6 Ensemble-Based Attacks 5 Conclusion References AIA: Attention in Attention Within Collaborate Domains 1 Introduction and Related Works 2 Methods 2.1 One Dimension Frequency Channel Attention 2.2 Joint Coordinate Attention 3 Experiments 3.1 Datasets and Initializations 3.2 Performances 3.3 Ablation Studies 4 Conclusion References Infrared and Near-Infrared Image Generation via Content Consistency and Style Adversarial Learning 1 Introduction 2 Related Works 2.1 IR/NIR Image Generation 2.2 Neural Style Transfer 3 Method 3.1 Framework Overview 3.2 Baseline Revisit 3.3 Fine-Grained Content Consistency Learning 3.4 Style Adversarial Learning 3.5 Model Optimization 4 Experiment 4.1 Implementation Details 4.2 Results of NIR Image Generation on RANUS Dataset 4.3 Results of IR Image Generation on LLVIP Dataset 4.4 Ablation Study 5 Conclusion References Adaptive Open Set Recognition with Multi-modal Joint Metric Learning 1 Introduction 2 Related Work 2.1 Deep Metric Learning 2.2 Open Set Recognition 3 Multi-modal Open Set Recognition 3.1 Motivation and Novelty 3.2 Entropy-Based Adaptive Weight Fusion 3.3 Joint Metric Learning in the Logit Space 3.4 Open Set Recognition by Scaled Fusion Logits 4 Experiments 4.1 Dataset and Preprocessing 4.2 Experimental Setup 4.3 Comparison to the State-of-the-Art Methods 4.4 Ablation Studies 5 Conclusion References Prior-Guided Multi-scale Fusion Transformer for Face Attribute Recognition 1 Introduction 2 Related Work 2.1 Face Attribute Recognition 2.2 Graph Convolution Network 2.3 Vision Transformer 3 Approaches 3.1 ARMM: Attribute Residual Mapping Module 3.2 MFM: Multi-scale Fusion Module 4 Experiments 4.1 Dataset 4.2 Evaluation Metrics 4.3 Implementation Detail 4.4 Comparison to the State-of-the-Arts 4.5 Ablution Study 4.6 Qualitative Evaluation 4.7 Quantitative Evaluation 5 Conclusion References KITPose: Keypoint-Interactive Transformer for Animal Pose Estimation 1 Introduction 2 Related Work 2.1 Human Pose Estimation 2.2 Animal Pose Estimation 3 The Proposed Method 3.1 FCN-Based Feature Extraction 3.2 Keypoint-Interactive Transformer 3.3 Trainable Joint Weights 4 Evaluation 4.1 AP10K Pose Estimation 4.2 ATRW Keypoint Detection 4.3 Ablation Study 5 Conclusion References Few-Shot Object Detection via Understanding Convolution and Attention 1 Introduction 2 Related Work 2.1 General Object Detection 2.2 Few-Shot Learning 2.3 Few-Shot Object Detection 3 Method 3.1 Problem Definition 3.2 Framework 3.3 Hybrid Dilation Convolution 3.4 Support Features Dynamic Fusion 4 Experiments 4.1 Experimental Setting 4.2 Experimental Results 4.3 Ablation Study 5 Conclusion References Every Corporation Owns Its Structure: Corporate Credit Rating via Graph Neural Networks 1 Introduction 2 Related Works 3 The Proposed Method: CCR-GNN 3.1 Problem Formulation 3.2 Architecture Overview 3.3 Corporation to Graph Layer(C2GL) 3.4 Graph Feature Interaction Layer (GFIL) 3.5 Credit Rating Layer (CRL) 4 Experiments 4.1 Data Set and Pre-processing 4.2 Comparison with Baseline Methods 5 Conclusions References Unsupervised Image Translation with GAN Prior 1 Introduction 2 Related Work 2.1 Generative Adversarial Network (GAN) 2.2 Image Translation 2.3 GAN Prior 3 Method 3.1 Two-stage Framework 3.2 Network Architecture 3.3 Loss Functions 4 Experiments 4.1 Implementation Details 4.2 Qualitative Evaluation 4.3 Quantitative Evaluation 5 Conclusion References An Adaptive PCA-Like Asynchronously Deep Reservoir Computing for Modeling Data-Driven Soft Sensors 1 Introduction 2 Adaptive PCA-Like ADRC 2.1 Introduction of ADRC 2.2 PCA-Like ADRC 2.3 Adaptive Algorithm for PCA-Like ADRC 3 Experiment 4 Conclusion References Double Recursive Sparse Self-attention Based Crowd Counting in the Cluttered Background 1 Introduction 2 Related Work 3 Method 3.1 Overall Network Structure 3.2 Double Recursive Sparse Self-attention Module 3.3 Feature Extraction Module Based on Feature Pyramid Transformer 3.4 Training Detail 4 Experiment 4.1 Evaluation Criteria 4.2 Benchmark Datasets and Ground-Truth Data 4.3 Ablation Experiment 4.4 Background False Detection Improvement 4.5 Overall Performance Evaluation 5 Conclusion References BiTMulV: Bidirectional-Decoding Based Transformer with Multi-view Visual Representation 1 Introduction 2 Related Work 2.1 Image Captioning 2.2 NMT 3 Methodology 3.1 Model Architecture 3.2 Multi-view Visual Representation Based Encoder 3.3 Bidirectional Decoder 3.4 Training and Objectives 4 Experiments 4.1 Experimental Setup 4.2 Ablation Study 4.3 Quantitative Analysis 4.4 Qualitative Analysis and Visualization 5 Conclusions References An Improved Lightweight Network Based on MobileNetV3 for Palmprint Recognition 1 Introduction 2 Related Work 2.1 Traditional Palmprint Recognition Methods 2.2 Large CNN-Based Palmprint Recognition Method 2.3 Lightweight CNN-Based Palmprint Recognition Method 3 Proposed Approach 3.1 Lightweight Network Based on MobileNetV3 3.2 Improvements of MobileNetV3 4 Experiments and Analysis 4.1 Analysis of Compression Factor Optimization 4.2 Analysis on Improving the Structure of Multi-layer Activation Learning Networks 4.3 Analysis of Channel Probability Calculation Function 5 Conclusion References A Radar HRRP Target Recognition Method Based on Conditional Wasserstein VAEGAN and 1-D CNN 1 Introduction 2 Related Work 2.1 HRRP Target Recognition Method 2.2 Variational Autoencoder (VAE) 2.3 Wasserstein Generative Adversarial Networks (WGAN) 3 Methods 3.1 Conditional Wasserstein Variational Autoencoders with GAN (CWVAEGAN) 3.2 CWVAEGAN-1DCNN 4 Experiment and Analysis 4.1 Dataset 4.2 Metrics 4.3 Experimental Procedure 4.4 Experiment Results and Analysis 5 Conclusions References Partial Least Square Regression via Three-Factor SVD-Type Manifold Optimization for EEG Decoding 1 Introduction 2 Method 2.1 Review of Partial Least Squares Regression 2.2 PLSRbiGr: PLSR via Optimization on Bi-Grassmann Manifold 3 Experiments and Results 3.1 EEG Decoding 3.2 Effects of Riemannian Preconditioning 4 Discussion and Conclusion References Single Deterministic Neural Network with Hierarchical Gaussian Mixture Model for Uncertainty Quantification 1 Introduction 2 Related Works 3 Methods 3.1 Hierarchical GMM for Classifier 3.2 Natural Gradient Updates for GMM Parameters 3.3 The Proposed Algorithm 4 Simulation Studies 4.1 Toy Example 4.2 Benchmark Problems 5 Conclusion References Exploring Masked Image Modeling for Face Anti-spoofing 1 Introduction 2 Related Work 2.1 Face Anti-spoofing 2.2 Masked Image Modeling 3 Methodology 3.1 Intuition and Motivation 3.2 The Proposed Method 4 Experiments 4.1 Datasets and Evaluation Metrics 4.2 Implementation Details 4.3 Experimental Results and Analysis 4.4 Visualization 5 Conclusion References Author Index