Nonlinear Dynamics of Time Delay Systems: Methods and Applications

Preface
Organization
Contents – Part I
Contents – Part II
Computer Vision
MARS: An Instance-Aware, Modular and Realistic Simulator for Autonomous Driving
	1 Introduction
	2 Method
		2.1 Scene Representation
		2.2 Compositional Rendering
		2.3 Towards Realistic Rendering
		2.4 Optimization
	3 Experiments
		3.1 Photorealistic Rendering
		3.2 Instance-Wise Editing
		3.3 The Blessing of Modular Design
		3.4 Ablation Results
	4 Conclusion
	References
Concealed Object Segmentation with Hierarchical Coherence Modeling
	1 Introduction
	2 Related Works
	3 Methodology
		3.1 Concealed Feature Encoder
		3.2 Hierarchical Coherence Modeling
		3.3 Reversible Re-calibration Decoder
		3.4 Loss Functions
	4 Experiments
		4.1 Camouflaged Object Segmentation
		4.2 Polyp Image Segmentation
		4.3 Transparent Object Detection
		4.4 Ablation Study and Further Analysis
	5 Conclusions
	References
ViT-MPI: Vision Transformer Multiplane Images for Surgical Single-View View Synthesis
	1 Introduction
	2 Related Work
	3 Approach
		3.1 MPI Representation and Rendering Using Single-View
		3.2 Vision Transformer Backbone
	4 Experiments
		4.1 Dataset and Training Objective
		4.2 Comparison
		4.3 Ablation Studies
	5 Conclusion
	References
Dual-Domain Network for Restoring Images from Under-Display Cameras
	1 Introduction
	2 Related Work
		2.1 UDC Image Enhancement
		2.2 Retinex-Based Visual Models
		2.3 Frequency Domain Enhancement Methods
	3 Methodology
		3.1 Retinex-Based Image Decomposition
		3.2 Amplitude-Phase Mutual Guided Block
		3.3 Multi-scale Hybrid Dilation Convolution Block
		3.4 Training Loss
	4 Experiments
		4.1 Datasets and Training Procedure
		4.2 Results
		4.3 Ablation Study
	5 Conclusion
	References
Sliding Window Detection and Distance-Based Matching for Tracking on Gigapixel Images
	1 Introduction
	2 Related Work
		2.1 Region Proposal Convolutional Detectors
		2.2 Multi-object Tracking
	3 Methodology
		3.1 Preliminary
		3.2 Multi-sclae Sliding Window
		3.3 Distance-Based Tracking Strategy
	4 Experiment
		4.1 Datasets and Metrics
		4.2 Implementation Details
		4.3 Comparsion with Other Method
		4.4 Ablation Study
	5 Conclusion
	References
Robust Self-contact Detection Based on Keypoint Condition and ControlNet-Based Augmentation
	1 Introduction
	2 Related Work
		2.1 Contact Detection
		2.2 3D Body Dataset
		2.3 Generation for Generalization
	3 Method
		3.1 Definition
		3.2 Self-contact Data Generation
		3.3 Method: RSCD
	4 Experiments
		4.1 Implementation Details
		4.2 Results
		4.3 Evaluation on Keypoint Condition
	5 Ablation Studies
		5.1 Generation to Generalization
	6 Limitations
	7 Conclusions
	References
Explicit Composition of Neural Radiance Fields by Learning an Occlusion Field
	1 Introduction
	2 Related Work
	3 Method
		3.1 Preliminaries
		3.2 Occlusion Field
		3.3 Composition Rendering Equation
		3.4 Model Training
	4 Experiment
		4.1 Data Preparation
		4.2 Qualitative Evaluation
		4.3 Comparison
		4.4 Ablation Studies
	5 Conclusion
	References
LEAD: LiDAR Extender for Autonomous Driving
	1 Introduction
	2 Related Work
		2.1 Depth Estimation
		2.2 Depth Completion
	3 Method
		3.1 Overview
		3.2 Self-supervised Teacher Network (STN)
		3.3 Propagative Probabilistic Generator (PPG)
		3.4 Probabilistic Derivation and Composition
		3.5 Network Training
	4 Experiment
		4.1 Hardware and Evaluation Dataset
		4.2 Qualitative Results
		4.3 Quantitative Results
	5 Conclusion
	References
Fast Hierarchical Depth Super-Resolution via Guided Attention
	1 Introduction
	2 Related Work
		2.1 Traditional Depth SR Method
		2.2 Deep Learning-Based Depth SR Method
	3 Proposed Method
		3.1 Framework Overview
		3.2 Structures of Main and Side Branches
		3.3 Guided Attention
	4 Experiments and Results
		4.1 Implementation Details
		4.2 Comparison with SOTA Methods
		4.3 Attention Analysis
		4.4 Running Time
	5 Conclusion
	References
A Hybrid Approach for Segmenting Non-ideal Iris Images Using CGAN and Geometry Constraints
	1 Introduction
	2 Related Work
	3 Methods
		3.1 Iris Segmentation Using CGAN
		3.2 Objective
		3.3 Network Architecture
		3.4 Refinement Using Geometry Constraints
	4 Experiments
		4.1 Datasets
		4.2 Evaluation Metrics
		4.3 Implementation Details
		4.4 Comparision with the State of the Art
	5 Conclusion
	References
3D-B2U: Self-supervised Fluorescent Image Sequences Denoising
	1 Introduction
	2 Related Work
		2.1 Image Denoising
		2.2 Self-supervised Denoising on Fluorescence Images
	3 Method
		3.1 Motivation
		3.2 Framework
		3.3 3D Global Masker
	4 Experimental Results
		4.1 Experimental Settings
		4.2 Synthetic Data Denoising
		4.3 Real Data Denoising
	5 Ablation
	6 Conclusion
	References
Equivariant Indoor Illumination Map Estimation from a Single Image
	1 Introduction
	2 Related Work
	3 Method
		3.1 Overall Architecture
		3.2 Point Cloud Generation
		3.3 Equivariant Illumination Estimation
		3.4 Loss
	4 Experiment
		4.1 Datasets and Preprocessing
		4.2 Comparison in Quality and Quantity
		4.3 Ablation Study
		4.4 Application on AR
	5 Conclusion
	References
Weakly-Supervised Grounding for VQA with Dual Visual-Linguistic Interaction
	1 Introduction
	2 Related Work
		2.1 Visual Question Answering
		2.2 VQA Grounding
	3 Method
		3.1 Architecture
		3.2 Language-Based Visual Decoder
		3.3 Pseudo Grounding Refinement
		3.4 Training Objectives
	4 Experiments
		4.1 Training Details
		4.2 Ablation Study
	5 Comparison with State-of-the-Arts
	6 Conclusion
	References
STU3: Multi-organ CT Medical Image Segmentation Model Based on Transformer and UNet
	1 Introduction
	2 Method
		2.1 Architecture Overview
		2.2 SWTB
		2.3 Local Part
		2.4 Global Part and RFFF
		2.5 Glfb
	3 Experiments
		3.1 Implementation Details
		3.2 Evaluating Metric
		3.3 Experimental Result
	4 Conclusion
	References
Integrating Human Parsing and Pose Network for Human Action Recognition
	1 Introduction
	2 Related Work
		2.1 Human Action Recognition
		2.2 Human Parsing
	3 IPP-Net
		3.1 Human Pose Learning
		3.2 Human Parsing Learning
		3.3 Integration
	4 Experiments
		4.1 Datasets
		4.2 Implementation Details
		4.3 Comparison with Related Methods
		4.4 Ablation Study
	5 Conclusions
	References
Lightweight Rolling Shutter Image Restoration Network Based on Undistorted Flow
	1 Introduction
	2 Related Works
	3 Approach
		3.1 Problem Formulation
		3.2 Architecture Overview
		3.3 Extract Feature Pyramid
		3.4 Generate Bidirectional Optical Flow
		3.5 Time Factor and Undistortion Flow
		3.6 Losses
	4 Experiment
		4.1 Datasets
		4.2 Evaluation Strategies
		4.3 Comparison with SOTA Methods
		4.4 Ablation Studies
	5 Conclusion
	References
An Efficient Graph Transformer Network for Video-Based Human Mesh Reconstruction
	1 Introduction
	2 Related Work
		2.1 Human Mesh Reconstruction
		2.2 Graph Convolutional Networks
	3 Method
		3.1 Overview of EGTR
		3.2 Temporal Redundancy Removal
		3.3 Spatial-Temporal Fusion
		3.4 Multi-branch Integration
		3.5 Loss Function
	4 Experiments
		4.1 Experimental Settings
		4.2 Comparison with State-of-the-Art Methods
		4.3 Qualitative Evaluation
		4.4 Ablation Analysis
	5 Conclusions
	References
Multi-scale Transformer with Decoder for Image Quality Assessment
	1 Introduction
	2 Related Work
		2.1 Traditional Blind IQA
		2.2 CNN-Based Blind IQA
		2.3 ViT-Based Blind IQA
	3 Proposed Method
		3.1 Overall Architecture
		3.2 Multi-scale Input
		3.3 Attention Aggregation in Transformer Encoder
		3.4 Decoder and Quality Score Generation
	4 Experimental Results
		4.1 Datasets and Evaluation Protocols
		4.2 Implementation Details
		4.3 Comparison of Quality Evaluation Results
		4.4 Ablation Experiment
	5 Conclusion
	References
Low-Light Image Enhancement via Unsupervised Learning
	1 Introduction
	2 Related Work
		2.1 Traditional Method
		2.2 Deep Learning Method
	3 Proposed Method
		3.1 Vision Transformer Discriminator
		3.2 Loss Function
	4 Experiment
		4.1 Datasets and Implementation Details.
		4.2 Performance Evaluation
		4.3 Ablation Study
	5 Conclusion
	References
GLCANet: Context Attention for Infrared Small Target Detection
	1 Introduction
	2 Related Work
		2.1 Infrared Small Target Detection
		2.2 Global Contextual Information
		2.3 Local Contextual Information
	3 Method
		3.1 Overall Architecture
		3.2 Global Context Extraction Module
		3.3 Local Context Attention Module
	4 Experiment
		4.1 Datasets and Evaluation Metrics
		4.2 Implementation Details
		4.3 Quantitative Results
		4.4 Visual Results
		4.5 Ablation Study
	5 Conclusion
	References
Fast Point Cloud Registration for Urban Scenes via Pillar-Point Representation
	1 Introduction
	2 Related Works
	3 Methodology
		3.1 Pillar-Point Based Feature Extractor
		3.2 Hierarchical Matching Scheme
		3.3 Pose Estimator
		3.4 Loss Function
	4 Experiments
		4.1 Implementation Details
		4.2 Results on KITTI and NuScenes
		4.3 Generalization to Apollo Southbay Dataset
		4.4 Ablation Study
	5 Conclusion
	References
PMPI: Patch-Based Multiplane Images for Real-Time Rendering of Neural Radiance Fields
	1 Introduction
	2 Related Work
		2.1 Neural Implicit Representations for View Synthesis
		2.2 Representations of Multiplane Images
	3 Representation of Patch-Based Multiplane Images
	4 Training of Adaptive PMPI
		4.1 Initialization of Patches
		4.2 Learning Appearance and Geometry
		4.3 Updating Structure of PMPI
	5 Real-Time Rendering with Customized CUDA Kernel
	6 Experiments
		6.1 Training Details of Our Method
		6.2 Comparisons
		6.3 Ablation Study
	7 Conclusions
	References
EFPNet: Effective Fusion Pyramid Network for Tiny Person Detection in UAV Images
	1 Introduction
	2 Related Works
	3 The Proposed Method
		3.1 Overview of the Proposed EFPNet
		3.2 Multi-dimensional Attention Module
		3.3 Effective Feature Fusion Module
	4 Experiments and Results
		4.1 Experimental Settings
		4.2 Visualization Analysis
		4.3 Comparison to State-of-the-arts
		4.4 Ablation Experiments
	5 Conclusions
	References
End-to-End Object-Level Contrastive Pretraining for Detection via Semantic-Aware Localization
	1 Introduction
	2 Related Work
		2.1 Self-supervised Learning for Classification
		2.2 Self-supervised Learning for Object Detection
		2.3 Selective Object COntrastive Learning (SoCo)
	3 Method
		3.1 Semantic-Aware Localization
		3.2 Center-Suppressed Sampling
		3.3 Multiple Crop
	4 Experiment
		4.1 Pretraining Settings
		4.2 Finetuning Settings
		4.3 Training Time and Space Cost
		4.4 Performance Comparison
		4.5 Ablation Study
	5 Conclusion
	References
PointerNet with Local and Global Contexts for Natural Language Moment Localization
	1 Introduction
	2 Our Approach
		2.1 Word Recurrence for Multimodal Clip Features
		2.2 Clip Recurrence and Global Video Context
		2.3 PointerNet-Based Moment Localization
		2.4 Training
	3 Experiments
		3.1 Ablation Study
		3.2 Comparison with State-of-the-art
	4 Conclusion
	References
Self-supervised Meta Auxiliary Learning for Actor and Action Video Segmentation from Natural Language
	1 Introduction
	2 Our Approach
		2.1 Feature Extraction Network
		2.2 Primary Segmentation Network
		2.3 Auxiliary Language Reconstruction Network
		2.4 Meta Auxiliary Training
	3 Experiments
		3.1 Ablation Study
		3.2 Comparison with the State-of-the-art
	4 Conclusion
	References
RsMmFormer: Multimodal Transformer Using Multiscale Self-attention for Remote Sensing Image Classification
	1 Introduction
	2 Related Works
	3 Methodology
		3.1 Overall Architecture
		3.2 Multi-scale Multi-Head Self-Attention (MSMHSA)
	4 Experiments and Analysis
		4.1 Experimental Setup
		4.2 Quantitative Analysis
		4.3 Ablation Study
		4.4 Visualization
	5 Conclusion
	References
Fashion Label Relation Networks for Attribute Recognition
	1 Introduction
	2 Related Work
	3 Methodology
		3.1 Framework
		3.2 Learning Objective
	4 Experiment
		4.1 Benchmark Dataset and Evaluation Metrics
		4.2 Implementation Details
		4.3 Comparison with the Benchmarking Methods
		4.4 Performance Analysis
		4.5 Results on Clothes Retrieval Datasets
	5 Conclusion
	References
A Modified Fuzzy Markov Random Field Incorporating Multiple Features for Liver Tumor Segmentation
	1 Introduction
	2 Proposed Method
		2.1 Feature Extraction
		2.2 The Modified Fuzzy Markov Random Field Model
		2.3 Post-processing
	3 Experimental Results
		3.1 Dataset and Evaluation Measures
		3.2 Experimental Details
		3.3 Results and Discussion
	4 Conclusion
	References
Weakly Supervised Optical Remote Sensing Salient Object Detection Based on Adaptive Discriminative Region Suppression
	1 Introduction
	2 Related Work
		2.1 Salient Object Detection in Optical Remote Sensing Images
		2.2 Weakly Supervised Salient Object Detection
	3 Method
		3.1 Overall Structure
		3.2 Local Activation Suppression Module
		3.3 Adaptive Fusion Module
		3.4 Multi-filter Directive Network
	4 Experiments
		4.1 Implementation Details
		4.2 Datasets and Evaluation Metrics
		4.3 Comparison with State-of-the-Art Methods
		4.4 Ablation Studies
	5 Conclusion
	References
SPCTNet: A Series-Parallel CNN and Transformer Network for 3D Medical Image Segmentation
	1 Introduction
	2 Method
		2.1 Architecture
		2.2 Transformer Block
		2.3 Multi-scale Feature Fusion (MSF)
		2.4 Loss Function
	3 Experiments
		3.1 Datasets
		3.2 Experimental Settings and Evaluation Metrics
		3.3 Ablation Study
		3.4 Quantitative Evaluation
		3.5 Qualitative Evaluation
	4 Conclusion
	References
LANet: A Single Stage Lane Detector with Lightweight Attention
	1 Introduction
	2 Related Work
	3 Proposed Method
		3.1 Anchor Representation
		3.2 Backbone and Feature Pooling
		3.3 Anchor-Frame Attention
		3.4 Prediction
		3.5 Non-maximum Supression (NMS)
		3.6 Model Training
	4 Experiments
		4.1 Tusimple
		4.2 CULane
		4.3 Ablation Study
	5 Conclusion
	References
Visible and NIR Image Fusion Algorithm Based on Information Complementarity
	1 Introduction
	2 Related Work
	3 The Proposed Algorithm
		3.1 Two-Scale Guided Image Decomposition
		3.2 Inter-band Information Complementarity Map Estimation
		3.3 Information Complementary Weight Model
	4 Experimental Results
		4.1 Objective Comparison
		4.2 Subject Comparison
	5 Conclusion
	References
Data Mining
End-to-End Optimization of Quantization-Based Structure Learning and Interventional Next-Item Recommendation
	1 Introduction
	2 Related Works
	3 Methodology
		3.1 Problem Formulation
		3.2 Causal Structure Learning in Recommendation
		3.3 Quantization-Based Structure Learning
		3.4 End-to-End Optimization
	4 Experiments
		4.1 Experimental Setup
		4.2 Experiments Results
		4.3 Parameter Analysis
		4.4 Ablation Study
		4.5 OOD Generalization
	5 Conclusion
	A  Appendix
		A.1  Experimental Setup
	References
Multi-trends Enhanced Dynamic Micro-video Recommendation
	1 Introduction
	2 Our Approach
		2.1 Problem Formulation
		2.2 Overview
		2.3 Implicit User Network
		2.4 Multi-trend Routing
		2.5 Multi-level Time Attention Mechanism
		2.6 Prediction
	3 Experiments
		3.1 Dataset
		3.2 Implementation Details
		3.3 Evaluation Metrics
		3.4 Competitors
		3.5 Results
		3.6 Recommendation Diversity
	4 Conclusion
	References
Parameters Efficient Fine-Tuning for Long-Tailed Sequential Recommendation
	1 Introduction
	2 Related Work
		2.1 Long-Tail
		2.2 Gradient Surgery
	3 Method
		3.1 Preliminaries
		3.2 Gradient Aggregation
		3.3 Plugin Network
	4 Experiments
		4.1 Experiment Settings
		4.2 Experiments and Results
	5 Conclusion
	A  Pseudo Code
	B  Experiments
		B.1  Experiments Settings
		B.2 Results
	References
Heterogeneous Link Prediction via Mutual Information Maximization Between Node Pairs
	1 Introduction
	2 Related Work
		2.1 Heterogeneous Graph Embedding
		2.2 Link Prediction
	3 Preliminaries
	4 Methodology
	5 Experiments
		5.1 Experiment Settings
		5.2 Results on Link Prediction
		5.3 Ablation Studies
		5.4 Parameters Experiments
	6 Conclusion
	References
Explainability, Understandability, and Verifiability of AI
ADAPT: Action-Aware Driving Caption Transformer
	1 Introduction
	2 Method
		2.1 Overview
		2.2 Model Design
	3 Experiment
		3.1 Implementation Details
		3.2 Main Results
		3.3 Accelerate the Inference Process
	4 Conclusion
	References
Structural Recognition of Handwritten Chinese Characters Using a Modified Part Capsule Auto-encoder
	1 Introduction
	2 Related Work
	3 Methodology
		3.1 Overall Framework
		3.2 Primitive Extraction
		3.3 Stroke Aggregation
		3.4 Character Recognition
	4 Experiments
		4.1 Implementation Details
		4.2 Datasets
		4.3 Effects of Primitive and Stroke Extraction
		4.4 Performance of Chinese Character Recognition
		4.5 Ablation Study
	5 Conclusion
	References
Natural Language Processing
Sequential Style Consistency Learning for Domain-Generalizable Text Recognition
	1 Introduction
	2 Method
		2.1 Base Network
		2.2 Sequential Style Consistency Learning
		2.3 Training and Inference
	3 Experiments
		3.1 Dataset and Metrics
		3.2 Implementation Details
		3.3 Model Selection
		3.4 Comparison Baselines
		3.5 Comparison Results
		3.6 Ablation Studies
	4 Conclusion
	References
MusicGAIL: A Generative Adversarial Imitation Learning Approach for Music Generation
	1 Introduction
	2 Musical Data Representation and Preprocessing
		2.1 Interactive Duet Model
		2.2 Pitch and Duration Encodings
	3 Generative Adversarial Imitation Learning
		3.1 MusicGAIL Framework
		3.2 Melodic Generator
		3.3 Style-Discriminator
	4 Experimental Results
		4.1 Dataset
		4.2 The Training Process of MusicGAIL
		4.3 Comparison and Evaluation
	5 Conclusion
	References
Unsupervised Traditional Chinese Herb Mention Normalization via Robustness-Promotion Oriented Self-supervised Training
	1 Introduction
	2 Approach
		2.1 Overall Result
		2.2 Ablation Study
		2.3 Analysis
	3 Conclusion
	References
Feature Fusion Gate: Improving Transformer Classifier Performance with Controlled Noise
	1 Introduction
	2 Related Works
		2.1 The Gating Mechanism of LSTM
		2.2 Mixup
	3 Proposed Methodology
	4 Experiments and Results
		4.1 Benchmark Datasets and Models
		4.2 Baselines
		4.3 Experimental Settings
		4.4 Overall Results
	5 Conclusion and Future Work
	References
Multi-round Dialogue State Tracking by Object-Entity Alignment in Visual Dialog
	1 Introduction
	2 Related Work
	3 Model
	4 Experiment
	5 Conclusions
	References
Multi-modal Dialogue State Tracking for Playing GuessWhich Game
	1 Introduction
	2 Related Work
	3 Model
	4 Experiment and Evaluation
	5 Conclusion
	References
Diagnosis Then Aggregation: An Adaptive Ensemble Strategy for Keyphrase Extraction
	1 Introduction
	2 Related Work
	3 Problem Definition
	4 Adaptive Ensemble Strategy via Cognitive Diagnosis
		4.1 Cognitive Diagnose for Keyphrase Extraction Algorithms
		4.2 Adaptive Ensemble Strategy
	5 Experiments
		5.1 Experimental Setup
		5.2 Evaluation Metrics
		5.3 Experimental Results
	6 Conclusion
	References
Author Index