Advances in Visual Computing. 17th International Symposium, ISVC 2022 San Diego, CA, USA, October 3–5, 2022 Proceedings, Part I

Preface
Organization
Keynote Talks
	Towards Scaling Up GANs
	Sensible Machine Learning for Geometry
	Designing Augmented Reality for the Future of Work
	The Future of Visual Computing via Foundation Models (Banquet Keynote Talk)
	3D Reconstruction: Leveraging Synthetic Data for Lightweight Reconstruction
	Human-AI Interaction in Visual Analytics: Designing for the “Two Black Boxes” Problem
	Contents – Part I
	Contents – Part II
Deep Learning I
Unsupervised Structure-Consistent Image-to-Image Translation
	1 Introduction
	2 Background and Related Work
	3 Method
		3.1 Encoder
		3.2 Generator
		3.3 Structure and Texture Disentanglement
		3.4 Objective Function
	4 Experiments
		4.1 Comparison to State-of-the-Art
	5 Applications
		5.1 Addressing Bias in Training Datasets
		5.2 Training Datasets for Semantic Segmentation of Satellite Images
	6 Discussion and Limitations
	7 Conclusions
	References
Learning Representations for Masked Facial Recovery
	1 Introduction
	2 Relevant Works
	3 Method
		3.1 Baseline Model
		3.2 Unmasking Model
		3.3 Datasets
		3.4 Implementation Details
	4 Experimental Results
	5 Conclusions
	References
Deep Learning Based Shrimp Classification
	1 Introduction
	2 Related Work
	3 Proposed Approach
		3.1 Acquisition
		3.2 Preprocessing
		3.3 Classification
	4 Experimental Results
	5 Conclusions
	References
Gait Emotion Recognition Using a Bi-modal Deep Neural Network
	1 Introduction
	2 Related Works
	3 Methodology
	4 Experimental Results
	5 Conclusion and Future Work
	References
Attacking Frequency Information with Enhanced Adversarial Networks to Generate Adversarial Samples
	1 Introduction
	2 Related Work
		2.1 Adversarial Samples
		2.2 Black-Box Attacks
		2.3 Frequency Features and Attacks
	3 Our Frequency Attack Approach
		3.1 Separate High and Low Frequency Information
		3.2 Dual Discriminators Support Attack
		3.3 Frequency Attack Framework
		3.4 Network Architecture
		3.5 Loss Function
	4 Experiments
		4.1 Evaluation Metric
		4.2 Ablation Study
		4.3 Transferability of FAF
		4.4 Attack Under Defenses
	5 Conclusion
	References
Visualization
Explainable Interactive Projections for Image Data
	1 Introduction
	2 Related Work
		2.1 Interactive Dimensionality Reduction
		2.2 Semantic Interaction
		2.3 Explainability in Deep Learning
	3 Tasks
		3.1 Define Custom Similarities Based on Prior Knowledge
		3.2 Link Human and Machine Defined Similarities
	4 Workflow and Methodology
		4.1 Initial State
		4.2 Interactions and Inverse Projection
		4.3 Visual Explanations
	5 Usage Scenario: Edamame Pods
	6 Discussion
	7 Conclusion
	References
MultiProjector: Temporal Projection for Multivariates Time Series
	1 Introduction
	2 Related Work
		2.1 Visualizing High Dimensional Temporal Datasets
		2.2 Dimension Reduction
	3 Methodology
		3.1 Clusterings
		3.2 Multidimensional Projections
		3.3 Visualizing the Time Dimension
		3.4 Multivariate Representations
	4 Use Cases
		4.1 Use Case 1: Monthly US Employment Rate
		4.2 Use Case 2: Monitoring Computer Metrics
		4.3 Use Case 3: Plant Genetics
		4.4 Discussion
	5 Conclusion
	References
Deep Learning Based Super-Resolution for Medical Volume Visualization with Direct Volume Rendering
	1 Introduction
	2 Related Work
		2.1 Image and Video Super-resolution
		2.2 Resolution Enhancement for Rendered Content
	3 Methodology
		3.1 Direct Volume Rendering Framework
		3.2 Network Architecture
	4 Dataset
	5 Evaluation
		5.1 Performance Gain with Additional Feature at the Input
		5.2 Performance Gain with Additional Previous Frames
		5.3 Upsampling Ratio
	6 Conclusion and Future Work
	References
Interactive Virtual Reality Exploration of Large-Scale Datasets Using Omnidirectional Stereo Images
	1 Introduction
	2 Related Work
		2.1 Image-Based Visualization
		2.2 Virtual Reality for Large-Scale Data Sets
	3 Science Drivers
		3.1 Cancer Cell Transport
		3.2 Graphene Superlubricity
	4 Cinema ODS Image Database
		4.1 Rendering
	5 Interactive Cinema ODS Viewer
	6 Evaluation
		6.1 Visualization Latency
		6.2 VR Frame Rate
		6.3 Qualitative Feedback
	7 Conclusion
	References
A Quantitative Analysis of Labeling Issues in the CelebA Dataset
	1 Introduction
	2 Related Work
	3 Incorrect Labels
		3.1 Contradicting and Conflicting Labels
		3.2 Mislabeling
	4 Inconsistent Labels
		4.1 Consistency
		4.2 Agreement
		4.3 Correlated Labels
	5 Conclusion
	References
Object Detection and Recognition
Recognition of Aquatic Invasive Species Larvae Using Autoencoder-Based Feature Averaging
	1 Introduction
	2 Related Work
		2.1 Aquatic Invasive Species
		2.2 Local Responses to Aquatic Invasive Species
		2.3 Classification with Image Sets
		2.4 Underwater Image Classification
		2.5 Autoencoders
	3 Methodology
		3.1 Solution Description
		3.2 Convolutional Autoencoder
		3.3 Classification Model
		3.4 Activation Functions
		3.5 Loss Functions
		3.6 Base Model
		3.7 Dataset
	4 Results
		4.1 Evaluation Metric
		4.2 Quantitative Analysis
		4.3 Comparative Analysis
	5 Conclusion
	References
Subspace Analysis for Multi-temporal Disaster Mapping Using Satellite Imagery
	1 Introduction
	2 Subspace Learning-Based Disaster Mapping
		2.1 Region Delineation
		2.2 Segmentation Fusion
		2.3 Subspace Learning for Disaster Mapping
	3 Determining the Changed and Unchanged Regions
	4 Experiments, Results and Discussion
		4.1 Experimental Setup
		4.2 Results and Discussion
	5 Conclusion
	References
Open-Set Plankton Recognition Using Similarity Learning
	1 Introduction
	2 Related Work
		2.1 Plankton Recognition
		2.2 Open-Set Classification
		2.3 Classification by Metric Learning
	3 Proposed Method
		3.1 Angular Margin Loss
	4 Experiments
		4.1 Data
		4.2 Description of Experiments
		4.3 Results
	5 Conclusions
	References
Sensor Fusion Operators for Multimodal 2D Object Detection
	1 Introduction
	2 Related Work
	3 Camera-LiDAR 2D Object Detector
	4 Sensor Fusion Operators
	5 Experimental Results
		5.1 Experimental Setting
		5.2 Evaluation of Early Sensor Fusion
		5.3 Evaluation of Mid-Level Sensor Fusion
		5.4 Complexity Analysis
	6 Conclusion
	References
Learning When to Say ``I Don\'t Know\"
	1 Introduction
	2 Preliminaries
	3 Related Work
	4 Proposed Method
	5 Experiments
		5.1 Synthetic Data
		5.2 Image Datasets
		5.3 Text Datasets
		5.4 Generalization from Validation to Test Data
		5.5 Alternative Confidence Interval Formulations
		5.6 Discussion
	6 Conclusion
	References
Multi-class Detection and Tracking of Intracorporeal Suturing Instruments in an FLS Laparoscopic Box Trainer Using Scaled-YOLOv4
	1 Introduction
	2 Related Works
	3 Methodology
		3.1 Scaled-YOLOv4 Architecture
		3.2 Measurement Algorithm
	4 Experimental Setup
		4.1 Dataset
		4.2 Software Implementation
	5 Results
	6 Discussion
	7 Conclusion and Future Work
	References
Deep Learning II
A New Approach to Visual Classification Using Concatenated Deep Learning for Multimode Fusion of EEG and Image Data
	1 Introduction
	2 Related Work
	3 Datasets
		3.1 EEG-ImageNet
		3.2 Visual Stimuli EEG Dataset: Real-World 3D Objects and Corresponding 2D Image Stimuli
	4 Data Encoding and Processing
		4.1 Classical Feature Extraction for EEG Data
		4.2 Classical Feature Extraction for Image Data
		4.3 Principal Component Analysis (PCA) Encoding
		4.4 Grayscale-Image Encoding for EEG Data
	5 Methods and Model Implementation
		5.1 Conventional Machine Learning Classifiers
		5.2 LSTM-Based EEG Model (LEM) ch17ourvisclasspaper
		5.3 CNN-Based Image Model (CIM) ch17ourvisclasspaper
		5.4 Grayscale-Image Encoded EEG Model (GEM)
		5.5 Concatenation-Based Models ch17ourvisclasspaper
	6 Experiments and Results
		6.1 Baseline Visual Classification for EEG and Image Data
		6.2 Classification Using Deep Learning Models
		6.3 Hemispherical Brain Region Classification Comparison
		6.4 Visual Classification Using Multimodal Deep Learning
		6.5 Visual Classification for Real Object Versus Image as Stimuli
	7 Discussion
	8 Conclusion
	References
Deep Learning-Based Classification of Plant Xylem Tissue from Light Micrographs
	1 Introduction
	2 Related Works
	3 Dataset and Problem Definition
	4 Methodology
		4.1 Data Augmentation and Pre-processing
		4.2 Cascading-Like Model
		4.3 Global Contextualization Approach
	5 Experiments and Results
		5.1 Model Evaluation Metric
		5.2 Baseline Results
		5.3 Results
	6 Discussion
	7 Conclusion
	References
VampNet: Unsupervised Vampirizing of Convolutional Networks
	1 Introduction
	2 Related Work
		2.1 Correlation-Based Feature Map Analysis
		2.2 Multitask Neural Networks
		2.3 Networks Merging
	3 Method
		3.1 Linearity Between Feature Maps
		3.2 Ranking Linearity Between Features
		3.3 Vampirizing a Feature Using a Convolutional Operator
		3.4 Vampirizing a Layer
		3.5 Automatic Selection of the Layer to Be Replaced
	4 Experiments
		4.1 Setup
		4.2 Linearity
		4.3 Trade-Off Selection Between the Accuracy and Computational Budget
	5 Conclusion
	References
RSI-Grad-CAM: Visual Explanations from Deep Networks via Riemann-Stieltjes Integrated Gradient-Based Localization
	1 Introduction
	2 Previous Work
	3 Methodology
		3.1 Motivation and Theoretical Background
		3.2 Riemann-Stieltjes Integration of Gradients
		3.3 Metrics
	4 Implementation and Testing
		4.1 Implementation
		4.2 Quantitative Evaluations
	5 Conclusions
	6 Future Work
	References
Deep Labeling of fMRI Brain Networks Using Cloud Based Processing
	1 Introduction
	2 Methods
		2.1 Data
		2.2 fMRI Resting State Analysis
		2.3 Neural Network Methods
	3 Results
		3.1 fMRI Resting State Analysis
		3.2 Neural Network Performance Comparison
	4 Discussion
	5 Conclusions
	References
Semantic Segmentation Using Neural Ordinary Differential Equations
	1 Introduction
	2 Related Work
	3 Method
		3.1 Baseline Network
		3.2 SegNode
	4 Experiments
		4.1 Setup
		4.2 Cityscapes
		4.3 CamVid
		4.4 LIP
		4.5 PASCAL-Context
		4.6 Results
		4.7 Empirical Computational Cost
		4.8 Trajectory Error
	5 Conclusion
	References
Video Analysis and Event Recognition
Graphing the Future: Activity and Next Active Object Prediction Using Graph-Based Activity Representations
	1 Introduction
	2 Related Work
	3 The Proposed Method - GTF
	4 Experiments
		4.1 Datasets
		4.2 Feature Extraction
		4.3 Evaluation Metrics
		4.4 Results
	5 Conclusions
	References
Detecting Fall Actions of Videos by Using Weakly-Supervised Learning and Unsupervised Clustering Learning
	1 Introduction
	2 Related Work
		2.1 Fall Detection
		2.2 Weakly Supervised Learning
		2.3 Unsupervised Clustering Learning
	3 Method
	4 Experiment
		4.1 Dataset
		4.2 Implementation
		4.3 Results and Analysis for Learning of ResVAE
		4.4 Results and Analysis for Learning of FCVAE
	5 Conclusion
	6 Future Work
	References
Multi-property Tensor-Based Learning for Abnormal Event Detection
	1 Introduction
		1.1 Related Work
		1.2 Our Contribution
	2 Intra/Inter Property Encoding
		2.1 Intra-property Encoding
		2.2 Inter-property Encoding
	3 Tensor-Based Unsupervised Learning for Inter-property Encoding
		3.1 Unsupervised Tensor-Based Learning
		3.2 The Rank-1 Canonical Decomposition of Network Parameters
		3.3 The Learning Algorithm
		3.4 Unsupervised Abnormal Event Detection
	4 Experimental Evaluation
	5 Conclusions
	References
Depth-Based vs. Color-Based Pose Estimation in Human Action Recognition
	1 Introduction
	2 Related Work
		2.1 Pose Estimation
		2.2 Action Recognition
	3 Methods
		3.1 Data Loading and Preprocessing
		3.2 Feature Extraction and Classification
	4 Experiments
		4.1 Datasets
		4.2 Evaluation Protocol
		4.3 Results
		4.4 Discussion
	5 Conclusions
	References
Cross-Domain Learning in Deep HAR Models via Natural Language Processing on Action Labels
	1 Introduction
	2 Related Work
	3 Proposed Method
		3.1 Dataset Label Association via NLP
		3.2 Dual-Dataset Learning Deep Architecture
		3.3 Factors that Affect Learning
	4 Experimental Setup
		4.1 Inter- and Intra-dataset Evaluation
		4.2 Feature Extraction
		4.3 Temporal Modelling Architectures
	5 Experimental Results
	6 Conclusions and Discussion
	References
Computer Graphics
Visualizing Data Flows in Computer Graphics Programs for Code Comprehension and Debugging
	1 Introduction
	2 Related Work
		2.1 OpenGL in Computer Science Education
		2.2 Program Comprehension and Debugging
		2.3 Benefits of Program Visualization for Code Comprehension and Debugging
		2.4 Program Visualization for Computer Graphics
	3 OpenGL Overview
		3.1 Program Structure
		3.2 Data Flows in OpenGL
		3.3 Explicit and Implicit Links
	4 Creating Data Flow Diagrams (DFDs)
		4.1 Vertex Attribute DFD
		4.2 Uniform Variable DFD
		4.3 Texture Sampler DFD
		4.4 Interactive Visualization
		4.5 When to Draw a Data Flow Diagram?
	5 A Debugging Case Study
	6 Discussion
	7 Conclusion and Future Work
	References
A Practical Algorithm for Degree-k Voronoi Domains of Three-Dimensional Periodic Point Sets
	1 Introduction: Motivations and Key Contributions
	2 Background Definitions from Computational Geometry
	3 The Geometric Structure of Degree-k Voronoi Domains
	4 Computing Degree-k Voronoi Domains of a Periodic Set
	5 Experiments on Degree-k Voronoi Domains for n=2,3
	References
End-to-End Deep Neural Network for Illumination Consistency and Global Illumination
	1 Introduction
	2 Proposed Method
		2.1 Generating Reflections and Shadows of Virtual Objects
		2.2 Combination of RGB and Geometric Data
	3 Experiment
		3.1 Creating a Dataset
		3.2 Training the Network
		3.3 Qualitative Results
		3.4 Quantitative Results
	4 Conclusion
	References
Pruning-Based Topology Refinement of 3D Mesh Using a 2D Alpha Mask
	1 Introduction
	2 Related Works
	3 Method
		3.1 General Overview
		3.2 Implementation Details
	4 Experiments
		4.1 Topological Refinement Evaluation - Qualitative Results
		4.2 2D and 3D-Based Quantitative Evaluation
	5 Limitations and Further Work
	6 Conclusion
	References
ST: Biomedical Imaging Techniques for Cancer Detection, Diagnosis and Management
ConnectedUNets++: Mass Segmentation from Whole Mammographic Images
	1 Introduction
	2 Related Works
	3 Methodology
		3.1 Architecture
		3.2 Dataset Preprocessing and Experimental Setup
		3.3 Evaluation Metrics
	4 Experimental Results and Discussion
	5 Conclusion
	References
Severity Classification of Ulcerative Colitis in Colonoscopy Videos by Learning from Confusion
	1 Introduction
	2 Related Work
	3 Methodology
		3.1 New Training Strategy
		3.2 Refine the Training Dataset and Train Final Model
	4 Experimental Results
		4.1 Testing and Comparison
		4.2 Severity Scores
	5 Concluding Remarks
	References
Detection and Classification of Lung Disease Using Deep Learning Architecture from X-ray Images
	1 Introduction
	2 Literature Review
	3 Methodology
		3.1 Dataset Description
		3.2 Data Processing
		3.3 Class Imbalance Problem
		3.4 Compound Scaling and Reduction Cells in CX-Ultranet
		3.5 Implementing Reduction Cells
	4 Results
	5 Discussion
	6 Conclusion
	References
PolypDEQ: Towards Effective Transformer-Based Deep Equilibrium Models for Colon Polyp Segmentation
	1 Introduction
	2 Related Work
		2.1 Medial Image Segmentation
		2.2 Implicit Deep Learning
	3 Methodology
		3.1 Multiscale Deep Equilibrium Models (MDEQs)
		3.2 SegFormer
		3.3 Our PolypDEQ
	4 Experiments and Discussion
		4.1 Benchmark Datasets
		4.2 Experiment Settings
	5 Results and Discussion
		5.1 Performance
		5.2 Time and Space Complexity
	6 Conclusion
	References
Author Index