Advances in Visual Computing: 16th International Symposium, ISVC 2021, Virtual Event, October 4-6, 2021, Proceedings, Part I (Lecture Notes in Computer Science)

✍ Scribed by George Bebis (editor), Vassilis Athitsos (editor), Tong Yan (editor), Manfred Lau (editor), Frederick Li (editor), Conglei Shi (editor), Xiaoru Yuan (editor), Christos Mousas (editor), Gerd Bruder (editor)

• Publisher: Springer
• Year: 2021
• Tongue: English
• Leaves: 635
• Edition: 1st ed. 2021
• Category: Library

✦ Synopsis

This two-volume set of LNCS 13017 and 13018 constitutes the refereed proceedings of the 16th International Symposium on Visual Computing, ISVC 2021, which was held in October 2021. The symposium took place virtually instead due to the COVID-19 pandemic.

The 48 papers presented in these volumes were carefully reviewed and selected from 135 submissions. The papers are organized into the following topical sections:
Part I: deep learning; computer graphics; segmentation; visualization; applications; 3D vision; virtual reality; motion and tracking; object detection and recognition.
Part II: ST: medical image analysis; pattern recognition; video analysis and event recognition; posters.

✦ Table of Contents

Preface
Organization
Keynote Talks
Embodied Perception in-the-Wild
Design Tools for Material Appearance
Guidance-Enriched Visual Analytics: Challenges and Opportunities
Learning and Accruing Knowledge over Time Using Modular Architectures
Combining Brain-Computer Interfaces and Virtual Reality: Novel 3D Interactions and Promising Applications
Direct Estimation of Appearance Models for Image Segmentation
Contents – Part I
Contents – Part II
Deep Learning I
Real-World Thermal Image Super-Resolution
1 Introduction
2 Related Work
2.1 RGB Image Super-Resolution
2.2 Thermal Image Super-Resolution
3 Dataset
4 Thermal RealSR
4.1 Realistic Degradation Using KernelGAN and Noise Injection
4.2 Super-Resolution Model
5 Experiments and Results
5.1 Evaluation Metrics
5.2 Comparison with the State of the Art
6 Conclusion
References
QR Code Style Transfer Method Based on Conditional Instance Regularization
1 Introduction
2 Related Works
2.1 QR Code Style Transfer System
2.2 Structure of the Style Transfer Network
3 Style Transfer Network Based on Conditional Instance Regularization
3.1 Conditional Instance Regularization
3.2 Residual Connected Module
3.3 Style Transfer Network Structure
3.4 Weighted Fusion Correction for Styled QR Codes
3.5 Artistic Style QR Code Dynamics
4 Experiment
4.1 Training
4.2 Single-Style Training
4.3 Multi-style Training
4.4 Comparison of Experiment Results
5 Conclusion
References
Multimodal Multi-tasking for Skin Lesion Classification Using Deep Neural Networks
1 Introduction
2 Methodology
2.1 Dataset
2.2 ABCD Rule Feature Extraction
2.3 Proposed Model
2.4 Class Balancing Techniques
3 Experiments
3.1 ABCD Rule for Multimodal Multi-tasking
3.2 Role of Segmentation in Lesion Classification
4 Results and Discussion
5 Conclusion
References
DeepSolfège: Recognizing Solfège Hand Signs Using Convolutional Neural Networks
1 Introduction
2 Related Work
3 Dataset
3.1 Labels
3.2 Preprocessing
4 Method
4.1 CNN Architecture
4.2 Training
4.3 Ablation Study
5 Evaluation
5.1 Real World Application
6 Conclusion
References
Image Prior Transfer and Ensemble Architectures for Parkinson's Disease Detection
1 Introduction
2 Background and Related Works
3 Proposed Method
3.1 Dataset
3.2 Preprocessing
3.3 Models
3.4 Ensemble Architecture 1
3.5 Ensemble Architecture 2
4 Experimental Results
5 Occlusion Analysis to Locate Relevant Regions
5.1 Occlusion Analysis for Modified ResNet
5.2 Occlusion Analysis for Ensemble Architecture - Model 1
6 Conclusion and Future Works
References
Computer Graphics I
BRDF Measurement of Real Materials Using Handheld Cameras
1 Introduction
2 BRDF Measurement Using Handheld Cameras
2.1 Use of Bivariate BRDF
2.2 BRDF Sampling
2.3 Dense BRDF Estimation
3 BRDF Measurement of Real Materials
3.1 Experimental Setup
3.2 Estimation Results
3.3 Measurement Time
4 Conclusion
References
SORGATE: Extracting Geometry and Texture from Images of Solids of Revolution
1 Introduction
2 Previous Work
3 Design
4 Implementation
5 Evaluation
6 Conclusions and Future Work
References
Putting Table Cartograms into Practice
1 Introduction
2 Related Work
3 TCarto: An Optimization Based Algorithm
4 Potential Applications
5 Experimental Results
6 Limitations and Directions for Future Research
References
Perceived Naturalness of Interpolation Methods for Character Upper Body Animation
1 Introduction
2 Prior Work
3 Methods
3.1 Studied Interpolations
3.2 Study Design
3.3 Experiment Design
3.4 Study Procedure
3.5 Data Collection and Analysis
4 Discussion
5 Conclusion and Future Work
References
Neuromuscular Control of the Face-Head-Neck Biomechanical Complex with Learning-Based Expression Transfer from Images and Videos
1 Introduction
2 Related Work
3 Musculoskeletal Model
3.1 Control
4 Expression Learning
4.1 Network Architecture
4.2 Training Data Generation
4.3 Network Training
4.4 Expression Transfer Pipeline
5 Experiments and Results
5.1 Facial Expression Datasets
5.2 Action Units and Muscle Activations
5.3 Head Orientation
5.4 Facial Expression Transfer
6 Conclusion and Future Work
References
Segmentation
Synthesized Image Datasets: Towards an Annotation-Free Instance Segmentation Strategy
1 Introduction
2 Related Work
3 Proposed Approach
3.1 Segmentation Algorithm
3.2 Synthesized Image Generation
4 Experimental Results
4.1 Case Study
4.2 Free Annotation Results
5 Conclusions
References
Holistically-Nested Structure-Aware Graph Neural Network for Road Extraction
1 Introduction
2 Related Work
3 Method
3.1 Road Border Detection
3.2 Road Structure Aware GNN
3.3 Element-Wise Attention
3.4 Joint Multi-task Training
4 Experimental Results
4.1 Evaluation Metrics
4.2 Results
5 Conclusions
References
Extraction and Merging of Stroke Structure of Chinese Characters
1 Introduction
2 Related Work
3 Methodology
3.1 The Path Network
3.2 Pixel Selection for Path Net
3.3 Stroke Merging Algorithm
4 Experiment Results
5 Conclusion
References
Analysis of Multi-temporal Image Series for the Preventive Conservation of Varnished Wooden Surfaces
1 Introduction
2 Related Works
2.1 Change and Damage Detection
2.2 A-Contrario Framework
3 Proposed Approach
3.1 Distance Matrix
3.2 Number of False Alarms
3.3 Maximal Clusters
4 Experiments
4.1 Dataset
4.2 Results
5 Conclusions
References
Visualization
Evaluating User Interfaces for a Driver Guidance System to Support Stationary Wireless Charging of Electric Vehicles
1 Introduction
2 Related Work
3 System
3.1 Goals
3.2 Overview
3.3 Visualization Types
3.4 Information Output Setup
4 User Study
4.1 Test Environment
4.2 Task
4.3 Data Acquisition
4.4 Results
5 Discussion
5.1 Precision
5.2 User Experience
5.3 Time
5.4 Observations
6 Conclusions
References
MOBA Coach: Exploring and Analyzing Multiplayer Online Battle Arena Data
1 Introduction
2 Related Work
3 MOBA Coach Tool
4 Evaluation
4.1 Analysis of the Results
4.2 Discussion of the Results
5 Conclusion and Future Work
References
JobNet: 2D and 3D Visualization for Temporal and Structural Association in High-Performance Computing System
1 Introduction
2 Related Work
3 Design and Implementation
3.1 Terms and Definitions
3.2 Design Rationale
4 Case Study
4.1 JobNet2D
4.2 JobNet3D
5 Conclusion
References
Evaluation and Selection of Autoencoders for Expressive Dimensionality Reduction of Spatial Ensembles
1 Introduction
2 Related Work
3 Study Setup, Metrics and Selection
4 Evaluation
5 Discussion and Outlook
References
Data-Driven Estimation of Temporal-Sampling Errors in Unsteady Flows-8pt
1 Introduction
2 Related Work
3 Temporal Subsampling Errors in Pathlines
3.1 Temporal Subsampling of Simulated Unsteady Flows
3.2 Data-Driven Modeling of Errors
4 Validation and Results
4.1 2D Flow Past a Cylinder
4.2 3D Lifted Ethylene Jet Flame
5 Conclusion
References
Applications
ReGenMorph: Visibly Realistic GAN Generated Face Morphing Attacks by Attack Re-generation
1 Introduction
2 Related Works
3 Methodology
3.1 The ReGenMorph Face Morphing Pipeline
3.2 Creating ReGenMorph Morphing Attacks
4 Experimental Setup
4.1 Database
4.2 Vulnerability Analyses
4.3 Detectability Analyses
5 Results
5.1 ReGenMorph Image Appearance
5.2 Vulnerability of Face Recognition to ReGenMorph
5.3 Detectability of ReGenMorph
6 Conclusion
References
Car Pose Estimation Through Wheel Detection
1 Introduction
2 Related Work
3 Wheel Detection Techniques
4 Pose Estimation
5 Evaluation
5.1 Evaluation Datasets
5.2 Algorithm Parameterization
5.3 Evaluation Metrics
5.4 Wheel Detection Accuracy and Time Performance
5.5 Pose Estimation Accuracy
6 Conclusion and Future Work
References
Improving Automatic Quality Inspection in the Automotive Industry by Combining Simulated and Real Data
1 Introduction
2 Related Work
3 Methods
3.1 Baseline
3.2 CycleGAN with Semantic Consistency
3.3 Detection System's Improvements
4 Experiments
5 Results and Discussion
5.1 Domain Mapping
5.2 Detector Fine-Tuning
6 Conclusion and Future Work
References
PW-MAD: Pixel-Wise Supervision for Generalized Face Morphing Attack Detection-8pt
1 Introduction
2 Related Work
3 Methodology
3.1 The Proposed PW-MAD
3.2 Baselines
4 Experimental Setup
4.1 The Dataset
4.2 Experiments and Evaluation Metrics
5 Results and Discussion
6 Conclusion
References
Integration of a BCI with a Hand Tracking System and a Motorized Robotic Arm to Improve Decoding of Brain Signals Related to Hand and Finger Movements
1 Introduction
2 System Assembly
3 Use-Cases
4 Conclusion
References
Deep Learning II
Fast Point Voxel Convolution Neural Network with Selective Feature Fusion for Point Cloud Semantic Segmentation
1 Introduction
2 Related Work
2.1 Volumentric Representation
2.2 Point-Based Representation
2.3 Efficiency of Current Models
3 Method
3.1 Point Voxel Convolution
3.2 Deep Supervision
4 Experiment
4.1 Object Classification
4.2 Shape Segmentation
4.3 Indoor Scene Segmentation
4.4 Ablation Study
5 Conclusion
References
Behaviour of Sample Selection Techniques Under Explicit Regularization
1 Introduction
2 Related Works
3 Experimentation
3.1 Observations
4 Conclusion
References
Adaptive Feature Norm for Unsupervised Subdomain Adaptation
1 Introduction
2 Related Work
2.1 Domain Adaptation
2.2 Subdomain Adaptation
2.3 Maximum Mean Discrepancy
3 Method
3.1 Local Maximum Mean Discrepancy
3.2 Architecture
3.3 Adaptive Feature Norm Loss
4 Experiment
4.1 Dataset
4.2 Setup
5 Results and Discussion
6 Conclusion
References
Normal Image Generation-Based Defect Detection by Generative Adversarial Network with Chaotic Random Images
1 Introduction
2 Fusion of Real Image Space and Latent Space
2.1 Generative Model-Based Defect Detection
2.2 Basic Idea
3 Proposed Method
3.1 Loss Function
3.2 Model Architecture
3.3 Random Image Generation
3.4 Anomaly Score
4 Experiments on Defect Detection
4.1 Settings
4.2 Results
4.3 Ablation Study
5 Conclusion
References
SPNet: Multi-shell Kernel Convolution for Point Cloud Semantic Segmentation
1 Introduction
2 Related Work
3 Methodology
3.1 Review on Kernel Point Convolution
3.2 SPConv
3.3 Feature Attention
3.4 Network Architecture
4 Experiments
4.1 Datasets
4.2 Implementation Details
4.3 Ablation Studies
5 Conclusions
References
Computer Graphics II
Procedural Modeling of the Great Barrier Reef
1 Introduction
2 Growth of the Great Barrier Reef: A Hypothesis
3 Problem Formulations
3.1 Terrain Representation
3.2 DLA Normal Distributions
3.3 Directional DLA Log-Normal Filter
3.4 Height-Aware Terrain Texture Synthesis
4 Experimental Results
4.1 Changing Input Terrains
4.2 Changing Aggregation Centers
4.3 Changing DLA Grid Sizes
4.4 Changing Textures
5 Conclusion
References
Art-Directable Cloud Animation
1 Introduction
2 Related Work
3 Methods
4 Results
5 Conclusion
References
Physically Based Rendering of Simple Thin Volume Natural Nanostructures
1 Introduction
2 Thin Volume Nanostructures and Layer Surface Modulations
3 Fourier Optics and Physically Based Rendering
3.1 Proposed Method
4 Experiments and Results
5 Conclusion
References
Deep Tiling: Texture Tile Synthesis Using a Constant Space Deep Learning Approach
1 Introduction
2 Related Work
2.1 Texture Synthesis
2.2 Texture Tiling
3 Deep Texture Tiling
4 Results and Evaluation
5 Conclusions and Future Work
References
BEAPS: Integrating Volumetric Dynamics in Virtual Agent Prototyping
1 Introduction
2 Related Work
3 Methodology
3.1 Nvidia FleX
3.2 KADAPT - A Virtual Agent Prototyping Test-Bed
4 Conclusion
References
3D Vision
IVS3D: An Open Source Framework for Intelligent Video Sampling and Preprocessing to Facilitate 3D Reconstruction
1 Introduction
2 Related Work
3 Approach
3.1 Architecture Overview
3.2 Baseline Plugins Already Developed
3.3 Batch Processing
4 Experiments
4.1 Experimental Setup for the Frame Selection
4.2 Semantic Segmentation for Masking Potentially Moving Objects
5 Conclusion and Future Work
References
3D Registration of Deformable Objects Using a Time-of-Flight Camera
1 Introduction
2 Related Work
3 Proposed Framework
3.1 Methodology
4 Experimental Results
4.1 Densely and Sparsely Textured Paper Sheets
4.2 Endoscopic Stereo Video
5 Conclusion and Future Work
References
Pose Estimation of Texture-Less Targets for Unconstrained Grasping
1 Introduction
2 Multi-task Learning for Initial Pose Estimation
3 Generation of Pose Dataset Based on 3D Reconstruction
4 Pose Optimization of Texture-Less Targets
5 Experiments
5.1 Object Detection and Edge Extraction Tests
5.2 Pose Accuracy Test
6 Conclusion
References
Virtual Reality
Wearable Augmented Reality System Using Head-Mounted Projector
1 Introduction
2 Related Work
2.1 AR Systems Using a Head-Mounted Display
2.2 AR Systems Using a Projector
2.3 Wearable AR Systems Using a Portable Projector
3 System Design
3.1 Wearable AR System Using Head-Mounted Projector
3.2 Image Projection Independent of the Shape of the Projection Surface
3.3 Work Space Expansion
4 Implementation and Evaluation
4.1 System Overview
4.2 Image Projection Independent of the Shape of the Projection Surface
4.3 Work Space Expansion
5 Applications
6 Limitataions
7 Conclusion
References
Generation of Virtual Reality Environment Based on 3D Scanned Indoor Physical Space
1 Introduction
2 Generation of Virtual Reality Environment Based on 3D Scanned Indoor Physical Space
2.1 Ceiling and Floor Removal
2.2 Clustering of Obstacle Points
2.3 Layout Optimization of Virtual Objects
2.4 Detection of Touchable Surfaces
3 Experiment
3.1 System
3.2 Generated Virtual Environment
4 Conclusion and Future Work
References
Omnichannel Retail Customer Experience with Mixed-Reality Shopping Assistant Systems
1 Introduction
2 Background and Previous Work
2.1 Optical-See Through Mixed Reality
2.2 MR Shopping Assistant System
2.3 Customer Experience and Hypotheses Development
3 MR Shopping Assistant System
4 Experiment Setup
5 Results
5.1 Regression Analysis
5.2 Qualitative Comments from the Participants
6 Discussion and Implications
7 Conclusion
Appendix
References
Motion and Tracking
MissFormer: (In-)Attention-Based Handling of Missing Observations for Trajectory Filtering and Prediction
1 Introduction and Related Work
2 MissFormer
3 Evaluation and Analysis
3.1 Synthetic Data
3.2 Real-World Data
4 Conclusion
References
Compressed Domain Consistent Motion Based Frame Scoring for IoT Edge Surveillance Videos
1 Introduction
2 Related Works
3 Background
3.1 MV Analysis General Template
3.2 Prior-Arts
4 Observations
5 Proposed RLV and Pruning Algorithm
5.1 Relative Variance Over Normalized MVs - RLV
5.2 Our Pruning Algorithm
6 Results
6.1 Dataset Details
6.2 RLV vs Others
6.3 Ablation Study: Results Comparing Neighborhood Choices
6.4 Ablation Study: Results by Applying Auxiliary Techniques
6.5 Execution Time Details
7 Conclusion
References
A Data-Driven Approach to Improve 3D Head-Pose Estimation
1 Introduction
2 Method
2.1 Problem Definition
2.2 Weight Initialization with Activation Analysis
2.3 Augmented Image Duplication
3 Experiments
3.1 Setup
3.2 Datasets
3.3 Optimizer
3.4 Results
3.5 AFLW2000 Occlusion, Motion-Blur, and Brightness Study
4 Conclusion
References
Object Detection and Recognition
Chicken Detection in Occlusion Scenes with Modified Single Shot MultiBox Detector
1 Introduction
2 Related Work
2.1 Object Detection
2.2 Non-Maximum Suppression
2.3 Attention Mechanisms in the CNN
2.4 Pedestrian Detection
3 Modified Convolutional Neural Network
3.1 Network Structure
3.2 DIoU-NMS
3.3 Data Augmentation
4 Experiment
4.1 Datasets
4.2 Experiment of Modified CBAM Structure
4.3 Experiment of Data Augmentation
5 Conclusion
References
PoseTED: A Novel Regression-Based Technique for Recognizing Multiple Pose Instances
1 Introduction
2 Proposed PoseTED
3 Experiments
3.1 Dataset
3.2 Experimental Setup and Training Details
3.3 Evaluation Matrics
4 Results and Discussions
4.1 Performance on COCO Dataset
4.2 Performance on MPII Dataset
5 Conclusion
References
Unsupervised Pixel-Wise Weighted Adversarial Domain Adaptation
1 Introduction
2 Related Work
2.1 Semantic Segmentation
2.2 Adversarial Domain Adaptation
3 Preliminaries
3.1 Supervised Semantic Segmentation
3.2 Unsupervised Output-Space Based Domain Adaptation
4 Pixel-Wise Weighted Adaptation Method
4.1 Generation Confidence Weighted Adaptation
4.2 Discrimination Confidence Weighted Aadaptation
4.3 Network Overview and Optimization
5 Experiments
5.1 Datasets
5.2 Implementation Details
5.3 Adaptation from GTA5 to Cityscapes
5.4 Adaptation from SYNTHIA to Cityscapes
5.5 Ablation Studies
6 Conclusion
References
A Quantum 3D Convolutional Neural Network with Application in Video Classification
1 Introduction
2 Related Work
3 Methodology
3.1 Quantum Background
3.2 Quantum Convolutional Neural Network
4 Experimental Evaluation
5 Discussion and Conclusion
References
Author Index