
Computer Vision - ECCV 2022 Workshops
Description
Alles über E-Books | Antworten auf Fragen rund um E-Books, Kopierschutz und Dateiformate finden Sie in unserem Info- & Hilfebereich.
The 367 full papers included in this volume set were carefully reviewed and selected for inclusion in the ECCV 2022 workshop proceedings. They were organized in individual parts as follows:
Part I:
W01 - AI for Space; W02 - Vision for Art; W03 - Adversarial Robustness in the Real World; W04 - Autonomous Vehicle Vision
Part II: W05 - Learning With Limited and Imperfect Data; W06 - Advances in Image Manipulation;
Part III: W07 - Medical Computer Vision; W08 - Computer Vision for Metaverse; W09 - Self-Supervised Learning: What Is Next?;
Part IV: W10 - Self-Supervised Learning for Next-Generation Industry-LevelAutonomous Driving; W11 - ISIC Skin Image Analysis; W12 - Cross-Modal Human-Robot Interaction; W13 - Text in Everything; W14 - BioImage Computing; W15 - Visual Object-Oriented Learning Meets Interaction: Discovery, Representations, and Applications; W16 - AI for Creative Video Editing and Understanding; W17 - Visual Inductive Priors for Data-Efficient Deep Learning; W18 - Mobile Intelligent Photography and Imaging;
Part V: W19 - People Analysis: From Face, Body and Fashion to 3D Virtual Avatars; W20 - Safe Artificial Intelligence for Automated Driving; W21 - Real-World Surveillance: Applications and Challenges; W22 - Affective Behavior Analysis In-the-Wild;
Part VI : W23 - Visual Perception for Navigation in Human Environments: The JackRabbot Human Body Pose Dataset and Benchmark; W24 - Distributed Smart Cameras; W25 - Causality in Vision; W26 - In-Vehicle Sensing and Monitorization; W27 - Assistive Computer Vision and Robotics; W28 - Computational Aspectsof Deep Learning;
Part VII: W29 - Computer Vision for Civil and Infrastructure Engineering; W30 - AI-Enabled Medical Image Analysis: Digital Pathology and Radiology/COVID19; W31 - Compositional and Multimodal Perception;
Part VIII: W32 - Uncertainty Quantification for Computer Vision; W33 - Recovering 6D Object Pose; W34 - Drawings and Abstract Imagery: Representation and Analysis; W35 - Sign Language Understanding; W36 - A Challenge for Out-of-Distribution Generalization in Computer Vision; W37 - Vision With Biased or Scarce Data; W38 - Visual Object Tracking Challenge.
More details
Other editions
Additional editions

Content
- Intro
- Foreword
- Preface
- Organization
- Contents - Part III
- W06 - Advances in Image Manipulation: Reports
- W06 - Advances in Image Manipulation: Reports
- Reversed Image Signal Processing and RAW Reconstruction. AIM 2022 Challenge Report
- 1 Introduction
- 2 AIM 2022 Reversed ISP Challenge
- 2.1 Datasets
- 2.2 Evaluation and Results
- 3 Proposed Methods and Teams
- 3.1 NOAHTCV
- 3.2 MiAlgo
- 3.3 CASIA LCVG
- 3.4 HIT-IIL
- 3.5 CS^2U
- 3.6 SenseBrains
- 3.7 HiImage
- 3.8 0noise
- 3.9 OzU VGL
- 3.10 PixelJump
- 3.11 CVIP
- 4 Conclusions
- A Appendix 1: Qualitative Results
- B Appendix 2: Teams and Affiliations
- References
- AIM 2022 Challenge on Instagram Filter Removal: Methods and Results
- 1 Introduction
- 2 Challenge
- 2.1 Challenge Data
- 2.2 Evaluation
- 2.3 Submissions
- 3 Results
- 3.1 Overall Results
- 3.2 Solutions
- 4 Teams and Affiliations
- 4.1 Organizers of AIM 2022 Challenge on Instagram Filter Removal
- 4.2 Fivewin
- 4.3 CASIA LCVG
- 4.4 MiAlgo
- 4.5 Strawberry
- 4.6 SYU-HnVLab
- 4.7 XDER
- 4.8 CVRG
- 4.9 CVML
- 4.10 Couger AI
- References
- Learned Smartphone ISP on Mobile GPUs with Deep Learning, Mobile AI & AIM 2022 Challenge: Report
- 1 Introduction
- 2 Challenge
- 2.1 Dataset
- 2.2 Local Runtime Evaluation
- 2.3 Runtime Evaluation on the Target Platform
- 2.4 Challenge Phases
- 2.5 Scoring System
- 3 Challenge Results
- 3.1 Results and Discussion
- 4 Challenge Methods
- 4.1 MiAlgo
- 4.2 Multimedia
- 4.3 ENERZAi
- 4.4 HITZST01
- 4.5 MINCHO
- 4.6 CASIA 1st
- 4.7 JMU-CVLab
- 4.8 DANN-ISP
- 4.9 Rainbow
- 4.10 SKD-VSP
- 4.11 CHannel Team
- 5 Additional Literature
- A Teams and Affiliations
- References
- Efficient Single-Image Depth Estimation on Mobile Devices, Mobile AI & AIM 2022 Challenge: Report*-4pt
- 1 Introduction
- 2 Challenge
- 2.1 Dataset
- 2.2 Local Runtime Evaluation
- 2.3 Runtime Evaluation on the Target Platform
- 2.4 Challenge Phases
- 2.5 Scoring System
- 3 Challenge Results
- 3.1 Results and Discussion
- 4 Challenge Methods
- 4.1 TCL
- 4.2 AIIA HIT
- 4.3 MiAIgo
- 4.4 Tencent GY-Lab
- 4.5 SmartLab
- 4.6 JMU-CVLab
- 4.7 ICL
- 5 Additional Literature
- A Teams and Affiliations
- References
- Efficient and Accurate Quantized Image Super-Resolution on Mobile NPUs, Mobile AI & AIM 2022 Challenge: Report
- 1 Introduction
- 2 Challenge
- 2.1 Dataset
- 2.2 Local Runtime Evaluation
- 2.3 Runtime Evaluation on the Target Platform
- 2.4 Challenge Phases
- 2.5 Scoring System
- 3 Challenge Results
- 3.1 Results and Discussion
- 4 Challenge Methods
- 4.1 Z6
- 4.2 TCLResearchEurope
- 4.3 ECNUSR
- 4.4 LCVG
- 4.5 BOE-IOT-AIBD
- 4.6 NJUST
- 4.7 Antins_cv
- 4.8 GenMedia Group
- 4.9 Vccip
- 4.10 MegSR
- 4.11 DoubleZ
- 4.12 Jeremy Kwon
- 4.13 Lab216
- 4.14 TOVB
- 4.15 Samsung Research
- 4.16 Rtsisr2022
- 4.17 Aselsan Research
- 4.18 Klab_SR
- 4.19 TCL Research HK
- 4.20 RepGSR
- 4.21 ICL
- 4.22 Just a Try
- 4.23 Bilibili AI
- 4.24 MobileSR
- 4.25 Noahtcv
- 5 Additional Literature
- A Teams and Affiliations
- References
- Power Efficient Video Super-Resolution on Mobile NPUs with Deep Learning, Mobile AI & AIM 2022 Challenge: Report
- 1 Introduction
- 2 Challenge
- 2.1 Dataset
- 2.2 Local Runtime Evaluation
- 2.3 Runtime Evaluation on the Target Platform
- 2.4 Challenge Phases
- 2.5 Scoring System
- 3 Challenge Results
- 3.1 Results and Discussion
- 4 Challenge Methods
- 4.1 MVideoSR
- 4.2 ZX VIP
- 4.3 Fighter
- 4.4 XJTU-MIGU SUPER
- 4.5 BOE-IOT-AIBD
- 4.6 GenMedia Group
- 4.7 NCUT VGroup
- 4.8 Mortar ICT
- 4.9 RedCat AutoX
- 4.10 221B
- 5 Additional Literature
- A Teams and Affiliations
- References
- Realistic Bokeh Effect Rendering on Mobile GPUs, Mobile AI & AIM 2022 Challenge: Report
- 1 Introduction
- 2 Challenge
- 2.1 Dataset
- 2.2 Local Runtime Evaluation
- 2.3 Runtime Evaluation on the Target Platform
- 2.4 Challenge Phases
- 2.5 Scoring System
- 3 Challenge Results
- 3.1 Results and Discussion
- 4 Challenge Methods
- 4.1 Antins_cv
- 4.2 ENERZAi
- 4.3 ZJUT-Vision
- 4.4 Sensebrain
- 4.5 VIC
- 4.6 MiAIgo
- 5 Additional Literature
- A Teams and Affiliations
- References
- AIM 2022 Challenge on Super-Resolution of Compressed Image and Video: Dataset, Methods and Results
- 1 Introduction
- 2 AIM 2022 Challenge
- 2.1 DIV2K ch8agustsson2017ntire Dataset
- 2.2 LDV 3.0 Dataset
- 2.3 Track 1 - Super-Resolution of Compressed Image
- 2.4 Track 2 - Super-Resolution of Compressed Video
- 3 Challenge Results
- 3.1 Track 1
- 3.2 Track 2
- 4 Teams and Methods
- 4.1 VUE Team
- 4.2 NoahTerminalCV Team
- 4.3 BSR Team ch8li2022multispspatch
- 4.4 CASIA LCVG Team ch8qin2022cidbnet
- 4.5 IVL Team
- 4.6 SRC-B Team
- 4.7 USTC-IR Team ch8li2022hst
- 4.8 MSDRSR Team
- 4.9 Giantpandacv
- 4.10 Aselsan Research Team
- 4.11 SRMUI Team ch8conde2022swin2sr
- 4.12 MVideo Team
- 4.13 UESTC+XJU CV Team
- 4.14 cvlab Team
- References
- W07 - Medical Computer Vision
- W07 - Medical Computer Vision
- Swin-Unet: Unet-Like Pure Transformer for Medical Image Segmentation
- 1 Introduction
- 2 Related Work
- 3 Method
- 3.1 Architecture Overview
- 3.2 Swin Transformer Block
- 3.3 Encoder
- 3.4 Bottleneck
- 3.5 Decoder
- 3.6 Skip Connection
- 4 Experiments
- 4.1 Datasets
- 4.2 Implementation Details
- 4.3 Experiment Results on Synapse Dataset
- 4.4 Experiment Results on ACDC Dataset
- 4.5 Ablation Study
- 4.6 Discussion
- 5 Conclusion
- References
- Self-attention Capsule Network for Tissue Classification in Case of Challenging Medical Image Statistics
- 1 Background
- 2 Related Work
- 2.1 CapsNet Architecture
- 2.2 Self-attention Mechanism
- 2.3 Self Attention/Capsule Networks for Tissue Classification
- 2.4 Capsule Networks with Self Attention Mechanism
- 3 The Proposed Model
- 3.1 Implementation Details
- 4 Experiments
- 4.1 Datasets
- 4.2 Performance Evaluation
- 5 Results
- 5.1 Qualitative Evaluation
- 5.2 Quantitative Evaluation and Comparison with Other Common Techniques
- 5.3 Ablation Study
- 5.4 Generalizability Across Domains
- 6 Computational Load
- 7 Discussion and Conclusion
- References
- ReLaX: Retinal Layer Attribution for Guided Explanations of Automated Optical Coherence Tomography Classification
- 1 Introduction
- 2 Literature Review
- 2.1 OCT Classification
- 2.2 CNN Visualization
- 2.3 Attempts at Explainable OCT Classification
- 3 Methodology
- 3.1 Classification Model
- 3.2 Gradient-Weighted Class Activation Mapping
- 3.3 Segmentation Model
- 3.4 Calculating Retinal Layer Attribution
- 4 Results
- 4.1 Data Set
- 4.2 Data Preprocessing
- 4.3 Results of the Classification Models
- 4.4 Preliminary Qualitative Interpretability Results
- 4.5 Full ReLaX Interpretability Results
- 4.6 Connection to Human Eyecare Professionals
- 4.7 Error Analysis
- 5 Discussion
- 6 Conclusion
- References
- Neural Registration and Segmentation of White Matter Tracts in Multi-modal Brain MRI
- 1 Introduction
- 2 Methods
- 2.1 Loss Function
- 2.2 Data Augmentation
- 2.3 Initialization
- 3 Experiments and Results
- 3.1 Data
- 3.2 Preprocessing
- 3.3 Implementation
- 3.4 Unregistration of T1w and DEC Data
- 3.5 Results
- 3.6 Ablation Studies
- 4 Conclusions
- References
- Complementary Phase Encoding for Pair-Wise Neural Deblurring of Accelerated Brain MRI
- 1 Introduction
- 2 Proposed Methodology
- 2.1 Registration
- 2.2 Architecture
- 3 Experiments
- 3.1 Dataset
- 3.2 Training Details
- 3.3 Results
- 4 Conclusions
- References
- Frequency Dropout: Feature-Level Regularization via Randomized Filtering
- 1 Introduction
- 2 Related Work
- 2.1 Image Filtering in CNNs
- 2.2 Dropout-Based Regularization
- 3 Background and Preliminaries
- 3.1 Image Filters
- 3.2 Spatial Dropout
- 4 Frequency Dropout
- 4.1 FD with Randomized Filtering
- 4.2 FD with Gaussian Filtering
- 5 Experiments and Results
- 5.1 Image Classification
- 5.2 Domain Adaptation
- 5.3 Ablation Study
- 6 Discussion and Conclusion
- References
- PVBM: A Python Vasculature Biomarker Toolbox Based on Retinal Blood Vessel Segmentation
- 1 Introduction
- 1.1 Prior Works
- 1.2 Research Gap and Objectives
- 2 Methods
- 2.1 Dataset
- 2.2 Digital Vasculature Biomarkers
- 3 Results
- 4 Discussion and Future Work
- References
- Simultaneous Detection and Classification of Partially and Weakly Supervised Cells
- 1 Introduction
- 2 Related Work
- 3 Method
- 3.1 Weakly Supervised Segmentation Using Partial Points
- 3.2 Multiclass Label Encoding
- 3.3 Network Architecture
- 3.4 Loss Functions
- 4 Experiments
- 4.1 Dataset
- 4.2 Evaluation Metrics
- 4.3 Implementation Details
- 4.4 Results
- 5 Conclusions
- References
- Deep-ASPECTS: A Segmentation-Assisted Model for Stroke Severity Measurement
- 1 Introduction
- 2 Related Work
- 3 Dataset
- 4 Methods
- 5 Results
- 5.1 How Does Volume of Infarct Influence the Segmentation Map?
- 5.2 How Do Different Models Segment MCA Territory?
- 5.3 Model Performance for Different ASPECT Score
- 5.4 Number of Parameters for Models and Their Computation Time
- 6 Inter-reader Variability
- 7 Conclusions
- References
- ExSwin-Unet: An Unbalanced Weighted Unet with Shifted Window and External Attentions for Fetal Brain MRI Image Segmentation
- 1 Introduction
- 2 Method
- 2.1 Window-Based Attention Block
- 2.2 External Attention Block
- 2.3 Unbalanced Unet Architecture
- 2.4 Adaptive Weighting Adjustments
- 2.5 Dual Loss Functions
- 3 Experimental Results
- 3.1 Datasets
- 3.2 Implement Details
- 3.3 Comparison with SOTA Methods
- 3.4 Ablation Studies
- 4 Discussions and Limitations
- 5 Conclusion
- References
- Contour Dice Loss for Structures with Fuzzy and Complex Boundaries in Fetal MRI
- 1 Introduction
- 2 Background and Related Work
- 2.1 Placenta Segmentation
- 2.2 Fetal Brain Segmentation
- 2.3 Boundary Loss Functions
- 3 Methods
- 3.1 Contours and Bands Extraction
- 3.2 Contour Dice Loss Computation
- 4 Experimental Results
- 4.1 Datasets and Annotations
- 4.2 Experimental Studies
- 5 Conclusions
- References
- Multi-scale Multi-task Distillation for Incremental 3D Medical Image Segmentation
- 1 Introduction
- 2 Related Work
- 3 Method
- 3.1 Incremental Medical Image Segmentation
- 3.2 Multi-task Knowledge Distillation
- 3.3 Multi-scale Contrastive Learning
- 3.4 Masking
- 3.5 Combine with Memory Based Approach
- 4 Experiments
- 4.1 Implementation Details
- 4.2 Comparison with Common Strategies
- 4.3 Ablation Study
- 4.4 Qualitative Analysis
- 5 Conclusion
- References
- A Data-Efficient Deep Learning Framework for Segmentation and Classification of Histopathology Images
- 1 Introduction
- 2 Background
- 2.1 Application of Artificial Intelligence for Autoimmune Diseases
- 2.2 Segmentation
- 2.3 Classification
- 3 Methodology
- 3.1 Segmentation
- 3.2 Classification
- 4 Experimental Details
- 4.1 Dataset
- 4.2 Segmentation
- 4.3 Classification
- 5 Results and Discussion
- 5.1 Segmentation
- 5.2 Classification
- 6 Conclusion
- A Appendix
- A.1 Expansion of Results
- A.2 Autoencoder with Efficientnet Encoder for Segmentation
- A.3 Metrics Description
- A.4 Effect of Different Weights
- A.5 Expansion on Experimental Details
- References
- Bounded Future MS-TCN++ for Surgical Gesture Recognition
- 1 Introduction
- 2 Related Work
- 3 Methods
- 3.1 Dataset
- 3.2 Architecture
- 3.3 Feature Extractor Implementation Details
- 4 Experimental Setup and Results
- 4.1 Experimental Setup
- 4.2 Evaluation Method
- 4.3 Evaluation Metrics
- 4.4 Experimental Studies
- 4.5 Results
- 5 Discussion and Conclusions
- References
- Anatomy-Aware Contrastive Representation Learning for Fetal Ultrasound
- 1 Introduction
- 2 Related Work
- 3 Fetal Ultrasound Imaging Dataset
- 4 Method
- 4.1 Problem Formulation
- 4.2 Anatomy-Aware Contrastive Learning
- 4.3 Implementation Details
- 5 Experiments on Transfer Learning
- 5.1 Evaluation on Standard Plane Detection
- 5.2 Evaluation on Recognition of First-Trimester Anatomies
- 5.3 Evaluation on Segmentation of NT and CRL
- 6 Analysis
- 6.1 Partial Fine-Tuning
- 6.2 Visualization of Feature Representations
- 6.3 Impact of Data Granularity on AWCL
- 6.4 Impact of Anatomy Ratio on AWCL
- 7 Conclusion
- References
- Joint Calibrationless Reconstruction and Segmentation of Parallel MRI
- 1 Introduction
- 2 Background
- 2.1 Forward Model
- 2.2 Calibrationless PMRI Recovery
- 2.3 End-to-End Multi-task Training Approaches
- 3 Proposed Framework
- 3.1 Image Domain Deep-SLR Formulation
- 3.2 Joint Reconstruction-Segmentation Framework
- 3.3 Few-Shot Learning for Semantic Segmentation
- 4 Implementation Details
- 4.1 Architecture of the CNNs and Training Details
- 4.2 State-of-the-Art Methods for Comparison
- 5 Experiments and Results
- 5.1 Calibration-Free PMRI
- 5.2 Segmentation Quality Comparison
- 5.3 Benefit of Shared Encoder Architecture
- 5.4 Comparison with State-of-the-Art
- 6 Discussion
- 7 Conclusion
- References
- Patient-Level Microsatellite Stability Assessment from Whole Slide Images by Combining Momentum Contrast Learning and Group Patch Embeddings
- 1 Introduction
- 1.1 CNN-Based MSI/MSS Classification from WSI Data
- 1.2 Self-supervision for Patch Embbedings
- 2 Related Work
- 2.1 CNN-Based MSS/MSI Classification
- 2.2 Self Supervision for Patch Embbedings
- 2.3 Hypothesis and Contributions
- 3 Data
- 4 Method and Model
- 4.1 Overview
- 4.2 Stage 1: Training a Self-supervised Feature Extractor for Patch-Level Embeddings
- 4.3 Stage 2: Training a Supervised Classifier on Patch Embedding Groups
- 4.4 Evaluation
- 5 Experiments and Results
- 5.1 Standard Dataset
- 5.2 Balanced Validation Set
- 6 Conclusions
- References
- Segmenting Glandular Biopsy Images Using the Separate Merged Objects Algorithm
- 1 Introduction and Related Work
- 2 Separation of Merged Objects
- 2.1 Line Representation
- 2.2 Optimization
- 2.3 Loss Function
- 2.4 Recursive Cutting Decision
- 3 Use Case: Colon Biopsy Crypt Segmentation
- 3.1 Data Set
- 3.2 Data Preparation
- 3.3 Separation Algorithm
- 4 Experimental Results
- 5 Conclusions and Future Work
- References
- qDWI-Morph: Motion-Compensated Quantitative Diffusion-Weighted MRI Analysis for Fetal Lung Maturity Assessment
- 1 Introduction
- 1.1 Background
- 1.2 Fetal Lung Maturity Assessment with qDWI
- 2 Method
- 2.1 DNN Architecture
- 2.2 Convergence Criteria
- 2.3 Bio-physically-informed Loss Function
- 3 Experiments
- 3.1 Clinical Data-Set
- 3.2 Experiment Goals
- 3.3 Experimental Setup
- 3.4 Implementation Details
- 4 Results
- 5 Conclusions
- References
- Estimating Withdrawal Time in Colonoscopies
- 1 Introduction
- 2 Related Work
- 3 Methods
- 3.1 The Estimation Pipeline
- 3.2 Unsupervised Visual Odometry
- 3.3 The Landmark Prediction Module
- 3.4 Representation Extraction
- 3.5 Feature Filtering
- 3.6 Per-frame Algorithmic Phase Classification
- 3.7 Change-Point Detection
- 4 Results
- 4.1 The Dataset
- 4.2 Hyperparameters
- 4.3 Results
- 4.4 Sensitivity Analysis
- 4.5 Analysis of Feature Importance
- 4.6 Ablation Studies
- 5 Conclusions
- References
- Beyond Local Processing: Adapting CNNs for CT Reconstruction
- 1 Introduction
- 2 Related Work
- 3 Adapting Radon Space Representation for CNNs
- 3.1 The Filtered Backprojection Algorithm
- 3.2 Localizing the Inverse Transformation
- 4 Experiments
- 4.1 CT Reconstruction Tasks
- 4.2 Reconstruction Methods
- 4.3 Results
- 5 Conclusion
- References
- CL-GAN: Contrastive Learning-Based Generative Adversarial Network for Modality Transfer with Limited Paired Data
- 1 Introduction
- 2 Related Work
- 3 Method
- 3.1 Intra-modality Training
- 3.2 Cross-modality Training
- 3.3 CL-GAN Training
- 4 Experiments
- 4.1 Dataset and Implementation Details
- 4.2 Experimental Results
- 5 Conclusion
- References
- IMPaSh: A Novel Domain-Shift Resistant Representation for Colorectal Cancer Tissue Classification
- 1 Introduction
- 2 The Proposed Method
- 2.1 Self-supervised Contrastive Learning
- 2.2 Learning Pretext-Invariant Representation
- 2.3 Momentum Contrast
- 2.4 Transfer Learning Task
- 3 Experiment
- 3.1 The Datasets
- 3.2 Experimental Settings
- 3.3 Comparative Experiments
- 3.4 Ablation Study
- 4 Results and Discussion
- 5 Conclusion
- References
- Surgical Workflow Recognition: From Analysis of Challenges to Architectural Study
- 1 Introduction
- 1.1 Related Work
- 2 Analysis of Challenges: Local vs. Global
- 3 Methodology
- 3.1 Visual Backbone - Stage 1
- 3.2 Temporal Models - Stage 2
- 4 Experimental Setup
- 4.1 Datasets
- 4.2 Architecture Settings
- 4.3 Evaluation Metrics and Baselines
- 5 Results and Discussion
- 6 Conclusion
- References
- RVENet: A Large Echocardiographic Dataset for the Deep Learning-Based Assessment of Right Ventricular Function
- 1 Introduction
- 2 Related Work
- 2.1 Datasets
- 2.2 Methods
- 3 Overview of the RVENet Dataset
- 3.1 Data Collection
- 3.2 Data Labeling
- 3.3 Composition of the Dataset
- 3.4 Data De-identification and Access
- 4 Benchmark Models
- 4.1 Methodology
- 4.2 Results
- 5 Discussion
- 5.1 Potential Clinical Application
- 5.2 Summary
- References
- W08 - Computer Vision for Metaverse
- W08 - Computer Vision for Metaverse
- Initialization and Alignment for Adversarial Texture Optimization*-4pt
- 1 Introduction
- 2 Related Work
- 3 Approach
- 3.1 Texture Initialization (TexInit)
- 3.2 Texture Smoothing (TexSmooth)
- 4 Experiments
- 4.1 Experimental Setup
- 4.2 Experimental Evaluation
- 5 Conclusion
- References
- SIGNet: Intrinsic Image Decomposition by a Semantic and Invariant Gradient Driven Network for Indoor Scenes
- 1 Introduction
- 2 Related Works
- 3 Methodology
- 3.1 Priors
- 3.2 Network Architecture
- 3.3 Dataset
- 3.4 Loss Functions and Training Details
- 4 Experiments
- 4.1 Ablation Study
- 4.2 Comparison to State of the Art
- 5 Conclusions
- References
- Implicit Map Augmentation for Relocalization
- 1 Introduction
- 2 Related Work
- 3 Method
- 3.1 Monocular Depth Pseudo Supervision
- 3.2 Active Sampling
- 3.3 Mapping and Relocalization
- 4 Experiments
- 4.1 Datasets
- 4.2 Implementation Details
- 4.3 Evaluation
- 4.4 Ablation Study
- 4.5 Limitations
- 5 Conclusions
- References
- Social Processes: Self-supervised Meta-learning Over Conversational Groups for Forecasting Nonverbal Social Cues*-4pt
- 1 Introduction
- 2 Related Work
- 3 Social Cue Forecasting: Task Formalization
- 3.1 Formalization and Distinction from Prior Task Formulations
- 4 Method Preliminaries
- 5 Social Processes: Methodology
- 6 Experiments and Results
- 6.1 Experimental Setup
- 6.2 Evaluation on Synthesized Behavior Data
- 6.3 Evaluation on Real-World Behavior Data
- 6.4 Ablations
- 7 Discussion
- References
- Photo-Realistic 360 Head Avatars in the Wild
- 1 Introduction
- 2 Related Work
- 3 Method
- 4 Data
- 5 Experiments
- 6 Conclusions
- References
- AvatarGen: A 3D Generative Model for Animatable Human Avatars
- 1 Introduction
- 2 Related Works
- 3 Method
- 3.1 Overview
- 3.2 Representations of 3D Avatars
- 3.3 Generative 3D Human Modeling
- 3.4 Geometry-Aware Human Modeling
- 3.5 Training
- 4 Experiments
- 4.1 Comparisons
- 4.2 Ablation Studies
- 4.3 Applications
- 5 Conclusion
- References
- INGeo: Accelerating Instant Neural Scene Reconstruction with Noisy Geometry Priors
- 1 Introduction
- 2 Related Works
- 3 Method
- 3.1 Can Pretrained Occupancy Grids Accelerate Training?
- 3.2 Initialize Occupancy Grids with Geometry Priors
- 4 Experiments
- 4.1 Experiments Settings
- 4.2 Training Speed Comparison with SotA
- 5 Conclusion
- References
- Number-Adaptive Prototype Learning for 3D Point Cloud Semantic Segmentation*-4pt
- 1 Introduction
- 2 Related Work
- 2.1 3D Point Cloud Semantic Segmentation
- 2.2 Mask Classification for Image Segmentation
- 3 Methodology
- 3.1 Overview of 3D Semantic Segmentation Paradigm
- 3.2 Number-Adaptive Prototype Learning
- 3.3 Model Architecture
- 3.4 Prototype Dropout Training Strategy
- 4 Experiments
- 4.1 Implementation Details
- 4.2 Results
- 5 Conclusions
- References
- Self-supervised 3D Human Pose Estimation in Static Video via Neural Rendering
- 1 Introduction
- 2 Related Work
- 3 Method
- 3.1 Template and Volume Rendering
- 3.2 Pose Regression and Shape Transformation
- 3.3 Kinematic Chain
- 3.4 Loss Function
- 4 Experiments
- 5 Results
- 6 Conclusion
- References
- Racial Bias in the Beautyverse: Evaluation of Augmented-Reality Beauty Filters*-4pt
- 1 Introduction
- 2 Related Work
- 3 The Implicit Racial Bias in AR-Based Beauty Filters
- 3.1 Experimental Setup
- 3.2 Results
- 4 Future Work and Conclusion
- References
- .26em plus .1em minus .1emLWA-HAND: Lightweight Attention Hand for Interacting Hand Reconstruction
- 1 Introduction
- 2 Related Work
- 2.1 Hand Reconstruction
- 2.2 Real-Time Hand Reconstruction
- 3 Formulation
- 3.1 Two-Hand Mesh Representation
- 3.2 Overview
- 4 Light Former Graph
- 4.1 Lightweight Feature Attention Module
- 4.2 Pyramid Cross Image and Graph Bridge Module
- 4.3 Lightweight Cross Hand Attention Module
- 4.4 Loss Function
- 5 Experiment
- 5.1 Experimental Settings
- 5.2 Datasets
- 5.3 Quantitative Results
- 6 Discussion
- 6.1 Conclusion
- 6.2 Limitation
- References
- Neural Mesh-Based Graphics
- 1 Introduction
- 2 Related Work
- 3 Method
- 3.1 Dual Feature Rasterization
- 3.2 Anisotropic Features
- 3.3 Split Neural Rendering
- 3.4 Hybrid Training Objective
- 4 Results
- 5 Conclusion
- References
- One-Shot Learning for Human Affordance Detection
- 1 Introduction
- 2 Related Work
- 3 Method
- 3.1 A Spatial Descriptor for Spatial Interactions
- 3.2 Human Affordances Detection
- 4 Experiments
- 5 Conclusion
- References
- Author Index
System requirements
File format: PDF
Copy protection: Watermark-DRM (Digital Rights Management)
System requirements:
- Computer (Windows; MacOS X; Linux): Use the free software Adobe Reader, Adobe Digital Editions, or any other PDF viewer of your choice (see eBook Help).
- Tablet/Smartphone (Android; iOS): Install the free app Adobe Digital Editions or another reading app for eBooks, e.g., PocketBook (see eBook Help).
- E-reader: Bookeen, Kobo, Pocketbook, Sony, Tolino and many more (only limited: Kindle).
The file format PDF always displays a book page identically on any hardware. This makes PDF suitable for complex layouts such as those used in textbooks and reference books (images, tables, columns, footnotes). Unfortunately, on the small screens of e-readers or smartphones, PDFs are rather annoying, requiring too much scrolling.
This eBook uses Watermark-DRM, a „soft” copy protection. This means that there are no technical restrictions to prevent illegal distribution. However, there is a personalised watermark embedded in the eBook that can be used to identify the purchaser of the eBook in the event of misuse and to provide evidence for legal purposes.
For more information, see our eBook Help page.