
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 VII
- W28 - Computational Aspects of Deep Learning
- W28 - Computational Aspects of Deep Learning
- EdgeNeXt: Efficiently Amalgamated CNN-Transformer Architecture for Mobile Vision Applications
- 1 Introduction
- 2 Related Work
- 3 EdgeNeXt
- 4 Experiments
- 4.1 Dataset
- 4.2 Implementation Details
- 4.3 Image Classification
- 4.4 ImageNet-21K Pretraining
- 4.5 Inference on Edge Devices
- 4.6 Object Detection
- 4.7 Semantic Segmentation
- 5 Ablations
- 6 Qualitative Results
- 7 Conclusion
- References
- Continual Inference: A Library for Efficient Online Inference with Deep Neural Networks in PyTorch
- 1 Introduction
- 2 Continual Inference Networks
- 3 Library Design
- 3.1 Principles
- 3.2 Core Modules
- 3.3 Composition Modules
- 4 Performance Comparisons
- 5 Conclusion
- References
- Hydra Attention: Efficient Attention with Many Heads
- 1 Introduction
- 2 Related Work
- 3 Hydra Attention
- 3.1 The Kernel Trick
- 3.2 Multi-head Attention
- 3.3 Adding Heads
- 3.4 The Hydra Trick
- 3.5 Relation to Other Works
- 4 Experiments
- 4.1 The Choice of Kernel
- 4.2 Visualizing Hydra Attention
- 4.3 Which Layers Can We Replace?
- 4.4 Results
- 5 Conclusion and Future Directions
- References
- BiTAT: Neural Network Binarization with Task-Dependent Aggregated Transformation
- 1 Introduction
- 2 Related Work
- 3 Weight Importance for Quantization-Aware Training
- 3.1 Problem Statement
- 3.2 Disentangling Weight Dependencies via Input-dependent Orthornormal Transformation
- 3.3 Cross-Layer Weight Correlation Impacts Model Performance
- 4 Task-Dependent Weight Transformation for Neural Network Binarization
- 4.1 Layer-Progressive Quantization with Block-Wise Weight Dependency
- 4.2 Cost-Efficient BiTAT via Aggregated Weight Correlation Using Reduction Matrix
- 5 Experiments
- 5.1 Quantitative Analysis
- 5.2 Qualitative Analysis
- 6 Conclusion
- References
- Power Awareness in Low Precision Neural Networks
- 1 Introduction
- 2 Related Work
- 2.1 Avoiding Multiplications
- 2.2 Quantization
- 3 Power Consumption of a Conventional DNN
- 3.1 Dynamic Power vs. Simplistic Power Consumption Indicators
- 3.2 Bit Toggling Simulation
- 4 Switching to Unsigned Arithmetic
- 5 Removing the Multiplier
- 5.1 Power Aware Weight Quantization
- 5.2 Power Consumption
- 6 Experiments
- 6.1 PANN at Post Training
- 6.2 PANN for Quantization Aware Training
- 7 Conclusion
- References
- Augmenting Legacy Networks for Flexible Inference
- 1 Introduction
- 2 Related Work
- 2.1 Static Architectures
- 2.2 Variable Architectures
- 2.3 Internal vs. External Routing Policy
- 3 Legacy Augmentation for Flexible Inference (LeAF)
- 3.1 LeAF Overview
- 3.2 Training LeAF
- 4 Evaluation
- 4.1 Methodology
- 4.2 LeAF-ResNet-50: On High End GPU
- 4.3 LeAF-ResNet-50 Latency on Mobile
- 4.4 Analysis as Function of Device
- 4.5 LeAF on MobileNetV2
- 5 Discussion and Conclusion
- References
- Deep Neural Network Compression for Image Inpainting
- 1 Introduction
- 2 Related Works
- 3 Proposed Algorithm
- 3.1 Channel Pruning for Image Inpainting
- 3.2 Knowledge Distillation for Image Inpainting
- 4 Experiments
- 4.1 Training Setting
- 4.2 Pruning Effectiveness
- 4.3 Knowledge Distillation Effectiveness
- 4.4 Ablation Study on the Proposed Pruning Method
- 4.5 Ablation Study on the Proposed Distillation Method
- 4.6 Memory Reduction and Speedup
- 5 Conclusion
- References
- QFT: Post-training Quantization via Fast Joint Finetuning of All Degrees of Freedom
- 1 Introduction
- 2 Prior Work
- 3 Methods
- 3.1 Quantization Description by Over-Parameterized Scale Tensors
- 3.2 QFT: Downscaled all-DoF QAT as a Simple PTQ Baseline
- 4 Experiments
- 5 Conclusions and Outlook
- References
- Searching for N:M Fine-grained Sparsity of Weights and Activations in Neural Networks
- 1 Introduction
- 2 Related Work
- 2.1 Unstructured and Structured Weight Pruning
- 2.2 N:M Fine-grained Structured Weight Pruning
- 2.3 Activation Pruning
- 2.4 Neural Architecture Search
- 3 Method
- 3.1 Pruning Search-Layer
- 3.2 Gumbel-Softmax
- 3.3 Training for N:M Sparse Activations Using STE
- 3.4 Controlling the NAS Choices Using Memory Bandwidth Penalty
- 3.5 Finalizing the Search-Layers Choices
- 4 Results
- 4.1 Image Classification
- 4.2 Ablation Study
- 4.3 NAS Pruning Choice Analysis
- 5 Conclusion
- References
- W29 - Computer Vision for Civil and Infrastructure Engineering
- W29 - Computer Vision for Civil and Infrastructure Engineering
- Image Illumination Enhancement for Construction Worker Pose Estimation in Low-light Conditions
- 1 Introduction
- 2 Related Work
- 2.1 Human Pose Estimation Methods in Construction
- 2.2 Image Illumination Enhancement
- 2.3 Construction Field Data Set Expansion
- 2.4 Research Challenges and Objectives
- 3 Methodology
- 3.1 Overall Framework
- 3.2 Proposed UIRE-Net
- 4 Experiments and Results
- 4.1 Data Sets
- 4.2 Comparison with State-of-the-art Methods
- 4.3 Real-world Evaluation
- 4.4 Ablation Studies
- 5 Conclusions
- References
- Towards an Error-free Deep Occupancy Detector for Smart Camera Parking System
- 1 Introduction
- 2 Related Work
- 2.1 Automatic Parking Occupancy Detection
- 2.2 Deep Model Uncertainty
- 2.3 Qualitative Comparison Between Existing Approaches and the Proposed System
- 3 Proposed Method and Dataset
- 3.1 Overall System Architecture
- 3.2 Dataset
- 3.3 Automatic Parking Occupancy Detector
- 3.4 Result Filter
- 4 Experiment
- 4.1 Experimental Setup
- 4.2 Detection Performance
- 4.3 Result Filter Performance
- 4.4 Optimal Routing and Parking Assignment
- 5 Conclusion & Future Work
- References
- CrackSeg9k: A Collection and Benchmark for Crack Segmentation Datasets and Frameworks
- 1 Introduction
- 2 Related Works
- 2.1 Rule Based Image Processing Techniques
- 2.2 Data Driven Methods
- 3 Dataset
- 3.1 Dataset Details
- 3.2 Dataset Refinement
- 3.3 Categorizing Dataset Based on Crack Types
- 4 Methodology and Experiments
- 4.1 Pipeline
- 4.2 Unsupervised Feature Extraction
- 4.3 Supervised Techniques
- 4.4 Using Semi-supervised DINO Features as Prior for Segmentation
- 4.5 Comparative Analysis of Results
- 5 Conclusion
- References
- PriSeg: IFC-Supported Primitive Instance Geometry Segmentation with Unsupervised Clustering
- 1 Introduction
- 2 Background
- 2.1 PCD-vs-DI Object Instance Detection and Segmentation
- 2.2 Gaps in Knowledge and Research Objectives
- 3 Proposed Solution
- 3.1 Scope and Overview
- 3.2 IFC-Based Instance Descriptor
- 3.3 Unsupervised Clustering
- 4 Implementation and Experiments
- 4.1 Data Acquisition and Pre-processing
- 4.2 Object Instance Segmentation
- 5 Discussion
- 6 Conclusions
- References
- Depth Contrast: Self-supervised Pretraining on 3DPM Images for Mining Material Classification
- 1 Introduction
- 1.1 Measurement System
- 1.2 Related Work
- 2 3DPM Dataset
- 3 Methodology and Experiments
- 3.1 Transfer Learning
- 3.2 Supervised Downstream Task
- 3.3 Depth Contrast - Self-supervised Learning
- 3.4 Experiments
- 4 Result and Discussion
- 4.1 Raw Image of Depth Maps Scales Representations and Improves Learning
- 4.2 CNN with Reduced Parameters Performs Better
- 4.3 Self-supervised Method Depth Contrast Significantly Improves Downstream Task Performance and Reduces Human Annotations
- 4.4 Self-supervised Method Depth Contrast Encourages Performance Generalization
- 5 Conclusion and Future Scope
- References
- Facilitating Construction Scene Understanding Knowledge Sharing and Reuse via Lifelong Site Object Detection
- 1 Introduction
- 2 Related Work
- 2.1 Object Detection
- 2.2 Metadata Standards in Construction Domains
- 2.3 Lifelong Learning
- 3 Lifelong Construction Resource Detection Benchmark
- 3.1 Open-source Datasets for Detecting Construction Resources
- 3.2 Label Space Transformation and Unification
- 4 Informativeness-based Lifelong Learning for Construction Resource Detection
- 4.1 Preliminaries
- 4.2 Informativeness-based Lifelong Object Detector
- 5 Experiments
- 5.1 Model Performance
- 6 Conclusions
- References
- Model-Assisted Labeling via Explainability for Visual Inspection of Civil Infrastructures
- 1 Introduction
- 2 Related Work
- 2.1 Weakly Supervised Segmentation and Localization
- 2.2 Annotation Tools
- 3 Model-Assisted Labeling Framework
- 3.1 Proposal Generation and Refinement
- 3.2 Interactive Aspects
- 4 Evaluation
- 4.1 Data Preparation
- 4.2 Classifier Training
- 4.3 Estimation of Time Saved
- 5 Conclusions
- References
- A Hyperspectral and RGB Dataset for Building Façade Segmentation
- 1 Introduction
- 2 Related Work
- 3 Dataset
- 4 Experiments
- 4.1 Metrics
- 4.2 Segmentation Models
- 4.3 Objective Function
- 4.4 Implementation
- 5 Results and Discussion
- 6 Conclusion
- References
- Improving Object Detection in VHR Aerial Orthomosaics
- 1 Introduction
- 2 Related Work
- 3 Dataset
- 4 Methods
- 4.1 Baseline
- 4.2 Neighbour NMS
- 4.3 Intersection over Area Filtering
- 4.4 ResnetYolo
- 4.5 RGBD Fusion
- 5 Experiments
- 5.1 Solar Panel Detection
- 5.2 Swimming Pool Detection
- 6 Conclusions
- References
- Active Learning for Imbalanced Civil Infrastructure Data
- 1 Introduction
- 2 Related Work
- 3 Background
- 3.1 Image Data
- 3.2 Instance Segmentation Annotations
- 4 Method
- 4.1 Active Learning for Heavily Imbalanced Data
- 4.2 Instance Segmentation to Classification Dataset Conversion
- 5 Evaluation
- 5.1 Civil Infrastructure Classification Dataset
- 5.2 Experiment Setup
- 5.3 Results
- 6 Conclusion
- References
- UAV-Based Visual Remote Sensing for Automated Building Inspection
- 1 Introduction
- 2 Related Works
- 2.1 Distance Between Adjacent Structures
- 2.2 Plan Shape and Roof Area Estimation
- 2.3 Roof Layout Estimation
- 3 Data Collection
- 4 Methodology
- 4.1 Distance Between Adjacent Buildings
- 4.2 Plan Shape and Roof Area Estimation
- 4.3 Roof Layout Estimation
- 5 Results
- 5.1 Distance Between Adjacent Buildings
- 5.2 Plan Shape and Roof Area Estimation
- 5.3 Roof Layout Estimation
- 6 Discussion
- 7 Conclusion
- References
- ConSLAM: Periodically Collected Real-World Construction Dataset for SLAM and Progress Monitoring
- 1 Introduction
- 2 Existing Datasets
- 3 Methodology
- 3.1 Sensors and Devices
- 3.2 Intrinsic Calibrations of the Sensors
- 3.3 Extrinsic Calibrations of the Sensors
- 3.4 Data Collection System of the Hand-Held Scanner
- 3.5 Ground-Truth Trajectories
- 4 Dataset: ConSLAM
- 4.1 Dataset Structure
- 4.2 Practical Application: Projecting LiDAR Points onto Corresponding Images
- 5 Conclusion and Future Direction
- References
- NeuralSI: Structural Parameter Identification in Nonlinear Dynamical Systems
- 1 Introduction
- 2 Related Work
- 3 Structural Problem - PDE Derivation
- 3.1 Problem Description
- 4 NeuralSI
- 4.1 Discretization of Space
- 4.2 The Proposed NeuralSI Schematic
- 4.3 Training Data Generation
- 4.4 Network Architecture and Training
- 5 Results and Performance
- 5.1 Results
- 5.2 Hyperparameter Investigation
- 6 Comparison of NeuralSI with a Direct Response Mapping Deep Neural Network and a PINN
- 7 Conclusion
- References
- A Geometric-Relational Deep Learning Framework for BIM Object Classification
- 1 Introduction
- 2 Related Work
- 2.1 3D Object Recognition
- 2.2 BIM Object Classification
- 2.3 BIM Object Datasets
- 3 Geometric-Relational Deep Learning Framework
- 3.1 Module Designs
- 3.2 Implementation Details
- 4 Implementation
- 4.1 Relational Models
- 4.2 Training Configurations
- 5 IFCNet++ Dataset
- 5.1 Dataset Overview
- 5.2 Relational Feature Design
- 5.3 Data Collection and Processing
- 6 Experiments
- 6.1 Testing Metrics
- 6.2 Confusion Rate
- 6.3 Corrected Classification Results
- 6.4 Computational Cost
- 6.5 Ablation Study
- 7 Conclusion
- References
- Generating Construction Safety Observations via CLIP-Based Image-Language Embedding
- 1 Introduction
- 2 Related Work
- 3 Methodology
- 3.1 Dataset Development
- 3.2 Construction CLIP Fine-Tuning
- 3.3 CLIP Prefix Captioning
- 4 Experiment
- 4.1 Construction CLIP Fine-Tuning
- 4.2 CLIP Prefix Captioning
- 5 Conclusion
- References
- W30 - AI-Enabled Medical Image Analysis: Digital Pathology and Radiology/COVID-19
- W30 - AI-Enabled Medical Image Analysis: Digital Pathology and Radiology/COVID-19
- Harmonization of Diffusion MRI Data Obtained with Multiple Head Coils Using Hybrid CNNs
- 1 Materials and Methods
- 1.1 Image Data and Preprocessing
- 1.2 Methods
- 1.3 Comparison Methods
- 2 Results
- 3 Discussion and Conclusion
- References
- CCRL: Contrastive Cell Representation Learning
- 1 Introduction
- 2 Previous Works
- 2.1 Self-supervised Learning
- 2.2 Cell Classification in Histopathology
- 3 Method
- 4 Experiments
- 4.1 Evaluation Metrics
- 4.2 Datasets
- 4.3 Data Preparation
- 4.4 Implementation Details
- 4.5 Results
- 4.6 Ablation Study
- 5 Discussion and Conclusion
- References
- Automatic Grading of Cervical Biopsies by Combining Full and Self-supervision
- 1 Introduction
- 2 Related Work
- 3 Dataset and Problem Setting
- 4 Proposed Architecture
- 4.1 Multiple Instance Learning and Attention
- 4.2 Self-supervised Learning
- 4.3 Cost-Sensitive Training
- 4.4 Mixed Supervision
- 5 Understanding the Feature Extractor with Activation Maximization
- 6 Experimental Setting
- 6.1 WSI Preprocessing
- 6.2 Data Splits for Cross-Validation
- 6.3 Feature Extractor Pre-training
- 6.4 Whole Slide Classification
- 6.5 Feature Visualization
- 7 Results
- 7.1 Self-supervised Fine-Tuning
- 7.2 Pre-training Policy Comparison
- 7.3 Number of Annotations vs Number of Epochs
- 7.4 Feature Visualization
- 8 Discussion
- References
- When CNN Meet with ViT: Towards Semi-supervised Learning for Multi-class Medical Image Semantic Segmentation
- 1 Introduction
- 2 Related Work
- 3 Methodology
- 3.1 CNN & ViT
- 3.2 Feature-Learning Module
- 3.3 Guidance Module
- 3.4 Objective
- 4 Experiments and Results
- 5 Conclusions
- References
- Using Whole Slide Image Representations from Self-supervised Contrastive Learning for Melanoma Concordance Regression
- 1 Introduction
- 2 Methods
- 2.1 Data Collection and Characteristics
- 2.2 Melanoma Concordance Regression Deep Learning Architecture
- 3 Results
- 3.1 Malignant Classification
- 3.2 Ablation Studies
- 4 Conclusions
- References
- Explainable Model for Localization of Spiculation in Lung Nodules
- 1 Introduction
- 2 Previous Work
- 3 Methodology
- 3.1 Dataset
- 3.2 Classifier Network
- 3.3 Attribution Method
- 3.4 Metrics
- 4 Results
- 5 Conclusions
- References
- Self-supervised Pretraining for 2D Medical Image Segmentation
- 1 Introduction
- 2 Related Work
- 2.1 Self-supervised Learning
- 2.2 Self-supervised Learning for Medical Image Processing
- 3 Methods and Data
- 3.1 Pretraining Approaches
- 3.2 Applied Self-supervised Learning Method
- 3.3 Cardiac Segmentation Dataset
- 3.4 Model Architecture
- 4 Experiments
- 4.1 Training Phases
- 4.2 Data-Efficient Learning
- 5 Results
- 5.1 Convergence and Stability
- 5.2 Data-Efficiency
- 5.3 Domain-Specific Pretraining Epochs
- 6 Conclusion
- References
- CMC_v2: Towards More Accurate COVID-19 Detection with Discriminative Video Priors
- 1 Introduction
- 2 Related Work
- 2.1 COVID-19 Detection
- 2.2 Advanced Network Architecture
- 3 Methodology
- 3.1 Recap of CMC_v1
- 3.2 Improving COVID-19 Detection with CMC_v2
- 4 Dataset
- 5 Experiments
- 5.1 Implementation Details
- 5.2 Evaluation Metrics
- 5.3 Ablation Studies on COVID-19 Detection Challenge
- 5.4 Results on COVID-19 Detection Challenge Leaderboard
- 5.5 Visualization Results
- 6 Conclusions
- References
- COVID Detection and Severity Prediction with 3D-ConvNeXt and Custom Pretrainings
- 1 Introduction
- 2 Related Work
- 3 Methods
- 3.1 ConvNeXt 3D
- 3.2 Pretraining
- 3.3 Approaches for Increased Robustness
- 4 Experiments
- 4.1 Preliminary Experiments
- 4.2 Comparison of Pretrainings
- 4.3 Challenge Submission Results
- 5 Conclusion
- References
- Two-Stage COVID19 Classification Using BERT Features
- 1 Introduction
- 2 Related Work
- 3 Data Preprocessing and Preparation
- 3.1 Slice Image Filtering and Lung Segmentation
- 3.2 Preparing Input for 3D CNN
- 4 Classification Networks
- 4.1 First Stage 3D CNN-BERT Network
- 4.2 Second Stage BERT Classification Network
- 4.3 COVID-19 Severity Classification
- 5 Experiment Results
- 5.1 Dataset
- 5.2 Implementation Details
- 5.3 Detection Challenge Results
- 5.4 Severity Challenge Results
- 6 Conclusions
- References
- PVT-COV19D: COVID-19 Detection Through Medical Image Classification Based on Pyramid Vision Transformer
- 1 Introduction
- 2 Related Work
- 2.1 CNN
- 2.2 The Vision Transformer
- 2.3 Pyramid Vision Transformer (PVT)
- 3 Methodology
- 4 Dataset
- 5 Experiments
- 5.1 Data Pre-processing
- 5.2 Implementation Details
- 6 Conclusions
- References
- Boosting COVID-19 Severity Detection with Infection-Aware Contrastive Mixup Classification
- 1 Introduction
- 2 Related Work
- 2.1 Segmentation of Lung Regions and Lesions
- 2.2 COVID-19 Severity Assessment
- 3 Methodology
- 3.1 Infection Segmentation
- 3.2 Infection-Aware COVID-19 Severity Classification
- 4 Dataset
- 4.1 COV19-CT-DB Database
- 4.2 COVID-19-CT-Seg Dataset
- 5 Experimental Results
- 5.1 Implementation Details
- 5.2 Evaluation Metrics
- 5.3 Results of Lung and Infection Segmentation
- 5.4 Results on the COVID-19 Severity Detection Challenge
- 6 Conclusions
- References
- Variability Matters: Evaluating Inter-Rater Variability in Histopathology for Robust Cell Detection
- 1 Introduction
- 2 Related Work
- 3 Evaluating Annotators and Variability
- 3.1 Anchor Annotator
- 3.2 Annotator Variability and Conformity Calculation
- 3.3 Annotator Grouping Based on the Conformity
- 4 Experimental Results
- 4.1 Dataset
- 4.2 Cell Detection Method and Experimental Details
- 4.3 Conformity of the Annotators
- 4.4 Impact of the Conformity of the Annotators in Cell Detection
- 5 Conclusions
- References
- FUSION: Fully Unsupervised Test-Time Stain Adaptation via Fused Normalization Statistics
- 1 Introduction
- 2 FUSION
- 3 Experiments
- 4 Results
- 5 Conclusion
- References
- Relieving Pixel-Wise Labeling Effort for Pathology Image Segmentation with Self-training
- 1 Introduction
- 2 Related Works
- 3 Methods
- 3.1 Self-training
- 3.2 Training
- 4 Data
- 5 Experimental Setup
- 6 Results
- 6.1 Self-training Performance at Fixed nl
- 6.2 Labeling a New Dataset: Sparsely or Exhaustively?
- 6.3 Experiments on Thyroid CytologyFNAB
- 7 Conclusion
- References
- CNR-IEMN-CD and CNR-IEMN-CSD Approaches for Covid-19 Detection and Covid-19 Severity Detection from 3D CT-scans
- 1 Introduction
- 2 Our Approaches
- 2.1 Covid-19 Detection
- 2.2 Covid-19 Severity Detection
- 3 Experiments and Results
- 3.1 The COV19-CT-DB Database
- 3.2 Experimental Setup
- 3.3 Covid-19 Recognition
- 3.4 Covid-19 Severity Detection
- 4 Discusssion
- 5 Conclusion
- References
- Representation Learning with Information Theory to Detect COVID-19 and Its Severity
- 1 Introduction
- 2 Related Works
- 3 Background and Proposed Method
- 3.1 Dependence Measure
- 4 Experimental Results
- 4.1 Datasets Description
- 4.2 Implementation and Training Details
- 4.3 COVID-19 Detection
- 4.4 COVID-19 Severity Detection
- 5 Conclusions
- References
- Spatial-Slice Feature Learning Using Visual Transformer and Essential Slices Selection Module for COVID-19 Detection of CT Scans in the Wild
- 1 Introduction
- 2 Related Work
- 3 Spatial-Slice Feature Learning
- 3.1 Essential Slices Set Selection Algorithm
- 3.2 Spatial Feature Learning
- 3.3 Slice Context Feature Learning
- 4 Experiments
- 4.1 Experimental Settings
- 4.2 Performance Evaluation
- 5 Conclusions
- References
- Multi-scale Attention-Based Multiple Instance Learning for Classification of Multi-gigapixel Histology Images
- 1 Introduction
- 2 Materials and Methods
- 2.1 Dataset
- 2.2 Tissue Classification
- 2.3 LMP1 Prediction
- 2.4 Metrics for Evaluation
- 3 Results
- 3.1 Tissue Classification
- 3.2 LMP1 Prediction
- 3.3 Model Interpretability
- 4 Conclusion
- References
- A Deep Wavelet Network for High-Resolution Microscopy Hyperspectral Image Reconstruction
- 1 Introduction
- 2 Related Works
- 3 Method
- 3.1 Spectral Grouping
- 3.2 Mutual Adaptation
- 3.3 Multi-scale Spatial Adaptation
- 4 Experiments
- 4.1 Dataset
- 4.2 Implementation Detail
- 4.3 Experimental Analysis
- 5 Conclusions
- References
- Using a 3D ResNet for Detecting the Presence and Severity of COVID-19 from CT Scans
- 1 Introduction
- 2 The COV19-CT-DB Database
- 3 Methods
- 3.1 Preprocessing
- 3.2 Neural Network Model
- 3.3 Loss
- 3.4 Training Procedure
- 3.5 Regularization and Data Augmentation
- 4 Results
- 5 Conclusion
- References
- AI-MIA: COVID-19 Detection and Severity Analysis Through Medical Imaging
- 1 Introduction
- 2 Related Work
- 3 The COV19-CT-DB Database
- 4 The Deep Learning Approach
- 4.1 3-D Analysis and COVID-19 Diagnosis
- 4.2 Pre-processing, Implementation Details and Evaluation Metrics
- 5 The COV19D Competition Results
- 5.1 COVID19 Detection Challenge Results
- 5.2 COVID19 Severity Detection Challenge Results
- 6 Conclusions and Future Work
- References
- Medical Image Segmentation: A Review of Modern Architectures
- 1 Introduction
- 2 Related Work
- 3 Experimental Setup
- 3.1 Datasets
- 3.2 Model Architectures
- 3.3 Metrics
- 3.4 Implementation Details
- 4 Results
- 5 Discussion
- 6 Conclusions
- References
- Medical Image Super Resolution by Preserving Interpretable and Disentangled Features
- 1 Introduction
- 1.1 Related Work:
- 1.2 Relevance of Interpretable Class Activation Maps
- 1.3 Our Contribution
- 2 Method
- 2.1 Overview
- 2.2 Feature Disentanglement
- 2.3 Explainability Loss Term
- 3 Experiments and Results
- 3.1 Dataset Description
- 3.2 Implementation Details
- 3.3 Quantitative Results
- 3.4 Ablation Studies
- 3.5 Classification Results on ChestXrays
- 3.6 Qualitative Results
- 4 Conclusions
- References
- Multi-label Attention Map Assisted Deep Feature Learning for Medical Image Classification
- 1 Introduction
- 2 Methods
- 2.1 Multi-label Activation Maps
- 2.2 Sample-wise Multi-label Loss Scores
- 2.3 Implementation Details
- 3 Results
- 3.1 Evaluation Metrics
- 3.2 Improved Classification Results
- 3.3 Comparison with Radiologist's Saliency Maps
- 3.4 Additional Results
- 4 Conclusion
- References
- Unsupervised Domain Adaptation Using Feature Disentanglement and GCNs for Medical Image Classification
- 1 Introduction
- 2 Method
- 2.1 Feature Disentanglement Network
- 2.2 Graph Convolutional Adversarial Network
- 3 Experiments and Results
- 3.1 Results for CAMELYON17 Dataset
- 3.2 Results on Chest Xray Dataset
- 3.3 T-SNE Visualizations
- 3.4 Results with L1 and L2 Loss
- 4 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.