Pattern Recognition and Computer Vision

Name: Pattern Recognition and Computer Vision | Second Chinese Conference, PRCV 2019, Xi'an, China, November 8-11, 2019, Proceedings, Part II
Brand: Springer
Price: 53.49 EUR
Availability: OnlineOnly

Second Chinese Conference, PRCV 2019, Xi'an, China, November 8-11, 2019, Proceedings, Part II

Zhouchen Lin Liang Wang Jian Yang Guangming Shi Tieniu Tan Nanning Zheng Xilin Chen Yanning Zhang(Herausgeber*in)

Springer (Verlag)

Erschienen am 31. Oktober 2019

XX, 813 Seiten

E-Book

PDF mit Wasserzeichen-DRM

Systemvoraussetzungen

978-3-030-31723-2 (ISBN)

53,49 €inkl. 7% MwSt.

Systemvoraussetzungen

für PDF mit Wasserzeichen-DRM

E-Book Einzellizenz

Als Download verfügbar

Beschreibung

Weitere Details

Weitere Ausgaben

Inhalt

Intro
Preface
Organization
Contents - Part II
Image/Video Processing and Analysis
Multiscale Entropy Analysis of EEG Based on Non-uniform Time
Abstract
1 Introduction
2 Method
2.1 Entropy Method
2.2 Multi-scale Method
3 Data and Results
3.1 Data
3.2 Experimental Results and Analysis
3.2.1 Entropy Value Analysis of Different Subjects
3.2.2 Different Test States KS Test Verification
3.2.3 Different Participant State SVM Classification
3.3 Multi-scale Method of Time Complexity Analysis
4 Conclusions
Acknowledgments
References
Recurrent Deconvolutional Generative Adversarial Networks with Application to Video Generation
1 Introduction
2 Related Work
3 Recurrent Deconvolutional Generative Adversarial Network
3.1 Recurrent Deconvolutional Network as Generator
3.2 Modified 3D Convolutional Neural Network as Discriminator
3.3 Learning
4 Experiments
4.1 Dataset and Implementation Details
4.2 Video Generation from Sentence
4.3 Video Classification
5 Conclusions and Future Work
References
Functional Brain Network Estimation Based on Weighted BOLD Signals for MCI Identification
Abstract
1 Introduction
2 Introduction
2.1 Data Acquisition Preprocessing
2.2 Functional Brain Network Estimation
2.3 Related Work
2.4 The Proposed Method
3 Experiment
3.1 Experimental Setting
3.2 Visualization of Function Brain Network
3.3 Classification Performance
3.4 Sensitivity to Network Modeling Parameters
4 Discussion
5 Conclusion
Acknowledgement
References
ESNet: An Efficient Symmetric Network for Real-Time Semantic Segmentation
1 Introduction
2 Related Work
3 ESNet
3.1 Network Overview
3.2 FCU Module
3.3 PFCU Module
3.4 Comparison of Network Complexity
4 Experiments
4.1 Dataset
4.2 Implementation Details
4.3 Evaluation Results
5 Conclusion Remark and Future Work
References
Assignment Problem Based Deep Embedding
1 Introduction
2 Related Work
3 Linear Assignment Problem Based Hard Mining
4 Experiment
4.1 Dataset
4.2 Implementation Details
4.3 Results
5 Conclusion
References
Auto Data Augmentation for Testing Set
1 Introduction
2 Related Work
3 Methods
3.1 Augmentation Environment
3.2 Controller Model
3.3 Compatible Implementation
4 Experiments
4.1 Datasets
4.2 Searching Augmentation Strategy on CIFAR-10
4.3 Searching Augmentation Strategy on Face Verification Datasets
4.4 The Transferability of Testing Phase Augmentation Strategy
4.5 Compatibility of Controller Model and Augmentation Environment
5 Conclusion
References
Dense Activation Network for Image Denoising
1 Introduction
2 Related Work
2.1 Image Denoising
2.2 Attention Mechanism
3 Proposed Method
3.1 Network Structure
3.2 Dense Information Fusion Net
3.3 Spatial Activation Net
4 Discussions
5 Experimental Results
5.1 Settings
5.2 Ablation Investigation
5.3 Comparisons with State-of-the-Arts
6 Conclutions
References
The Optimal Graph Regularized Sparse Coding with Application to Image Representation
1 Introduction
2 The Relative Work
3 The Proposed Method
3.1 Motivation
3.2 Constrained Laplacian Rank (CLR)
3.3 The Objective Function of the Proposed Method
4 Experimental Results
4.1 Yale Face Dataset
4.2 ORL Face Dataset
4.3 FERENT Face Dataset
4.4 The Analysis of the Parameters
5 Conclusion
References
Robust Embedding Regression for Face Recognition
Abstract
1 Introduction
2 Notation and Definition
3 Robust Embedding Regression
3.1 Motivation
3.2 The Objective Function of RER
3.3 The Optimal Solution
3.4 Convergence
4 Experiments
4.1 Details of Face Datasets
4.2 Experimental Setting and Results
4.3 Parameters Selection
5 Conclusion
Acknowledgments
References
Deep Feature-Preserving Based Face Hallucination: Feature Discrimination Versus Pixels Approximation
1 Introduction
2 Related Works
3 Proposed Method
3.1 Super-Resolution Network
3.2 Discriminative Network
3.3 Encoder Network
3.4 Feature Discriminative Network
3.5 Overall Training Loss
4 Experiments
4.1 Training Protocol
5 Experimental Results and Analysis
6 Conclusion and Future Work
References
Lung Parenchymal Segmentation Algorithm Based on Improved Marker Watershed for Lung CT Images
Abstract
1 Introduction
2 Related Work
3 Proposed Segmentation Approach
3.1 Original Lung CT Image Loading Processing
3.2 Convert HU Value
3.3 Generate Markup
3.4 Segmentation of Lung Parenchyma
4 Experimental and Analysis
4.1 Lung CT Image Data Set and Experimental Environment Description
4.2 Performance
4.3 Experimental Results and Discussion
5 Summary
References
Fine Grain Lung Nodule Diagnosis Based on CT Using 3D Convolutional Neural Network
1 Introduction
2 Method
2.1 Sensitivity Analysis of Attributes
2.2 Normalization
2.3 Architecture of ASMB3DCNN
3 Experiments and Results
3.1 Datasets
3.2 Nodule Attributes Prediction Results
3.3 Malignancy Suspiciousness Estimation
3.4 False Positive Reduction
3.5 Fine Grain Lung Nodule Diagnosis
4 Conclusions
References
Segmentation Guided Regression Network for Breast Cancer Cellularity
1 Introduction
2 Related Works
2.1 CNNs in Semantic Segmentation
2.2 CNNs in Image Classification
3 Segmentation Guided Regression Network
3.1 The Architecture of Segmentation Guided Regression Network
3.2 Training
4 Experimental Results
4.1 Training Strategy
4.2 Metrics and Results
5 Conclusion
References
Automatic Inspection of Yarn Locations by Utilizing Histogram Segmentation and Monotone Hypothesis
Abstract
1 Introduction
2 Our Method
2.1 Calculating Accumulated Partial Derivatives
2.2 Histogram Segmentation
3 Numerical Results
3.1 Automatic Selection of Parameter
3.2 Detection of the Warp and Weft Yarn
3.3 Detection of the Fabric Structure
4 Conclusion
Acknowledgements
References
Membranous Nephropathy Identification Using Hyperspectral Microscopic Images
1 Introduction
2 Membranous Nephropathy Hyperspectral Data
3 Membranous Nephropathy Identification Framework
3.1 Filtering Preprocessing
3.2 Projection Transformation
3.3 Deep Neural Network Feature Extractor
4 Experimental Results and Analysis
4.1 Parameter Tuning
4.2 Classification Performance and Analysis
5 Conclusion
References
A Level Set Method Combined with Gaussian Mixture Model for Image Segmentation
1 Introduction
2 Related Works
2.1 The CV Model
2.2 Improved CV Model
3 Experiments and Analysis
4 Conclusion
References
Nonstandard Periodic Gait Energy Image for Gait Recognition and Data Augmentation
Abstract
1 Introduction
2 Related Work
2.1 Gait Cycle Detection
2.2 Gait Recognition Method Based on GEI
2.3 Data Augmentation
3 Method
3.1 Overview
3.2 Synthesis of NP-GEI
3.3 Network Architectures
4 Experiments
4.1 Experiment 1: Effectiveness of NP-GEI for Gait Recognition
4.2 Experiment 2: Robustness of CNN to GEI Recognition of Cross-Frame Number from a Single Perspective
4.3 Experiments 3: Robustness of CNN to GEI Recognition of Cross-Frame Number in Multi-view
4.4 Experiments 4: Experiments on the Effectiveness of Data Augment Based on NP-GEI
5 Conclusion
References
A Temporal Attentive Approach for Video-Based Pedestrian Attribute Recognition
1 Introduction
2 Dataset
3 Approach
3.1 Network Architecture
3.2 Temporal-Attention Strategy
4 Experiments
4.1 Settings
4.2 Comparison with Other Approaches
4.3 Ablation Study
5 Conclusion
References
An Effective Network with ConvLSTM for Low-Light Image Enhancement
1 Introduction
2 Related Work
2.1 Traditional Methods
2.2 Deep Learning Based Methods
3 Our Method
3.1 Base Network
3.2 LSTM Module
3.3 Loss Function
4 Experiment
4.1 The Datasets
4.2 Implementation Details
4.3 The Results
4.4 The Ablation Study
5 Conclusion
References
Self-Calibrating Scene Understanding Based on Motifnet
1 Introduction
2 Model
3 Experiment
4 Conclusion
References
BDGAN: Image Blind Denoising Using Generative Adversarial Networks
1 Introduction
2 Related Work
2.1 Blind Image Denoising
2.2 Generative Adversarial Networks
2.3 Pretrained Models
3 The Proposed Method
3.1 Network Architecture
3.2 Multi-scale Discriminators
3.3 Loss Function
4 Training Details
5 Experiments
5.1 Dataset
5.2 Evaluation
6 Conclusion
References
Single Image Reflection Removal Based on Deep Residual Learning
1 Introduction and Related Work
2 Deep Residual Learning Network with GAN
2.1 The Generative Adversarial Framework
2.2 Residual Learning
2.3 Design of GAN for Single Image Reflection Removal
2.4 Multi-part Balanced Loss
3 Experiment
3.1 Dataset
3.2 Implementation and Analysis
3.3 Comparison to State-of-the-art Methods
4 Conclusion
References
An Automated Method with Attention Network for Cervical Cancer Scanning
1 Introduction
2 Related Work
3 Methodology
4 Experiments
4.1 Attention or Without Attention
4.2 Comparison of Different Classification Networks.
4.3 The Result of Transfer Learning
4.4 Comparison with Other Methods
5 Conclusion
References
Graph-Based Scale-Aware Network for Human Parsing
1 Introduction
2 Related Work
2.1 Semantic Segmentation
2.2 Human Parsing
3 Graph-Based Scale-Aware Network
3.1 Graph-Based Part Reasoning Layer
3.2 Scale-Aware Context Embedding Layer
3.3 Edge Module
4 Experiments
4.1 Data
4.2 Implementation Details
4.3 Experimental Results
5 Conclusion
References
Semi-supervised Lesion Detection with Reliable Label Propagation and Missing Label Mining
1 Introduction
2 Related Work
3 Method
3.1 Slice-Level Label Propagation
3.2 Missing Ground-Truth Mining
4 Experiments
4.1 Experimental Setup
4.2 Results and Discussion
4.3 Ablation Study
5 Conclusion
References
Image Aesthetic Assessment Based on Perception Consistency
1 Introduction
2 Related Work
2.1 Holistic Perspective
2.2 Fine-Gained Details Perspective
3 Our Method
3.1 Overview
3.2 Subject-Focused Subnet
3.3 Detail-Oriented Subnet
4 Experimental Result
4.1 Experimental Setting
4.2 Experimental Result on Subject-Focus Subnet
4.3 Experimental Result on Detail-Oriented Subnet
4.4 VP-Net Experience Result
5 Conclusion
References
Image De-noising by an Effective SURE-Based Weighted Bilateral Filtering
Abstract
1 Introduction
2 New Pseudo-Median Bilateral Filtering with Edge Preserving
2.1 Robust Estimators of Kernel Functions
2.2 New Pseudo-Median Bilateral Filtering with Tukey's Kernel
3 Weighted Bilateral Filtering with SURE-Robust Estimation
3.1 Robust Bilateral Filter
3.2 SURE-Optimally Bilateral Filtering
3.3 SURE and Edge-Preserving Weighted Bilateral Filtering
4 Experiments
4.1 Weighted Kernel Function
4.2 Experiments with Additive Gaussian Noise
4.3 Experiments on Mixed Noise
5 Conclusion
References
Automatic Detection of Pneumonia in Chest X-Ray Images Using Cooperative Convolutional Neural Networks
1 Introduction
2 Related Works
3 Methods
3.1 Network Architecture of Co-CNN
3.2 Implementation Process of Co-CNN
3.3 Dataset
4 Results
5 Discussion
6 Conclusion
References
Siamese Spatial Pyramid Matching Network with Location Prior for Anatomical Landmark Tracking in 3-Dimension Ultrasound Sequence
Abstract
1 Introduction
2 Methods
2.1 End-to-End Siamese Spatial Pyramid Matching Network
2.2 Patch-Based Matching Based Tracking
2.3 Temporal Consistency Model for Landmark Tracking
3 Experiment and Result
3.1 Implementation Details
3.2 Performance Evaluation and Results Analysis
3.3 Ablation Study
4 Conclusion and Discussion
Acknowledgement
References
Local Context Embedding Neural Network for Scene Semantic Segmentation
1 Introduction
2 Related Work
3 Local Context Embedding Network
3.1 Network Architecture
3.2 Local Contextual Information Embedding
4 Experiments
4.1 Result Comparisons
4.2 Ablations Studies of Our Network
5 Conclusions and Future Work
References
Retinex Based Flicker-Free Low-Light Video Enhancement
1 Introduction
2 Related Works
2.1 Center-Surrounded Retinex Based Enhancement Method
2.2 Optical Flow
3 Proposed Method
3.1 Algorithm Introduction
3.2 Intra-frame Brightness Enhancement
3.3 Inter-frame Brightness Continuity
4 Experimental Results and Analysis
4.1 Subjective Assessment
4.2 Objective Assessment
4.3 Flickering-Artifact Assessment
5 Conclusion
References
Transfer Learning for Rigid 2D/3D Cardiovascular Images Registration
1 Introduction
2 Methods
2.1 2D/3D Registration
2.2 Registration with Transfer Learning
2.3 Evaluation of Image Registration Accuracy
3 Experiments
3.1 Data Preparation
3.2 Experimental Environment
3.3 Rigid Registration Without Transfer Learning
3.4 Rigid Registration with Transfer Learning
4 Results
4.1 Rigid Registration Without Transfer Learning
4.2 Optimal Parameter Selection with Transfer Learning
4.3 Transfer Learning for Different Patients
5 Discussion and Conclusion
References
Temporal Invariant Factor Disentangled Model for Representation Learning
1 Introduction
2 Related Work
3 Temporal Invariant Factor Disentangled Model
3.1 Calculating KL[q||p]
3.2 Calculating V(zST)
3.3 Final Formulation
4 Optimization
4.1 Prior Distribution
4.2 Inferred Posterior Distribution
4.3 Implementation with Deep Neural Networks
5 Experiments
5.1 Dataset
5.2 Training Details
5.3 Evaluations
5.4 Quantitative Analysis
5.5 Qualitative Analysis
6 Conclusion
References
A Multi-frame Video Interpolation Neural Network for Large Motion
Abstract
1 Introduction
2 Related Work
3 Our Proposed Method
3.1 Estimation of Approximate Optical Flows
3.2 Flow Refine Network
3.3 Training Loss
4 Experiments
4.1 Implementation Details and Datasets
4.2 Ablation Experiments
4.3 Comparison with State-of-the-Art Methods
5 Conclusion
Acknowledgement
References
One-Shot Video Object Segmentation Initialized with Referring Expression
1 Introduction
2 Related Work
3 Proposed Method
3.1 The Framework of Our Method
3.2 Referring Expression for Semantic Segmentation
3.3 Semi-supervised Video Object Segmentation
4 Experimental Validation
4.1 Evaluation of Results on DAVIS Dataset
4.2 The Effect of Mark Accuracy to Segmentation Results
4.3 Training Iterations and Timing
4.4 Improve Accuracy by Annotating Different Frames
5 Conclusions
References
Scalable Receptive Field GAN: An End-to-End Adversarial Learning Framework for Crowd Counting
1 Introduction
2 Related Work
3 Proposed Method
3.1 SRFGAN Framework
3.2 Residual Structure for Crowd Counting
3.3 Image-Level Prior
3.4 Comparison on Receptive Field
3.5 Loss Function
4 Experiments
4.1 Evaluation Metric
4.2 Implementation Details
4.3 Ablation Study
4.4 Evaluation and Comparison
5 Conclusions
References
Lightweight Video Object Segmentation Based on ConvGRU
Abstract
1 Introduction
2 Related Work
2.1 Convolutional Gated Recurrent Unit
2.2 MobileNet
3 Method
3.1 Problem Formulation
3.2 Network Architecture
3.3 Training Details
4 Experiment
5 Conclusion
Acknowledgements
References
Crowd Counting via Conditional Generative Adversarial Networks
1 Introduction
2 Related Work
2.1 Early Traditional Methods
2.2 CNN-Based Methods
3 Method
3.1 Network Architecture
3.2 Loss Function
4 Experiments
4.1 Crowd Counting Datasets
4.2 Density Map for Training
4.3 Training Details
4.4 Evaluation Metrics
4.5 Results and Analysis
5 Conclusion
References
Gemini Network for Temporal Action Localization
1 Introduction
2 Related Work
3 Method
3.1 Proposal Generation
3.2 Subnet I
3.3 Subnet II
3.4 Proposal Type and Multi-task Loss
4 Experiments
4.1 Experimental Settings
4.2 Model Analysis
4.3 Comparison with State-of-the-Art Methods
5 Conclusion
References
SS-GANs: Text-to-Image via Stage by Stage Generative Adversarial Networks
1 Introduction
2 SS-GANs for Text-to-Image Generation
2.1 The First Stage: Coarse-Grained Generation
2.2 The Second Stage: Transitional Stage
2.3 The Third Stage: Fine-Grained Generation
2.4 Objective Function
3 Hard Sample Selection and Coordinated Training
3.1 Hard Sample Selection
3.2 Coordinated Training
4 Experiments
4.1 Datasets
4.2 Evaluation Metric
4.3 Experimental Comparison
4.4 Ablation Studies
5 Conclusion and Future Work
References
Face Super-Resolution via Discriminative-Attributes
Abstract
1 Introduction
2 The Proposed Algorithm
2.1 Generator Network
2.2 Classifier Network
2.3 Discriminator Network
2.4 Loss Function
3 Experimental Results and Discussions
3.1 Experiment Settings
3.2 Experimental Results and Discussions
4 Conclusion
Acknowledgments
References
RefineNet4Dehaze: Single Image Dehazing Network Based on RefineNet
Abstract
1 Introduction
2 Related Work
3 RefineNet4Dehaze
3.1 RefineNet4Dehaze
3.2 Ablation Analysis for RefineNet
3.3 MSE and Perceptual Combined Loss
4 Experiments and Results
4.1 Datasets
4.2 Evaluation Metrics
4.3 Results
5 Conclusion
Acknowledgement
References
Level Set Image Segmentation Based on Non-independent and Identically Distributed
Abstract
1 Introduction
2 The Proposed Approach
2.1 Level Set Method
2.2 Non-independent and Identically Distributed
2.3 Combination of CV Model and Non-IID
3 Experimental Results and Analysis
3.1 Simple Image Segmentation
3.2 Image Segmentation with Weak Edges
3.3 Segmentation of Complex Images
4 Conclusion
Acknowledgments
References
KSLIC: K-mediods Clustering Based Simple Linear Iterative Clustering
1 Introduction
2 KSLIC Algorithm
2.1 Distance Measure
2.2 Calculation of Centers
3 Experiments and Results
3.1 Evaluation Metrics
3.2 Parameter Settings of Algorithms
3.3 Results and Analysis
4 Conclusion
References
Social Behavior Recognition in Mouse Video Using Agent Embedding and LSTM Modelling
1 Introduction
2 Proposed Method
2.1 Agent Embedding
2.2 Interaction Modelling
3 Experiment Setup
3.1 Datasets
3.2 Data Preparation
3.3 Embedding Training
3.4 LSTM Training and Evaluation
3.5 Two Stream System
4 Results
5 Conclusions
References
Unsupervised Global Manifold Alignment for Cross-Scene Hyperspectral Image Classification
1 Introduction
2 Method
2.1 Notation
2.2 Overall Objective Function
2.3 Optimized Iterative Method
3 Experiments and Results
3.1 Dataset
3.2 Illustrate How We Use Unsupervised Learning
4 Conclusions
References
Poleward Moving Aurora Recognition with Deep Convolutional Networks
1 Introduction
1.1 Related Work
2 Our Approach
3 Experiments
4 Conclusions
References
Robust Hyperspectral Image Pan-Sharpening via Channel-Constrained Spatial Spectral Network
1 Introduction
2 Robust Channel-Constrained Spatial Spectral Network
2.1 Network Architecture
2.2 Robust Deep Feature Extraction
2.3 Loss Function
2.4 Implementation Details
3 Experiments
3.1 Database Description and Parameter Setting
3.2 Experimental Results
3.3 Running Time
3.4 Effectiveness of FODFE in RCSSN
3.5 Conclusion
References
Ensemble Transductive Learning for Skin Lesion Segmentation
1 Introduction
2 Transductive Skin Lesion Segmentation
2.1 Transductive Learning
2.2 Ensemble Transductive Learning
3 Experimental Evaluations
3.1 Experimental Setup
3.2 Effectiveness of Transductive Segmentation
3.3 Robustness of Transductive Segmentation over Model Structures
3.4 Effectiveness of Single Transductive Segmentation
3.5 Influence of Hyper-parameters
4 Conclusion
References
MobileCount: An Efficient Encoder-Decoder Framework for Real-Time Crowd Counting
1 Introduction
2 Related Work
2.1 Crowd Counting
2.2 FCN Decoder Architectures
2.3 Light-Weight Networks
3 Method
3.1 Network Architecture
3.2 Loss Functions
3.3 Ground Truth Generation
4 Experiments
4.1 Implementation Details
4.2 Datasets
4.3 Evaluation Metrics
4.4 Results and Analysis
5 Conclusion
References
Multi-scale Densely 3D CNN for Hyperspectral Image Classification
Abstract
1 Introduction
2 Multi-scale Densely 3D CNN
2.1 3D Dense Block
2.2 MSD 3D-CNN Architecture
3 Experimental Results and Discussion
3.1 Datasets
3.2 Results and Analysis
4 Conclusion
Acknowledgment
References
No-Reference Image Quality Assessment via Multi-order Perception Similarity
1 Introduction
2 Methodology
2.1 High and Low Order Maps Construction
2.2 Feature Similarity Extraction
2.3 Multi-scale and Multi-channel Application
2.4 Regression
3 Experimental Results and Analyses
4 Conclusion
References
Blind Quality Assessment for DIBR-Synthesized Images Based on Chromatic and Disoccluded Information
1 Introduction
2 Proposed Method
2.1 Chromatic Maps
2.2 Disoccluded Regions
2.3 Weighted Maps
2.4 Regression Model
3 Experimental Results
3.1 Databases
3.2 Evaluation Indicators
3.3 Performance Comparison
3.4 Parameters Sensitivity
3.5 Comparison of Different Disoccluded Regions Extraction Strategies
4 Conclusion
References
Gait Recognition with Clothing and Carrying Variations Based on GEI and CAPDS Features
Abstract
1 Introduction
2 Feature Modeling
2.1 GEI Feature Modeling
2.2 CAPDS Feature Modeling
3 Network Architecture
3.1 The Architecture of PRN
3.2 The Architecture of TSPN
4 Framework of the Proposed Method
5 Experiments and Analysis
5.1 Setups
5.2 Comparison with Other Networks
5.3 Generalization Experiment with a Small Training Set
5.4 Contribution of the Features
6 Conclusions
Acknowledgement
References
Stage-by-Stage Based Design Paradigm of Two-Pathway Model for Gaze Following
1 Introduction
2 Related Works
3 Method
3.1 Problem Statement
3.2 Stage-by-Stage Based Design Paradigm of Two-Pathway Model
3.3 Loss Function
4 Experiments
4.1 Experiments on GazeFollow
5 Conclusion
References
Multi-modal Feature Fusion Based on Variational Autoencoder for Visual Question Answering
1 Introduction
2 Related Work
2.1 Visual Question Answering
2.2 Attention Mechanism
2.3 Variational Autoencoder
3 Proposed Method
3.1 Multi-modal Feature Fusion
3.2 Variational Attention Mechanism
4 Experiment
4.1 Datasets
4.2 Configurations
4.3 Results
5 Conclusion
References
Local and Global Feature Learning for Subtle Facial Expression Recognition from Attention Perspective
1 Introduction
2 Related Work
2.1 Feature Merging in Face-Related Tasks
2.2 Attention Mechanism
3 Method
3.1 Local Attention Learning
3.2 Self Attention Learning
3.3 Feature Fusion
3.4 Loss Function
4 Experiments
4.1 Datasets and Protocols
4.2 Implementation Details
4.3 Results
5 Conclusion
References
Multi-label Chest X-Ray Image Classification via Label Co-occurrence Learning
1 Introduction
2 Related Works
3 Method
3.1 Notations and Problem Definition
3.2 Achitecture of LCL-Net
4 Experiments
4.1 Dataset
4.2 Implementation Details
4.3 Parameter Analysis
4.4 Comparison to State-of-the-Art Methods
4.5 Qualitative Results
5 Conclusion
References
Asymmetric Pyramid Based Super Resolution from Very Low Resolution Face Image
1 Introduction
2 Related Work
3 Proposed Method
3.1 Pyramidal Network for Large Upscale Factor
3.2 Dense Block Unit
3.3 Asymmetric Structure
3.4 Optimization and Implementation Details
4 Experiment Result
5 Conclusion
References
A Hybrid Pan-Sharpening Approach Using Nonnegative Matrix Factorization for WorldView Imageries
1 Introduction
2 Hybrid Framework
3 Proposed Approach
3.1 NMF
3.2 Nonlinear Fitting
3.3 Fusion
4 Experiment Result and Analysis
4.1 Parameter Learning
4.2 Visual Analysis
4.3 Quantitative Assessment
5 Conclusion
References
Distinguishing Individual Red Pandas from Their Faces
1 Introduction
2 Related Work
3 Method
3.1 Overview
3.2 Red Panda Face Detection
3.3 Red Panda Face Alignment
3.4 Red Panda Identification
4 Experiments
4.1 Database
4.2 Identification Accuracy by Different Features
4.3 Performance Before and After Discarding Correlated Images
4.4 Impact of Face Alignment
5 Conclusion
References
Facial Expression Recognition: Disentangling Expression Based on Self-attention Conditional Generative Adversarial Nets
Abstract
1 Introduction
2 Related Works
3 The Proposed Method: DESA-CGAN
3.1 Neutral Face Regeneration
3.2 The Self-attention Layer
3.3 The Classification of Facial Expression
4 Experiments
4.1 Visualization of Regenerated Neutral Faces
4.2 Expression Recognition Results
5 Conclusion
Acknowledgement
References
Image Enhancement of Shadow Region Based on Polarization Imaging
Abstract
1 Introduction
2 DOP Invariance and DOP-Chromatic Consistency Priors
2.1 Polarization Imaging Model in Shadow Region
2.2 DOP Invariance of Different Light Intensities
2.3 DOP-Chromatic Consistency in Local Region
3 Image Enhancement Method Based on Polarizing Imaging
3.1 Image Enhancement Method Based on DOP Invariance Prior
3.2 Noise-Suppression Method Based on DOP-Chromatic Consistency Prior
4 Experiments
5 Evaluation
6 Conclusion
Acknowledgements
References
Multi-scale Convolutional Capsule Network for Hyperspectral Image Classification
Abstract
1 Introduction
2 Related Work
2.1 Multi-scale CNN
2.2 Capsule Network
3 Proposed Approach
3.1 Network Design
3.2 Parameter Setting
4 Experiments
4.1 Datasets
5 Conclusions
Acknowledgment
References
Dark Channel Prior Guided Conditional Generative Adversarial Network for Single Image Dehazing
Abstract
1 Introduction
2 Related Work
3 Proposed Method
3.1 Siamese like Generator
3.2 Multi-scale Discriminator
3.3 Overall Loss Function
4 Experiments
4.1 Dataset
4.2 Implementation Details
4.3 Ablation Study
4.4 Comparisons with State-of-the-Art Methods
5 Conclusion
References
A Fast Region Growing Based Superpixel Segmentation for Hyperspectral Image Classification
Abstract
1 Introduction
2 Proposed Method
2.1 FRGSS for HS Images Superpixel Segmentation
2.2 TASIS for HS Images Classification
3 Experimental Results
3.1 Performance of the FRGSS Superpixel Segmentation
3.2 Performance of the TASIS Classification Strategy
4 Conclusion
Acknowledgements
References
Complexity Reduction for Depth Map Coding in 3D-HEVC
1 Introduction
2 3D-HEVC Intra Depth Coding
2.1 Fast DMMs Selection
2.2 CU Partition
3 Proposed Fast Algorithm
3.1 Fast Depth Map Intra Mode Decision
3.2 Early Terminated CU Split Decision
4 Experiment Results
5 Conclution
References
Super Resolution via Residual Restructured Dense Network
Abstract
1 Introduction
2 Related Work
2.1 ResNet
2.2 DenseNet
3 Residual Restructured Dense Network
3.1 Bottleneck Layer
3.2 Residual Restructured Dense Block(RRDBlock)
3.3 Residual Restructured Dense Network(RRDN)
4 Experiments
4.1 Settings
4.2 Ablation Investigation
4.3 Comparison with State-of-the-Art Models
5 Conclusion
References
Author Index

Systemvoraussetzungen

Als PDF speichern Als Link merken