Pattern Recognition. ICPR International Workshops and Challenges: Virtual Event, January 10–15, 2021, Proceedings, Part III (Image Processing, Computer Vision, Pattern Recognition, and Graphics)

This document was uploaded by one of our users. The uploader already confirmed that they had the permission to publish it. If you are author/publisher or own the copyright of this documents, please report to us by using this DMCA report form.

Simply click on the Download Book button.

Yes, Book downloads on Ebookily are 100% Free.

Sometimes the book is free on Amazon As well, so go ahead and hit "Search on Amazon"

This 8-volumes set constitutes the refereed of the 25th International Conference on Pattern Recognition Workshops, ICPR 2020, held virtually in Milan, Italy and rescheduled to January 10 - 11, 2021 due to Covid-19 pandemic. The 416 full papers presented in these 8 volumes were carefully reviewed and selected from about 700 submissions. The 46 workshops cover a wide range of areas including machine learning, pattern analysis, healthcare, human behavior, environment, surveillance, forensics and biometrics, robotics and egovision, cultural heritage and document analysis, retrieval, and women at ICPR2020.

Author(s): Alberto Del Bimbo (editor), Rita Cucchiara (editor), Stan Sclaroff (editor), Giovanni Maria Farinella (editor), Tao Mei (editor), Marco Bertini (editor), Hugo Jair Escalante (editor), Roberto Vezzani (editor)
Publisher: Springer
Year: 2021

Language: English
Pages: 792

Foreword by General Chairs
Preface
Challenges
ICPR Organization
Towards AI Ethics and Explainability (ICPR EDL-AI Workshop Plenary Talk)
Contents – Part III
EDL-AI - Explainable Deep Learning/AI
Preface
Organization
General Chairs
Program Committee Chairs
Program Committee
Publication Chairs
Panel Chair
Additional Reviewers
A Multi-layered Approach for Tailored Black-Box Explanations
1 Introduction
2 Context and Requirements
2.1 Context
2.2 Requirements
3 From Contexts to Explanations
3.1 From Context to Requirements
3.2 From Requirements to Technical Options
4 IBEX at Work: Application to Case Studies
4.1 Explanations to Enhance Trust
4.2 Explanations to Take Actions
5 Related Works
6 Conclusion
References
Post-hoc Explanation Options for XAI in Deep Learning: The Insight Centre for Data Analytics Perspective
1 Introduction
2 Post-hoc Factual Explanations: Images
2.1 The Method: COLE
2.2 Results: Factual Image-Based Explanations
3 Post-hoc Counterfactual Explanations: Images
3.1 The Method: PIECE
3.2 Results: Counterfactual Image-Based Explanations
4 Post-hoc Semi-factual Explanations: Images
4.1 Method and Results: PIECE for Semi-factuals
5 Post-hoc,Counterfactual Explanations: Time-Series
5.1 The Method: Native-Guide for Time-Series Counterfactuals
5.2 Results: Native-Guide for Counterfactuals
6 Future Directions
References
Expert Level Evaluations for Explainable AI (XAI) Methods in the Medical Domain
1 Introduction
2 Eye-Tracking Experiments and Data Collection
2.1 Data Collection Protocol
3 Explainable AI Methods
3.1 SIDU
3.2 GRAD-CAM
4 Comparison Metrics for XAI Methods
4.1 Area Under ROC Curve (AUC)
4.2 Kullback-Leibler Divergence (KL-DIV)
5 Experimental Evaluation and Results
5.1 Training CNN Models
5.2 Results and Discussion
6 Concluding Remarks
References
Samples Classification Analysis Across DNN Layers with Fractal Curves
1 Introduction
2 Previous Works
2.1 Visualization for the Interpretation of Deep Neural Networks
2.2 Hilbert Curve in Information Visualization
3 Proposed Method
3.1 Domain Level
3.2 Abstraction Level
3.3 Technique Level
3.4 Algorithms Level
4 Experimental Protocol
4.1 Scenarios
4.2 Implementation and Execution Infrastructure
5 Results and Discussion
6 Conclusion
References
Random Forest Model and Sample Explainer for Non-experts in Machine Learning – Two Case Studies
1 Introduction
2 Integrated Random Forest Model and Sample Explainer - RFEX
2.1 RFEX Model Explainer
2.2 RFEX Sample Explainer
3 RFEX Explanation of Early Mortality Prediction for COVID-19 Patients
3.1 RFEX Model Explainer for the Prediction of COVID-19 Mortality from the Data for Day 0
3.2 RFEX Model Explainer for Early Prediction of COVID-19 Mortality from the Data for Day -7
4 RFEX Explanation of Classification of Human Nervous System Cell Type Clusters
4.1 RFEX Model Explainer Applied to JCVI Data
4.2 RFEX Sample Explainer Applied to JCVI Data
References
Jointly Optimize Positive and Negative Saliencies for Black Box Classifiers
1 Introduction
2 Related Works
2.1 Class Activation Map
2.2 Backpropagation-Based Method
2.3 Perturbation-Based Method
2.4 Mask-Based Method
3 Joint Mask Method
3.1 Baseline Works
3.2 Deletion and Negative Saliency
3.3 Integration of Preservation and Deletion Masks
4 Experiments
4.1 Comparison with the Other Perspectives
4.2 Comparison with Other Saliency Methods
4.3 Sanity Check
5 Conclusions
References
Low Dimensional Visual Attributes: An Interpretable Image Encoding
1 Introduction
2 Related Work
3 Low Dimensional Visual Attributes
4 Label-Limited Classification Evaluation
5 Interpretability
5.1 Learned Part and Prototypes
5.2 Crowd-Sourced Experiments
6 Conclusion
References
Explainable 3D-CNN for Multiple Sclerosis Patients Stratification
1 Introduction
2 Materials and Methods
2.1 Population, Data Acquisition and Image Processing
2.2 Network Architecture
2.3 Training, Validation and Testing
2.4 CNN Visualization
2.5 LRP Heatmaps Analysis
3 Results
4 Discussion
5 Conclusions
References
Visualizing the Effect of Semantic Classes in the Attribution of Scene Recognition Models
1 Introduction
2 Method
2.1 Preliminaries
2.2 Score Deviation
2.3 Score Deviation Map
2.4 Class-Wise Statistics
2.5 Particularization to Scene Recognition
3 Experimental Results
3.1 Implementation Details
3.2 Score Deviation Maps
3.3 Relevant, Irrelevant and Distracting Semantic Classes
4 Conclusions
References
The Impact of Activation Sparsity on Overfitting in Convolutional Neural Networks
1 Introduction
1.1 Related Work
1.2 Contribution
2 Methods
2.1 Activation Sparsity Definition
2.2 Activation Sparsity Visualisation
2.3 Activation Sparsity Regularisation
2.4 Experimental Design
3 Results
3.1 Relationship Between Overfitting and Activation Sparsity
3.2 Spatial Analysis of Activation Sparstiy
3.3 Activation Sparsity Regularisation Results
3.4 Activation Sparsity in Common Deep Architectures
4 Conclusion
References
Remove to Improve?
1 Introduction
2 Previous Work
3 Background and Definitions
4 Methodology
5 Experiments
5.1 Class-Wise Accuracy Changes
5.2 Identification of Filters for Each Class
5.3 Class-Wise Filter Overlap and Semantic Similarity
5.4 Analysis of Groups
5.5 Nearest Neighbours
5.6 Class-Wise Correlation Between Pruned Filters
6 Discussion and Conclusions
References
Explaining How Deep Neural Networks Forget by Deep Visualization
1 Introduction
2 Related Work
3 Approach
3.1 Catastrophic Forgetting Dissector - CFD
3.2 Critical Freezing
4 Experiments
5 Conclusion and Future Work
References
Deep Learning for Astrophysics, Understanding the Impact of Attention on Variability Induced by Parameter Initialization
1 Introduction
2 Related Work
3 Proposed Architecture and Performance
3.1 -PhysNet Architecture
3.2 Experiments
4 Understanding the Impact of Dual Attention
5 Conclusion
References
A General Approach to Compute the Relevance of Middle-Level Input Features
1 Introduction
2 Related Works
3 Middle-Level Relevance
3.1 Decoder by Super-Pixel Segmentation
3.2 Decoder by Sparse Dictionary Learning Methods
4 Experimental Assessment
4.1 Qualitative Results
4.2 Quantitative Evaluation
5 Conclusions
References
Evaluation of Interpretable Association Rule Mining Methods on Time-Series in the Maritime Domain
1 Introduction
2 Related Work
3 Foundations
3.1 Classification
3.2 Data Preprocessing
3.3 Association Rule Mining—ARM
4 Methods
4.1 Scalable Bayesian Rule Lists—SBRL
4.2 Rule-Based Regularization Method
4.3 Gini Regularization Method
5 Evaluation Metrics
6 Experimental Results
6.1 Datasets
6.2 Experimental Setup
6.3 Results
7 Conclusion and Future Work
References
Anchors vs Attention: Comparing XAI on a Real-Life Use Case
1 Introduction
2 Related Works
2.1 EXplainable Artificial Intelligence
2.2 Evaluate Explanations
3 Experiments
3.1 LEGO
3.2 YELP
4 Evaluating Explanations
4.1 Quantitative Analysis
4.2 Qualitative Analysis
5 Conclusion
References
Explanation-Driven Characterization of Android Ransomware
1 Introduction
2 Background on Android
2.1 Android Ransomware
3 Explanation Methods
4 Ransomware Detection and Explanations
4.1 Detector Design
4.2 Explaining Android Ransomware
5 Experimental Analysis
5.1 Setting
5.2 Preliminary Evaluation
5.3 Explanation Analysis
6 Contributions, Limitations, and Future Work
References
Reliability of eXplainable Artificial Intelligence in Adversarial Perturbation Scenarios
1 Introduction
2 Related Works
3 Methods
4 Results
5 Conclusions
References
AI Explainability. A Bridge Between Machine Vision and Natural Language Processing
1 Introduction
2 Background
3 Link Between Image and Text in Explainability
4 Potential Benefits to NLP Community
4.1 Word-Sense Disambiguation
4.2 Text Argumentation Theory
4.3 Sentiment Analysis
4.4 Topical Modelling
4.5 Automatic Textual Summarization
5 Conclusion
References
Recursive Division of Image for Explanation of Shallow CNN Models
1 Introduction
2 Shallow vs Deep Architectures
3 Explanation Methods
4 Recursive Division
5 Experimental Modelling
5.1 UEC FOOD 100
5.2 UEC FOOD 256
5.3 Crack Dataset
5.4 Performance
6 Conclusions
References
EgoApp 2020 - 2nd Workshop on Applications of Egocentric Vision 2020
The Second Workshop on Applications of Egocentric Vision (EgoApp)
Workshop Description
Organization
Organizing Committee
Program Committee
Camera Ego-Positioning Using Sensor Fusion and Complementary Method
1 Introduction
2 Related Work
2.1 Visual Ego-Positioning
2.2 Sensor Fusion of Camera and IMU
2.3 Complementary Ego-Positioning
3 Sensor Fusion of Camera and IMU
3.1 Method
3.2 Camera-IMU System Calibration
4 Complementary Ego-Positioning
4.1 3D Point Registration
4.2 Complementary Fusion
5 Experiments
6 Conclusion
References
ATSal: An Attention Based Architecture for Saliency Prediction in 360 Videos
1 Introduction
2 Related Work
2.1 2D Dynamic Saliency Models
2.2 360 Heuristic Approaches
2.3 360 Data-Driven Approaches
3 Proposed Model
3.1 Attention Mechanism
3.2 Expert Models
3.3 Loss Function for the Attention Stream
4 Experiments
4.1 Experimental Setup
4.2 Results
4.3 Performance Study
5 Conclusion
References
Rescue Dog Action Recognition by Integrating Ego-Centric Video, Sound and Sensor Information
1 Introduction
2 Related Work
2.1 Third-Person Activity Recognition
2.2 First-Person Activity Recognition
2.3 Dog-Centric Activity Modeling
3 Dataset
3.1 11 Dog Activity Classes
4 Method
4.1 The Detail of the Image/Sound/Sensor-Based Four-Stream CNN
5 Experiments
5.1 Selection of Sound Stream Network
5.2 Selection of Sound Window Size
5.3 Selection of Sensor Data Network
5.4 Selection of Sensor Window Size
5.5 Experiments by Integration of All Modalities
6 Conclusions
References
Understanding Event Boundaries for Egocentric Activity Recognition from Photo-Streams
1 Introduction
2 Related Work
3 Activity Recognition from Event Boundaries
3.1 Boundaries Detection
3.2 Event-Based Activity Recognition
4 Experimental Setup
4.1 Dataset
4.2 Implementation
4.3 Evaluation Metrics
5 Results
5.1 Generic Vs Personalized Learning
5.2 Random Forest Based Models Vs. Deep Models
5.3 LSTM Vs. BLSTM Temporal Models
5.4 Event Clustering Vs. No Segmentation
6 Conclusions
References
Egomap: Hierarchical First-Person Semantic Mapping
1 Introduction
2 Relationship with Previous Work
2.1 Related Approaches
2.2 Datasets
3 Model
3.1 Hierarchical Map
3.2 View Representation
4 Motion Analysis for Transition Detection
5 Inference
5.1 View Inference via Recursive Bayes
5.2 Station Inference
6 Learning
6.1 View Update
6.2 View Creation
6.3 Station Update
6.4 Station Creation
6.5 Transition Matrix Updates
7 Experiments and Metrics
7.1 Dataset
7.2 Dictionary Learning and Parameter Tuning
7.3 Map Evaluation
7.4 Baseline and Ablation
8 Results
8.1 Visualisation
8.2 Quantitative Evaluation
9 Conclusions
References
ETTAC 2020 - Workshop on Eye Tracking Techniques, Applications and Challenges
Preface
Organization
General Chairs
Program Committee
Additional Reviewers
Ultrasound for Gaze Estimation
1 Introduction
2 Materials and Methods
2.1 Bench-Top Setup
2.2 Data Analysis
3 Results
4 Discussion
References
Synthetic Gaze Data Augmentation for Improved User Calibration
1 Introduction
2 Related Works
3 Working Framework
3.1 Image Databases
3.2 Image Conditioning
3.3 Network Architecture
3.4 Implementation Details
4 Subject Calibration
5 Experiments
6 Results
6.1 Number of Training Images and Regressor Estimation
6.2 U2Eyes and ImageNet Methods
7 Conclusions
References
Eye Movement Classification with Temporal Convolutional Networks
1 Introduction
2 Related Work
2.1 Threshold-Based Methods
2.2 Probabilistic Methods
2.3 Data-Driven Methods
3 Model Architecture
4 Evaluation
4.1 Materials
4.2 Dataset
4.3 Feature Extraction
4.4 Metrics
4.5 Training and Evaluation
5 Results
6 Discussion
7 Conclusion
References
A Web-Based Eye Tracking Data Visualization Tool
1 Introduction
2 Related Work
3 Data Handling
3.1 Data Validation
3.2 Clustering
3.3 Heatmap Data
3.4 Caching
4 Visualization Techniques
4.1 AOI Timeline
4.2 Gaze Plot
4.3 Heatmap
4.4 Scarf Plot
4.5 General Interactions
5 Web Application Architecture
6 Use Case: Metro Map of Antwerp
7 Discussion and Limitations
7.1 Front-End and Back-End Decisions
7.2 Performance
7.3 Uploading
7.4 Interactions
7.5 Interface
7.6 Views
8 Conclusion and Future Work
References
Influence of Peripheral Vibration Stimulus on Viewing and Response Actions
1 Introduction
2 Experiment
2.1 Visual Stimuli
2.2 Experimental Procedure
2.3 Subjects
3 Results
3.1 Percentage Correct of the Task of Peripheral Field of Vision Viewing
3.2 Hierarchical Bayesian Modelling
3.3 Analysis of Microsaccades
4 Discussion
5 Conclusion
References
Judging Qualification, Gender, and Age of the Observer Based on Gaze Patterns When Looking at Faces
1 Introduction
2 Methods and Material
3 Results
3.1 Observer Related Variables
3.2 Image Properties
3.3 Classification
4 Summary
References
Gaze Stability During Ocular Proton Therapy: Quantitative Evaluation Based on Eye Surface Surveillance Videos
1 Introduction
2 Material and Methods
2.1 Patient Data
2.2 Automatic Pupil Detection Algorithm
2.3 Validation
2.4 Evaluation of Pupil Position Stability
3 Results
3.1 Validation
3.2 Evaluation of Pupil Position Stability
4 Discussion
References
Predicting Reading Speed from Eye-Movement Measures
1 Introduction
1.1 Eye Movements and Reading Speed
1.2 Effects of Inter-letter Spacing Modulation on Reading
2 Methods
2.1 Participants
2.2 Apparatus
2.3 Stimuli and Experimental Procedure
2.4 Data Analysis
3 Results
3.1 Correlation Analyses
3.2 Regression Using All Eye-Movement Measures (Set 1)
3.3 Regression Using Different Subsets of Eye-Movement Measures (Set 2–4)
4 Discussion
References
Investigating the Effect of Inter-letter Spacing Modulation on Data-Driven Detection of Developmental Dyslexia Based on Eye-Movement Correlates of Reading: A Machine Learning Approach
1 Introduction
2 Materials and Methods
2.1 Materials
2.2 Methods
3 Results
4 Discussion
References
FAPER - International Workshop on Fine Art Pattern Extraction and Recognition
Preface
Organization
Chairs
Program Committee
A Brief Overview of Deep Learning Approaches to Pattern Extraction and Recognition in Paintings and Drawings
1 Introduction
2 Main Datasets and Deep Learning Approaches
2.1 Datasets
2.2 Deep Learning Approaches
3 Main Research Trends
3.1 Artwork Attribute Prediction
3.2 Object Recognition and Detection
3.3 Content Generation
4 Concluding Remarks and Future Directions
References
Iconographic Image Captioning for Artworks
1 Introduction
2 Related Work
3 Experimental Setup
3.1 Iconclass Caption Dataset
3.2 Vision-Language Model
4 Results
4.1 Quantitative Results
4.2 Qualitative Analysis
5 Conclusion
References
Semantic Analysis of Cultural Heritage Data: Aligning Paintings and Descriptions in Art-Historic Collections
1 Introduction
2 Related Work
3 Challenges for Image and Text Alignment in the Cultural Heritage Domain
4 Alignment Approach
4.1 Word Encodings
4.2 Vocabulary Augmentation
4.3 Neural Style Transfer
5 Results and Discussion
5.1 Experimental Setup
5.2 Datasets
5.3 Results
6 Conclusion
References
Insights from a Large-Scale Database of Material Depictions in Paintings
1 Introduction
2 Dataset
3 Using Computer Vision to Analyze Paintings
3.1 Extracting Polygon Segments with Interactive Segmentation
3.2 Detecting Materials in Unlabeled Paintings
4 Using Paintings to Build Better Recognition Systems
4.1 Learning Robust Cues for Finegrained Fabric Classification
4.2 Benchmarking Unsupervised Domain Adaptation
5 Conclusion
References
An Analysis of the Transfer Learning of Convolutional Neural Networks for Artistic Images
1 Introduction
2 Related Work
2.1 Deep Transfer Learning for Art Classification Problems
2.2 Deep Convolutional Neural Network Understanding
2.3 Datasets
3 Analyzing CNNs Trained for Art Classification Tasks
3.1 From Natural to Art Images
3.2 Training from Scratch
3.3 Classification Performance
3.4 Quantitative Evaluation of the CNNs Modification
3.5 From One Art Dataset to Another
4 Conclusion
References
Handwriting Classification for the Analysis of Art-Historical Documents
1 Introduction
2 Related Work
3 Method
3.1 Handwriting Synthesis
3.2 Handwriting Classification Networks
4 Results and Discussion
4.1 Datasets
4.2 Experimental Setup
4.3 Results on the GANwriting Dataset
4.4 Results on the 5CHPT Dataset
4.5 Results on the WPI Dataset
4.6 Classification of an Additional Unseen Class
5 Conclusion
References
Color Space Exploration of Paintings Using a Novel Probabilistic Divergence
1 Introduction
2 The Color Theory
3 RGB Color Space
4 Probabilistic Divergence Measure
5 A Novel Probabilistic Divergence Measure
6 Experiments and Results
6.1 Classic Masters
6.2 Modern Masters
6.3 Abstract Masters
7 Related Works
8 Conclusion
References
Identifying Centres of Interest in Paintings Using Alignment and Edge Detection
1 Introduction
2 Case Studies
3 Step I. Finding the Original Image
4 Step II. Aligning the Painting and the Original
5 Step III. Micro-transformations
6 Step IV. Deconstructing Possible Meanings
7 A Second Case Study
8 Conclusions
References
Attention-Based Multi-modal Emotion Recognition from Art
1 Introduction
2 Related Work
2.1 Modalities in Emotion Recognition
2.2 Emotion Recognition from Art
2.3 Proposed Multi-modal Fusion Model
2.4 Image Feature Representation
2.5 Text Feature Representation
2.6 Attention Layer
2.7 Classification Layer
3 Experiment and Results
3.1 Dataset
3.2 Training Details
3.3 Baseline
3.4 Results
4 Conclusion
References
Machines Learning for Mixed Reality
1 Introduction
2 Milan Cathedral Survey. A Brief Overview
3 Multi-level Multi-resolution Classification
3.1 Methodology
3.2 Classification
4 Mixed Reality System
4.1 MR to Support Maintenance Works in Complex Architecture
4.2 The Developed Prototype
5 Conclusion and Future Works
References
From Fully Supervised to Blind Digital Anastylosis on DAFNE Dataset
1 Introduction
2 Related Works
3 The DAFNE Challenge
3.1 The Supervised Approach
4 Blind Digital Anastylosis
4.1 Blind Digital Anastylosis Like a Hard Jigsaw Puzzle Problem
4.2 A Preliminary Approach
5 Experimental Results
6 Conclusions
References
Restoration and Enhancement of Historical Stereo Photos Through Optical Flow
1 Introduction
2 Method Description
2.1 Auxiliary Image Point-Wise Transfer
2.2 Color Correction
2.3 Data Fusion
2.4 Refinement
3 Evaluation
3.1 Dataset
3.2 Compared Methods
3.3 Results
4 Conclusion and Future Work
References
Automatic Chain Line Segmentation in Historical Prints
1 Introduction
2 Related Work
3 Method
3.1 Line Segmentation Network
3.2 Line Detection and Parameterization
4 Evaluation
4.1 Dataset
4.2 Implementation Details and Evaluation Metrics
4.3 Results
5 Conclusion
References
Documenting the State of Preservation of Historical Stone Sculptures in Three Dimensions with Digital Tools
1 Introduction
2 Overview and Methods
3 Application and Results
4 Discussion and Conclusions
References
FBE2020 - Workshop on Facial and Body Expressions, micro-expressions and behavior recognition
Workshop on Facial and Body Expressions, micro-expressions and behavior recognition (FBE2020)
Organization
FBE2020 Workshop Chairs
Website Chair
Program Committee
Additional Reviewer
FBE 2020 Organizers
Motion Attention Deep Transfer Network for Cross-database Micro-expression Recognition
1 Introduction
2 Related Work
3 Methodology
3.1 Motion Attention Representation
3.2 Deep Transfer Network
4 Experiments
4.1 Databases
4.2 Implementation Details
4.3 Results
4.4 Ablation Analysis
5 Conclusion
References
Spatial Temporal Transformer Network for Skeleton-Based Action Recognition
1 Introduction
2 Spatial Temporal Transformer Network
2.1 Spatial Self-Attention (SSA)
2.2 Temporal Self-Attention (TSA)
2.3 Two-Stream Spatial Temporal Transformer Network
3 Model Evaluation
3.1 Datasets
3.2 Experimental Settings
3.3 Results
4 Comparison with State-of-the-Art
5 Conclusions
References
Slow Feature Subspace for Action Recognition
1 Introduction
2 Related Work
2.1 Slow Feature Analysis for Action Recognition
2.2 Subspace-Based Methods for Action Recognition
3 Proposed Method and Framework
3.1 Subspace Representation of Slowly Varying Components
3.2 Proposed Framework for Action Recognition
4 Experimental Results and Discussions
4.1 Experiment with KTH Action Dataset
4.2 Experiments with Isolated SLR500 Dataset
5 Conclusion and Future Work
References
Classification Mechanism of Convolutional Neural Network for Facial Expression Recognition
1 Introduction
2 Manifolds in Deep Learning
2.1 Encoder and Decoder
2.2 Traditional Expression Recognition Method
3 Proposed Model
3.1 Model Design
4 Classification Mechanism Analysis of the Network
4.1 Deconvolution Visualization
4.2 Facial Action Units
5 Experiments and Results
5.1 Dataset and Implementation
5.2 Criteria for Distance Measurement
5.3 Results
6 Conclusions
References
Applying Delaunay Triangulation Augmentation for Deep Learning Facial Expression Generation and Recognition
1 Introduction
2 Dataset
3 Augmented Dataset
3.1 Alignment, Centering and Cropping
3.2 Computing Delaunay Triangulation and Transform
3.3 Limitations
4 Facial Expression Generation Task
5 Facial Expression Recognition Task
6 Discussion
7 Future Work
8 Conclusion
References
Deformable Convolutional LSTM for Human Body Emotion Recognition
1 Introduction
2 Method
2.1 Input Preprocessing
2.2 Network Architecture
3 Experiments
3.1 Dataset
3.2 Comparison Between Deformable ConvLSTM and ConvLSTM
3.3 Comparison Between Other Methods
4 Conclusion
References
Nonlinear Temporal Correlation Based Network for Action Recognition
1 Introduction
2 Related Work
3 Proposed Network
3.1 Spatial-Temporal Separable Convolution
3.2 Correlation Based Feature Learning
3.3 NTE Block Design
3.4 Nonlinear Temporal Networks
4 Experiments
4.1 Datasets
4.2 Experimental Setup
4.3 Ablation Study
4.4 Study on the Change of Accuracy
4.5 Comparison with State-of-art on Mini-Kinetics-200 Datasets
4.6 Comparison with State-of-art on UCF-101 and HMDB-51
5 Conclusion
References
Author Index