Progress in Pattern Recognition, Image Analysis, Computer Vision, and Applications: 25th Iberoamerican Congress, CIARP 2021, Porto, Portugal, May ... Papers (Lecture Notes in Computer Science)

✍ Scribed by João Manuel R. S. Tavares (editor), João Paulo Papa (editor), Manuel González Hidalgo (editor)

• Publisher: Springer
• Year: 2022
• Tongue: English
• Leaves: 493
• Category: Library

✦ Synopsis

This book constitutes the proceedings of the 25th Iberoamerican Congress on Progress in Pattern Recognition, Image Analysis, Computer Vision, and Applications, CIARP 2021, which took place during May 10–13, 2021. The conference was initially planned to take place in Porto, Portugal, but changed to a virtual event due to the COVID-19 pandemic. The 45 papers presented in this volume were carefully reviewed and selected from 82 submissions. They were organized in topical sections as follows: medical applications; natural language processing; metaheuristics; image segmentation; databases; deep learning; explainable artificial intelligence; image processing; machine learning; and computer vision.

✦ Table of Contents

Preface
Organization
Contents
Medical Applications
Predicting the Use of Invasive Mechanical Ventilation in ICU COVID-19 Patients
1 Introduction
2 State of the Art
3 Data Set
4 Methods
4.1 Data Preparation
4.2 Data Pre-processing
4.3 Modelling
5 Results
6 Conclusions
References
A Coarse to Fine Corneal Ulcer Segmentation Approach Using U-net and DexiNed in Chain
1 Introduction
2 Related Works
3 Materials and Methods
3.1 Evaluated CNN Architectures
3.2 Image Dataset
3.3 Evaluation Metrics
4 Proposed Method
5 Results and Discussion
6 Conclusion
References
Replacing Data Augmentation with Rotation-Equivariant CNNs in Image-Based Classification of Oral Cancer
1 Introduction
2 Methodology
3 Oral Dataset
4 Experiments
5 Results
6 Conclusions and Future Work
References
A Multitasking Learning Framework for Dermoscopic Image Analysis
1 Introduction
2 Network Architecture and Learning Details
2.1 Learning Details
3 Experimental Design and Results
3.1 Dataset and Implementation Details
3.2 Experiments and Analysis
4 Conclusions
References
An Evaluation of Segmentation Techniques for Covid-19 Identification in Chest X-Ray
1 Introduction
2 Proposed Method
3 Experimental Setup
3.1 Parameters
3.2 Data Augmentation
3.3 Evaluation Metrics
4 Results and Discussion
4.1 Segmentation Performance
4.2 COVID-19 Identification Scores
4.3 Models Interpretability with LIME
5 Conclusion
References
A Study on Annotation Efficient Learning Methods for Segmentation in Prostate Histopathological Images
1 Introduction
2 Related Work
2.1 Segmentation
2.2 Cancer Detection in WSIs
2.3 Unsupervised Representation Learning
3 Methodology
3.1 U-Net
3.2 Transfer Learning
3.3 Self-supervised Learning
4 Evaluation
4.1 Dataset
4.2 Segmentation Results
5 Conclusion
References
Natural Language Processing
Data-Augmented Emoji Approach to Sentiment Classification of Tweets
1 Introduction and Background
1.1 Bidirectional Encoder Representations from Transformers (BERT)
2 Methodology
2.1 Datasets
2.2 Additional Pre-training
2.3 Data Augmentation
2.4 Emoji Extraction
2.5 Model Architecture
2.6 Training Protocol
3 Results
3.1 Evaluation Metrics
3.2 Pre-training Results
3.3 Data Augmentation Results
3.4 Fine-Tuning Results
4 Conclusions
References
Detecting Hate Speech in Cross-Lingual and Multi-lingual Settings Using Language Agnostic Representations
1 Introduction
2 Related Works
3 Proposal
4 Experiments
4.1 Dataset
4.2 Models and Evaluation Metrics
5 Results
5.1 Mono-lingual
5.2 Multi-lingual
5.3 Cross-Lingual
6 Conclusions
References
Prediction of Perception of Security Using Social Media Content
1 Introduction
2 Materials and Methods
2.1 Proposed Model
2.2 Estimating Model Parameters
2.3 Predicting Future Tweets
2.4 Experimental Settings
3 Results
4 Conclusions
References
Metaheuristics
Fine-Tuning Dropout Regularization in Energy-Based Deep Learning
1 Introduction
2 Related Works
3 Theoretical Background
3.1 Restricted Boltzmann Machines
3.2 Dropout-Based Restricted Boltzmann Machines
4 Methodology
4.1 Proposed Approach
4.2 Experimental Setup
4.3 Datasets
5 Experimental Results
5.1 Restricted Boltzmann Machines
5.2 Deep Belief Networks
6 Conclusion
References
Enhancing Hyper-to-Real Space Projections Through Euclidean Norm Meta-heuristic Optimization
1 Introduction
2 Hypercomplex Representation
2.1 Minkowski p-norm
3 Meta-heuristic Optimization
4 Methodology
4.1 Hypercomplex Optimization
4.2 Last Iteration Optimization
4.3 Benchmarking Functions
4.4 Experimental Setup
5 Experimental Results
5.1 Overall Discussion
5.2 Computational Burden
5.3 How Does p Influence Projections?
6 Conclusion
References
Using Particle Swarm Optimization with Gradient Descent for Parameter Learning in Convolutional Neural Networks
1 Introduction
2 Gradient-Based Learning in Neural Networks
3 Particle Swarm Optimization
4 Literature Review
5 Experimental Setup
5.1 MNIST Database
5.2 Evaluation
5.3 Hyperparameters
6 Model
7 Results
8 Conclusions and Future Work
References
Image Segmentation
Object Delineation by Iterative Dynamic Trees
1 Introduction
2 Iterative Dynamic Trees
2.1 Object Delineation by Image Foresting Transform
2.2 The IDT Algorithm
3 Experimental Results
4 Conclusion
References
Low-Cost Domain Adaptation for Crop and Weed Segmentation
1 Introduction
2 Background
3 Methodology
4 Results
5 Conclusions and Future Work
References
Databases
MIGMA: The Facial Emotion Image Dataset for Human Expression Recognition
1 Introduction
2 Related Works
3 Methodology: Proposed Dataset Environmental Protocol
4 Results
4.1 Dataset Properties
4.2 Dataset Statistical Analysis
4.3 Case-Study: Dataset Performance in a Convolutional Neural Network Framework
5 Conclusion and Discussions
References
Construction of Brazilian Regulatory Traffic Sign Recognition Dataset
1 Introduction
2 Related Works
3 Proposed Architecture
3.1 Image Pre-processing
3.2 The Dataset
3.3 CNN Architecture
4 Results and Discussion
5 Conclusion and Future Work
References
Japanese Kana and Brazilian Portuguese Manuscript Database
1 Introduction
2 The Dataset
3 Related Databases
4 Experimental Settings
4.1 Feature Extraction
4.2 Classifiers
5 Experimental Results and Discussion
5.1 Writer Identification
5.2 Syllabary Identification
6 Concluding Remarks
References
Skelibras: A Large 2D Skeleton Dataset of Dynamic Brazilian Signs
1 Introduction
2 Related Work
3 Corpus de Libras Dataset
4 Skelibras Dataset
5 Baseline Classifiers
6 Experiments
7 Conclusion
References
Deep Learning
Cricket Scene Analysis Using the RetinaNet Architecture
1 Introduction
2 Problem Background
2.1 Related Works
3 Dataset and Experiment Setup
4 RetinaNet Architecture
5 Results
6 Discussion and Critique
7 Conclusion
References
Texture-Based Image Transformations for Improved Deep Learning Classification
1 Introduction
2 Related Work
3 Proposed Method
4 Datasets
5 Evaluation and Results
5.1 Results for KTH-TIPS2-b Dataset
5.2 Results for Virus Dataset
6 Conclusion
References
Towards Precise Recognition of Pollen Bearing Bees by Convolutional Neural Networks
1 Introduction
2 Related Works
3 Convolutional Neural Networks Architectures
3.1 Transfer Learning
4 Pollen Bearing Bees Dataset
5 Experimental Setup
5.1 Colour Preprocessing Techniques
6 Results and Discussion
7 Conclusion
References
Web Application Attacks Detection Using Deep Learning
1 Introduction
2 Background and Related Work
3 A Two-Step Learning Approach for Anomaly Detection
3.1 Pre-training a HTTP Language Model
3.2 One-Class Classification
4 Results
5 Conclusion and Further Work
References
Less Is More: Accelerating Faster Neural Networks Straight from JPEG
1 Introduction
2 JPEG Compression
3 Related Work
4 Speeding up CNN Models Designed for DCT Input
4.1 Reducing the Number of Channels
4.2 Reducing the Number of Layers
5 Experiments and Results
5.1 Effects of Reducing the Number of Channels
5.2 Effects of Reducing the Number of Layers
6 Conclusion
References
Optimizing Person Re-Identification Using Generated Attention Masks
1 Introduction
2 Proposed Methodology
2.1 Network Architecture
2.2 Loss
3 Experimental Settings
3.1 Data
3.2 Data Augmentation
3.3 Training
4 Results and Discussion
5 Conclusion
References
Self-supervised Bernoulli Autoencoders for Semi-supervised Hashing*-8pt
1 Introduction
2 Related Works
3 Methods
3.1 Generative Model and Bernoulli Autoencoders
3.2 Parametrization by Neural Nets
3.3 Unsupervised Training
3.4 Semi-supervised Training
3.5 Efficient Implementation
4 Experiments
5 Conclusions
References
Explainable Artificial Intelligence
Interpretable Concept Drift
1 Introduction
2 Related Works
3 Visualizing Drift in Decision Trees
3.1 Node Frequency Analysis
3.2 Node Accuracy Analysis
4 Interpretable Drift Detector
5 Experiments
6 Conclusion
References
Interpreting a Conditional Generative Adversarial Network Model for Crime Prediction
1 Introduction
2 Related Work
3 Methodology
3.1 Dataset
3.2 Conditional Generative Adversarial Network
3.3 Training and Evaluation Metrics
3.4 SHAP Values and Model Interpretation
4 Results
4.1 cGAN
4.2 Analysis of SHAP Values
5 Conclusions and Future Work
References
Interpreting Decision Patterns in Financial Applications
1 Introduction
2 Background - Interpretable AI in Finance
2.1 Interpretability Approaches
2.2 Interpretability Models
3 Proposed Approach
4 Experimental Setup
4.1 Dataset Description
4.2 Evaluation Metrics
4.3 Models
5 Experimental Results and Analysis
6 Conclusions and Future Work
References
Image Processing
Metal Artifact Reduction Based on Color Mapping and Inpainting Techniques
1 Introduction
2 MAR Based on Color Mapping and Inpainting Techniques
2.1 Tone Mapping
2.2 Metallic Artifacts Classification
2.3 Artifacts Geometry Evaluation
2.4 Inpainting
2.5 3D Reconstruction
3 Results
3.1 Tone Mapping Enhancement
3.2 Structuring Element Analysis
3.3 Reconstruction
4 Discussion
5 Conclusion
References
New Improvement in Obtaining Monogenic Phase Congruency
1 Introduction
2 Monogenic Phase Congruency
3 Incorrect Edge Detection in MPC
4 Materials
5 Problem Solution
6 Experimental Results and Analysis
7 Conclusions
References
Machine Learning
Evaluating the Construction of Feature Descriptors in the Performance of the Image Data Stream Classification
1 Introduction
2 Background
3 Related Works
4 Experimental Method
5 Experimental Evaluation
5.1 Experimental Results
5.2 Statistical Analysis
6 Conclusion
References
Clustering-Based Partitioning of Water Distribution Networks for Leak Zone Location
1 Introduction
2 Materials
2.1 WDN Partitioning Strategies
3 Methodology
3.1 Class Formation
3.2 a DBSCAN Variation
4 Case Study
4.1 Modena WDN
4.2 Data Generation
4.3 Sensor Configuration
5 Results and Discussion
5.1 Topology-Based Clustering Methods
5.2 Hydraulics-Based Clustering Methods
5.3 Effect of the Variable Used
6 Conclusions
References
Bias Quantification for Protected Features in Pattern Classification Problems
1 Introduction
2 Fuzzy-Rough Set Theory
3 Similarity Function and Bias Quantification Measure
4 Experiments, Results and Discussion
5 Concluding Remarks
References
Regional Commodities Price Volatility Assessment Using Self-driven Recurrent Networks
1 Introduction
2 Commodities Prices Prediction Models
2.1 Problem Formulation
2.2 Recurrent Neural Network Architecture
2.3 Training Procedure
3 Experiments
3.1 Data
3.2 Hyperparameters Selection
3.3 International Shock Simulation
4 Conclusions
References
Semi-supervised Deep Learning Based on Label Propagation in a 2D Embedded Space
1 Introduction
2 Proposed Pipeline
2.1 Deep Feature Learning
2.2 Feature Space Projection
2.3 Label Propagation
3 Experiments and Results
3.1 Experimental Set-Up
3.2 Datasets
3.3 Implementation Details
3.4 Experimental Results
4 Discussion
5 Conclusion
References
Iterative Creation of Matching-Graphs – Finding Relevant Substructures in Graph Sets
1 Introduction and Related Work
2 Graphs and Graph Edit Distance - Basic Definitions
3 Matching-Graphs
3.1 Creating Matching-Graphs
3.2 Iterative Building of Matching-Graphs
4 Experimental Evaluation
4.1 Experimental Setup
4.2 Test Results and Discussion
5 Conclusions and Future Work
References
Semi-Autogeonous (SAG) Mill Overload Forecasting
1 Introduction
2 Prediction of Overloads
2.1 Related Work
3 Proposed Method
3.1 Feature Selection
3.2 Convolutional Neural Networks and Gram Matrices
3.3 Unbalanced Classes and Snowball Method
3.4 Summarize as a Whole Framework
4 Experiments and Results
5 Conclusions and Future Works
References
Novel Time-Frequency Based Scheme for Detecting Sound Events from Sound Background in Audio Segments
1 Introduction
2 Literature Review
3 Theoretical Framework
3.1 Wavelet Transform
3.2 Mel-Frequency Cepstral Coefficients
3.3 Support Vector Machine
3.4 K-Nearest Neighbor Classifier
4 Proposed Method
4.1 The US-SED Dataset
4.2 Pre-processing
4.3 Feature Extraction
4.4 Detection
5 Simulation Results
6 Conclusion
References
Computer Vision
Generalized Conics with the Sharp Corners
1 Introduction
2 Generalized Conics
3 Generalized Conics from Distance Transform
4 Multifocal Ellipse with Corners
5 Changing the Angle of the Egg-Shape Corner
6 Multifocal Hyperbola with Corners
7 Changing the Angle of the Hyperbolic Shape Corner
8 Shape Representation with the Generalized Conics
9 Conclusion
References
Automatic Face Mask Detection Using a Hide and Seek Algorithm
1 Introduction
2 Related Work
2.1 General Object Detection
2.2 Convolutional Neural Networks
3 Proposed Approach
4 Experimentation
4.1 Dataset
4.2 Evaluation Metrics
5 Result Analysis
5.1 Visual Intuition Development
5.2 Bias Elimination
5.3 Confusion Matrix and It's Analysis
5.4 Computational Power Comparison
5.5 Invariant to Orientation of Face
6 Discussion
6.1 Facial Point Choice
7 Conclusion and Future Works
References
A Feature Extraction Approach Based on LBP Operator and Complex Networks for Face Recognition
1 Introduction
2 Complex Networks
3 Methodology
3.1 Materials
4 Results and Discussion
5 Conclusions
References
End-to-End Deep Sketch-to-Photo Matching Enforcing Realistic Photo Generation
1 Introduction
2 Proposed Methodology
2.1 Network Architecture
2.2 Loss
3 Experimental Settings
3.1 Data
3.2 Pre-processing
3.3 Training
4 Results and Discussion
4.1 Realistic Generation Performance
5 Conclusion
References
Forensic Analysis of Tampered Digital Photos
1 Introduction
2 State of the Art
2.1 Digital Forensics
2.2 Multimedia Manipulation Techniques
2.3 Techniques Used to Detect Photos Manipulation
3 Architecture
3.1 General Architecture
3.2 Autopsy Module Architecture
4 Experimental Setup
4.1 Datasets
4.2 Evaluation Metrics
5 Results Analysis
6 Conclusion
References
COVID-19 Lung CT Images Recognition: A Feature-Based Approach
1 Introduction
2 Proposed Description and Proposed Classification Algorithm
2.1 Lung Segmentation
2.2 Histogram Computation
2.3 Feature Extraction
3 Performance Assessment
3.1 Patients Dataset
3.2 Classification Procedure and Results
4 Conclusions
References
A Topologically Consistent Color Digital Image Representation by a Single Tree
1 Introduction
2 Related Works
3 Building the CRIT
4 Conclusions and Future Work
References
Author Index