Progress in Pattern Recognition, Image Analysis, Computer Vision, and Applications: 26th Iberoamerican Congress, CIARP 2023, Coimbra, Portugal, ... Part I (Lecture Notes in Computer Science)

✍ Scribed by Verónica Vasconcelos (editor), Inês Domingues (editor), Simão Paredes (editor)

• Publisher: Springer
• Year: 2023
• Tongue: English
• Leaves: 764
• Category: Library

✦ Synopsis

This 2-volume set, LNCS 14469 and 14470, constitutes the proceedings of the 26th Iberoamerican Congress on Progress in Pattern Recognition, Image Analysis, Computer Vision, and Applications, CIARP 2023, which took place in Coimbra, Portugal, in November 2023.
The 61 papers presented were carefully reviewed and selected from 106 submissions. And present research in the fields of pattern recognition, artificial intelligence, and related areas.

✦ Table of Contents

Preface
Organization
Contents – Part I
Contents – Part II
Deblur Capsule Networks
1 Introduction
2 Related Works
3 Deblur Capsule Networks
3.1 Blur Type Classification
3.2 Point Spread Function Reconstruction
3.3 Image Deep Regularized Deconvolution
4 Experiments and Analysis
4.1 DbCN Optimization Procedure
4.2 Synthetic Camera Motion Blur
4.3 Synthetic Multi-domain Blur
4.4 Ablation Study
5 Conclusions and Future Works
References
Graph Embedding of Almost Constant Large Graphs
1 Introduction
2 Graphs and Graph Embedding
3 GraphFingerprint: An Embedding for Almost Constant Graphs
3.1 Algorithm Input Parameters
3.2 Local Substructures
3.3 GraphFingerprint Definition
3.4 GraphFingerprint Examples
4 Experimental Section
4.1 From Metal-Oxide Nanocompound to GraphFingerprint
4.2 Toxicity Prediction Based on Global Features
4.3 Toxicity Prediction Based on GraphFingerprints
4.4 Toxicity Prediction Based on Global Features and GraphFingerprints
5 Conclusions
6 Future Work
References
Feature Importance for Clustering
1 Introduction
2 Cluster Analysis
3 Proposed Methods
3.1 Prototype-Based Feature Importance
3.2 SHAP-Based Feature Importance
4 Experimental Simulations
5 Concluding Remarks
References
Uncovering Manipulated Files Using Mathematical Natural Laws
1 Introduction
2 Related Work
3 Benford's Law Fundamentals
3.1 Benford's Law Statement
4 Dataset
5 Benford's Law-Based Method
5.1 Pre-processing
5.2 Processing
5.3 Median Absolute Deviation
5.4 Evaluation Metrics
6 Results
6.1 Analysis of Results
7 Conclusions and Future Work
References
History Based Incremental Singular Value Decomposition for Background Initialization and Foreground Segmentation
1 Introduction
2 Related Work
2.1 Identification of SFOs
2.2 Background Dependency
3 Methodology
3.1 Notation and Preliminaries
3.2 Computation of Incremental SVD
3.3 History Based Incremental SVD (hi-SVD)
4 Experimental Results
4.1 Datasets
4.2 SFO Status Identification Experiment
4.3 Foreground Segmentation Experiment
5 Conclusion
References
Vehicle Re-Identification Based on Unsupervised Domain Adaptation by Incremental Generation of Pseudo-Labels
1 Introduction
2 Related Work
3 Proposed Method
3.1 DBSCAN Pseudo-Labels
3.2 Fine-Tuning
3.3 Unsupervised Domain Adaptation
4 Experimental Validation
4.1 Dataset and Evaluation Environment
4.2 Implementation Details
4.3 Ablation Study for Eps-Neighborhood in DBSCAN
4.4 Ablation Study for Increasing the Number of Cycles in the Generation of Pseudo-Labels
5 Conclusions
References
How to Turn Your Camera into a Perfect Pinhole Model
1 Introduction
2 Methods
2.1 Gaussian Processes
2.2 Constructing an Ideal Pinhole Camera
2.3 The Datasets
3 Results
3.1 Collineation Assumption
3.2 Reprojection Error
3.3 Distortion Removal
4 Discussion
5 Conclusion
A Zhang's Method
B Simplified Zhang's Method
References
Single Image HDR Synthesis with Histogram Learning
1 Introduction
2 Method
2.1 LDR2EDR by Histogram and Resolution difference
2.2 EDR2HDR by Cumulative Histogram Learning
2.3 Fine-Tuning with Reinforcement Learning
3 Experiments
4 Conclusion
References
But That's Not Why: Inference Adjustment by Interactive Prototype Revision
1 Introduction
2 Prototype-Based Learning
3 Interactive Prototype Revision
4 Results
5 Conclusion
References
Teaching Practices Analysis Through Audio Signal Processing
1 Introduction
2 Related Work
3 Dataset Description
4 Classroom Activity Detection
4.1 Training and Testing Data
4.2 Unsupervised Diarization Approach
4.3 Supervised Audio Classification
4.4 Experiments and Results
5 Additional Tools for English Lessons Analysis
5.1 Language Detection
5.2 Key Phrases Matching
5.3 User Interface for Education Technicians
6 Conclusions and Further Work
References
Time Distributed Multiview Representation for Speech Emotion Recognition
1 Introduction
2 Related Work
3 Proposed Strategy
3.1 Database Description
3.2 General Architecture
3.3 Step 1 - Initial Procedures
3.4 Steps 2 and 3 - Algorithms and Combiner
3.5 Steps 4 and 5 - LSTM and Emotion Classification
4 Experimental Results
4.1 Experiment 1 - Results for All RAVDESS Database
4.2 Experiment 2 - Database Divided by Intensity
4.3 Experiment 3 - LOSO Protocol
5 Discussions
6 Conclusion
References
Detection of Covid-19 in Chest X-Ray Images Using Percolation Features and Hermite Polynomial Classification
1 Introduction
2 Materials and Methods
2.1 Image Database
2.2 Methodology
3 Results and Discussion
3.1 Feature Evaluation
3.2 Performance of the HP Classifier
4 Conclusion
References
Abandoned Object Detection Using Persistent Homology
1 Introduction
2 From Background Subtraction to Simplicial Complex
3 Surveillance Points
4 Filtration
5 Persistent Homology and Topological Signature
6 Detecting Abandoned Objects
7 Experimental Results
8 Conclusion and Future Works
References
Interactive Segmentation with Incremental Watershed Cuts
1 Introduction
2 Watershed Cuts
3 Semi-supervised Watershed Cut Algorithm with Interactions
3.1 Tree-Node Marking
3.2 Pixel Labeling
3.3 Incremental Workflow
4 Experiments
4.1 Experiment with User Generated Seeds
4.2 Experiment with Randomly Generated Seeds
5 Conclusion
References
Supervised Learning of Hierarchical Image Segmentation
1 Introduction
2 Ultrametric Dataset
3 Model
4 Evaluation Metrics
5 Experiments
6 Conclusion
References
Unveiling the Influence of Image Super-Resolution on Aerial Scene Classification
1 Introduction
2 Super-Resolution
2.1 Super-Resolution Convolution Neural Network (SRCNN)
2.2 Modified 3D Residual-in-Residual Dense Block (m3DRRDB)
2.3 SwinIR Transformer
3 Scene Classification
4 Experiments
4.1 Datasets
4.2 Experimental Settings
5 Results and Discussion
5.1 Experiment 1: Ranking of Super-Resolution Methods
5.2 Experiment 2: Impact of Super-Resolution on Aerial Scene Classification
6 Conclusion
References
Weeds Classification with Deep Learning: An Investigation Using CNN, Vision Transformers, Pyramid Vision Transformers, and Ensemble Strategy
1 Introduction
2 Related Work
3 Methodology
3.1 Datasets
3.2 Models
3.3 Ensemble
3.4 Evaluation
3.5 Training Configuration
3.6 Execution Environment
4 Results and Discussion
5 Conclusions
References
Leveraging Question Answering for Domain-Agnostic Information Extraction
1 Introduction
2 Related Work
3 Question Answering for Information Extraction
3.1 Models and Questions
3.2 Approach
3.3 Visual Explainability on Evaluation
4 Case Studies
4.1 Application to Toxicology Analysis
4.2 Application to Finance
5 Conclusion
References
Towards a Robust Solution for the Supermarket Shelf Audit Problem: Obsolete Price Tags in Shelves
1 Introduction
2 Dataset Collection
3 Proposed Approach
3.1 Image-to-Text Track
3.2 Localization Track
3.3 Selection Track
3.4 Report Generation
4 Results
5 Conclusions and Future Works
References
A Self-Organizing Map Clustering Approach to Support Territorial Zoning
1 Introduction
2 Related Work
2.1 Zoning with Self-Organizing Maps
2.2 Clustering Ordinal Categorical Data
3 Material and Methods
3.1 The Alto Taquari Basin - MS/MT, Brazil
3.2 Clustering Categorical Ordinal Data
3.3 Clustering Assessing
4 Results and Discussion
5 Conclusions
References
Spatial-Temporal Graph Transformer for Surgical Skill Assessment in Simulation Sessions
1 Introduction
2 Proposed Approach
2.1 Spectral Graph Convolutional Networks
2.2 Transformer Encoder
2.3 Surgical Skill Classifier
3 Experimental Results
3.1 Dataset
3.2 Data Preprocessing
3.3 Results
4 Conclusion
References
Deep Learning in the Identification of Psoriatic Skin Lesions
1 Introduction
2 Background and Literature Review
3 Methodology
3.1 Understanding the Problem
3.2 Deep Learning
3.3 Dataset
3.4 Classification Architectures
3.5 Training
4 Results
4.1 CLAHE
4.2 Type of Input Image
4.3 Data Augmentation
5 Discussion
6 Mobile Application
7 Conclusion
References
WildFruiP: Estimating Fruit Physicochemical Parameters from Images Captured in the Wild
1 Introduction
2 Related Work
2.1 Detection Methods
2.2 Segmentation Methods
2.3 Methods for Estimating Fruit Ripeness in Images
3 Proposed Method
3.1 Fruit Detection and Segmentation
3.2 Image Alignment
3.3 Determination of Physicochemical Parameters
4 Dataset
5 Experiments
5.1 Implementation Details
5.2 Metrics
5.3 Performance of the Proposed Approach
5.4 Impact of Alignment Phase
5.5 Impact of Model Architecture
5.6 Hard Samples
6 Conclusion and Future Work Prospects
References
Depression Detection Using Deep Learning and Natural Language Processing Techniques: A Comparative Study
1 Introduction
2 Related Work
3 Methodology
3.1 Dataset
3.2 Data Pre-processing
3.3 Manual Validation of Label
3.4 Exploratory Data Analysis
3.5 Feature Generation
3.6 Experimental Setup
3.7 Evaluation
4 Results
5 Discussion
6 Conclusion
References
Impact of Synthetic Images on Morphing Attack Detection Using a Siamese Network
1 Introduction
2 Related Work
3 Database
3.1 SYN-MAD-2022
3.2 SOTA Databases
3.3 Metrics
4 Method
5 Experiments and Results
5.1 Exp1 - Train SYN-MAD/TEST SDD-Bechmark
5.2 Exp2 - Train SYN-MAD/TEST SOTA
5.3 Exp3 - Train SOTA/TEST SDD-Benchmark
5.4 Exp4 - Train Mix/TEST SDD-Benchmark
6 Conclusions
References
Face Image Quality Estimation on Presentation Attack Detection
1 Introduction
2 Relatwd Work
2.1 Face PAD
2.2 Face Image Quality Assessment
2.3 Face Image Quality Applied to PAD
2.4 Datasets
3 Method
3.1 Training of PAD Algorithms
4 Experiments and Results
4.1 FIQA Effect on Filtering PAs
4.2 PAD Performance Versus Input Face Quality
4.3 FIQA Filtering of the Training Dataset
5 Conclusions and Future Work
References
Knowledge Distillation of Vision Transformers and Convolutional Networks to Predict Inflammatory Bowel Disease
1 Introduction
2 Methodology and Data
2.1 Dataset
2.2 Experimental Setup
2.3 Pre-processing
2.4 Data Augmentation
2.5 Deep Learning Models
2.6 Knowledge Distillation
2.7 Evaluation
3 Results
4 Discussion
5 Conclusion
References
Analysis and Impact of Training Set Size in Cross-Subject Human Activity Recognition
1 Introduction
2 Related Work
3 Materials and Methods
3.1 Data Collection
3.2 Data Processing
3.3 Model Architecture
3.4 Performance Evaluation
4 Results
5 Discussion
6 Conclusion
References
Efficient Brazilian Sign Language Recognition: A Study on Mobile Devices
1 Introduction
2 Related Works
3 Methodology
3.1 Dataset
3.2 Data Preprocessing
3.3 Proposed Model
3.4 Training Details
3.5 Hardware Specifications
3.6 Experiments Conducted
4 Results
5 Discussion
6 Conclusion
References
Presumably Correct Undersampling
1 Introduction
2 Presumably Correct Decision Sets
3 Presumably Correct Undersampling
4 Experimental Simulations
5 Concluding Remarks
References
Leveraging Longitudinal Data for Cardiomegaly and Change Detection in Chest Radiography
1 Introduction
2 Related Work
3 Methodology
3.1 Dataset
3.2 Experimental Settings
3.3 Rigid CXR Alignment
3.4 Longitudinal Data Augmentation
3.5 Evaluation and Metrics
4 Results and Discussion
4.1 Rigid CXR Alignment
4.2 Pathology and Change Prediction
4.3 Longitudinal Data Augmentation
5 Conclusions
References
Self-supervised Monocular Depth Estimation on Unseen Synthetic Cameras
1 Introduction
2 Related Work
2.1 Self-supervised Monocular Depth Estimation
2.2 Restrictions of Existing Methods
2.3 Approaches to Address Camera Restrictions
3 Method
3.1 Baseline Algorithm
3.2 Baseline Algorithm + Adversarial Training
4 Experiments
4.1 Setup
4.2 Analysis of the Results
5 Conclusion/Discussion
References
Novelty Detection in Human-Machine Interaction Through a Multimodal Approach
1 Introduction
2 Related Work
2.1 Terminology
2.2 Existing Modeling Architectures
3 Methodology
4 Datasets
5 Experiments
5.1 Distance-Based Experiment
5.2 Distribution and Density-Based Experiments
5.3 Non Threshold-Based Models
6 Performance Optimization
6.1 NCM-Based Algorithm
6.2 Distribution and Density-Based Approaches
7 Conclusions
References
Filtering Safe Temporal Motifs in Dynamic Graphs for Dissemination Purposes
1 Introduction
2 Related Work
3 Fundamental Concepts
3.1 Temporal Graph
3.2 Temporal Motifs
3.3 Transitive Reduction
4 Filtering Motifs on Temporal Graphs
5 Filtering Transitive Edges
6 Experiments
6.1 Experimental Setup
6.2 Evaluation and Discussion
7 Conclusion
References
Graph-Based Feature Learning from Image Markers
1 Introduction
2 Related Works
3 Graph-Based Feature Learning from Image Markers
3.1 Superpixel Graphs
3.2 Kernel Computation
3.3 User-Guided Decoder for Object Detection
4 Experiments
4.1 Architecture Parameters
4.2 Bounding Boxes Computation
4.3 Evaluation Measures
4.4 Results
5 Conclusion
References
Seabream Freshness Classification Using Vision Transformers
1 Introduction
2 State-of-the-Art Review
2.1 Traditional Systems
2.2 Deep Learning Based Systems
2.3 Discussion
3 Seabream Freshness Dataset
3.1 Image Acquisition
3.2 Equipment
3.3 Storage
4 Proposed Fish-Freshness Classification System
4.1 Image Preprocessing and Segmentation
4.2 Image Augmentation
4.3 Feature Extraction and Classification
5 Results and Discussion
5.1 Segmentation Model
5.2 Feature Extraction and Classification Model
6 Conclusions and Future Work
References
Explaining Semantic Text Similarity in Knowledge Graphs
1 Introduction
2 Related Work
3 Preliminary Definitions
3.1 Knowledge Graphs
3.2 Saliency Scores
3.3 Integrated Gradients
3.4 Explanations
4 Methodology
5 Experiments and Results
5.1 Experiment 1. Discriminating Power of the STS Metrics
5.2 Experiment 2. Linking Children to Parents
5.3 Explanations for the Concept Linking Task
6 Conclusions
References
Active Supervision: Human in the Loop
1 Introduction
2 Related Work
3 Proposal
4 Experiments
5 Discussion
6 Conclusion
References
Condition Invariance for Autonomous Driving by Adversarial Learning
1 Introduction
2 Related Work
2.1 Object Detection
2.2 Domain Adaptation and Feature Invariance
2.3 Domain Adaptation in Autonomous Driving
3 Proposal
4 Implementation
4.1 Data
4.2 Model
4.3 Training and Testing
4.4 Performance Metrics
5 Results and Discussion
6 Conclusion
References
YOLOMM – You Only Look Once for Multi-modal Multi-tasking
1 Introduction
2 State of the Art
2.1 Model Selection
2.2 Dataset Selection
3 Implementation
3.1 Multi-modality
3.2 Lidar Segmentation
3.3 Training Process
4 Experiments and Results
5 Conclusion
References
Classify NIR Iris Images Under Alcohol/Drugs/Sleepiness Conditions Using a Siamese Network
1 Introduction
1.1 Related Work
1.2 Method
1.3 Metrics
1.4 Database
1.5 Experiment and Results
2 Comparison with SOTA
3 Visualisation
4 Conclusion
References
Bipartite Graph Coarsening for Text Classification Using Graph Neural Networks
1 Introduction
2 Related Work
3 Proposed Method
3.1 Graph Creation and Coarsening Step
3.2 Graph Representation Learning
4 Experiments
4.1 Datasets
4.2 Hyperparameter Settings
4.3 Experimental Results
5 Concluding Remarks
References
Towards Robust Defect Detection in Casting Using Contrastive Learning
1 Introduction
2 Related Work
3 Method
4 Experiments
4.1 Dataset Description
4.2 Experimental Procedure and Preliminary Results
4.3 Robust Defect Detection
4.4 Discussion
5 Conclusions
References
Development and Testing of an MRI-Compatible Immobilization Device for Head and Neck Imaging
1 Introduction
2 Materials and Methods
2.1 The Immobilization Device
2.2 Software Developed to Analyse MRI Images
2.3 Digital Algorithm Tests
2.4 Movement Phantom Algorithm Tests
3 Results
4 Discussion
5 Conclusions
References
DIF-SR: A Differential Item Functioning-Based Sample Reweighting Method
1 Introduction
2 Background
2.1 Group Fairness Analysis
2.2 Item Response Theory
2.3 Differential Item Functioning
2.4 Related Work
3 Proposed Method: DIF-SR
3.1 Base Classifier Predictions—Step 1
3.2 Item Modeling—Step 2
3.3 IRT Calibration—Step 3
3.4 Sample Weighting—Step 4
4 Experimental Setting
4.1 Datasets
4.2 Classification Algorithms
4.3 Sample Reweighting Methods
4.4 Evaluation Measures
5 Results
6 Conclusion
References
IR-Guided Energy Optimization Framework for Depth Enhancement in Time of Flight Imaging
1 Introduction
2 Related Work
3 IR Image as Guidance
4 Proposed Model
4.1 Image Energy Function
4.2 Spatial Error Energy Term
4.3 Conditional Entropy Energy Term
4.4 Image Energy Minimization
5 Data and Preprocessing
6 Experiments and Results
7 Conclusions
References
Multi-conformation Aproach of ENM-NMA Dynamic-Based Descriptors for HIV Drug Resistance Prediction
1 Introduction
2 Materials and Methods
2.1 Dataset
2.2 Computational Prediction
3 Results and Discussion
4 Conclusions and Further Work
References
Replay-Based Online Adaptation for Unsupervised Deep Visual Odometry
1 Introduction
1.1 Unsupervised Deep Visual Odometry
1.2 Online Adaptation
2 Methodology
2.1 Self-supervision
2.2 Replay-Based Online Adaptation
2.3 Model Setup
3 Evaluation Protocol
4 Results
5 Conclusion
References
Facial Point Graphs for Stroke Identification
1 Introduction
2 Related Works
3 Theoretical Background
3.1 Graph Neural Networks
3.2 Graph Attention Networks
4 Experimental Methodology
4.1 Experimental Dataset
4.2 Pre-processing and Feature Extraction
4.3 Classification and Evaluation
4.4 Proposed Model
5 Results
6 Discussion and Conclusion
References
Fast, Memory-Efficient Spectral Clustering with Cosine Similarity
1 Introduction
2 Methodology
2.1 Learning from a Single Batch of Data
2.2 How to Choose the Batch Size s
2.3 Out of Sample Extension
3 Experiments
3.1 Real-World Benchmark Data Sets
3.2 Experimental Setup
3.3 Choosing the Optimal Sample Size
3.4 Method Comparisons
4 Conclusions and Future Work
References
An End-to-End Deep Learning Approach for Video Captioning Through Mobile Devices
1 Introduction
2 Related Works
3 Video Captioning Frameworks
3.1 FW1: Multiple Image Captioning with Audio Classification
3.2 FW2: Video Captioning with Audio Classification
4 Experiment Setup
4.1 Datasets and Backbones
4.2 Hardware, Software and Training Settings
5 Results and Discussion
5.1 Image Feature Extractors Comparison
5.2 Qualitative Assessment of Video Descriptions
5.3 Strategies Resources Consumption
6 Conclusion and Future Works
References
Stingless Bee Classification: A New Dataset and Baseline Results
1 Introduction
2 Related Work
3 Methodological Design
3.1 Data Acquisition
3.2 Classifier Description
3.3 Experimental Setup
4 Results and Discussions
5 Conclusion and Future Work
References
Author Index