Pattern Recognition and Image Analysis: 11th Iberian Conference, IbPRIA 2023, Alicante, Spain, June 27–30, 2023, Proceedings (Lecture Notes in Computer Science)

This book constitutes the refereed proceedings of the 11th Iberian Conference on Pattern Recognition and Image Analysis, IbPRIA 2023, held in Alicante, Spain, in June 27–30, 2023. The 56 papers accepted for these proceedings were carefully reviewed and selected from 86 submissions. They deal with Machine Learning, Document Analysis, Computer Vision, 3D Computer Vision, Computer Vision Applications, Medical Imaging & Applications, Machine Learning Applications. Preface Organization Plenary Talks Distribution Shift: The Key Bottleneck for Pattern Recognition in Practice? Data Science Against COVID-19 AI for Sustainable Earth Sciences Invited Tutorials Machine Learning for Computational Photography A Brief History of Unsupervised Machine Translation: from a Crazy idea to the Future of MT? Continual Visual Learning: Where are we? Contents Machine Learning CCLM: Class-Conditional Label Noise Modelling 1 Introduction 2 Related Works 3 The Label Noise Modelling Obstacle 4 Proposed Methodology 5 Experiments and Results 5.1 Datasets and Implementation Details 5.2 Results 6 Conclusions References Addressing Class Imbalance in Multilabel Prototype Generation for k-Nearest Neighbor Classification 1 Introduction 2 Methodology 2.1 Candidate Selection Stage 2.2 Prototype Merging Policies 3 Experimental Set-Up 3.1 Datasets 3.2 Classification Method and PG Parameterization 3.3 Evaluation Metrics 4 Results 4.1 Multi-objective Optimization 4.2 Statistical Significance Analysis 5 Conclusions References Time Series Imputation in Faulty Systems 1 Introduction 2 Related Work 3 Methods 3.1 Dataset and Feature Engineering 3.2 Generation of Synthetic Missing Data 3.3 Algorithms Implementation and Evaluation 4 Results and Discussion 5 Conclusion and Future Work References DARTS with Degeneracy Correction 1 Introduction 2 Search Space 3 Continuous Transformation 3.1 DARTS Algorithm Limits 4 DARTS Improvement 4.1 Batch Norm 4.2 Discretization Error 4.3 Hypergradient Approximation 4.4 Reduction of the Number of Hyperparameters 4.5 Convex Depth Regularization 4.6 Customized Discretization Operator 4.7 Degeneracy Correction 5 Experiments 5.1 Settings 5.2 Results 6 High Number of Incomings Edges 7 Conclusion References A Fuzzy Logic Inference System for Display Characterization 1 Introduction 2 Fuzzy Logic Modelling Approach 2.1 Fuzzy Modelling and Identification Toolbox 2.2 Training a Fuzzy System with FMID for Display Characterization 2.3 Trained System Interpretation 3 Experimental Comparison and Discussion 4 Conclusions References Learning Semantic-Visual Embeddings with a Priority Queue 1 Introduction 2 Related Work 3 Learning with SGD + Priority Queue 3.1 Low Variance Loss Estimate 3.2 Robust Gradient Estimation 4 Experiments 4.1 Datasets 4.2 Training Details 4.3 Baselines and Metrics 4.4 Results and Discussion 4.5 Ablation Study 5 Conclusion References Optimizing Object Detection Models via Active Learning 1 Introduction 2 Background and Related Work 2.1 Object Detection Models 2.2 Active Learning 2.3 Active Learning Approaches 3 Methodology 4 Experimental Setup 4.1 Case Study 4.2 Experimental Design 4.3 Performance Metrics 5 Experimental Results and Analysis 6 Conclusions and Future Work References Continual Vocabularies to Tackle the Catastrophic Forgetting Problem in Machine Translation 1 Introduction 2 Related Work 3 Models 3.1 Compositional Embeddings 3.2 Transformer Architecture 4 Experimental Setup 4.1 Datasets 4.2 Training Details 4.3 Evaluation Metrics 5 Experimentation 5.1 Effects of the Vocabulary on the Catastrophic Forgetting Problem 5.2 On the Continual Vocabulary Problem 5.3 Tackling the Catastrophic Forgetting Problem 6 Conclusions References Evaluating Domain Generalization in Kitchen Utensils Classification 1 Introduction 2 Background 3 Methodology 3.1 Domain Generalization Methods 4 Experimental Setup 4.1 The Kurcuma Dataset 4.2 Implementation Details 5 Results 6 Conclusions References Document Analysis Segmentation of Large Historical Manuscript Bundles into Multi-page Deeds 1 Introduction 2 The JMDB Series of Notarial Record Manuscripts 3 Proposed Approach 3.1 Problem Statement and Proposed Approach 3.2 Individual Page Image Type Classification 4 Evaluation Measures 5 Data Set and Experimental Setup 5.1 Data Set 5.2 Empirical Settings 6 Results 7 Conclusion References A Study of Augmentation Methods for Handwritten Stenography Recognition 1 Introduction 2 Related Work 2.1 Handwritten Stenography Recognition 2.2 Data Augmentation Methods for Handwritten Text Recognition 3 Study Design 3.1 Examined Augmentations 3.2 Dataset 3.3 Model Architecture 3.4 Experimental Settings 4 Results and Discussion 5 Conclusions References Lifelong Learning for Document Image Binarization: An Experimental Study 1 Introduction 2 Methodology 2.1 Frozen Encoder 2.2 Vector Quantized-Variational AutoEncoder 2.3 Discrete Key-Value Bottleneck 2.4 Training Considerations 3 Experimental Setup 3.1 Datasets 3.2 CL Scenarios 3.3 Metrics 3.4 Implementation Details 4 Results 5 Conclusions References Test-Time Augmentation for Document Image Binarization 1 Introduction 2 Methodology 3 Experimental Setup 3.1 Corpora 3.2 Metrics 3.3 Implementation Details 4 Results 5 Conclusions References A Weakly-Supervised Approach for Layout Analysis in Music Score Images 1 Introduction 2 Methodology 3 Experimental Setup 3.1 Corpora 3.2 Metrics 3.3 Experiments 3.4 Implementation Details 4 Results 4.1 Evaluation Scenario I: Staff Detection 4.2 Evaluation Scenario II: End-to-End Recognition 4.3 Overview of the Pipeline Quality 5 Conclusions References ResPho(SC)Net: A Zero-Shot Learning Framework for Norwegian Handwritten Word Image Recognition 1 Introduction 2 Related Work 3 Methodology 3.1 Pipeline and Network Design 3.2 Problem Description 3.3 PHOS and PHOC Representation 3.4 ResPhoscNet Architecture 3.5 Loss Function 4 Experiments and Results 4.1 Training Set-Up 4.2 Datasets 4.3 Results on Norwegian Data 4.4 Comparison of Results on Norwegian Data and IAM Data 5 Conclusion References Computer Vision DeepArUco: Marker Detection and Classification in Challenging Lighting Conditions 1 Introduction 2 Related Works 3 Datasets 3.1 Flying-ArUco Dataset 3.2 Shadow-ArUco Dataset 4 Proposed method 4.1 Bounding-Box-Level Detection 4.2 Corner Regression 4.3 Marker Decoding 5 Experiments and Results 5.1 Metrics and Methodology 5.2 Implementation Details 5.3 Experimental Results and Analysis 5.4 Working Examples and Limitations 6 Conclusions and Future Work References Automated Detection and Identification of Olive Fruit Fly Using YOLOv7 Algorithm 1 Introduction 2 Related Work 2.1 Image Processing Techniques 2.2 Learning Techniques 3 Proposed Framework 3.1 Dataset Description 3.2 Data Preparation 3.3 Data Augmentation 3.4 YOLOv7 Architecture 3.5 YOLOv7 Training and Testing 4 Results Evaluation 5 Conclusion and Future Work References Learning to Search for and Detect Objects in Foveal Images Using Deep Learning 1 Introduction 2 Related Work 3 System Overview 3.1 Fixation Prediction Module 3.2 Target Detection Module 3.3 Dual Task Model 4 Implementation 4.1 Dataset 4.2 Training 4.3 Prediction 5 Results 5.1 Two Stage-Pipeline 5.2 Dual Task 6 Conclusions and Future Work References Relation Networks for Few-Shot Video Object Detection 1 Introduction 2 Related Work 3 Proposed Method 3.1 Problem Definition 3.2 FSVDet: Few-Shot Video Object Detection 3.3 Single Image Spatio-Temporal Training 3.4 Proposal Feature Aggregation 3.5 Loss Function 3.6 Confidence Score Optimization (CSO) 4 Experiments 5 Conclusions References Optimal Wavelength Selection for Deep Learning from Hyperspectral Images 1 Introduction 2 Related Work 3 Experimental Study 3.1 Experimental Setup 3.2 Deep Learning Vision Task 4 Experimental Analysis 4.1 Baseline Results 4.2 Greedy Selection 4.3 Bayesian Optimization 5 Discussion 5.1 Expert Interpretation of Optimal Wavelengths 5.2 Final Comparison of Techniques 6 Conclusions References Can Representation Learning for Multimodal Image Registration be Improved by Supervision of Intermediate Layers? 1 Introduction 2 Background 2.1 Representation Learning 2.2 Contrastive Learning 3 Method 3.1 Additional Supervision of the BN Latent Representation 3.2 Implementation Details 4 Evaluation 4.1 Datasets 4.2 Evaluation Metrics 5 Results 6 Exploration of the Embedding Space 7 Discussion 8 Conclusions References Interpretability-Guided Human Feedback During Neural Network Training 1 Introduction 2 Related Work 3 Proposal 3.1 Sampling 3.2 Interpretability 3.3 User Feedback 4 Experiments 4.1 Data 4.2 Model Architecture 4.3 Training Phase 5 Results 6 Discussion 7 Conclusion References Calibration of Non-Central Conical Catadioptric Systems from Parallel Lines 1 Introduction 1.1 Related Work 2 Background 2.1 Plücker Coordinate Lines 2.2 Lines in Conical Mirror Systems 3 Calibration Method 4 Experimental Results 4.1 Real and Synthetic Images 5 Conclusion References S2-LOR: Supervised Stream Learning for Object Recognition 1 Introduction 2 Related Work 3 Supervised Stream Learning for Object Recognition(S2-LOR) 3.1 Initialization Setup 3.2 Operation Mode 4 Experiments 4.1 The Benchmark: CORe50 4.2 Benchmark Configuration for S2-LOR 4.3 Experiment Results over CORE50 Dataset 5 Conclusions References Evaluation of Regularization Techniques for Transformers-Based Models 1 Introduction 2 Literature Review 3 Methodology 4 Setup and Experiments 5 Results and Discussion 6 Conclusions References 3D Computer Vision Guided Depth Completion Using Active Infrared Images in Time of Flight Systems 1 Introduction 2 Related Work 3 Proposed Approach 4 Experiments and Datasets 4.1 Experiments Based on the ADTF3175 Dataset 4.2 Experiments Based on RGB-D Middlebury Dataset 4.3 Selected Baseline Methods 5 Results and Discussion 5.1 Comparison to Baseline Methods 5.2 Edge Recovery Analysis 6 Conclusion References StOCaMo: Online Calibration Monitoring for Stereo Cameras 1 Introduction 2 Related Work 3 Methods 3.1 The StOCaMo Method 4 Experiments 4.1 Loss Function Shape 4.2 Decalibration Detection 4.3 Calibration Monitoring on Synthetic Decalibration 4.4 Long-Time SLAM Stability with Calibration Monitoring 4.5 Algorithmic Efficiency 5 Conclusion References Smart-Tree: Neural Medial Axis Approximation of Point Clouds for 3D Tree Skeletonization 1 Introduction 2 Method 2.1 Dataset 2.2 Skeletonization Overview 2.3 Neural Network 2.4 Skeletonization Algorithm 3 Results 3.1 Metrics 3.2 Quantitative Results 3.3 Qualitative Results 4 Conclusion and Future Work References A Measure of Tortuosity for 3D Curves: Identifying 3D Beating Patterns of Sperm Flagella 1 Introduction 2 Definitions 2.1 Computation of the Slope Changes of a 3D Polygonal Curve 2.2 Computation of the Torsions of a 3D Polygonal Curve 2.3 Definition and Computation of Accumulated Angles 3 The Proposed Measure of Tortuosity 3.1 Independence of Translation, Rotation, and Scale 3.2 Invariance Under Starting Point 3.3 Invariance Under Mirror Imaging 3.4 The Extrema of 3D Tortuosity 3.5 Normalized 3D Tortuosity 4 Results 4.1 Analysis of Closed Curves by Means of Their Tortuosity 4.2 Tortuosity of the 3D Flagellar Beat of Human Sperm 5 Conclusion References The ETS2 Dataset, Synthetic Data from Video Games for Monocular Depth Estimation 1 Introduction 2 Related Works 2.1 Real World Datasets 2.2 Synthetic Datasets 3 Obtaining Data from the Video Game 3.1 The Data Capture Tool 4 The ETS2 Dataset 5 Validating the Dataset 5.1 Training the Neural Networks with ETS2 5.2 Results 5.3 Discussion About Domain Gap 6 Conclusion References Computer Vision Applications Multimodal Human Pose Feature Fusion for Gait Recognition 1 Introduction 2 Related Work 3 Methodology 3.1 Pose Representations 3.2 Fusion Strategies 4 Experiments and Results 4.1 Datasets 4.2 Implementation Details 4.3 Baseline Results 4.4 Study of Early Fusion Strategies 4.5 Study of Late Fusion Strategies 4.6 Comparison to the State of the Art Using Pose 5 Conclusions References Proxemics-Net: Automatic Proxemics Recognition in Images 1 Introduction 2 Related Work 3 Proposed Method 3.1 Overview of the Model: Proxemics-Net 3.2 Evaluated Backbones 4 Experiments 4.1 Dataset: Proxemics 4.2 Metrics 4.3 Implementation Details 4.4 Experimental Results 5 Conclusions and Future Work References Lightweight Vision Transformers for Face Verification in the Wild 1 Introduction 2 Related Work 3 Proposed Approach 3.1 Vision Transformers 3.2 Knowledge Distillation 4 Experiments and Results 4.1 Datasets 4.2 Training Vision Transformers 4.3 Improving Lightweight Vision Transformers 4.4 Comparison with the State of the Art 5 Conclusions References Py4MER: A CTC-Based Mathematical Expression Recognition System 1 Introduction 2 Related Work 3 Problem Formulation 4 Data Set Characteristics 5 Model Architecture 6 Evaluation 6.1 Training and Evaluation Protocol 6.2 Results 7 Conclusion References Hierarchical Line Extremity Segmentation U-Net for the SoccerNet 2022 Calibration Challenge - Pitch Localization 1 Introduction 2 Related Work 3 Pitch Line and Line Extremity Segmentation 3.1 Modified U-Net Network Structure 3.2 Network Training 3.3 Extremity Coordinates Prediction 4 Experiments 5 Results 6 Conclusion References Object Localization with Multiplanar Fiducial Markers: Accurate Pose Estimation 1 Introduction 2 Proposed Method: DoducoPose 2.1 Markers Design 2.2 Marker Detection 2.3 Marker Mapping 2.4 Pose Estimation 2.5 Pose Refinement 3 Experiments and Results 4 Conclusions and Future Work References Real-Time Unsupervised Object Localization on the Edge for Airport Video Surveillance 1 Introduction 2 Related Work 2.1 Objects Localization Solutions 2.2 Implementations on Embedded Systems 3 Methodology 3.1 Pipeline Description 3.2 Hardware Optimizations 4 Experiments and Results 4.1 Dataset 4.2 Implementation Details 4.3 Performance Evaluation 4.4 Experimental Results 5 Conclusions References Identifying Thermokarst Lakes Using Discrete Wavelet Transform–Based Deep Learning Framework 1 Introduction 1.1 Background 2 Methods 2.1 Data Collection 2.2 Data Pre-processing 2.3 Model Framework 2.4 Training and Testing 3 Results and Model Evaluation 3.1 Confusion Matrices 3.2 Receiver Operating Characteristic (ROC) Curves 3.3 Summary of Metrics 4 Conclusion and Discussion References Object Detection for Rescue Operations by High-Altitude Infrared Thermal Imaging Collected by Unmanned Aerial Vehicles 1 Introduction 2 Background and Related Work 2.1 Neural Network Object Detection Methods 2.2 Infrared Object Detection 2.3 UAV Infrared Thermal Datasets 2.4 Object Detection Metrics 3 Methodology 3.1 Exploratory Data Analysis (EDA) 3.2 Model Selection 3.3 Experimental Workflow 4 Results and Their Discussion 5 Conclusion References Medical Imaging and Applications Inter vs. Intra Domain Study of COVID Chest X-Ray Classification with Imbalanced Datasets 1 Introduction 2 Methodology 2.1 Initialization 2.2 Data Augmentation 2.3 Imbalanced Data 2.4 Domain Adaptation 3 Experimental Setup 3.1 Datasets 3.2 Network Architecture 3.3 Metrics 4 Results 4.1 Initialization and Data Augmentation 4.2 Dealing with Imbalanced Data 4.3 Domain Adaptation 4.4 Discussion 5 Conclusions References Automatic Eye-Tracking-Assisted Chest Radiography Pathology Screening 1 Introduction 2 Materials and Methods 2.1 Datasets 2.2 Heatmap Generation 2.3 Eye-Tracking-Assisted Pathology Screening 2.4 Experiments 3 Results 3.1 Heatmap Reconstruction 3.2 Pathology Classification 3.3 Model Explainability 4 Conclusions References Deep Neural Networks to Distinguish Between Crohn's Disease and Ulcerative Colitis 1 Introduction 2 Related Work 3 Methodology 3.1 Dataset 3.2 Pre-Processing 3.3 Experimental Setup 3.4 Convolutional Neural Networks 3.5 Evaluation 4 Results 5 Conclusion References Few-Shot Image Classification for Automatic COVID-19 Diagnosis 1 Introduction 2 Related Work 3 Methodology 3.1 L-RPN: Lung-Aware Region Proposal Network 3.2 Support Set Ensemble 3.3 Misdiagnosis-Sensitive Learning 4 Experiments 4.1 Implementation Details 4.2 Results 5 Conclusions References An Ensemble-Based Phenotype Classifier to Diagnose Crohn's Disease from 16s rRNA Gene Sequences 1 Introduction 2 Materials and Methods 2.1 Dataset 2.2 EPheClass Pipeline for Phenotype Classification 3 Results 4 Discussion and Conclusion References Synthetic Spermatozoa Video Sequences Generation Using Adversarial Imitation Learning 1 Introduction 2 Related Works 3 Adversarial Imitation Learning on Spermatozoa Sequence Generation 3.1 Background 3.2 Spermatozoa Parametric Model 3.3 Image Generation 3.4 Sequence Generation 4 Experiments 4.1 Experimental Setup 4.2 Image Generation Analysis 4.3 Sequence Generation Analysis 5 Conclusion References A Deep Approach for Volumetric Tractography Segmentation 1 Introduction 2 3D Tractography Segmentation 2.1 Network Architecture 2.2 Loss Function 2.3 Pre-processing Proposals 3 Experimentation 3.1 Experimental Dataset 3.2 Results 4 Conclusions References MicrogliaJ: An Automatic Tool for Microglial Cell Detection and Segmentation 1 Introduction 2 Related Work 3 Materials and Methods 3.1 Animals 3.2 Tissue Preparation 3.3 Iba-1 Immunohistochemistry 3.4 Image Acquisition 3.5 ImageJ Manual Threshold Method 4 MicrogliaJ 5 Results and Discussion 6 Conclusions and Further Work References Automated Orientation Detection of 3D Head Reconstructions from sMRI Using Multiview Orthographic Projections: An Image Classification-Based Approach 1 Introduction 2 Related Work 3 Proposed Method 4 Experiments and Results 4.1 Data 4.2 Experimental Setup and Evaluation Metrics 4.3 Orientation Detection by Projection Image Classification 4.4 Comparison Against State-of-the-Art PCR Methods 5 Conclusions References Machine Learning Applications Enhancing Transferability of Adversarial Audio in Speaker Recognition Systems 1 Introduction 2 Related Work 3 Methodology 4 Experiments and Results 4.1 VoxCeleb Dataset 4.2 Target Models 4.3 Experiments 4.4 Results 5 Conclusion References Fishing Gear Classification from Vessel Trajectories and Velocity Profiles: Database and Benchmark 1 Introduction 2 Database and Methods 2.1 Database 2.2 Data Curation 2.3 Feature Extraction 3 Experiments and Results 4 Conclusions References Multi-view Infant Cry Classification 1 Introduction 1.1 Multi-view Learning 2 Proposed Method 2.1 Problem Statement 2.2 Method 3 Experimental Results 3.1 Data Sets 3.2 Feature Extraction 3.3 Experimental Setting 3.4 Multi-view Based Classification 3.5 Ablation Study 4 Conclusions References Study and Automatic Translation of Toki Pona 1 Introduction 2 Toki Pona 3 Dataset 4 Our Implementation 4.1 Model from Scratch 4.2 Transfer Learning with OPUS 5 Results 5.1 Model from Scratch 5.2 Transfer Learning with OPUS 6 Conclusion References Detecting Loose Wheel Bolts of a Vehicle Using Accelerometers in the Chassis 1 Introduction 2 Test Drives and Data Collection 3 Time Series Anomaly Detection 4 Anomalous Structure-Borne Sound Detection 4.1 Acoustic Features 4.2 Anomalous Sound Detection Algorithms 4.3 Metrics 5 Experiments 5.1 Feature Extraction 5.2 Training and Hyperparameters 5.3 Results and Discussion 6 Summary and Outlook References Clustering ECG Time Series for the Quantification of Physiological Reactions to Emotional Stimuli 1 Introduction 2 Materials and Methods 2.1 Experimental Protocol and ECG Acquisition 2.2 ECG Pre-processing and Feature Extraction 2.3 Detection of Time Instants with a Significant Group Response 2.4 Cluster Analysis of Individual Responses 3 Results and Discussion 4 Conclusions References Predicting the Subjective Responses' Emotion in Dialogues with Multi-Task Learning 1 Introduction 2 Related Work 3 Deep Learning Architectures for Emotional Response Prediction 3.1 Single-Task (ST) Architecture 3.2 Multi-Task (MT) Architecture 3.3 Implementation Details 4 Experiments and Results 4.1 PELD Dataset 4.2 Results 5 Conclusion References Few-Shot Learning for Prediction of Electricity Consumption Patterns 1 Introduction 2 Preliminaries 2.1 Dataset 2.2 Precision Metrics 3 Definition of the Problem 3.1 Models 3.2 Graph Neural Networks 3.3 Embeddings and Weight Pools 4 Methodology 4.1 Training Standard and Fine-Tuned Model 4.2 Generating Test Splits 5 Experimental Results 6 Conclusions References Author Index