Image Analysis and Processing - ICIAP 2023 Workshops: Udine, Italy, September 1115, 2023, Proceedings, Part I (Lecture Notes in Computer Science, 14365)

The two-volume set LNCS 14365 and 14366 constitutes the papers of workshops hosted by the 22nd International Conference on Image Analysis and Processing, ICIAP 2023, held in Udine, Italy, in September 2023. In total, 72 workshop papers and 10 industrial poster session papers have been accepted for publication. Part I of the set, volume 14365, contains 10 papers from the industrial poster session, and 31 papers from the following workshops:– Advances in Gaze Analysis, Visual attention and Eye-gaze modelling (AGAVE)– Beyond Vision: Physics meets AI (BVPAI)– Automatic Affect Analysis and Synthesis (3AS)– International Contest on Fire Detection (ONFIRE)– Recent Advances in Digital Security: Biometrics and Forensics (BioFor)– Computer Vision for Environment Monitoring and Preservation (CVEMP)– Generation of Human Face and Body Behavior (GHB) Preface Organization Contents – Part I Contents – Part II Industrial Poster Session Instance Segmentation Applied to Underground Infrastructures 1 Introduction 2 Pipes Recognition 2.1 Dataset Creation 2.2 Dataset Description 2.3 Deep Learning Architecture 2.4 Classification Overhead 3 Experiments 3.1 Training 3.2 Instance Segmentation Results 3.3 Classification Refinement Results 4 Conclusion References Generating Invariance-Based Adversarial Examples: Bringing Humans Back into the Loop 1 Introduction 2 Adversarial Examples 2.1 Sensitivity-Based Adversarial Examples 2.2 Invariance-Based Adversarial Examples 3 Human Perception 4 Experiments 4.1 Stimuli 4.2 Human Subjects 4.3 Eye Tracking Experimental Set-Up 4.4 Creation of Occlusion-Based Adversarial Examples 5 Results 6 Adaptive Approaches 7 Conclusion References MARS: Mask Attention Refinement with Sequential Quadtree Nodes for Car Damage Instance Segmentation 1 Introduction 2 Related Works 3 Proposed Method 4 Experiments 4.1 Experimental Setup 4.2 Comparison with SOTA 4.3 Implementation Details 5 Conclusions References On-Device Learning with Binary Neural Networks 1 Introduction 2 Related Literature 2.1 Continual Learning 2.2 Binary Neural Networks 3 On-Device CWR Optimization 3.1 Gradients Computation 3.2 Quantization Strategy 4 Experiments 5 Conclusion References Towards One-Shot PCB Component Detection with YOLO 1 Introduction 2 Background 3 Methodology 3.1 Dataset Generation 3.2 Improving YOLOv5 Architecture for Small Object Detection 3.3 Training Procedure and Evaluation Metrics 4 Experimental Results 4.1 Experiments with Original YOLOv5 4.2 Experiments with Improved YOLOv5 5 Conclusion References Wildfires Classification: A Comparative Study 1 Introduction 2 Datasets for Fire and Smoke Analysis 3 Fire Analysis Techniques 4 Wildfires Classification 4.1 Selected Deep Models 4.2 Wildfires Dataset 4.3 Deploying the Model on Embedded Systems 5 Experimental Results 5.1 Results on the FireNet Dataset 5.2 Results on Our Wildfires Dataset 6 Conclusions and Future Works References A General Purpose Method for Image Collection Summarization and Exploration 1 Introduction and Background 2 Proposed Method 2.1 Group Selection 2.2 Clustering of Pictures Within a Group 2.3 Best Picture Selection 3 Datasets 3.1 Automatic Triage for a Photo Series 3.2 Camera Scene Detection Dataset 4 Experimental Setup 4.1 Data Setup 4.2 Evaluation Metrics 4.3 Experimental Details 5 Results 5.1 Subjective Results 6 Conclusions References Automated Identification of Failure Cases in Organ at Risk Segmentation Using Distance Metrics: A Study on CT Data 1 Introduction 2 Methods and Materials 2.1 Dice and Hausdorff Distance Metrics 2.2 Dataset 2.3 Model Architecture 2.4 Experimental Setup 3 Results and Discussion 4 Conclusion References Digitizer: A Synthetic Dataset for Well-Log Analysis 1 Introduction 2 Generating Synthetic Well-Log Curves 3 Dataset Analysis 4 Conclusions References CNN-BLSTM Model for Arabic Text Recognition in Unconstrained Captured Identity Documents 1 Introduction 2 Related Works 3 Problem Statement 3.1 Identity Document Reading 3.2 Arabic Text Recognition 4 Proposed Method 4.1 Architecture Description 4.2 Training Parameters 4.3 Data Pre-processing 5 Experiments 5.1 Dataset 5.2 Evaluation Metrics 5.3 Model Training 5.4 Model Evaluation 6 Conclusion References Advances in Gaze Analysis, Visual attention and Eye-gaze modelling (AGAVE) Detection and Localization of Changes in Immersive Virtual Reality 1 Introduction 2 Experiment and Computational Model 2.1 Task 2.2 Measures 2.3 Apparatus and Procedure 2.4 Participants 2.5 Computational Model 3 Results 3.1 Change Localization Experiment 3.2 The Proposed Model Accounts for the Patterns of Human Data 4 Discussion and Conclusion References Pain and Fear in the Eyes: Gaze Dynamics Predicts Social Anxiety from Fear Generalisation 1 Introduction 2 Background and Hypotheses 3 Method 3.1 Participants and Procedure 3.2 Proposed Model and Data Analyses 4 Results 5 Conclusions References Eye Gaze Analysis Towards an AI System for Dynamic Content Layout 1 Introduction 2 AI System 3 Experiment 3.1 Overview 3.2 Set-Up 3.3 Clips 3.4 Data Analysis 4 Results 4.1 Graphic Durations 4.2 Fixation Durations 4.3 Fixation Scores 4.4 Region of Interest Hypothesis 5 Discussion 5.1 Target 1 5.2 Target 2 5.3 Targets 3 and 4 5.4 Target 5 5.5 Target 6 6 Conclusions References Beyond Vision: Physics Meets AI (BVPAI) A Variational AutoEncoder for Model Independent Searches of New Physics at LHC 1 Introduction 2 The Physics Use-Case: An Effective Field Theory Interpretation of Vector Boson Scattering 2.1 Same Sign WW Scattering 2.2 Modeling the Anomalies: The SM as an Effective Field Theory 3 Variational AutoEncoders 3.1 The VAE Architecture 3.2 Anomaly Detection with VAEs 4 Embedding a Classification Step in the Training to Optimize for Discrimination 4.1 Results 5 Conclusions and Future Perspectives References Adaptive Voronoi Binning in Muon Radiography for Detecting Subsurface Cavities 1 Introduction to Muon Radiography 1.1 Cosmic Rays 1.2 Muon Radiography 2 The MIMA Detector and Its Application at the Temperino Mine 3 Adaptive Tessellation Through Voronoi Binning 4 Conclusion References Optimizing Deep Learning Models for Cell Recognition in Fluorescence Microscopy: The Impact of Loss Functions on Performance and Generalization 1 Introduction 2 Related Works 3 Methods 3.1 Model Training 4 Results 5 Discussion References A New IBA Imaging System for the Transportable MACHINA Accelerator 1 Introduction 2 Methods 3 Tests and Results 4 Conclusions References Abstracts Embeddings Evaluation: A Case Study of Artificial Intelligence and Medical Imaging for the COVID-19 Infection 1 Introduction 2 Related Works 3 Methodology 3.1 Dataset 3.2 Label Assignment 3.3 Models 3.4 Performance Metrics 4 Results and Discussions 5 Conclusions References Pigments and Brush Strokes: Investigating the Painting Techniques Using MA-XRF and Laser Profilometry 1 Introduction 2 Materials and Methods 3 Results and Discussion 4 Conclusions References Automatic Affect Analysis and Synthesis (3AS) Pain Classification and Intensity Estimation Through the Analysis of Facial Action Units 1 Introduction 2 Background and Rationales 3 Materials and Methods 3.1 Datasets 3.2 Methods 3.3 Results 4 Conclusion References Towards a Better Understanding of Human Emotions: Challenges of Dataset Labeling 1 Introduction 2 Related Work 3 Proposed Emotion Taxonomy 4 Experiments 4.1 Relabeling the AffectNet Dataset 4.2 Results 5 Conclusions References Video-Based Emotion Estimation Using Deep Neural Networks: A Comparative Study 1 Introduction 2 Related Work 3 The OMG-Emotion Dataset 4 The Video Emotion Estimation Pipeline 4.1 Pre-processing 4.2 Backbone 4.3 Temporal Aggregation 4.4 Multi-layer Perceptron 4.5 Loss Functions 5 Experiments and Results 5.1 Ablation Analysis 5.2 Model Comparison 5.3 State-of-the-Art Comparison 6 Conclusions A Appendix References International Contest on Fire Detection (ONFIRE) ONFIRE Contest 2023: Real-Time Fire Detection on the Edge 1 Introduction 2 Related Works 3 Contest Dataset and Task 4 Evaluation Metrics 5 Conclusion References FIRESTART: Fire Ignition Recognition with Enhanced Smoothing Techniques and Real-Time Tracking 1 Introduction 2 Materials and Methods 2.1 Dataset 2.2 Convolutional Neural Networks 2.3 Vision Transformers 2.4 The Proposed Approach 2.5 Metrics 3 Experimental Results and Discussion 3.1 Experimental Setup 3.2 Experimental Results 4 Conclusions References Rapid Fire Detection with Early Exiting 1 Introduction 2 Related Works 2.1 Fire Detection 2.2 Advantages of Deep Learning 2.3 Early Exiting 3 Methodology 3.1 Dataset 3.2 Model Architectures 4 Preliminary Results 5 Conclusions References Recent Advances in Digital Security: Biometrics and Forensics (BIOFORM) Morphing-Attacks Against Binary Fingervein Templates 1 Introduction 2 Morphing of Binary Finger Vein Templates 3 Experimental Settings 3.1 Assessment Criteria 3.2 Data and Recognition Software 4 Experimental Results 4.1 Threat Evaluation 4.2 Detecting Morphed Templates 5 Conclusion and Future Work References A Robust Approach for Crop Misalignment Estimation in Single and Double JPEG Compressed Images 1 Introduction 2 Proposed Approach 3 Experimental Results 4 Conclusions References Human-in-the-Loop Person Re-Identification as a Defence Against Adversarial Attacks 1 Introduction 2 Related Work 3 The HITL Approach as an Adversarial Defence 4 Experiments 5 Conclusions References Generalized Deepfake Detection Algorithm Based on Inconsistency Between Inner and Outer Faces 1 Introduction 2 Related Works on Deepfake Detection 3 The Proposed Method 4 Experimental Analysis and Results 4.1 Datasets 4.2 Experimental Setup 4.3 Results 5 Conclusions References Real-Time Multiclass Face Spoofing Recognition Through Spatiotemporal Convolutional 3D Features 1 Introduction 2 Proposed Method 3 Evaluation 3.1 Dataset and Evaluation Metrics 3.2 Implementation Details 3.3 Binary Classifier 3.4 Multiclass Classifier 3.5 Comparison with the State-of-the-Art 3.6 Ablation Study 4 Conclusion References Computer Vision for Environment Monitoring and Preservation (CVEMP) Enhancing Air Quality Forecasting Through Enhancing Air Quality Forecasting Through Deep Learning and Continuous Wavelet Transform 1 Introduction 2 Related Work 3 Methodology and Data 3.1 1D Analysis 3.2 2D Data Generation 3.3 2D Analysis 4 Experimental Setup 4.1 Dataset 4.2 Implementation Details 4.3 Performance Measures 5 Results 6 Conclusion References Optimize Vision Transformer Architecture via Efficient Attention Modules: A Study on the Monocular Depth Estimation Task 1 Introduction 2 Related Works 2.1 Depth Estimation 2.2 Efficient Vision Transformers 3 Proposed Method 3.1 Meta-METER 3.2 Pyra-METER 4 Results 4.1 Experimental Setup 4.2 Performance Analysis 4.3 Inference Time Analysis 5 Conclusions References Assessing Machine Learning Algorithms for Land Use and Land Cover Classification in Morocco Using Google Earth Engine 1 Introduction 2 Related Work 3 Methodology 3.1 Study Area and Dataset 3.2 Data Preparation 3.3 Machine Learning Algorithms 3.4 Evaluation of Accuracy 4 Results and Discussion 5 Conclusion References An Application of Artificial Intelligence and Genetic Algorithm to Support the Discovering of Roman Centuriation Remains 1 Introduction 2 Related Work 3 Methodology 3.1 Edges Detection 3.2 Lines Identification 3.3 Grid Alignment 4 Implementation 4.1 The U-Net Convolutional Neural Network 4.2 Image Processing 4.3 Genetic Algorithm 5 Experimental Results 6 Conclusions References Convolutional Neural Networks for the Detection of Esca Disease Complex in Asymptomatic Grapevine Leaves 1 Introduction 2 Materials and Methods 2.1 Leaf Samples 2.2 Hyperspectral Image Acquisition and Processing 3 Experiments and Results 4 Discussion 5 Conclusions and Future Work References ArcheoWeedNet: Weed Classification in the Parco archeologico del Colosseo 1 Introduction 2 Dataset Collection and Analysis 3 Methods 4 Results 5 Conclusions References Automatic Alignment of Multi-scale Aerial and Underwater Photogrammetric Point Clouds: A Case Study in the Maldivian Coral Reef 1 Introduction 2 Data Acquisition and Monocular Photogrammetric Point Clouds Generation 3 Manual Registration and Its Use as Ground Truth 4 The Automatic Registration Pipeline 5 Conclusions References Generation of Human Face and Body Behavior (GHB) Upsampling 4D Point Clouds of Human Body via Adversarial Generation 1 Introduction 2 Related Work 3 Upsampling of Time Varying Point Clouds 3.1 Proposed Architecture 4 Experiments 4.1 Implementation Details and Used Dataset 4.2 Results 4.3 Ablation Study 5 Conclusions References Decoding Deception: Understanding Human Discrimination Ability in Differentiating Authentic Faces from Deepfake Deceits 1 Introduction 1.1 Forms of Facial Manipulation 2 Methodology 2.1 Generative Adversarial Networks (GANs) 2.2 Conditional GANs (CGANs) 2.3 Faceswap 2.4 Lip-Sync 2.5 Dataset 2.6 Experimental Procedure 3 Results and Discussion 4 Conclusion References Generative Data Augmentation of Human Biomechanics 1 Introduction 2 Methods 2.1 Experimental Dataset 2.2 Data Pre-processing 2.3 Model Architecture 2.4 Model Training 2.5 Fine-Tuning 2.6 Evaluation 3 Results 3.1 Fine-Tuning 4 Discussion References Avatar Reaction to Multimodal Human Behavior 1 Introduction 2 Overview 3 Proposed Method 3.1 Motion Database 3.2 Data Capture 3.3 Action Prediction and Recognition 3.4 Facial Expression Recognition 4 Avatar Environment 5 Conclusions and Future Work References Author Index