Intelligent Systems and Pattern Recognition: Second International Conference, ISPR 2022, Hammamet, Tunisia, March 24–26, 2022, Revised Selected Papers ... in Computer and Information Science, 1589)

✍ Scribed by Akram Bennour (editor), Tolga Ensari (editor), Yousri Kessentini (editor), Sean Eom (editor)

• Publisher: Springer
• Year: 2022
• Tongue: English
• Leaves: 417
• Edition: 1st ed. 2022
• Category: Library

✦ Synopsis

This volume constitutes selected papers presented during the Second International Conference on Intelligent Systems and Pattern Recognition, ISPR 2022, held in Hammamet, Tunisia, in March 2022. Due to the COVID-19 pandemic the conference was held online. 
The 22 full papers and 10 short papers presented were thoroughly reviewed and selected from the 91 submissions. The papers are organized in the following topical sections: computer vision; data mining; pattern recognition; machine and deep learning.

✦ Table of Contents

Preface
Organization
Contents
Computer Vision
3D Dense & Scaled Reconstruction Pipeline with Smartphone Acquisition
1 Introduction
2 Related Work
3 3D Reconstruction Pipeline
3.1 Data Capture
3.2 Feature Points Extraction and Correspondence
3.3 Triangulation
3.4 Pose Estimation
3.5 Bundle Adjustment
3.6 Point Cloud Densification
3.7 Mesh Reconstruction
3.8 Texture Mapping
3.9 Scaling
3.10 Model Cleaning
4 Experimental Results and Limitations
4.1 Results
4.2 Limitations
5 Conclusion and Future Work
References
A Genetic Model for Medical Images Reproduction
1 Introduction
2 Proposed Method
2.1 Principle
2.2 Modeling Steps
3 Results and Discussion
4 Conclusion
References
A New Study of Needs and Motivations Generated by Virtual Reality Games and Factor Products for Generation Z in Bangkok
1 Introduction
1.1 Desire and Motivations for Playing Virtual Reality Games
1.2 Generation Z in Bangkok
2 Theoretical Framework and Bases for Study Measures
2.1 Uses and Gratifications
2.2 Concepts of Game Types Affecting Gaming Motivation
2.3 Demographic Concepts that Affecting the Gaming Motivation
3 Research Questions
4 Methods
4.1 Prototype
4.2 Survey Design
4.3 Sampling Frame
4.4 Sampling Techniques
4.5 Variables
5 Results
6 Discussion
7 Conclusion and Future Work
References
A Hybrid Method for Window Detection on High Resolution Facade Images
1 Introduction
2 Related Work
3 Dataset
4 Window Detection Network
5 Refinement of the Windows by Finding the Edges
6 Evaluation
7 Conclusion and Future Work
References
Neuro-Fuzzy Predictive Approach for Visual Analytics Evaluation of Medical Data
1 Introduction
2 Visual Analytics Evaluation
3 Learning Phase
3.1 Vague Nature of User’s Language
3.2 Fuzzy Intensifier Learning Process
3.3 Known Terms Learning Process
3.4 SOM Creation
3.5 Learning Phase
3.6 Execution Phase
4 Experimentation and Results
4.1 Step 1: Evaluation Questionnaire
4.2 Step 2: Fuzzy Logic Application
5 Conclusion
References
Improved Cerebral Images Semantic Segmentation Using Advanced Approaches of Deep Learning
1 Introduction
2 Related Work
3 Used Concepts
4 Proposed Approach
4.1 GAN for Generating Synthesis Images of the Minority Data Class
4.2 CycleGAN and UNet for Data Segmentation
5 Experiments and Results
5.1 Dataset
5.2 Augmentation of the Minority Data Class Using GAN
5.3 Data Segmentation Using CycleGAN and U-Net
5.4 Evaluation
5.5 Discussion
6 Conclusion
References
Self-supervised Learning for COVID-19 Detection from Chest X-ray Images
1 Introduction
2 Related Work
3 Method
3.1 Self-supervised Pre-training Based on Contrastive Learning
3.2 Supervised Fine-Tuning
4 Experiments and Results
4.1 Dataset
4.2 Implementation Details and Evaluation Metric
4.3 Experimental Results
5 Conclusion
References
Data Mining
Deep Learning-Based Segmentation of Connected Components in Arabic Handwritten Documents
1 Introduction
2 Related Works
3 Proposed Use of AR2U-Net
3.1 AR2U-Net Architecture
3.2 Post-processing
4 Experimental Results and Discussion
4.1 The Used Database
4.2 AR2U-Net Training
4.3 Analysis
5 Conclusion and Prospects
References
Classifying the Human Activities of Sensor Data Using Deep Neural Network
1 Introduction
2 Related Work
3 DNN Structure
3.1 Multilayer Perceptron (MLP)
3.2 MLP Training
4 Methodology
5 Data Set Description
6 Evaluation Metrics
7 Results
8 Conclusion
References
Exploratory Analysis of Driver and Vehicle Factors Associated with Traffic Accidents in Morocco
1 Introduction
2 Study Methodology
2.1 Exploratory Data Analysis
2.2 Data Preparation
2.3 Descriptive Analysis
2.4 Exploratory Analysis
3 Results and Discussion
4 Conclusion and Perspective
References
Building a Multilingual Corpus of Tweets Relating to Algerian Higher Education
1 Introduction
2 Data Source
3 Data Collection
4 Data Preprocessing
5 Sentiment Polarity Annotation
6 Conclusion and Future Work
References
Recursive Feature Elimination Technique for Technical Indicators Selection
1 Introduction
2 Related Work
3 Proposed Method
4 Experimental Results
5 Conclusion
References
Document-Based Knowledge Discovery with Microservices Architecture
1 Introduction
2 Research Context
2.1 Example Scenario
2.2 Research Questions
3 Background and Related Works
4 Conceptual Approach
4.1 The Microservices Specification
4.2 Data Model
4.3 Communication
5 Evaluation and Assessment
5.1 Implementation
5.2 Evaluation
6 Conclusion
References
Pattern Recognition
Feature Selection for Credit Risk Classification
1 Introduction
2 Related Literature
2.1 Feature Selection
2.2 Feature Selection for Credit Risk Classification
3 Methodology
3.1 Wrapper Algorithm Based on Importance Measure by Permutation
3.2 Logistic Regressiom with LASSO Regularization
3.3 Classification Based Measures
4 Experiments and Results
4.1 Data Pre-processing
4.2 Data Transformation
4.3 Subdivision into ``Training-Test'' Datasets
4.4 Results
4.5 Classification Accuracy
5 Conclusion and Future Work
References
Parameter Identification and Validation of Multi-innovation Least Squares Lithium Battery for Second-Order Battery Model
1 Introduction
2 Lithium-Ion Battery Model
2.1 Second-Order RC Equivalent Circuit Model
2.2 System Equation of State
2.3 Lithium Battery Identifiable Model
3 Multi-innovation Least Squares
3.1 Experimental Validation and Analysis
3.2 Experimental Platform
3.3 Experimental Procedure
3.4 Experimental Results and Analysis
4 Conclusion
References
Bat Echolocation Call Detection and Species Recognition by Transformers with Self-attention
1 Introduction
2 Related Work
3 Methods
3.1 Bat Echolocation Call Detection
3.2 Bat Species Recognition
4 Experiments
4.1 Quality Metrics
4.2 Data
4.3 Results
5 Conclusion
References
Scheduling Techniques for Liver Segmentation: ReduceLRonPlateau vs OneCycleLR
1 Introduction
2 Related Work
3 Methodology
3.1 Dataset and Pre-processing Techniques
3.2 Training Environment and Parameters
3.3 Test Records and Evaluation Metrics
4 Results and Discussion
5 Conclusion and Future Work
References
An Explainable Predictive Model for the Geolocation of English Tweets
1 Introduction
2 Theoretical Background
2.1 Overview of the DeepGeoloc Model
2.2 Overview of Explainable Artificial Intelligence for Natural Language Processing
3 Methodology
4 Results and Discussion
5 Conclusion
References
Removing Redundancies in Binary Images
1 Introduction
2 Motivations and Definitions
2.1 Graph-Based Representation
2.2 Image Pyramid
2.3 Combinatorial Pyramid
3 Structurally Redundant Edges
3.1 Selecting the Contraction Kernel
3.2 Redundant Edges
3.3 Removing Redundant Edges in Parallel
4 Memory Consumption
5 Comparisons and Results
6 Conclusion and Future Works
References
Neural Machine Translation of Low Resource Languages: Application to Transcriptions of Tunisian Dialect
1 Introduction
2 Related Work
3 PARASAR: PARAllel Speech Corpus of ARabic Language
3.1 PARASAR TD Data
3.2 TD-MSA Parallel Corpus
3.3 Data Preprocessing
3.4 TD-MSA Transcriptions Distribution
4 TD-MSA NMT Models
4.1 Seq2seq Models
4.2 Transformer Models
5 Experimental Results and Discussion
5.1 Data Augmentation
5.2 NMT Experiments
5.3 Discussion
6 Conclusion
References
SPIRAL: SPellIng eRror Parallel Corpus for Arabic Language
1 Introduction
2 Why Building a Corpus is Important in NLP?
3 Related Work
4 The Main Guidelines for Building the SPIRAL Corpus
4.1 Collecting Digital Arabic Words
4.2 Process of Error Generation
5 Results and Discussion
6 Conclusion
References
Machine and Deep Learning
VMs Migration Mechanism for Underloaded Machines in Green Cloud Computing
1 Introduction
2 Related Work
3 VMs Migration Mechanism
3.1 Case 1: An Underloaded Resource
3.2 Case 2: All Resources are Underloaded
3.3 Case 3: Large Number of Underloaded Resources
3.4 Case 4: Resources with Load = Max and Others with Loads < Min
3.5 Case 5: Independent Requests
3.6 Case 6: Dependent Requests
3.7 Case 7: Free Load not Available on the Activated Resources
3.8 Case 8: Overloaded Resources After Migration
3.9 Case 9: Part of Load will be Released
3.10 Case 10: The Resource will not be Underloaded
3.11 Case 11: VMs with Similar Execution Time
4 Comparison Between Related Work and our Mechanism
5 Experimentation
6 Conclusion
References
TunBERT: Pretrained Contextualized Text Representation for Tunisian Dialect
1 Introduction
2 Related Works
3 TunBERT
3.1 Training Setup
3.2 Pre-training Dataset
4 Evaluation
4.1 Sentiment Analysis
4.2 Tunisian Dialect Identification
4.3 Reading Comprehension Question-Answering
5 Experiments and Discussion
5.1 Tunisian Sentiment Analysis
5.2 Tunisian Dialect Identification
5.3 Reading Comprehension Question-Answering
5.4 Discussion
6 Conclusion
References
Road Recognition for Autonomous Vehicles Based on Intelligent Tire and SE-CNN
1 Introduction
2 Related Works
3 Intelligent Tire System
4 Methods
4.1 Data Preprocessing
4.2 Squeezing-and-Excitation Block
4.3 Proposed SE-CNN
5 Experiment
6 Results and Discussion
6.1 Implementation Details
6.2 Performance Evaluation and Comparison
6.3 Ablation Study
6.4 Discussion
7 Conclusion
References
Malicious Packet Classification Based on Neural Network Using Kitsune Features
1 Introduction
2 Related Work
3 Preliminaries
3.1 Problem Setting
3.2 Feature Extraction of Kitsune
4 Methodology
5 Experiment
5.1 CSE-CIC-IDS2018 Dataset
5.2 Labeling and Feature Extraction
5.3 Experiments and Results
6 Discussion
7 Conclusion
References
An Hybrid Deep Learning Approach for Prediction and Binary Classification of Student’s Stress
1 Introduction
2 Related Works of Stress Prediction
3 Proposed Stress Prediction Framework
3.1 Data Pre-processing
3.2 Autoencoder Element Extraction for Classification
3.3 LSTM Anticipating Architecture
4 Experimental Outcomes
5 Conclusion
References
A Novel Deep Convolutional Neural Network Architecture for Customer Counting in the Retail Environment
1 Introduction
2 Related Work
3 Proposed Approach
3.1 Modified Residual Block
3.2 Transposed Convolution
3.3 Dilated Spatial Attention Block
3.4 Implementation Details
4 Experimental Corpus and Data Augmentation
5 Experimentation and Discussion
6 Conclusion and Future Work
References
Transfer Learning for the Classification of Small-Cell and Non-small-Cell Lung Cancer
1 Introduction
2 Lung Cancer
3 Related Works
4 Method
4.1 Dataset
4.2 Data Augmentation
4.3 Proposed Approach
5 Results and Evaluation
6 Conclusion
References
Attentional Conditional Generative Adversarial Network for Ambient Occlusion Approximation
1 Introduction
2 Related Works
3 Preliminaries
3.1 Adversarial Neural Network
3.2 Attention Mechanism
4 Proposed Method
4.1 Architecture
4.2 Attention Mechanism for Ambient Occlusion
4.3 Loss Function
5 Experiments
5.1 Data Preparation
5.2 Qualitative and Quantitative Evaluation
5.3 Ablation Stady
6 Conclusion and Future Works
References
An Improvement of CNN Model for Traffic Sign Recognition and Classification
1 Introduction
2 Literature Survey
3 The Proposed Model
3.1 Data Acquisition
3.2 The Model Description
3.3 Data Preprocessing
3.4 Compilation
3.5 Train and Evaluation of the Model
3.6 Evaluation of the Model
4 Discussion
5 Conclusion
References
Social Media Sentiment Classification for Tunisian Dialect: A Deep Learning Approach
1 Introduction
2 Related Work
2.1 The Lexicon Based Approach
2.2 Machine Learning Based Approach
2.3 The Hybrid Approach
3 Tunisian Dialect Description
3.1 Description
3.2 Difficulties and Challenges
4 Our Proposed Approach
4.1 Dataset Description
4.2 Transliteration from Tunisian Arabizi to Tunisian Arabic
4.3 Dataset Pre-processing
4.4 Dataset Annotation
4.5 Vectorization
4.6 Sentiment Classification with Deep Learning Techniques
5 Experiments and Discussion
5.1 Setting Parameters
5.2 Experiments
6 Conclusion and Perspectives
References
Soil Moisture Prediction Based on Satellite Data Using a Novel Deep Learning Model
1 Introduction
2 Related Work
3 Methods
3.1 Data
3.2 Model Architecture
3.3 Evaluation Metrics
3.4 Deployment
4 Experiments
4.1 Optimization
4.2 Model Training
4.3 Results and Discussion
5 Conclusion and Further Work
References
Author Index