Computer Vision and Machine Intelligence: Proceedings of CVMI 2022

✍ Scribed by Massimo Tistarelli, Shiv Ram Dubey, Satish Kumar Singh, Xiaoyi Jiang

• Publisher: Springer
• Year: 2023
• Tongue: English
• Leaves: 777
• Series: Lecture Notes in Networks and Systems, 586
• Category: Library

✦ Synopsis

This book presents selected research papers on current developments in the fields of computer vision and machine intelligence from International Conference on Computer Vision and Machine Intelligence (CVMI 2022). The book covers topics in image processing, artificial intelligence, machine learning, deep learning, computer vision, machine intelligence, etc. The book is useful for researchers, postgraduate and undergraduate students, and professionals working in this domain.

✦ Table of Contents

Organization
Preface
Contents
Editors and Contributors
Efficient Voluntary Contact-Tracing System and Network for COVID-19 Patients Using Sound Waves and Predictive Analysis Using K-Means
1 Introduction
2 Literature Review
3 System Design or Methodology
4 Proposed Network Protocol
4.1 Solution to Objective 1: Communication Protocol
4.2 Solution to Objective 2: Predictive Analysis
5 Opportunity and Threat Analysis
6 Limitations of Study
7 Conclusion and Future Scope
References
Direct De Novo Molecule Generation Using Probabilistic Diverse Variational Autoencoder
1 Introduction
1.1 Contributions
1.2 Organization
2 Related Work
3 Representation and Dataset
3.1 Molecule Representation
3.2 Dataset
4 Experimental Setup
4.1 Model Architecture
4.2 Training of dVAE
4.3 Interpolation and Diversity
5 Results and Discussion
5.1 Compared Model
5.2 Evaluation Matrices
6 Conclusion
References
Automated Molecular Subtyping of Breast Cancer Through Immunohistochemistry Image Analysis
1 Introduction
2 Related Works
3 Proposed Method
3.1 Dataset
3.2 Segmentation
3.3 Classification and Molecular Subtyping
4 Experimental Results and Discussion
4.1 Experimental Setup
5 Conclusion
6 Compliance with Ethical Standards
References
Emotions Classification Using EEG in Health Care
1 Introduction
2 Proposed System
2.1 The EEG Headset
2.2 Wavelet Decomposition and Feature Extraction
2.3 Classification of Emotions Using Different Classifiers
3 Experiments and Results
3.1 Dataset Description and Experimental Protocol
3.2 Emotion Classification Results Using RF
3.3 Comparative Analysis of SVMs Using Different Kernels
3.4 Comparative Analysis at Different Brain Lobes
3.5 Comparative Analysis by Varying Time
3.6 Comparative Performance Analysis
4 Conclusion and Future Scope
References
Moment Centralization-Based Gradient Descent Optimizers for Convolutional Neural Networks
1 Introduction
2 Proposed Moment Centralization-Based SGD Optimizers
3 Experimental Setup
3.1 CNN Models Used
3.2 Datasets Used
3.3 Hyperparameter Settings
4 Experimental Results and Analysis
5 Conclusion
References
Context Unaware Knowledge Distillation for Image Retrieval
1 Introduction
2 Related Work
3 Context Unaware Knowledge Distillation
3.1 ResNet Overview
3.2 Student Model
3.3 Knowledge Distillation
4 Experiments
4.1 Datasets and Evaluation Metrics
4.2 Training and Results
5 Conclusions
References
Detection of Motion Vector-Based Stegomalware in Video Files
1 Introduction
2 Literature Survey
2.1 Steganography in Video Files
2.2 Steganalysis in Video Files
3 Video Compression and Decompression
3.1 Motion Estimation and Motion Compensation
3.2 Block Matching Algorithms
3.3 Variable-Sized Macroblocks
3.4 Sub-pixel Motion Estimation
4 Proposed Design
4.1 Dataset and Preprocessing
4.2 Feature Extraction
4.3 Dimensionality Reduction
5 Implementation Results
6 Conclusion
References
Unsupervised Description of 3D Shapes by Superquadrics Using Deep Learning
1 Introduction
2 Related Work
3 Superquadrics and Our New Model
3.1 Setup of Our Network
3.2 New Model for the Loss Function
3.3 Comparison to the Previous Model for the Loss Function
4 Experimental Evaluation
5 Conclusion
References
Multimodal Controller for Generative Models
1 Introduction
2 Related Work
3 Multimodal Controller
4 Experiments
4.1 Image Generation
4.2 Image Creation from Novel Data Modalities
5 Conclusion
References
TexIm: A Novel Text-to-Image Encoding Technique Using BERT
1 Introduction
2 Related Works
2.1 Representation of Text
2.2 Compression of Text
2.3 Visualization and Analysis of Text
3 Proposed Methodology
3.1 Pre-processing of Text
3.2 Representation of Text
3.3 Dimension Reduction
3.4 Normalization and Feature Scaling
3.5 RGB Feature Matrix Generation
3.6 Text to RGB Sequence Conversion
3.7 Reshaping RGB Matrix and Image Generation
4 Experimental Setup
4.1 Data
4.2 Implementation Details and Hyperparameters
5 Results and Analysis
5.1 Demonstration of TexIm
5.2 Representation of the TexIm Embeddings
5.3 Evaluation of Efficacy
6 Discussion
6.1 Compression of Text
6.2 Maximum Supported Vocabulary Size
6.3 Dealing with Unseen Text
7 Conclusion
References
ED-NET: Educational Teaching Video Classification Network
1 Introduction
2 Related Work
3 Dataset Development and Description
4 Methodology
4.1 Problem Formulation
4.2 Model Architectures
5 Experimental Setup
5.1 Implementation Details
5.2 Evaluation Metrics
6 Result and Discussion
7 Conclusion
References
Detection of COVID-19 Using Machine Learning
1 Introduction
2 Related Work
3 Method
3.1 Dataset
3.2 K-Nearest Neighbors Algorithm
3.3 Support Vector Machine
3.4 Decision Trees
3.5 XGBoost
4 Proposed Model
5 Results and Discussions
5.1 Performance Metrics
5.2 Performance Evaluation
5.3 Linear Regression
6 Conclusion
References
A Comparison of Model Confidence Metrics on Visual Manufacturing Quality Data
1 Introduction
2 Related Work
2.1 Plain Network Output
2.2 Intermediate Layer Distributions
2.3 Distribution of Input Data
3 Experimental Setup
3.1 Instantaneous Change Scenario
3.2 Continuous Drift Scenario
4 Selected Methods
5 Results and Discussion
6 Conclusion and Further Research
References
High-Speed HDR Video Reconstruction from Hybrid Intensity Frames and Events
1 Introduction
2 Related Works
2.1 HDR Image-Based Reconstruction
2.2 HDR Video Reconstruction
2.3 Event-Based Interpolation
3 Hybrid Event-HDR
4 Experiments, Datasets, and Simulation
5 Results
6 Conclusion
References
Diagnosis of COVID-19 Using Deep Learning Augmented with Contour Detection on X-rays
1 Introduction
1.1 Preprocessing Algorithms
1.2 Convolution Neural Network
2 Literature Survey
3 Methodology
4 Results
5 Conclusion
References
A Review: The Study and Analysis of Neural Style Transfer in Image
1 Introduction
1.1 Motivation
1.2 Key Challenges
2 The Literature Survey
2.1 Simple Style Transfer
2.2 Neural Style Transfer
3 Experimental Results
4 Research Gaps
5 Challenges and Issues
6 Conclusion and Future Scope
References
A Black-Box Attack on Optical Character Recognition Systems
1 Introduction
1.1 Related Works
2 Problem Definition
2.1 Adversarial Example
3 Proposed Method
3.1 Additive Perturbations
3.2 Erosive Perturbations
3.3 ECoBA: Efficient Combinatorial Black-Box Adversarial Attack
4 Simulations
4.1 Data Sets
4.2 Models
4.3 Results
5 Conclusion
References
Segmentation of Bone Tissue from CT Images
1 Introduction
2 Literature Review
3 Dataset
4 Proposed Methodology
4.1 Contrast Stretching
4.2 Region Growing
4.3 Outlier Removal
5 Performance Metrics
6 Result and Inference
7 Conclusion
8 Limitations and Future Work
References
Fusion of Features Extracted from Transfer Learning and Handcrafted Methods to Enhance Skin Cancer Classification Performance
1 Introduction
2 Literature
3 Proposed Methodology
3.1 Preprocessing
3.2 Feature Extraction Techniques
3.3 Classification Technique
4 Experimental Setup and Result Analysis
4.1 Evaluation Metrics
4.2 Dataset Description
4.3 Experimental Results and Analysis
5 Conclusion
References
Investigation of Feature Importance for Blood Pressure Estimation Using Photoplethysmogram
1 Introduction
2 Database
3 Methodology
3.1 Preprocessing
3.2 Feature Extraction
3.3 Blood Pressure Estimation Using Regression Models
3.4 Feature Analysis
4 Result
5 Conclusion
References
Low-Cost Hardware-Accelerated Vision-Based Depth Perception for Real-Time Applications
1 Introduction
2 Related Work
2.1 Feature Engineering Methods
2.2 Learned Approach (Neural Networks)
3 System Design
3.1 Point Cloud Generation
3.2 3D Object Tracking Using Bayesian Inference
4 Dataset
5 Experimental Evaluation
5.1 Disparity Generation
5.2 Formula Student Driverless Simulator
5.3 Bayesian Prediction Error
6 Real-World Testing
7 Conclusion and Future Work
References
Development of an Automated Algorithm to Quantify Optic Nerve Diameter Using Ultrasound Measures: Implications for Optic Neuropathies
1 Introduction
2 Methods and Materials
2.1 Image Pre-processing and Retina Globe Detection
2.2 Optic Nerve Localization and Segmentation
2.3 Optic Nerve Diameter Measurement
3 Result
4 Discussion
5 Conclusion
References
APFNet: Attention Pyramidal Fusion Network for Semantic Segmentation
1 Introduction
2 Literature Review
3 The Proposed Network
3.1 Encoder Network
3.2 Decoder Network
4 Fusion Architectures
4.1 Fusion Architecture 1 (Summation)
4.2 Fusion architecture 2 (concatenation)
4.3 Fusion architecture 3 (concatenation at AASPP)
5 Experimental Results and Observations
5.1 Dataset
5.2 Experimental Setup
5.3 Evaluation Metrics
5.4 Overall Experimental Results
5.5 Experimental Analysis of Under Different Illuminations
5.6 Adaptation Study
6 Conclusions and Future Directions
References
Unsupervised Virtual Drift Detection Method in Streaming Environment
1 Introduction
2 Related Work
3 Proposed Work
3.1 Proposed Drift Detection Method
4 Evaluation and Results
4.1 Case Study on Iris Dataset
4.2 Datasets
4.3 Experimental Results and Analyses
5 Conclusion
References
Balanced Sampling-Based Active Learning for Object Detection
1 Introduction
2 Related Works
3 Method
3.1 Problem Definition
3.2 CALD
3.3 Proposed Method
4 Experiments
4.1 Dataset Used
4.2 Metrics Used
4.3 Model
4.4 Results
5 Conclusion
References
Multi-scale Contrastive Learning for Image Colorization
1 Introduction
2 Related Work
2.1 Generative Adversarial Network
2.2 Pix2Pix
2.3 CycleGAN
2.4 Contrastive Learning
3 Proposed Multi-scale Contrastive Learning Technique
4 Experimental Setup
4.1 Models Used
4.2 Datasets Used
4.3 Metrics Used
5 Experimental Results and Analysis
6 Conclusion
References
Human Activity Recognition Using CTAL Model
1 Introduction
2 Related Work
3 Database and Model Structure
3.1 UCF50 Dataset
3.2 Model Architecture
4 Environment and Experiment
4.1 Experimental Environment
4.2 Experiment
5 Result and Comparison
6 Result Discussion and Comparison to SoTA Method
7 Conclusion
References
Deep Learning Sequence Models for Forecasting COVID-19 Spread and Vaccinations
1 Introduction
2 Literature Survey
2.1 Research Gap and Motivation
3 Proposed Approach, Details of Implementation and Methodology
4 Model Evaluation Metrics
5 Dataset Details
5.1 Daily Total Confirmed Cases
5.2 Daily Positive Tests
5.3 Total Individuals Vaccinated
6 Results
6.1 Prediction of Daily Total Confirmed Cases
6.2 Prediction of Daily Positive Tests
6.3 Prediction of Total Individuals Vaccinated
7 Discussion
8 Conclusion
References
Yoga Pose Rectification Using Mediapipe and Catboost Classifier
1 Introduction
2 Related Works
3 Methodology
3.1 Pre-processing Part of the System
3.2 Key Point Extraction and Angle Calculation
3.3 Training Models for Yoga Identification
3.4 Identifying Improper Body Part
4 Type of System
4.1 Static System
4.2 Dynamic/Real-Time System
5 Results
6 Conclusion
7 Future Work
References
A Machine Learning Approach for PM2.5 Estimation for the Capital City of New Delhi Using Multispectral LANDSAT-8 Satellite Observations
1 Introduction
2 Related Works
3 LANDSAT-8 Data Description
4 Study Area
5 Problem Formulation and Methodology
6 Results and Discussion
7 Conclusion and Future Work
References
Motion Prior-Based Dual Markov Decision Processes for Multi-airplane Tracking
1 Introduction
2 Related Work
2.1 MAT
2.2 Motion Modeling
3 Proposed Method
3.1 Problem Formulation
3.2 msMDP
3.3 tsMDP
4 Experiments
4.1 Implementation Details
4.2 Dataset
4.3 Benchmark Evaluation
4.4 Ablation Study
5 Conclusion
References
URL Classification on Extracted Feature Using Deep Learning
1 Introduction
2 Related Work
2.1 Malicious URL Classification Based on Signature
2.2 Machine Learning-Based URL Classification
3 Data Preparation
3.1 Dataset Used
3.2 Feature Extraction
3.3 Data Exploration
3.4 Over-Sampling
3.5 Train-Test Split and Hardware Used
4 Technology Used
5 Experimental Results
6 Conclusion and Future Work
References
Semi-supervised Semantic Segmentation for Effusion Cytology Images
1 Introduction
2 Semi-supervised Learning
3 SSL for Semantic Segmentation
3.1 Pseudo-Label Generation
3.2 Network Training
4 Experiments and Results
4.1 Data Set
4.2 Metrics
4.3 Baseline
4.4 Semi-supervised
4.5 Classification
5 Conclusion
References
Image Augmentation Strategies to Train GANs with Limited Data
1 Introduction
2 Literature Survey
3 Methodology
3.1 Architecture
3.2 Dataset
3.3 Data Augmentation
3.4 Generative Adversarial Networks Model
4 Results and Analysis
5 Conclusion and Future Work
References
Role of Deep Learning in Tumor Malignancy Identification and Classification
1 Introduction
2 Literature Review
3 Discussion
4 Conclusion
References
Local DCT-Based Deep Learning Architecture for Image Forgery Detection
1 Introduction
2 Review of Literature
3 Proposed Method
4 Algorithm
5 Experiments
5.1 Dataset
5.2 Performance Analysis
6 Conclusion
References
Active Domain-Invariant Self-localization Using Ego-Centric and World-Centric Maps
1 Introduction
2 Related Work
3 Approach
3.1 VPR Model
3.2 CNN Output-Layer Cue
3.3 CNN Intermediate Layer Cue
3.4 Reciprocal Rank Transfer
3.5 Training NBV Planner
4 Experiments
4.1 Settings
4.2 Results
5 Conclusions
References
Video Anomaly Detection for Pedestrian Surveillance
1 Introduction
2 Related Work
3 Proposed Methodology
3.1 Data Collection
3.2 Pre-processing
3.3 Feature Addition
3.4 Training
3.5 Testing
4 Results
5 Conclusion and Future Scope
References
Cough Sound Analysis for the Evidence of Covid-19
1 Introduction
2 Related Works
3 Method
3.1 Dataset Collection
3.2 Feature Extraction and Selection
3.3 CNN Architecture
4 Experiments
4.1 Validation
4.2 Our Results
4.3 Dropout
4.4 Comparative Analysis
5 Discussion
6 Conclusion and Future Work
References
Keypoint-Based Detection and Region Growing-Based Localization of Copy-Move Forgery in Digital Images
1 Introduction
2 Related Works
3 Working and Implementation of the Proposed Approach
3.1 SIFT Keypoint Extraction and DBSCAN Clustering
3.2 Building NtimesN Blocks Around Extracted Keypoints
3.3 Calculating the Distance Between Each Pair of Blocks Within the Cluster
3.4 Region Growing on the Previously Detected Blocks
4 Experiment Results
5 Conclusion
References
Auxiliary Label Embedding for Multi-label Learning with Missing Labels
1 Introduction
2 Related Work
3 Problem Formulation and Algorithm Proposed
3.1 Learnt Label Correlations Embedding
3.2 Instance Similarity
3.3 Optimization
4 Experiments
4.1 Datasets
4.2 Multi-label Metrics
4.3 Baselines
4.4 Empirical Results and Discussion
5 Conclusion
References
Semi-supervised Semantic Segmentation of Effusion Cytology Images Using Adversarial Training
1 Introduction
2 Adversarial Network-Based Semi-supervised Image Segmentation Methodology
2.1 Network Architecture
2.2 Loss Function
3 Experiments and Result
3.1 Dataset
3.2 Experiment 1
3.3 Experiment 2
4 Conclusion
References
RNCE: A New Image Segmentation Approach
1 Introduction
2 Literature Survey
3 Proposed Architecture
3.1 Data Acquisition
3.2 Model Selection
3.3 Proposed Model
3.4 Loss Function and Training
4 Experimental Results and Comparisons
4.1 Accuracy
4.2 Mean Intersection-Over-Union
5 Conclusion
References
Cross-Media Topic Detection: Approaches, Challenges, and Applications
1 Introduction
2 Topic Detection Approaches
2.1 Graph-Based Methods
2.2 Machine Learning-Based Methods
2.3 Deep Learning-Based Methods
3 Challenges in Cross-media Topic Detection
4 Topic Detection Applications
5 Conclusion
References
Water Salinity Assessment Using Remotely Sensed Images—A Comprehensive Survey
1 Introduction
2 Data Sources
3 Water Salinity Prediction Approaches
3.1 Empirical Methods
3.2 Regression Analysis for Salinity Prediction
3.3 Machine Learning-Based Approaches
3.4 Deep Learning Approaches
4 Discussion
5 Concluding Remarks
References
Domain Adaptation: A Survey
1 Introduction
2 Datasets Used for Domain Adaptation
2.1 Office 31
2.2 Caltech
2.3 Office Home
2.4 MNIST and MNIST-M
3 Methods of Deep Domain Adaptation
3.1 Discrepancy Based
3.2 Adversarial Based
3.3 Reconstruction Based
3.4 Combination Based
3.5 Transformation Based
4 Conclusions
5 Future Works
References
Multi-branch Deep Neural Model for Natural Language-Based Vehicle Retrieval
1 Introduction
2 Related Work
3 Proposed Methodology
4 Experimental Results and Discussion
5 Conclusion and Future Work
References
Kullback–Leibler Distance-Based Fuzzy K-Plane Clustering Approach for Noisy Human Brain MRI Image Segmentation
1 Introduction
2 Preliminaries and Related Work
2.1 Concept of Relative Entropy (a.k.a. Kullback–Leibler Distance) for Fuzzy Sets
2.2 kPC Method
3 Proposed Kullback–Leibler Distance-Based Fuzzy K-Plane Clustering Approach
4 Dataset and Experimental Results
4.1 Datasets
4.2 Performance Metrics
4.3 Results on BrainWeb: Simulated Brain MRI Database
4.4 Results on IBSR: Real Clinical Brain MRI Database
4.5 Results on MRBrainS18: Real Brain Challenge MRI Database
5 Conclusion and Future Direction
References
Performance Comparison of HC-SR04 Ultrasonic Sensor and TF-Luna LIDAR for Obstacle Detection
1 Introduction
2 Sensor Characteristic and Technical Specification
3 Experimental Setup
3.1 Interfacing of Ultrasonic Sensor with Arduino UNO
3.2 Interfacing TF-Luna with Arduino UNO
4 Results and Discussion
5 Conclusion
References
Infrared and Visible Image Fusion Using Morphological Reconstruction Filters and Refined Toggle-Contrast Edge Features
1 Introduction
2 Preliminaries
2.1 Grayscale Morphological Reconstruction Filters
2.2 Toggle-Contrast Filter
3 Proposed Fusion Scheme
3.1 Feature Extraction Using Open (Close) and Toggle-Contrast Filters
3.2 Feature Comparison
3.3 Feature Cumulation and Refinement
3.4 Final Fusion
4 Experimental Analysis and Discussion
4.1 Subjective Evaluation
5 Conclusion
References
Extractive Text Summarization Using Statistical Approach
1 Introduction
2 Literature Survey
3 Proposed Work
3.1 Dataset
4 Experiments and Results
4.1 Range of LFW and GFW Used
4.2 Length of Generated Summary
4.3 Comparison with State-of-the-art Literature
5 Conclusion and Future Work
References
Near-Infrared Hyperspectral Imaging in Tandem with Machine Learning Techniques to Identify the Near Geographical Origins of Barley Seeds
1 Introduction
2 Material and Methods
2.1 Barley Samples
2.2 Hyperspectral Image Acquisition and Calibration
2.3 Image Processing and Extraction of Sample-Wise Spectra
2.4 Classification Models’ Development and Validation
3 Results and Discussion
4 Conclusions
References
MultiNet: A Multimodal Approach for Biometric Verification
1 Introduction
1.1 Biometric Review
1.2 Multimodal Biometric System
2 Literature Review
3 Dataset
3.1 Fingerprint Dataset
3.2 Iris Dataset
4 Methodology
4.1 Network Architecture
4.2 Data Preprocessing
4.3 Fusion Approach
5 Experimental Setup
6 Results and Analysis
7 Conclusion
References
Synthesis of Human-Inspired Intelligent Fonts Using Conditional-DCGAN
1 Introduction
1.1 Handwritten Fonts
1.2 Random Fonts
1.3 Intelligent Fonts
1.4 Motivation
1.5 Our Contribution
2 Literature Review
3 Dataset
3.1 About Dataset
3.2 Dataset Collection Process
4 Preliminary Knowledge
4.1 Generative Adversarial Network (GAN)
4.2 Conditional GAN (cGAN)
4.3 Deep Convolutional GAN (DCGAN)
5 Proposed Methodology
5.1 Random Font-cDCGAN (R-cDCGAN)
5.2 Intelligent Font-cDCGAN (I-cDCGAN)
5.3 Sentence Formation from Generated Alphabets
6 Results and Analysis
6.1 Model Training and Parameters of R-cDCGAN and I-cDCGAN
6.2 Results
7 Performance Evaluation
7.1 Fréchet Inception Distance (FID)
7.2 Between Class FID (BCFID)
7.3 Within Class FID (WCFID)
7.4 Experiments
8 Conclusion and Future Work
References
Analysis and Application of Multispectral Data for Water Segmentation Using Machine Learning
1 Introduction
2 Satellite Data and Study Site
3 Methodology Used
3.1 Data Processing
3.2 Band Reflectance Analysis
3.3 BandNet
3.4 Multispectral Image Analysis
4 Implementation Details
5 Results and Discussion
6 Conclusion
References
MangoYOLO5: A Fast and Compact YOLOv5 Model for Mango Detection
1 Introduction
2 Literature Review
3 Methodology
3.1 Data Pre-processing
3.2 MangoYOLO5 Network Architecture
4 Result Discussion and Experimental Analysis
4.1 Analysis Using Performance Evaluation Metrics
5 Conclusion and Future Scope
References
Resolution Invariant Face Recognition
1 Introduction
2 Related Works
2.1 Super-Resolution Techniques
2.2 Resolution Invariant Techniques
3 Methodology and Workflow
3.1 Data Pre-processing
3.2 Network Architecture
4 Experiments and Result
4.1 Dataset
4.2 Implementation Details
4.3 Results
5 Conclusion
References
Target Detection Using Transformer: A Study Using DETR
1 Introduction
2 Related Work
2.1 The Transformer Model
2.2 The DETR Model [11]
3 Target Detection with Transformer
3.1 Dataset
3.2 Customizations
3.3 Training
4 Results
5 Conclusion
References
Document Image Binarization in JPEG Compressed Domain Using Dual Discriminator Generative Adversarial Networks
1 Introduction
2 Related Literature
3 Proposed Model
3.1 Image Pre-processing
3.2 Network Architecture
3.3 Total GAN Loss
4 Experiment and Results
4.1 DIBCO Dataset
4.2 Results
5 Conclusion
References
Author Index