Pattern Recognition and Artificial Intelligence

Name: Pattern Recognition and Artificial Intelligence | Third International Conference, ICPRAI 2022, Paris, France, June 1-3, 2022, Proceedings, Part I
Brand: Springer
Price: 160.49 EUR
Availability: OnlineOnly

Third International Conference, ICPRAI 2022, Paris, France, June 1-3, 2022, Proceedings, Part I

Mounîm El Yacoubi Eric Granger Pong Chi Yuen Umapada Pal Nicole Vincent(Editor)

Springer (Publisher)

Published on 1. June 2022

XXIII, 704 pages

E-Book

PDF with digital watermarking

System requirements

978-3-031-09037-0 (ISBN)

€160.49incl. 7% vat

System requirements

for PDF with digital watermarking

E-Book Single Licence

Available for download

Description

More details

Other editions

Content

Intro
Preface
Organization
Contents - Part I
Contents - Part II
Computer Vision
Identifying, Evaluating, and Addressing Nondeterminism in Mask R-CNNs
1 Introduction
1.1 Motivation
1.2 Terminology
1.3 Contributions
2 Nondeterminism Introduced in Training
2.1 Random Number Generators
2.2 Model Structure
2.3 CUDA Algorithms and PyTorch
3 Nondeterminism Introduced by Hardware
3.1 Floating-Point Operations
3.2 Atomic Operations
3.3 Parallel Structure
4 Experimental Setup
4.1 System Architecture
4.2 Measuring Performance
4.3 Model Training Process
4.4 Configuring Settings
5 Experimental Results
5.1 Data Collected from Models Trained on GPU
5.2 Data Collected from Models Trained on CPU
6 Discussion of Results
6.1 Impact of Embedded Randomness on Model Precision
6.2 Increase in Training Time After Configuring Randomness
6.3 Impact of Seed Sensitivity
6.4 Conclusion
References
Structure-Aware Photorealistic Style Transfer Using Ghost Bottlenecks
1 Introduction
2 Related Work
3 Proposed Method
3.1 Workflow
3.2 Architecture
3.3 Loss Function
4 Experiments
4.1 Qualitative Results
4.2 Quantitative Results
5 Conclusion
References
Analysing the Impact of Vibrations on Smart Wheelchair Systems and Users
1 Introduction
2 Methodology
3 Results
4 Conclusion
References
Visual Microfossil Identification via Deep Metric Learning
1 Introduction
2 Background
3 Dataset
4 Experimental Setup
5 Results
6 Taxonomic Reflection
7 Conclusion
References
QAP Optimisation with Reinforcement Learning for Faster Graph Matching in Sequential Semantic Image Analysis
1 Introduction
2 Reinforcement Learning for Sequential Graph Matching
2.1 Neural Network and Graphs
2.2 Sequential One-to-one Matching by Reinforcement Learning
2.3 Sequential Refinement: Many-to-one-or-none Matching
3 Experiments
3.1 Datasets
3.2 Evaluation Protocol
3.3 Results
4 Conclusion
References
Towards a Unified Benchmark for Monocular Radial Distortion Correction and the Importance of Testing on Real-World Data
1 Introduction
2 Related Work
3 Data Generation
3.1 Lens Distortion Models
3.2 Preprocessing Strategy
3.3 Synthetically Distorted Datasets
4 Experiments
4.1 Implementation Details
4.2 Radial Distortion Estimation
4.3 Influence of Pre-trained Weights
4.4 Explainability of the Results
5 Conclusion
References
Comparing Artificial Intelligence Algorithms in Computer Vision: The Weapon Detection Benchmark
1 Introduction
2 State of Art
3 Datasets
4 Methods
5 Results
6 Discussion
7 Conclusion
References
Metrics for Saliency Map Evaluation of Deep Learning Explanation Methods
1 Introduction
2 Existing Metrics
2.1 DAUC and IAUC
2.2 Limitations
3 Score Aware Metrics
3.1 The Sparsity Metric
3.2 The DC and IC Metrics
3.3 Limitations
4 Benchmark
5 Discussion
6 Conclusion
References
Image Classification via Multi-branch Position Attention Network
1 Introduction
2 The Proposed Method
2.1 Position Channel Attention Module (PCAM)
2.2 Local Spatial Attention Module (LSAM)
2.3 Integrating Attention Modules into Backbone Networks
3 Experiments
3.1 Implementation Details
3.2 Compared to State-of-the-Art Methods
3.3 Reports of Confusion Matrix
3.4 Ablation Study
4 Conclusion
References
UGQE: Uncertainty Guided Query Expansion
1 Introduction
2 Related Work
3 Method
4 Experiments
5 Conclusion
References
Controlling the Quality of GAN-Based Generated Images for Predictions Tasks
1 Introduction and Related Work
2 Background
3 Materials and Methods
4 Results
5 Conclusion
References
A Framework for Registration of Multi-modal Spatial Transcriptomics Data
1 Introduction
2 Methodology
2.1 Generation of Expression Image
2.2 Framework for Registration
3 Experimental Results
4 Conclusions
References
Extracting and Classifying Salient Fields of View from Microscopy Slides of Tuberculosis Bacteria
1 Introduction
1.1 Importance of Microscopy
2 Related Work
3 Proposed Method
3.1 Discrimination Enhanced Representation Extraction
3.2 The Learning
4 Evaluation
4.1 Data
4.2 Competitor Models
4.3 Hyper-parameter Learning
4.4 Results and Discussion
5 Conclusion
References
Segmentation
Deep Learning for Fast Segmentation of E-waste Devices' Inner Parts in a Recycling Scenario
1 Introduction
2 Related Work
3 Methods
3.1 Dataset
3.2 Precision
3.3 Inference Time
3.4 Instance Segmentation
3.5 Tracking
3.6 Training
4 Results
4.1 Precision and Speed
4.2 Tracking
5 Conclusion
References
Improving Semantic Segmentation with Graph-Based Structural Knowledge
1 Introduction
2 Proposed Method
2.1 Graph Definitions
2.2 Graph Matching
2.3 Modelling Spatial Relationships
3 Application to Segmentation of 3D MRI
4 Conclusion
References
Application of Rail Segmentation in the Monitoring of Autonomous Train's Frontal Environment
1 Introduction
2 Related Works
2.1 Dataset
2.2 Segmentation
3 Rail Segmentation
3.1 RailSet
3.2 Experiments
4 Conclusion
References
DR-VNet: Retinal Vessel Segmentation via Dense Residual UNet
1 Introduction
2 Related Works
3 Proposed Method
4 Experiments
5 Conclusion
References
.26em plus .1em minus .1emDealing with Incomplete Land-Cover Database Annotations Applied to Satellite Image Time Series Semantic Segmentation
1 Introduction
2 Proposed Method
2.1 Deep-STaR Principle ch18ChelaliKPV21
2.2 From Classification to Segmentation
2.3 Pixel Classification Improvement
3 Experimental Study
3.1 Data
3.2 Parameter Values
3.3 Evaluation
4 Conclusion
References
On the Feasibility and Generality of Patch-Based Adversarial Attacks on Semantic Segmentation Problems
1 Introduction
2 Adversarial Attacks
3 Patch-Based Segmentation Attacks on a Simple Dataset
3.1 Real-Life Images of Simple Shapes
4 Patch-Based Segmentation Attacks on Cityscapes
5 A Complexity Analysis of Patch Based Attacks
5.1 Upper Bound of Linear Regions
6 Conclusion
References
Unsupervised Cell Segmentation in Fluorescence Microscopy Images via Self-supervised Learning
1 Introduction
2 Method
3 Experimental Results
4 Conclusion
References
Document
Hypercomplex Generative Adversarial Networks for Lightweight Semantic Labeling
1 Introduction and Related Work
2 Proposed Model
2.1 Hypercomplex Layers
2.2 Model Architecture
3 Experiments
4 Conclusion and Future Work
References
Learning Document Graphs with Attention for Image Manipulation Detection
1 Introduction
2 Related Work
3 Methods
4 Experiments
References
Modular StoryGAN with Background and Theme Awareness for Story Visualization
1 Introduction
2 Related Works
3 Modular StoryGAN
3.1 Ideas for Background and Theme Awareness
3.2 Proposed Design Architecture
3.3 The Objective Function for Learning
4 Evaluation
4.1 PororoSV Dataset
4.2 Quantitative Evaluation Metrics
4.3 User Assessment: Background and Theme Awareness
5 Results
6 Conclusion
References
Is On-Line Handwriting Gender-Sensitive? What Tells us a Combination of Statistical and Machine Learning Approaches
1 Introduction
2 Data Collection and Extracted Features
2.1 Data Collection
2.2 Extracted Features
2.3 Feature Selection
3 Experiments
4 Conclusion
References
Using Convolutional Neural Network to Handle Word Shape Similarities in Handwritten Cursive Arabic Scripts of Pashto Language
1 Introduction
2 Background
3 Procedure
3.1 Experimental Setup
3.2 Data Base
4 Analysis
5 Conclusion
References
Discriminatory Expressions to Improve Model Comprehensibility in Short Documents
1 Introduction
2 Related Work
2.1 Dimensionality Reduction
2.2 Interpretability Measures
3 Discriminatory Expressions
3.1 Ranking Method: CF-ICF
4 Experiments
5 Results and Discussion
5.1 Classification Performance
5.2 Generalisation Capacity
5.3 Comprehensibility
6 Conclusions
References
Scene Text Recognition: An Overview
1 Introduction
2 Scene Text Detection Based on Deep Learning
2.1 Regression-Based Scene Text Detection
2.2 Segmentation-Based Scene Text Detection
2.3 Hybrid-Based Scene Text Detection
3 Scene Text Recognition Based on Deep Learning
3.1 CTC-Based Scene Text Recognition
3.2 Attention-Based Scene Text Recognition
4 End-to-End Scene Text Spotting
5 Future Direction
References
An Encoder-Decoder Approach to Offline Handwritten Mathematical Expression Recognition with Residual Attention
1 Introduction
1.1 Handwritten Mathematical Expression Recognition
1.2 Our Main Work
2 Related Studies
2.1 HMER Method
2.2 Attention
3 Methodology
3.1 DenseNet Encoder
3.2 Residual Attention Modules
3.3 DATWAP Architecture
3.4 Label Smoothing
4 Experiments
4.1 Experimental Setup
4.2 Ablation Experiment
4.3 Comparison with Other HMER Systems and Case Study
5 Conclusion
References
Hologram Detection for Identity Document Authentication
1 Introduction
2 Single Image Analysis
2.1 Pixel Level
2.2 Local Level
3 Video Analysis
3.1 Two Images Analysis
3.2 Hologram Extraction
4 Experimental Study
5 Conclusion
References
A New Deep Fuzzy Based MSER Model for Multiple Document Images Classification
1 Introduction
2 Related Work
3 The Proposed Method
3.1 Candidate Component Detection
3.2 Distance Feature Extraction from Candidate Components
3.3 Deep CNN for Classification of Multiple Documents
4 Experimental Results
4.1 Ablation Study
4.2 Experiments on Classification of Multiple Documents
4.3 Experiments for Validating Classification Method
5 Conclusion and Future Work
References
Video - 3D
Multi-view Monocular Depth and Uncertainty Prediction with Deep SfM in Dynamic Environments
1 Introduction
2 Related Work
3 Approach
3.1 Networks
4 Training
5 Experiments
6 Conclusion
References
Fourier Domain CT Reconstruction with Complex Valued Neural Networks
1 Introduction
2 Projections and the Fourier Slice Theorem
3 Problem Statement
4 Reconstruction with Complex Valued Neural Networks
4.1 The Complex Valued Regression Model
4.2 The Training Phase
5 Evaluation and Results
5.1 Training Data and Hyperparameters
5.2 Results
6 Conclusion
References
Seeking Attention: Using Full Context Transformers for Better Disparity Estimation
1 Introduction
2 Related Work
2.1 Vision Transformers
2.2 Stereo Disparity Estimation
3 Methods
3.1 Feature Extraction
3.2 Full Context Stereo Transformers
4 Experiments and Results
4.1 Datasets and Metrics
4.2 Training
4.3 Experiments
4.4 Results and Comparison
5 Conclusion
References
Inpainting Applied to Facade Images: A Comparison of Algorithms
1 Introduction
2 Related Work
3 Algorithms
3.1 Local, Diffusion-Based Inpainting Algorithms
3.2 Global Inpainting Algorithms Not Relying on Deep Learning
3.3 Deep Learning-Based Global Inpainting Algorithms
4 Ground Truth, Training Data, and Network Training
5 Results
6 Conclusions
References
Self-distilled Self-supervised Depth Estimation in Monocular Videos
1 Introduction
2 Related Work
3 Depth Estimation Method in Monocular Videos
3.1 Preliminaries
3.2 Self-distillation via Prediction Consistency
3.3 Filtering Pseudo-labels
3.4 Consistency Enforcement Strategies
3.5 Additional Considerations
4 Evaluation
4.1 Experimental Setup
4.2 Self-distillation via Prediction Consistency
4.3 Filtering Pseudo-labels
4.4 Consistency Enforcement Strategies
4.5 State-of-the-Art Comparison
5 Conclusions
References
Improving UWB Image Reconstruction for Breast Cancer Diagnosis by Doing an Iterative Analysis of Radar Signals
1 Introduction
1.1 Breast Cancer Diagnosis
2 Breast Cancer Detection by UWB Imaging
3 Iterative Method Adopting Form of MLEM (Maximum-Likelihood Expectation-Maximization) Algorithm
4 Experiments and Simulations
4.1 Calibration
4.2 Results of DAS, DMAS. Iterative DAS and Iterative DMAS
5 Conclusions
References
Personalized Frame-Level Facial Expression Recognition in Video
1 Introduction
2 Methodology
2.1 Facial Expression Recognition in Videos
2.2 Classification of the Frame-Level Features
2.3 Proposed Approach
3 Experimental Study
3.1 Dataset
3.2 Training Details
3.3 Results
4 Conclusion
References
3D Reconstruction of Medical Image Based on Improved Ray Casting Algorithm
1 Introduction
2 Method
2.1 Resampling Interpolation Algorithm
2.2 The Bounding Box in 3D Reconstruction
2.3 Data Synthesis of Sampling Points
2.4 The Fusion Algorithm
3 Result
3.1 Experimental Equipment and Environment
3.2 Data
3.3 Objective Evaluation Method
3.4 Experimental Results and Analysis
3.5 Summary
References
Lateral Ego-Vehicle Control Without Supervision Using Point Clouds
1 Introduction and Related Work
2 Method
2.1 Point Cloud Generation
2.2 Camera Pose Estimation
2.3 Point Cloud Augmentation
2.4 Label Generation
2.5 Model Architecture
3 Experiments
4 Discussion
5 Conclusion
References
RDMMLND: A New Robust Deep Model for Multiple License Plate Number Detection in Video
1 Introduction
2 Related Work
3 The Proposed Method
3.1 Vehicle Detection
3.2 License Plate Number Region Enhancement
3.3 Multiple License Plate Number Detection
4 Experimental Results
4.1 Ablation Study
4.2 Experiments on Multiple License Plate Number Detection
4.3 Experiments on Natural Scene Text Detection
5 Conclusion and Future Work
References
Generative Target Update for Adaptive Siamese Tracking
1 Introduction
2 Related Work
3 Proposed Adaptive Siamese Tracker
4 Results and Discussion
5 Conclusion
References
From Synthetic to One-Shot Regression of Camera-Agnostic Human Performances
1 Introduction
2 Related Work
3 Synthetic Training Data Generation
4 Model
5 Experiments
6 Conclusion
References
Feature
Computation of 2D Discrete Geometric Moments Through Inclusion-Exclusion
1 Introduction
2 Background Notions
2.1 Geometric (Cartesian) Moments
2.2 Green's Theorem
3 Related Work
3.1 Decomposition-Based Algorithms
3.2 Boundary-Based Algorithms
4 Our Formula
4.1 The Proposed Formula
4.2 Proof of Correctness
5 Experimental Validation
5.1 Comparison with Other Boundary-Based Formulas
5.2 Comparison with the Decomposition-Based Approach
6 Summary and Future Work
References
Feature Subset Selection for Detecting Fatigue in Runners Using Time Series Sensor Data
1 Introduction
2 Predicting Fatigue in Runners
2.1 Dataset
2.2 Personalised Vs Global Classification
3 Feature Selection for Sensor Data
3.1 Correlation Based Feature Subset Selection for Time Series
4 Evaluation
4.1 MSTS for Feature Subset Selection on Individuals
4.2 Generic Model Selection
5 Conclusions and Future Work
References
Compositing Foreground and Background Using Variational Autoencoders
1 Introduction
2 Related Work
3 Model
3.1 VAE
3.2 BFVAE
3.3 LSR-GAN
4 Experiments
4.1 BFVAE with LSR-GAN
4.2 Substitute Attributes for Foreground
5 Discussion
References
Unsupervised Representation Learning via Information Compression
1 Introduction
2 Related Work
3 Model
3.1 Glimpsing Network
3.2 Model Architecture and Training
4 Experiments
4.1 Quantitative Comparison
4.2 Downstream Task
4.3 Generalization
4.4 Ablation Study
4.5 CLEVR and Multi-dSprites
5 Conclusion
References
Remote Sensing Scene Classification Based on Covariance Pooling of Multi-layer CNN Features Guided by Saliency Maps
1 Introduction
2 Related Works
3 Proposed Method: ELCP Guided by Saliency Maps (EL-SCP)
3.1 Saliency Map Generation
3.2 Weighted Covariance Matrix Estimator
4 Experiments
5 Conclusion
References
Attention Embedding ResNet for Pest Classification
1 Introduction
1.1 The Challenge of Pest Classification
1.2 Researches on Classification for Pest
1.3 Pest Recognition Datasets
1.4 Our Work
2 Related Studies
2.1 ResNet
2.2 Attention Mechanism
3 Methodology
3.1 ECA: Efficient Channel Attention
3.2 CCO: Coordinate Attention
4 Experiment
4.1 Experimental Settings
4.2 Architecture
4.3 Comparison with Other Methods
4.4 Ablation Experiment
4.5 Visualization of Experiment
5 Conclusion
References
Multi Layered Feature Explanation Method for Convolutional Neural Networks
1 Introduction and Related Work
2 Multi Layered Feature Explanation Method
2.1 Reminder of Feature Explanation Method (FEM)
2.2 Principles of Multi Layered FEM (MLFEM)
2.3 Fusion Operators
2.4 Implementation of MLFEM on CNN Classifiers
3 Evaluation of MLFEM Explanations
4 Experiments and Results
4.1 Datasets
4.2 Results
5 Conclusion
References
Bayesian Gate Mechanism for Multi-task Scale Learning
1 Introduction
2 Related Work
2.1 Multi-task Learning
2.2 Bayesian Deep Learning
3 Method
3.1 Overview
3.2 Multiple Scale-Task Decoder
3.3 Bayesian Knowledge Gating Module
4 Experiment
4.1 Experimental Setup
4.2 Network Performance
4.3 Ablation Study
5 Conclusion
References
ASRSNet: Automatic Salient Region Selection Network for Few-Shot Fine-Grained Image Classification
1 Introduction
2 Related Work
2.1 Fine-Grained Image Classification
2.2 Meta-learning Based FSFGIC Methods
2.3 Metric-learning Based FSFGIC Methods
3 Methodology
3.1 Problem Definition
3.2 Automatic Salient Regions Selection for FSFGIC
4 Experiment
4.1 Datasets
4.2 Experimental Setup
4.3 Performance Comparison
5 Conclusion
References
Face Age Progression with Attribute Manipulation
1 Introduction
2 Related Works
2.1 Face Aging
2.2 Facial Attribute Manipulation
3 Proposed Architecture for Face Age Progression with Attribute Manipulation
3.1 Overview of Our Proposed FAWAM Model
3.2 Face Aging Module
3.3 Attribute Manipulation Module
4 Experimental Results
4.1 Datasets
4.2 Implementation Details
4.3 Face Aging Module Results
4.4 Attribute Manipulation Module Results
4.5 FAWAM Results
5 Conclusions
References
Random Dilated Shapelet Transform: A New Approach for Time Series Shapelets
1 Introduction
2 Background
2.1 Time Series Classification
2.2 Shapelets
3 Proposed Method
3.1 Dilated Shapelets
3.2 Shapelet Occurrence Feature
3.3 Random Dilated Shapelet Transform (RDST)
4 Experiments
4.1 Sensitivity Analysis
4.2 Scalability
4.3 Comparative Study
4.4 Interpretability
5 Conclusions and Future Work
References
Guiding Random Walks by Effective Resistance for Effective Node Embedding
1 Introduction
2 Related Work
3 Guiding Random Walks with Effective Resistance
3.1 Preliminaries
3.2 Effective Resistance as a Walk Guide
3.3 Combining Guided and Regular Random Walks
4 Experimentation
4.1 Datasets
4.2 Evaluation Tasks
4.3 Results and Discussion
5 Conclusion and Future Work
References
An Oculomotor Digital Parkinson Biomarker from a Deep Riemannian Representation
1 Introduction
2 Proposed Method
2.1 Convolutional Module and Symmetric Pooling Representation
2.2 Riemmanian Module Structure
3 Experimental Setup
3.1 Dataset Description
3.2 Network Configuration
4 Evaluation and Results
5 Conclusions and Future Work
References
Pruning Neural Nets by Optimal Neuron Merging
1 Introduction
1.1 Related Work
2 Background: Pruning and Merging
2.1 Error Propagation in Pruning
2.2 Pruning-and-Merging
3 New Method: Optimal Pruning-and-merging
3.1 Solution of Problem 1, Step I
3.2 Solution of Problem 1, Step II
4 Nuclear Norm Regularization for Better Pruning
5 Experiments
5.1 Prune and Merge Evaluation
5.2 Prune, Merge and Fine-tune
5.3 Nuclear Norm Regularization
6 Conclusions
References
Correction to: Learning Document Graphs with Attention for Image Manipulation Detection
Correction to: Chapter "Learning Document Graphs with Attention for Image Manipulation Detection" in: M. El Yacoubi et al. (Eds.): Pattern Recognition and Artificial Intelligence, LNCS 13363, https://doi.org/10.1007/978-3-031-09037-0_22
Author Index

System requirements

Save as PDF Copy link into clipboard

Schweitzer Fachinformationen

Pattern Recognition and Artificial Intelligence

Description

More details

Other editions

Additional editions

Content

System requirements