Pattern Recognition and Information Processing: 15th International Conference, PRIP 2021, Minsk, Belarus, September 21–24, 2021, Revised Selected ... in Computer and Information Science)

This book constitutes the refereed proceedings of the 15th International Conference on Pattern Recognition and Information Processing, PRIP 2021, held in Minsk, Belarus, in September 2021. Due to the COVID-19 pandemic the conference was held online. The 17 revised full papers were carefully reviewed and selected from 90 submissions. The papers present a discussion on theoretical and applied aspects of computer vision, recognition of signals and images, the use of distributed resources, and high-performance systems. Preface Organization Contents Classification of Histology Images Based on a Compact 3D Representation Abstract 1 Introduction 2 Method 2.1 Cluster Data Preparation 2.2 Tensor-Based Feature Extractor and Classifier 3 Experiments and Results 3.1 Experimental Setup and Results 4 Conclusions Acknowledgments References Smart Tiling for Program Optimization and Parallelization Abstract 1 Introduction 2 Dependencies Between Operators in Programs 3 Graph Models of Iteration Space 4 Parallelization Task 5 Methods of Transformation of Iteration Space 6 Cyclic Partitioning of Iteration Space 7 Smart Tiling 8 The Experiments 9 Conclusions References Digest of Blockchain Technologies to Design System for Big Image Data Provenance and Security Abstract 1 Introduction 2 BC Surveys 3 Blockchain 3.1 Transactions 3.2 Block Structure 3.3 BC Architecture 3.4 Smart Contract 4 Information Security with BC 4.1 Security Threats 4.2 Consensus Mechanisms 5 Digest of BC Technologies 5.1 Implementations of BC and Other DLT Systems 5.2 Summary of BC Technologies 6 BC for Big Image Data 6.1 Implementation 6.2 Performance 6.3 Image Data Hashing 7 Conclusions References Formalisation of Motion Description in Microscopy Images Abstract 1 Introduction 2 Formalization of a Dynamic Object 3 Formalization of Motion 4 Motion of a Set of Dynamic Objects in a Microscopic Image Sequences 5 Monitoring of Dynamic Objects Motion 6 Practical Results 7 Conclusion Acknowledgment References Predicting Events by Analyzing the Results of the Work of Predictive Models Abstract 1 Introduction 2 “Success Probability” of Prediction 3 Selection Best Pairs 4 The Example of Dynamic Prediction 5 Updating a Prediction Model Under New Data 6 “Approximate Coincidences” of Predictions 7 {{\varvec n}}-Dimensional Models 8 On the Further Development of this Theory 9 Conclusion References Formalization of People and Crowd Detection and Tracking for Smart Video Surveillance Abstract 1 Introduction 2 Formalization of Person Motion Detection Problem 3 Formalization of Crowd Motion Detection Problem 4 Formalization of Person and Crowd Tracking Problem 4.1 Single Person Tracking 4.2 Multiple Person Tracking 4.3 Crowd Tracking 5 Experimental Results 5.1 People Detection and Tracking Results 5.2 Crowd Motion Detection and Tracking 6 Conclusion References Investigation of the GAN-SSL Classifier Properties for Identification Expertise Abstract 1 Introduction 2 Related Works 3 Model Framework 3.1 Problem Definition 3.2 GAN-SSL Architecture 3.3 Learning Algorithm 4 GAN-SSL Experimental Research 4.1 Classification for Model Data 4.2 Classification for Petrol Identification Expertise 4.3 Properties of Generator 4.4 Advantages of Semi-supervised Learning for Classification 5 Conclusion References Comparing the Performance of Classical and Deep Learning Methods on Small Image Datasets Abstract 1 Introduction 1.1 The Motivation 1.2 The Context of Test Images 2 Materials 2.1 Histopathology Images 2.2 Computed Tomography Images 3 Traditional Methods 3.1 Traditional Methods of Histology Image Classification 3.2 Traditional Methods of CT Image Classification 4 CNN-Based Methods 4.1 CNN-Based Methods of Histology Image Classification 4.2 CNN-Based Methods of CT Image Classification 5 Experimental Arrangements 6 Results 6.1 Results of Histology Image Classification 6.2 Results of CT Image Classification 7 Conclusions References Generative Autoencoders for Designing Novel Small-Molecule Compounds as Potential SARS-CoV-2 Main Protease Inhibitors Abstract 1 Introduction 2 Methods 2.1 Training Set Preparation 2.2 3D Structures Generation for Generated Molecules 2.3 Molecular Docking 2.4 Deep Learning 2.5 Deep Learning-Based Compounds Generation 3 Results 3.1 Overview of General Results 3.2 Results of Experiments by Models and Generation Modes 3.3 An Experiment with Setting Different Binding Free Energy Thresholds 3.4 Experiment with Gaussian Noise Utilization 3.5 Models and Generation Modes Comparison 3.6 Results Discussion 4 Conclusion Acknowledgments References Mask R-CNN-Based System for Automated Reindeer Recognition and Counting from Aerial Photographs Abstract 1 Introduction 2 Principles and Approaches to Recognition of Natural Objects 3 Recognition of Animals in Aerial Images from Reference Images Using Artificial Neural Networks 3.1 Convergent Neural Networks as an Image Recognition Tool 3.2 Training the Network for Reindeer Recognition in Aerial Images 3.3 The Web Interface of the System and the Results of Its Validation on an Independent Data Set 4 Conclusion Acknowledgements References Retinal Image Analysis Approach for Diabetic Retinopathy Grading Abstract 1 Introduction 2 Development of Our Technology for Retina Image Analysis 2.1 The Scheme of Our Technology 2.2 Quality Image Analysis 2.3 Image Preprocessing 3 Experimental Environment 4 Methodology of Experiments 4.1 Experimental Details 4.2 Machine Learning Model Development and Evaluation 5 Discussions 6 Conclusions Acknowledgements References Comparison of Deep Learning Preprocessing Algorithms of Nuclei Segmentation on Fluorescence Immunohistology Images of Cancer Cells Abstract 1 Introduction 2 Materials and Methods 3 Results and Discussions 4 Conclusion References Simulation Modelling and Machine Learning Platform for Processing Fluorescence Spectroscopy Data Abstract 1 Introduction 2 Methodology 2.1 Processing Fluorescence Data Using Simulation Modelling and Machine Learning Algorithms 2.2 Review of the Computational Tools for a Digital Platform 2.3 Conception of the Digital Platform 3 Results 4 Conclusions References A Bottom-Up Method for Pose Detection of Multiple People on Real-Time Video Abstract 1 Introduction 2 Problem Review 3 Solution Review 3.1 Image Preprocessing 3.2 Simultaneous Body Parts Detection and Association 3.3 Probability Mapping 3.4 Body Parts Compatibility Fields Calculation 3.5 Multiple People Affine Fields Processing 4 Solution Testing and Results 4.1 Test Results on MPII Multi-person Dataset 4.2 Test Results on COCO Keypoints Challenge Dataset References Authentication System Based on Biometric Data of Smiling Face from Stacked Autoencoder and Concatenated Reed-Solomon Codes 1 Introduction 2 Autoencoders and Error Correcting Codes 2.1 Autoencoders 2.2 Error Correcting Codes 3 Proposed System 3.1 System Structure 3.2 Algorithms for RS Codes Decoding 4 Experiments Performed 4.1 SAE and RS Simulation 4.2 Security Issues 5 Conclusions References Detection of Features Regions of Syndrome in Multiple Sclerosis on MRI Abstract 1 Introduction 2 Dataset Preparing 2.1 Properties of Multiple Sclerosis Images 2.2 Annotation Software 2.3 Correction of Tumor Node Shape 2.4 Preparation of Training Images 3 Definition of CNN Model 3.1 Model on Base DenseNet 3.2 Model on Base U-Net 3+ 4 Neural Network Training 5 Quality Assessment of Network 6 Discussion and Conclusion Acknowledgment References Automatic Tuning of the Motion Control System of a Mobile Robot Along a Trajectory Based on the Reinforcement Learning Method Abstract 1 Introduction 2 Reinforcement Learning Elements 2.1 The Description of the Problem 2.2 Learning Process 2.3 Controller Design of the Reinforcement Learning 2.4 Reward Function 2.5 Creation of a Neural Network of Actor and Critic 3 The Obtained Results, Their Significance and Comparison with Former Work 4 Conclusion References Author Index