ENHANCING WRITER IDENTIFICATION AND WRITER RETRIEVAL WITH CenSurE AND VISION TRANSFORMERS Mykyta Shupyliuk, Vitalii Martovytskyi, Yuri Romanenkov Technology Audit and Production Reserves, 2025 The object of research is the process of writer identification based on handwritten text. Despite significant progress, existing methods for author identification from handwritten text have limitations that prevent them from achieving maximum accuracy and reliability. This paper focuses on optimizing and improving the efficiency of writer identification from handwritten text by integrating image preprocessing methods, feature detection, and modern machine learning architectures. To this end, a functional model was developed that uses the CenSurE algorithm to detect key points and extract relevant image areas, and then the Vision Transformer model to identify the writer based on these extracted features. To reduce the variability of the results, experimental validation was conducted using a dual search and classification methodology. The use of the public CVL dataset increases reproducibility and helps in comparative analysis. The findings indicate that the implementation of the proposed approach leads to an improvement in the identification accuracy during retrieval, surpassing the results of other studies. This is evidenced by increased accuracy values of hard top k and soft top k by 1% and mean average precision by 2%. In addition, findings indicate significant performance improvement in the feature detection preprocessing stage. This improvement is quantitatively supported by reductions in both the average time per item and total processing duration by 39%, alongside the increase in total count of patches extracted by 70%. The results obtained contribute to increasing the reliability of automated handwriting analysis systems, especially for the task of writer identification. This achievement is a valuable tool for graphologists and forensic document experts, supporting such critical tasks as the forensic authorship process.
KNOWLEDGE SYSTEMATIZATION ABOUT THE CHARACTERISTICS OF EXISTING TECHNOLOGICAL MEANS FOR ASSISTING PEOPLE WITH VISUAL IMPAIRMENTS Vitalii Serdechnyi, Olesia Barkovska, Andriy Kovalenko, Vitalii Martovytskyi, Anton Havrashenko Advanced Information Systems, 2025 The work is devoted to a detailed review of the main aspects of physical vulnerability for people with visual impairments, as well as technological means for navigation and adaptation to the surrounding environment, which can significantly enhance their sense of safety and security. The relevance of the topic is justified by its large social focus, because such systems help people with visual impairments to socialize more easily and ensure greater inclusion. This is particularly important in urban environments where insufficient attention is paid to inclusivity and the comfort of visually impaired individuals (e.g., lack of audible traffic lights, tactile paving, etc.). The subject of the article is the study of hardware components that ensure the functionality of support systems for people with visual impairments. The goal of this paper is to systematize knowledge about existing technological tools for people with visual impairments and to analyze the hardware characteristics of the components of such solutions. The task of this work is to examine the psychophysiological factors and aspects of physical vulnerability for people with visual impairments, review existing assistance systems for visually impaired individuals, identify the hardware base required for creating a “vision” system of the surrounding environment, and analyze the characteristics of sensors considering the external conditions in which visually impaired people may find themselves. The objectives are achieved through the use of methods such as comparative analysis, classification and categorization, and a systematic review of the literature in the relevant problem domain. The results of the work include a proposed classification of assistive devices for people with visual impairments, which encompasses the following classes: navigation applications and devices; sensory systems for obstacle and object detection; wearable devices with augmented reality (AR) features; “vision” systems for the surrounding environment; and text recognition systems. The evaluation and analysis of the advantages and disadvantages of devices in each of these classes demonstrate that a new solution should meet the criteria of compactness, wearability, energy efficiency, ease of use, and high accuracy in detecting environmental conditions, obstacles, and objects on the user's path. Conclusions. To ensure data complementarity in tasks of detecting moving objects in intelligent assistance systems for visually impaired individuals, the optimal approach is to combine multiple sensors using the Multisensor Fusion methodology. Specifically, this involves high-resolution cameras that provide detailed scene imaging and LiDARs that ensure precise distance measurement and 3D modeling of the environment. Such an approach compensates for the limitations of individual sensors and provides a more comprehensive understanding of the scene, improving data quality through the integration of diverse information sources. Further research will focus on conducting experimental research aimed at practically justifying the joint use of cameras, audio sensors, and LiDARs for obtaining heterogeneous data that provide the most comprehensive depiction of the environment surrounding visually impaired individuals.
RESEARCH ON MACHINE LEARNING METHODS FOR DETECTING OBJECTS IN DIFFICULT SHOOTING CONDITIONS Vitalii Serdechnyi, Olesia Barkovska, Andriy Kovalenko, Anton Havrashenko, Vitalii Martovytskyi Radioelectronic and Computer Systems, 2025 The subject matter of the article is research into machine learning methods for object detection in images and videos under complex urban conditions, particularly under poor lighting, the presence of precipitation, high scene complexity, and limited computational resources. The goal of this research is to identify the most effective deep learning models based on convolutional neural networks for object detection tasks under challenging imaging conditions, considering the practical requirements for accuracy and processing speed. The tasks to be solved are: analysis of object detectors (YOLO v8–11, DETR, SSD, Mask R-CNN, Faster R-CNN, RetinaNet); preparation of a dataset with real weather conditions and pedestrian environments in Ukraine; experimental evaluation of selected detectors using the metrics mAP@0.5, mAP@.5:.95, Recall, Precision, IoU, FPS, and F1-Score; and analysis of the obtained results. The methods used are: convolutional neural networks, automated image annotation, comparative analysis of quality metrics (F1-score, mAP@0.5:.95, Precision, Recall, IoU, FPS), and manual correction of annotations. The following results were obtained: the YOLOv10-m and YOLOv11-m models demonstrated the best quality indicators under conditions of limited visibility and varying lighting. The YOLOv11-m model was the most balanced in terms of accuracy and speed across all tested conditions - snow, rain, and sunshine. YOLOv11-m is recommended as the baseline model for implementation in real-time systems, particularly in intelligent assistants for people with visual impairments. Conclusions: The scientific novelty of the results obtained is as follows: 1) a comprehensive evaluation of modern deep learning architectures for object detection (YOLOv8–v11, Faster R-CNN, SSD, Mask R-CNN, DETR, RetinaNet) was carried out under non-laboratory conditions, including real weather scenarios such as snow, rain, and poor lighting, which are typical for urban environments in Eastern Europe; 2) the software tool for automated model evaluation was developed, allowing simultaneous testing of multiple architectures and visualization of performance metrics (F1-score, mAP@0.5, mAP@.5:.95, IoU, Precision, Recall, FPS) with support for manual annotation correction and comparative model analysis; 3) it was experimentally established that the YOLOv11-m model demonstrates the best balance of accuracy and inference speed across various complex imaging conditions, justifying its recommendation as a baseline model for real-time vision-based assistive systems.
DEVISING AN APPROACH TO PERSONALITY IDENTIFICATION BASED ON HANDWRITTEN TEXT USING A VISION TRANSFORMER Mykyta Shupyliuk, Vitalii Martovytskyi, Nataliia Bolohova, Yuri Romanenkov, Serhii Osiievskyi, et al. Eastern European Journal of Enterprise Technologies, 2025 The object of this study is the approach to personality identification based on handwritten text using machine learning methods. Increasing the accuracy of personality identification and automating feature extraction could make it possible to perform more accurate analysis of handwritten text. A functional model has been built, and an experimental study of the proposed approach was conducted. The results of the study showed that the proposed approach increased the overall accuracy of identification, compared to other studies, as evidenced by the obtained accuracy values with the lowest indicator of 94.84 % for Friendliness and the highest 99.48 % for Conscientiousness. The accuracy indicator also improved compared to other models, as evidenced by the average accuracy value, which increased from 0.65 to 0.94. Such results were obtained through the use of the "Vision Transformer" method, which makes it possible to remove the need for feature extraction as a separate step, and the scale-invariant feature transformation approach made it possible to extract relevant image patches. An experimental validation was conducted using retrieval and classification approaches, which minimizes the variability of the results. The use of the Big Five model and the CVL dataset improves the accessibility of the study for comparison and reproducibility. In practice, handwriting analysis is widely used in forensics, for personnel selection, as well as in other areas of activity. The results increase the reliability of automated handwriting analysis systems in the area of personality identification, which could help graphologists and handwriting experts in their work both to assess personality traits and to determine whether a certain handwritten text belongs to a specific person
DEVELOPING A RISK MANAGEMENT APPROACH BASED ON REINFORCEMENT TRAINING IN THE FORMATION OF AN INVESTMENT PORTFOLIO Vitalii Martovytskyi, Volodymyr Argunov, Igor Ruban, Yuri Romanenkov Eastern European Journal of Enterprise Technologies, 2023 Investments play a significant role in the functioning and development of the economy. Risk management is an integral part of the formation of the investment portfolio. This means that an investor must be willing to take on a certain level of risk in order to receive a certain level of return. However, when forming an investment portfolio, an investor faces such problems as market unpredictability, asset correlation, incorrect asset allocation. Therefore, when forming an investment portfolio, an investor should carefully study all possible risks and try to minimize them. The object of research is an approach to risk management in the formation of an investment portfolio using the method of reinforcement training. The basic principles of formation of the investment portfolio and determination of risks are described. The application of the method of reinforcement training for building a model of risk management of investment portfolio is considered. The process of selecting optimal investment assets based on alternative data sources that minimize risks and maximize profits is also considered. A functional model of the process of risk optimization in the formation of an investment portfolio based on machine learning methods has been developed. The functional model constructed makes it possible to build a process of risk optimization, including asset selection, risk comparison and assessment, to form an investment portfolio and monitor its risks. The study results showed that the proposed approach to the formation of the investment portfolio increased the total growth of the investment portfolio by 0.4363 compared to the base model. Also, the volatility indicator improved compared to the market, as evidenced by the percentage difference between the initial and final cash amount, which increased from 128.98 to 295.57
Self-healing Systems Monitoring Igor Ruban, Vitalii Martovytskyy, Olesia Barkovska Studies in Systems Decision and Control, 2022
DEVISING AN APPROACH TO THE IDENTIFICATION OF SYSTEMUSERS BY THEIR BEHAVIOR USING MACHINE LEARNING METHODS Vitalii Martovytskyi, Оleksandr Sievierinov, Oleksii Liashenko, Yuri Koltun, Serhii Liashenko, et al. Eastern European Journal of Enterprise Technologies, 2022 One of the biggest reasons that lead to violations of the security of companies’ services is obtaining access by the intruder to the legitimate accounts of users in the system. It is almost impossible to fight this since the intruder is authorized as a legitimate user, which makes intrusion detection systems ineffective. Thus, the task to devise methods and means of protection (intrusion detection) that would make it possible to identify system users by their behavior becomes relevant. This will in no way protect against the theft of the data of the accounts of users of the system but will make it possible to counteract the intruders in cases where they use this account for further hacking of the system. The object of this study is the process of protecting system users in the case of theft of their authentication data. The subject is the process of identifying users of the system by their behavior in the system. This paper reports a functional model of the process of ensuring the identification of users by their behavior in the system, which makes it possible to build additional means of protecting system users in the case of theft of their authentication data. The identification model takes into consideration the statistical parameters of user behavior that were obtained during the session. In contrast to the existing approaches, the proposed model makes it possible to provide a comprehensive approach to the analysis of the behavior of users both during their work (in a real-time mode) and after the session is over (in a delayed mode). An experimental study on the proposed approach of identifying users by their behavior in the system showed that the built patterns of user behavior using machine learning methods demonstrated an assessment of the quality of identification exceeding 0.95
Mathematical Model of User Behavior in Computer Systems Vitalii Martovytskyi, Igor Ruban, Oleksandr Sievierinov, Andrii Nosyk, Valentyn Lebediev 2020 IEEE International Conference on Problems of Infocommunications Science and Technology Pic S and T 2020 Proceedings, 2021 Various electronic keys (tokens, cards, etc.) are also quite common in the quality of identifiers. But it should be noted that in recent years, more and more widespread identification systems that use biometric characteristics of man in solving the problem of access to information systems. Therefore, the task of developing models of behavior of users of computer systems, which take into account both dynamic and static properties of the behavior of users, as well as their possible ones, is urgent. The paper presents a comprehensive approach to the analysis of user behavior in order to identify anomalies in its work. This approach should take into account the dynamic and statistical properties of the behavior, as well as possible changes in behavior that are not associated with anomalies.
Method of Detecting FDI Attacks on Smart Grid Vitalii Martovytskyi, Igor Ruban, Hennadiy Lahutin, Irina Ilina, Volodymyr Rykun, et al. 2020 IEEE International Conference on Problems of Infocommunications Science and Technology Pic S and T 2020 Proceedings, 2021
Measuring Vulnerabilities in Threat Modelling with Risk Matrix Andrii Hapon, Volodymyr Fedorchenko, Vitalii Martovytskyi, Volodymyr Rykun, Oleksandr Sievierinov, et al. 2021 IEEE 8th International Conference on Problems of Infocommunications Science and Technology Pic S and T 2021 Proceedings, 2021
Functional Model of Computer Networks Security Information Dmytro Holubnychyi, Vitalii Martovytskyi, Igor Ruban, Oleksandr Sievierinov, Valentyn Lebediev, et al. 2021 IEEE 8th International Conference on Problems of Infocommunications Science and Technology Pic S and T 2021 Proceedings, 2021
Efficiency of image convolution Kirill Smelyakov, Mykyta Shupyliuk, Vitalii Martovytskyi, Dariia Tovchyrechko, Oleksandr Ponomarenko Proceedings of the International Conference on Advanced Optoelectronics and Lasers Caol, 2019
Method of neural network recognition of ground-based air objects Igor Ruban, Kirill Smelyakov, Martovytskyi Vitalii, Pribylnov Dmitry, Nataliia Bolohova Proceedings of 2018 IEEE 9th International Conference on Dependable Systems Services and Technologies Dessert 2018, 2018
Search by Image. New Search Engine Service Model Kirill Smelyakov, Denys Sandrkin, Igor Ruban, Martovytskyi Vitalii, Yury Romanenkov 2018 International Scientific Practical Conference on Problems of Infocommunications Science and Technology Pic S and T 2018 Proceedings, 2018
