Кафедра інформатики (ІНФ)
Постійний URI для цієї колекції
Перегляд
Останні подання
Публікація Evaluating language models on low-resource pairs(ХНУРЕ, 2025) Bodenchuk-Pastukhov, Y. V.This work is devoted to the evaluation of the Facebook M2M100_418M and Alirezamsh Small100 models for low-resource language pairs. For this study, parallel corpora were selected for the following language pairs: Japanese-Ukrainian, Korean-Ukrainian, Turkish-Ukrainian, Vietnamese-Ukrainian, and Chinese-Ukrainian. The models were assessed based on their performance in translating these language pairs. Evaluation metrics included BLEU and ChrF scores, which measure the quality of the translations. Additionally, differences between the target and translated sentences were analyzed. The study aims to highlight the strengths and weaknesses of each model when working with low-resource languages. A comparative analysis of the results provides insights into the effectiveness of these models. The findings can be useful for future improvements in machine translation for underrepresented language pairs.Публікація Ієрархічні моделі семантичної сегментації зображень в інформаційних технологіях(ХНУРЕ, 2025) Єнацький, О. О.In recent years, the rapid growth of visual data in fields such as medicine, security, and autonomous systems has highlighted the need for more advanced image processing techniques. Semantic segmentation, a fundamental task in computer vision, plays a crucial role in extracting meaningful information from images by dividing them into regions based on their content. This process is particularly important for applications like autonomous driving, medical diagnostics, satellite image analysis, and intelligent video surveillance systems.Публікація Квантові вейвлет перетворення(ХНУРЕ, 2025) Гальченко, К. Р.The paper describes the promising directions of using quantum wavelet transforms (QWT) in the field of image processing, as well as the components of these transforms. The use of QWT allows for efficient image compression while preserving detail and more accurate color space information. For example, the quantum wavelet transform can be used to encode images in quantum states with subsequent restoration.Публікація Розробка сервісу для пошуку новин на основі уподобань користувача(ХНУРЕ, 2025) Галета, В. Ю.The work focuses on the analysis and development of a Telegram bot designed for personalized news search based on user preferences. By integrating user-provided preferences and channel subscriptions, the bot employs text analysis and machine learning techniques to identify relevant news items and notify users. The study examines the challenges in implementing real-time data processing and ensuring scalability of the service.Публікація Програмний контроль часу використання інструментальних засобів(ХНУРЕ, 2025) Гаєвий, А. О.The work focuses on the development of a desktop application to enhance self-control and reduce dependence on instant gratification sources, such as YouTube Shorts and games. Using WPF and C#, the application allows users to set time limits on unproductive programs, track progress in achieving goals, and foster habits for independent productivity. Experimental use demonstrates its ef-fectiveness in minimizing distractions and improving focus.Публікація Дослідження роботи алгоритму ML.NET для обробки користувацьких рекомендацій(ХНУРЕ, 2025) Верепа, Д. С.This work explores the ML.NET algorithm for processing user recommendations. It addresses the challenge of large-scale data and diverse user preferences. The research employs matrix factorization and collaborative filtering techniques. The MatrixFactorizationTrainer is central in learning latent factors for users and items. An Alternating Least Squares (ALS) method is used to optimize predictions. Key hyperparameters such as latent dimensionality, iteration count, and regularization are tuned. The study demonstrates ML.NET’s capacity to build scalable, effective recommendation systems.Публікація Методи статистичного аналізу даних(ХНУРЕ, 2025) Бочарніков, І. В.The thesis discusses the basic concepts and methods of testing statistical hypotheses. The concepts of statistical, null and alternative hypotheses, their content and purpose are described. The classification of statistical significance criteria, including parametric, non-parametric and consistency criteria, is considered. The main distributions used to test hypotheses are presented, including the normal distribution, χ², Student’s and Fisher’s distributions. Particular attention is paid to the practical application of criteria for testing the normal distribution in a sample. The conditions for testing the hypothesis and the methods of analysis used are presented.Публікація Аналіз особливостей сучасних застосунків для пошуку рецептів за продуктами харчування(ХНУРЕ, 2025) Боцвін, О. С.This work analyzes the features of modern applications that help users find recipes based on the ingredients they have at home. As more people look to make meals without shopping for extra items, these apps have become important tools. The study looks at apps like Cookpad, retsepty.online.ua, My Fridge Food, and Supercook, comparing how they work and what users like or dislike about them. While these apps make it easier to cook with what’s on hand, they have some issues, like including extra ingredients users didn’t choose and being tricky to navigate. This analysis highlights the advantages and disadvantages of current applications, which can be improved when creating new apps.Публікація Аналіз особливостей Character.AI: проблема довготривалої пам’яті у генеративному штучному інтелекті(ХНУРЕ, 2025) Богдан, Н. І.This work analyzes the capabilities and limitations of the Character.AI platform, which utilizes generative artificial intelligence to create interactive characters. The study highlights the growing popularity of AI-driven character generation among writers, game developers, and enthusiasts of interactive storytelling. Character.AI allows users to engage in virtual conversations with AI-generated personalities, enhancing creative experiences. The key advantages of Character.AI include its flexibility in character creation, interactive dialogue adaptation, support for role-playing scenarios, free accessibility, and a vast selection of pre-existing characters. These features make it a valuable tool for creative applications. However, despite its advantages, the platform has several notable shortcomings.Публікація Особливості існуючих вебзастосунків для розпізнавання та контролювання правильності виконання фізичних вправ(ХНУРЕ, 2025) Голубович, Є. Д.This work is devoted to analyzing and developing an accessible and convenient tool for personal training, which uses and combines the latest technologies for advanced control of physical activity. The relevance of the task in the modern world is described. The web application, the development of which is described in this work, is designed to solve this problem by using computer vision methods to analyze keypoints of the body during exercise. This will allow users to inde-pendently control the quality of their training in real time without the constant presence of a trainer.Публікація Порівняння ефективності стиснення електронних книг у різних форматах(ХНУРЕ, 2025) Захаров, В. В.This work is devoted to assessing the efficiency of e-book compression in various formats. A comparative analysis of EPUB, MOBI, and PDF formats was conducted based on file size reduction for different literary works. The study demonstrated that EPUB provides the most efficient compression, consistently producing the smallest file sizes. MOBI exhibits moderate compression efficiency, while PDF results in significantly larger files due to its document-oriented structure. The findings highlight EPUB as the optimal format for minimizing storage requirements.Публікація Інтеграція криптогаманця MetaMask для проведення web3 транзакцій(ХНУРЕ, 2025) Залізко, В. Я.This project focuses on integrating a cryptocurrency wallet for Web3 transactions in decentralized applications (dApps). Using JavaScript and React, we provide a secure and convenient way for users to connect to their crypto wallets, such as MetaMask. This allows them to make transactions, check their balances, and interact with smart contracts on the Ethereum blockchain. This system not only improves security and user experience, but also opens up integration opportunities with DeFi platforms, NFT marketplaces, and other Web3 services.Публікація Вирішення проблеми публікації фейкових вакансій в інтернеті(ХНУРЕ, 2025) Білоцерківська, В. А.Fake job postings pose a significant threat to job seekers, eroding trust in the online recruitment environment. This paper outlines a comprehensive strategy that integrates automated text analysis, manual moderation, and robust legal frameworks to identify and remove fraudulent vacancies. The methodology leverages machine learning algorithms to detect anomalies in job postings, while user feedback mechanisms help continuously refine the detection process. Additionally, educational initiatives aim to equip candidates with the knowledge to recognize and avoid suspicious offers, thereby creating a safer and more transparent job market. Building on these approaches, the proposed model emphasizes multilevel verification and prompt incident reporting to bolster system integrity and user confidence.Публікація Зрівняння розробки для iOS та Android в контексті сучасних мобільних технологій(ХНУРЕ, 2025) Бєгунова, В. Д.The work compares the main differences between developing mobile appli-cations for iOS and Android platforms. It analyzes the features of each platform’s ecosystem, including the choice of programming languages, development envi-ronments, and user interface design requirements. It discusses the benefits and challenges that developers face, including issues of performance, stability, testing, and adaptation of applications to different devices. It also focuses on the features of publishing applications on the App Store and Google Play. The study allows us to better understand what factors influence the choice of approach to developing mobile applications depending on business needs and user expectations.Публікація Сканування та аналіз пристроїв локальної мережі(ХНУРЕ, 2025) Балабуха, І. О.This study presents a methodology for scanning and analyzing local network devices using ARP requests and Nmap. The approach enables accurate identification of active devices, including MAC addresses, operating systems, and open ports. The use of Scapy and Nmap ensures efficient data collection and network visualization. Device classification based on MAC addresses helps determine hardware manufacturers and device types. The proposed method enhances network security, simplifies monitoring, and improves administrative efficiency. It is applicable for corporate network analysis, threat detection, and unauthorized access prevention.Публікація Розробка веб-застосунку для електронного медичного запису(ХНУРЕ, 2025) Малахова, А. Р.This paper presents the development of a web-based doctor appointment scheduling system designed to optimize patient-doctor interactions. The system allows users to book medical consultations, manage appointments, and track visit history. The application is built with React for the front end and Node.js for the back end, utilizing a REST API for seamless communication with a PostgreSQL database. The architecture follows a client-server model with role-based access control for patients, doctors, and administrators. Security features such as JWT authentication and data encryption ensure the protection of sensitive medical information. The proposed solution distinguishes itself through an efficient appointment management system, administrative tools for scheduling control, and GDPR-compliant data handling, making it a robust alternative to existing healthcare platforms.Публікація Розгляд питання створення конспекту лекцій на основі відео та презентації(ХНУРЕ, 2025) Максімов, Г. Р.Video content is a major source of information, yet time constraints often prevent users from fully engaging with it. This work presents an innovative application that automatically generates comprehensive lecture notes by analyzing video, audio, and presentation materials. The system utilizes advanced techniques such as OpenAI Whisper for transcription, Silero VAD for silence detection, FFmpeg for key frame extraction, and modern large language models for summarization. Unlike existing solutions, our approach effectively integrates visual cues from key frames to produce contextually rich summaries.Публікація Методи семантичного аналізу відеоданих(ХНУРЕ, 2025) Макаров, Д. С.Semantic analysis of video data involves extracting meaningful information and understanding the context and objects within the video content. With the rapid growth of video data in recent years, there has been a significant interest in developing techniques for automatic interpretation. This report explores various methods for semantic analysis, including object recognition, activity recognition, scene understanding, and deep learning-based approaches. The use of convolutional neural networks (CNNs) and recurrent neural networks (RNNs) for analyzing temporal and spatial data is discussed, alongside challenges such as data complexity, scalability, and real-time processing.Публікація Класифікація зображень на основі квантування ознак(ХНУРЕ, 2025) Мазур, Є. В.The results of an experimental study of a highspeed image classification method that performs classification using cluster representation weights for a reference database are presented. The BRISK detector was used to obtain keypoint descriptors. Offline and online versions of the classifier were developed. The results showed a significant acceleration of classification compared to the traditional method based on linear search.Публікація Застосування нейромережевого підходу для оживлення зображень обличчя(ХНУРЕ, 2025) Льолін, Д. В.In this work, methods for animating 3D facial models using the neural network approach of the First Order Motion Model (FOMM) and the integration of animation data into Blender are discussed. Key points are used to create a motion map that allows the transmission of facial expressions and gestures to the 3D model. The obtained animation data is synchronized with Shape Keys in Blender, providing natural movement of the model without the need for complex rigging.