Перегляд за Автор "Zalutska, O."

Зараз показуємо 1 - 5 з 5
  • Ескіз
    Документ
    An Approach to Using MobileNet CNN-model for Gesture Recognition
    (2024) Mazurets, O.; Zalutska, O.; Tyschenko, O.; Bohdanova, A.; Мазурець, Олександр Вікторович
    As a result of the work performed, one of the types of MobileNet architecture of CNN artificial neural networks was applied in practice to solve the problem of gesture recognition in real time. The obtained results are important not only in the scientific sense, but also for practical application, because neural networks help people with various diseases, such as Parkinson's disease, speech or hearing impairment, tunnel syndrome, to facilitate their interaction with the computer, to carry out effective studying in schools and universities and, of course, socializing. So, these are just the first steps to simplify people's lives as much as possible and make it bright, despite certain limitations
  • Ескіз
    Документ
    An Approach to Using the mBERT Deep Learning Neural Network Model for Identifying Emotional Components and Communication Intentions
    (2024) Мазурець, Олександр Вікторович; Blazhuk, V.; Mazurets, O.; Zalutska, O.
    The problem of using the mBERT deep learning neural network model for identifying emotional components and communication intentions was considered. For this purpose, a method of determining the emotional components and communication intentions of text messages using natural language processing tools was developed, which is capable of determining the emotional components and dominant emotion of a text message and forming an expert opinion regarding the communication intentions based on the determined dominant emotion with justification in the form of a list of emotionally colored words and phrases To solve the task of identifying emotional components from text messages, the mBERT model revealed several key advantages, the main one of which is a deep understanding of the context, thanks to bidirectional learning and a multi-headed attention mechanism. This allows to capture complex emotional connections between words in the text, even if the context depends on the order of the words. Another feature of the mBERT model is its high adaptability to new words thanks to the sequential segmentation of words into tokens. This is especially relevant when analyzing text messages from social networks, where informal vocabulary is mostly used.
  • Ескіз
    Документ
    Method for Determining the Person Emotional State in Real Time by Neural Networks Tools
    (2024) Hladun, O.; Molchanova, M.; Zalutska, O.
    The paper proposes method for determining the person emotional state in real time by means of neural networks, which allows for transformation of input data in the form of trained neural network model of convolutional architecture and video stream into output data that includes information about person's emotional state, presented in the form of emotional tags that correspond to emotions: joy, sadness, anger, disgust, fear, surprise and neutral.
  • Ескіз
    Документ
    Research on the effectiveness of classifying the remains of destroyed buildings using MobileNetV3 neural network architecture
    (2025) Hladun, O.; Zalutska, O.; Klimenko, V.; Mazurets, O.; Мазурець, Олександр Вікторович
    The study investigates the effectiveness of the MobileNetV3 neural network architecture in classifying the remains of destroyed buildings, a task of increasing relevance due to the widespread destruction of infrastructure caused by military actions, natural disasters, and industrial accidents. A software application with a graphical interface was developed to enable interactive analysis of photo data using pre-trained models. The system allows users to classify construction debris into multiple categories with high accuracy and reliability. Experimental results demonstrated strong performance metrics, including an overall accuracy of 95% and high values for precision, recall, and F1-score across ten classes of construction materials. The model showed excellent discriminative capability, as evidenced by ROC curves with AUC values close to 1.00. The solution holds promise for practical applications such as automated sorting of construction waste and monitoring of damage zones, contributing to more efficient disaster response and recovery operations.
  • Ескіз
    Документ
    Software architecture of information system for exchanging LLM thematic prompts
    (2025) Murava, V.; Zalutska, O.; Didur, V.; Mazurets, O.; Мазурець, Олександр Вікторович
    This paper presents an object-oriented software solution for calculating material requirements in garment production based on digital sketches. The system models garment components—such as sleeves, collars, and panels— as reusable classes, enabling accurate fabric estimates and modular functionality. A class diagram structures key elements, including user input forms, data storage, and calculation algorithms. The approach enhances flexibility, supports customisation, and reduces waste by integrating with production planning. Future applications include expansion to machine learning, IoT-based manufacturing, and cross-domain material optimisation.