Перегляд за Автор "Didur, V."

Зараз показуємо 1 - 4 з 4
  • Ескіз
    Документ
    Approach to Using Cloud Services for Visual Analytics of Neural Network Analysis of Texts Emotional Tonality
    (2024) Yurchenko, D.; Mazurets, O.; Didur, V.; Molchanova, M.; Мазурець, Олександр Вікторович
    For the neural network analysis of the emotional tonality of messages, it is proposed to use a hybrid architecture neural network that combines the simultaneous advantages of the CNN and BiLSTM architectures. A neural network architecture was developed for determining emotional tonality, and the LIME model of interpreted model-agnostic explanations was used to visually explain the results of the neural network analysis of emotional tonality. This approach will make it possible to use all the advantages of neural network solutions, but to have an understanding for the user of what influenced these solutions.
  • Ескіз
    Документ
    Recurrent Neural Network Model Architecture for Detecting a Tendency to Atypical Behavior Of Individuals by Text Posts
    (2024) Sobko, О.; Mazurets, О.; Didur, V.; Chervonchuk, I.; Мазурець, Олександр Вікторович
    The architectures for two recurrent neural network models that are the input data of the method for detecting a tendency to atypical behavior in a text message were presented. The architecture of the RNN model for detecting a tendency to atypical behavior has 2 outputs responsible for the presence of a tendency to atypical behavior or the absence of a tendency to atypical behavior, and the architecture of the RNN model for determining the type of mental disorder provides 10 outputs, each of which corresponds to the type of mental disorder that may affect the tendency to atypical behavior.
  • Ескіз
    Документ
    Research on the effectiveness of neural network detection of plots with the destroyed buildings remains
    (2025) Didur, V.; Molchanova, M.; Mazurets, O.; Мазурець, Олександр Вікторович
    This paper explores the effectiveness of neural network approaches for detecting and classifying plots containing the remains of destroyed buildings using aerial imagery. The proposed method integrates a YOLO-based object detector and a Vision Transformer for multi-class classification of structural debris such as concrete, metal, brick, and wood. The system achieves high accuracy (97%) and demonstrates strong performance across key classification metrics. This research highlights the critical role of deep learning in accelerating post-disaster damage assessment, supporting emergency response, cultural heritage preservation, and long-term urban resilience planning.
  • 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.