Object-Oriented Intelligent System for Automated Control of Smoking by Video Data
Вантажиться...
Дата
2025
Назва журналу
Номер ISSN
Назва тому
Видавець
Анотація
This paper uses deep learning to present an object-oriented intelligent system for automated smoking detection in video surveillance. The system is based on a Vision Transformer (ViT) model, fine-tuned to identify smoking behaviour with high accuracy by analysing visual cues such as hand-to-mouth gestures and smoke. A modular architecture enables flexibility and scalability, using abstract classes and interfaces for components like preprocessing, patch embedding, classification, and visualisation. The system supports dataset management and performance monitoring via an interactive interface. It is designed for real-time applications and can be deployed in public spaces and industrial environments, contributing to health and safety compliance. Integration with smart city infrastructure and environmental sensors allows for proactive responses. The approach is extendable to detecting other behaviours, promoting broader public health and safety goals.
Опис
Ключові слова
Бібліографічний опис
Kok I.A., Kadynska V.D., Zalutska O.O., Mazurets O.V. Object-Oriented Intelligent System for Automated Control of Smoking by Video Data. Current scientific goals, approaches and challenges. Proceedings of IV International Scientific and Theoretical Conference. June 13, 2025. Dresden, Federal Republic of Germany. Pp. 156-164