Інформаційна система виявлення курців у потоковому відео з використанням нейромережевих моделей виявлення обʼєктів

Abstract

An information system for real-time smoker detection from streaming video was developed. A specialized dataset was created on the Roboflow platform, and multiple object detection models, including YOLO and Roboflow, were evaluated. The best-performing model was integrated into a mobile application developed in Kotlin. The system enables automated identification of smoking events in public spaces and can be used for public safety and surveillance purposes.