<![CDATA[Image processing technology continues to advance and is widely used for visual identification of human activities, including monitoring smoking behavior in no-smoking areas. This study develops an automated smoking activity detection and recognition system based on digital image processing, combining YOLO (You Only Look Once) for object detection and a CNN (Convolutional Neural Network) as an image classifier. YOLO detects and crops human objects, while the CNN classifies smoking and non-smoking activities based on visual features. The preprocessed dataset contains 560 valid images per class (smoking and not smoking). Training results show 96.09% accuracy on the training set and 94.44% on the validation set, with stable loss, while model evaluation yields 94.44% accuracy, 92.55% precision, 96.67% recall, 94.57% F1-score, and Average Precision (AP), indicating excellent classification performance. The model can also detect smoking activities in real-time images and camera feeds, demonstrating the effectiveness of combining YOLO and a CNN for automated detection, with potential applications in no-smoking areas.]]>
Copyrights © 2026
