Sistem Pendeteksi Asap Rokok Berbasis Website Menggunakan Metode YOLOv8

Abstract

Violations of smoke-free area regulations remain a common issue because monitoring activities are generally performed manually and require continuous supervision. This condition encourages the development of an automated system capable of detecting cigarette smoke to improve monitoring effectiveness. This study aims to design and implement a web-based cigarette smoke detection system using the You Only Look Once version 8 (YOLOv8) algorithm. The research employed the Research and Development (R&D) method, including literature review, dataset collection and annotation using Roboflow, YOLOv8 model training, implementation into a Flask-based web application, and system evaluation through Black Box Testing. The developed system provides user authentication, a dashboard, image upload detection, real-time webcam detection, detection history, evidence storage, and alarm notifications when cigarette smoke is detected. Implementation results demonstrate that the YOLOv8 model was successfully integrated into the web application and capable of detecting cigarette smoke from both images and real-time video streams. All system functions operated according to specified functional requirements, while detection performance was influenced by lighting conditions, object distance, and background characteristics. The proposed system is expected to support more effective monitoring of smoke-free areas by utilizing computer vision technology.