Year: 2025
Type: article
Abstract: A unique tracking machine for fire and smoke detection is critical to ensure human protection and safety. Although cutting-edge fire alarm systems offer realistic solutions, there may nevertheless be an urgent need for greater accurate detection techniques. Fires can cause extensive harm; consequently, early detection is important. Convolutional Neural Networks (CNNs) are deep learning techniques that have been advanced to understand smoke and fire in images and video frames. The YOLO (You Only Look Once) model has shown incredible ability, especially the advanced YOLOv8 version. YOLOv8 offers faster and greater accurate detection of smoke and fire capabilities. In this examine, we advise using YOLOv8 for flame detection and check its performance in comparison to standard shallow learning models based on fuzzy common sense, color, motion, and shape. The tough Smoke and fire dataset, which incorporates a wide range of actual-world images, was used to assess the fashions. The outcomes exhibit that YOLOv8 outperforms conventional techniques in terms of model length, accuracy, and detection pace. With an average common Precision (mAP) of 95.3%, it gives an effective solution for smart flame and smoke detection.
Authors: Ali Al-Khafagi, Audia Saburi
Institution: Islamic University Babylon Campus
Language: English
Subfield: Safety, Risk, Reliability and Quality
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W4412430661