Real-Time Video-Based Fire Detection Using Deep Learning Techniques: A Study of YOLO and CNN Architectures

Abstract

Early and accurate fire detection is vital for preventing severe loss of life and property. Traditional fire detection systems based on sensors often suffer from delays and limited coverage. With advancements in computer vision and deep learning, video-based fire detection has become a promising alternative. This paper explores real-time fire detection in video streams using deep learning models, particularly YOLO (You Only Look Once) and Convolution Neural Networks (CNNs). We present a comprehensive methodology, covering dataset preparation, preprocessing, model training, and evaluation. The strengths and limitations of each approach are discussed, and experimental results demonstrate the effectiveness of YOLO and CNNs for timely and reliable fire detection.