International Journal of Advanced Research in Computer and Communication Engineering

Abstract: Fire detection is a crucial safety application aimed at minimizing the risk of human casualties and property loss. Conventional systems primarily rely on smoke or heat sensors, which often fail to detect fires in open or smoke-free environments. This research proposes an intelligent, Machine Learning-based vision system for real-time fire detection using computer vision techniques. The proposed model leverages a Convolutional Neural Network (CNN) trained on diverse datasets of fire and non-fire images to accurately classify fire instances from live video streams. Implemented in Python using OpenCV for image acquisition and TensorFlow/Keras for deep learning inference, the system triggers an alarm alert when fire is detected. Experimental results demonstrate over 92% detection accuracy, robust performance across varying lighting conditions, and minimal false positives. The system’s low computational cost and high responsiveness make it ideal for integration into smart surveillance, industrial safety, and IoT-based monitoring systems.

Keywords: Fire Detection, Machine Learning, Computer Vision, Convolutional Neural Network (CNN), OpenCV, TensorFlow, Real-Time Monitoring, Safety System.

Downloads: | DOI: 10.17148/IJARCCE.2025.141110