Real-Time Fire Object Detection System Using Machine Learning

Department of CSE, Vignan’s Foundation for Science, Technology and Research Vadlamudi, Guntur, Andhra Pradesh, India

¹ Email; akuthotabhargavi864@gmail.com

Abstract

The spread of forest fires presents one of the major concerning ecosystems, human security, and property. This paper introduces a fire object detection system that employs machine learning algorithms to enhance early detection of fire breakout and response to the same. The computer vision and deep learning algorithms allow the system to identify features related to fire objects and actions in images and video feeds. This set of scenarios under various fire conditions, environmental conditions, and backgrounds was curated for training a CNN. In terms of evaluating the model’s robustness in real applications across various settings, the metrics were defined by accuracy, precision, recall, and F1 scores. The proposed system is designed for alerting emergency responders within time so that quicker intervention may be made to possibly mitigate the devastating effects of wildfires. Future research will be the integration of the system into real-time surveillance systems and exploring added sensory data to increase the detection capabilities.