Early Fire Detection using Deep Learning

Abstract

Fires are man-made calamities that cause environmental, cultural, and economic damage. Rapid fire detection and fully autonomous response are critical and valuable to disaster management systems in decreasing these losses. Deep learning is a novel concept based on neural networks that has yielded exceptional results in a wide range of fields, including computer vision. As a result, we present an early fire detection system for CCTV security cameras that uses customized convolutional neural networks to detect fire in a variety of indoor and outdoor scenarios in this research. We used CNN to build our model and obtained some very good results in the evaluation measures. We developed a fire detection threshold to aid in the evaluation of several automatic fires in the area. We seek to overcome the shortcomings of present systems by developing a reliable and accurate system capable of recognizing fires as early as possible while function-ing in a wide range of situations, thereby saving numerous lives and resources.