Automatic Fire Detection, based on early recognition using deep features.

Abstract

Automatic fire detection has the potential to provide a 24/7 monitoring system for various environments at a relatively low cost. Fire detectors are one of the most reliable ways to detect fire. Once a fire is detected, the system may perform different functions, including alarming the protection system, discharging the fire extinguisher system, or activating the fire sprinkler system. It depends mainly on smoke detector sensors which may not provide reliable results based on different environmental factors. With the help of computer vision, a fire detection system can automatically notify the occupants of a building and emergency services of a fire while it is still in its initial stages. Also, for outdoor environments, such a fire detection system can automatically monitor a large area and alert the appropriate channels at the first sign of a fire. Deep learning techniques have been proven effective in various computer vision tasks. This paper presents a Deep learning method for fire detection that ensures quicker response time and better management of fire incidents which help identify fires early, improving response times and potentially reducing damage. Experiments show promising results of the proposed method compared to the state-of-the-art.