Light-weight convolutional neural network for fire detection
Date
2021-07
Journal Title
Journal ISSN
Volume Title
Type
Article
Publisher
Institute of Electrical and Electronics Engineers Inc.
Series Info
ICEEM 2021 - 2nd IEEE International Conference on Electronic Engineering.;3 July 2021 Article number 94803782nd IEEE International Conference on Electronic Engineering, ICEEM 2021
Scientific Journal Rankings
Abstract
Fire disasters damage the economy across the globe and cause many casualties among civilians and firefighters. In this paper, a deep learning architecture based on the convolutional neural network (CNN) is proposed to detect fires efficiently. We trained the network on 9247, picked high-resolution images containing fire and other ones without any fire, and investigated the effect of CNN depth on its classification accuracy. In this proposed work, we achieved 98% accuracy on the testing set, which is so far better than the previous state-of-the-art and will eventually minimize fire disasters and reduce the damage caused by human resources. © 2021 IEEE.
Description
Scopus
Keywords
Classification, Computer Vision, Convolution Neural Network (CNN), Fire Detection, Supervised Learning