PyroVision: A Deep Learning Based Model for Wildfire Detection in Satellite Imagery

Abstract

The significance of wildlife ecosystems emphasizes the necessity of effective methods for detecting and preventing wildfires which are also extremely dangerous due to their quick spread and catastrophic effects. Wildfire persistence is a result of both human activity and climate change. Current methods for preventing disasters frequently fall short of providing all-encompassing answers. In response, our research presents a novel method for detecting wildfires using satellite imagery. The accuracy and specificity of current methods are limited, even with recent advancements. As a result, we propose Pyro Vision utilizing Convolutional Neural Network (CNN) with attention mechanisms with an impressive accuracy of 95.51 %. Our method is not only good at locating individual wildfires, but it also works well at identifying regions. This complete approach to wildfire monitoring offers an effective way to improve environmental safety and reduce the effects of these calamities.