Version 8 of YOLO for Wildfire Detection

Large-scale forest fires can cause irreversible damage to the ecosystem, prompting research into different ways to prevent and monitor these events from happening. One possible prevention method is monitoring areas most susceptible to fires and using computer vision techniques to detect these events as quickly as possible while they are still small-scale, accelerating the response of responsible authorities, and hence reducing environmental damage and restoration costs. Convolutional neural networks (CNNs) is currently enjoying the best accuracy among other methods (e.g. feature modeling) for wildfire detection from images. This paper applies version 8 of YOLO to reduce computational costs while maintaining the high detection capability.