Abstract
This paper presents an overview of the current state-of-the-art methods in Unmanned Aerial Vehicle (UAV)-assisted detection of fire and smoke using artificial intelligence. The existing methods are reviewed, with a focus on deep learning methods and architectures. Furthermore, an experiment comparing the performance of two different deep learning models for object detection using the D-Fire dataset is performed. Results of the experiment indicate that while Faster Region-based Convolutional Neural Network (Faster R-CNN) has higher sensitivity to fire and smoke, You Only Look Once (YOLO) version 8 nano (YOLOv8n) has more stable and efficient predictions, making it more suitable for real-time monitoring of fire with UAVs. Finally, the challenges in the current methods for detecting fire and smoke with UAVs and deep learning are discussed, as well as suggestions for future research in these areas.