Comparative Analysis of YOLOv8 and Faster R-CNN in Fire and Smoke Detection

The increasing prevalence of wildfires poses significant risks to the environment, human life, and property. The heightened intensity and frequency of these fires underscore the need for effective early detection systems to enhance response efforts and reduce potential damage. Early detection is vital for preventing catastrophic events at the moment a fire ignites. This paper explores the utilization of Faster R-CNN and Yolo for the efficient and accurate detection of fire and smoke, leveraging advancements in machine learning and computer vision. Paper presents the Comparative Analysis of YOLOv8(small, medium, and large) variants and Faster R-CNN in fire and smoke detection. Results obtained for fire detection based on Precision, Recall, Accuracy, mAP@50, F1 score is as follows: Yolov8s(0.831, 0.856, 0.7272, 0.892, 0.8420), Yolov8m(0.817, 0.819, 0.6957, 0.864, 0.8205), Yolov8l (0.825, 0.841, 0.7264, 0.88, 0.8415) and Faster R-CNN (0.751, 1, 0.835, 0.740, 0.858). Faster R-CNN demonstrated better accuracy as compared to Yolov8.The performance of Yolo and Faster R-CNN exhibits variability due to differences in architectural design, methodologies employed for object detection, and the inherent trade-offs between speed and accuracy.