View article

This paper presents a comprehensive evaluation of YOLO architectures for smoke and wildfire detection, including YOLOv5, YOLOv6, YOLOv7, YOLOv8, and YOLO-NAS. We aim to assess their effectiveness in early detection of wildfires. The Foggia dataset is used for this, and performance metrics such as Recall, Precision, F1-score, and mean Average Precision are employed. Our methodology trains each architecture for 300 epochs, focusing on recall for its relevance in this area. The ‘best models’ are evaluated on the Foggia test set and further tested with a challenging, custom-assembled dataset from independent online sources to assess real-world performance. Results show that YOLOv5, YOLOv7, and YOLOv8 exhibit a balanced performance across all metrics in both validation and testing. YOLOv6 performs slightly lower in recall during validation but achieves a good balance on testing. YOLO-NAS variants …