This paper explores the use of recent YOLO-based models for detecting and localizing smoke and wildfires in ground and aerial images. The authors evaluate the performance of YOLOv5, YOLOv7, YOLOv8, and YOLOv5u models using the D-Fire and WSDY datasets. The D-Fire dataset includes a variety of images with smoke, fire, and smoke-like objects, while the WSDY dataset focuses solely on smoke images. The YOLOv7x model achieved the best performance with an mAP score of 80.40%, outperforming baseline models. YOLOv8s obtained a high mAP of 98.10% in identifying and localizing wildfire smoke. The lightweight models (YOLOv5n, YOLOv5s, YOLOv8n, YOLOv8s, YOLOv5nu, and YOLOv5su) demonstrated fast processing speeds and lower power consumption compared to larger models. The YOLO models effectively addressed challenging scenarios, including the detection of small smoke and fire areas, varying shapes, sizes, and intensities, complex backgrounds, and visual similarities between smoke, fog, clouds, and fire, lighting, and sun glare. The paper concludes that YOLO models are effective for real-time fire detection and can enhance wildfire management strategies, reducing environmental damage and economic losses. Future work will focus on optimizing the models for edge devices.This paper explores the use of recent YOLO-based models for detecting and localizing smoke and wildfires in ground and aerial images. The authors evaluate the performance of YOLOv5, YOLOv7, YOLOv8, and YOLOv5u models using the D-Fire and WSDY datasets. The D-Fire dataset includes a variety of images with smoke, fire, and smoke-like objects, while the WSDY dataset focuses solely on smoke images. The YOLOv7x model achieved the best performance with an mAP score of 80.40%, outperforming baseline models. YOLOv8s obtained a high mAP of 98.10% in identifying and localizing wildfire smoke. The lightweight models (YOLOv5n, YOLOv5s, YOLOv8n, YOLOv8s, YOLOv5nu, and YOLOv5su) demonstrated fast processing speeds and lower power consumption compared to larger models. The YOLO models effectively addressed challenging scenarios, including the detection of small smoke and fire areas, varying shapes, sizes, and intensities, complex backgrounds, and visual similarities between smoke, fog, clouds, and fire, lighting, and sun glare. The paper concludes that YOLO models are effective for real-time fire detection and can enhance wildfire management strategies, reducing environmental damage and economic losses. Future work will focus on optimizing the models for edge devices.