FireSense: AI-Based Wildfire and Smoke Detection in Forests

Forests are an essential part of the environment which plays a vital role in sustaining ecological balance and providing habitat to different wildlife. Wildfires are one of the serious threats to forests which create a devastating impact on the natural ecosystem. Early identification of smoke and wildfires will help to reduce damage and support a quick reaction. Earlier, wildfire and smoke detection in forests were very difficult due to the challenging forest terrain, low light conditions, and difficult weather conditions. With the development in technology, several computer vision methods are combined with satellite and Unmanned Aerial Vehicle (UAV) data to correctly identify wildfires, but accuracy is still a concern with these models. This paper introduced a novel method to detect wildfire and smoke in forests utilizing fine-tuned YOLOv11n using an aerial vehicle dataset. The mAP of the fine-tuned YOLOv11n is 88.6%, with a precision of 100% and an F1-score of 85%. The number of parameters has decreased by 30% compared to its previous models, and the model size is only 5.5MB.