Detection Of Forest Fire Using YOLOv8 Algorithm

The incidences of fire and smoke in forests can result in significant damage and even casualties. Many recent detection methods lack generalization because they are mainly designed to suit specific data types and scenes. In this project, a Fire and Smoke Detection System is developed for forests using YOLOv8, which is an object detection algorithm. The main objective is to achieve efficient and accurate identification of occurrences of fire and smoke in images or video files. It requires a heavy amount of data which must be prepared by collecting and annotating the images that contain fire or smoke. YOLOv8 is the model used for training the collected dataset along with specific class configurations set on the classes of interest. In addition to this core detection ability, it also includes required evaluation metrics to measure the Model performance. The evaluation of precision and robustness is done through precision, recall, and F1-score for fire and smoke detection. Finally, the project also focuses on developing a Graphical User Interface (GUI) to visualize the prediction. A user can feed in the images or video streams. GUI will show the predictions. Bounding boxes will surround the regions, detected as fire and smoke. The novelty of this work lies in the integration of YOLOv8 with a user-friendly interface for real-time visualization, enabling efficient detection and interpretation of fire and smoke occurrences in forest environments.