FE-YOLOv8n: A YOLOV8 Enhancement Algorithm for Effective Fire Extinguisher Detection

People often rely on optical devices to identify the source of fire in a timely manner in order to prevent fires, but it is also crucial to promptly activate firefighting equipment. At present, there are models on the market that combine neural networks such as CNN with location algorithms for real-time detection of fire extinguisher positions, which can make real-time detection and alarm through cloud services. However, there is a lack of algorithms that can accurately identify the type of fire extinguisher and specific occlusion situations. Firstly, this article proposes using a model, based on YOLOv8n, to detect several main types of fire extinguishers (which can be determined by shape). Secondly, we propose a target detection algorithm based on CNN (Convolutional Neural Network) that can effectively detect specific fire extinguishers: FE-YOLOv8. The dataset consists of CSDM, Billbill videos, and on-site images, and specific image enhancements have been made for different types of fire extinguishers.The trained model achieved a mAP of 90%, outperformed the original, and increased by 1.2 percentage points.