Face Mask Wearing Detection Based on YOLOv5

Abstract In recent years, COVID-19 has swept the world, and people in crowded public places are usually large. order to reduce risk of virus transmission, stop spread epidemic cross-infection, wearing masks correctly become an important measure prevent virus. Aiming at time-consuming laborious situation manually, this paper proposes a mask detection method based on yolov5. The input layer is mainly used for mosaic data enhancement, that is, adaptive anchor box image scaling technology; Yolov5 backbone adopts focus CSP (cross stage partial) structure; neck spp (spatial pyramid pooling) module FPN (feature networks) + pan (pixel aggregation network) output ciou bounding loss function_Loss average index NMS (non maximum suppression). This uses 8000 preprocessed images as set trains 200 epochs get final model. algorithm visually displays training test results through tensor board, inputs pictures captured by camera into model detect whether face wears mask. accuracy, recall mean accuracy (map) 94.8%, 89.0% 93.5% respectively, which higher than yolov3 yolov4 algorithms.