Mask detection device and method based on improved YOLOv8

The invention relates to an improved YOLOv8-based mask detection device and method, and belongs to the technical field of image processing, and the method comprises the steps: A, building an image data set containing a human face and a mask, and carrying out the marking of the position and category of the mask; b, adopting Mosaic data enhancement to increase the number of data sets with different levels; the corresponding mask position marks are combined and adjusted according to the same mode; c, the YOLOv8 model is improved, a Slim-Neck structure is adopted in a neck network, a DyHead detection head is adopted in a detection layer, and a YOLOv8-SLIM-DYHEAD model is constructed; step D, a YOLOv8-SLIM-DYHEAD model is trained by using the data set after data enhancement, model parameters are optimized by minimizing a loss function, and an optimal YOLOv8-SLIM-DYHEAD model is obtained; e, data to be detected are input into the trained YOLOv8-SLIM-DYHEAD model, and the wearing condition of the mask is detected; a