Face mask detection based on Transfer learning and PP-YOLO

Abstract

Taking into account the accuracy and reasoning speed of face mask detection tasks, this paper proposes a detection model PP-YOLO-Mask based on PP-YOLO through transfer learning, data augmentation and model compression methods. In a public scene, the mask detection mAP of this model is 86.69%, and the single picture recognition speed is 11.842ms. Experimental results show that compared with YOLOv3 and FasterRCNN, the model has faster accuracy and detection speed.