Accurate and Fast Mask Recognition Based on Multiple Color Areas Detection and Face Landmarks Locating

Using deep learning technologies to assist mask recognition in personal healthy activities detection is recom-mended. However, there are some disadvantages in directly using deep learning to make the mask recognition, such as the low recognition rate in detecting the blurry human faces far away or detecting the masks from the human side faces. For the fast moving human faces, the mask detection also has the challenge. In this paper, a novel idea about using the color analysis from the special area of the human face to detect the mask wearing, which combines the face landmarks detection, location and color detection. Using the special color analysis to detect the mask wearing will save more computed resources and get fast performance. What more, for the human face far away, this method still remains a good performance compared with the pure learning mask detection. A novel multiple areas detection and comparison method that performed well is proposed in this paper. The related code is open source in GitHub.