Recognises the outline of the human head in the input image, including the face, hair and ears, and hairpiece area, without the neck, returning a transparent image (4 channels) with only the head area visible, suitable for single person scenes, and multiple person scenes. The input of images with a more visible portrait will work better.