For a long time, understanding human faces has been the grounds for research for computer vision engineers. The first application for this research came in the form of face recognition features. To identify a face in an input image or a video frame, our algorithm should first detect the location of the face. It will then cause a bounding box to frame a face in the image, as follows:
Once we have the bounding boxes, the obvious next step is to identify facial key points with more granular details inside the boxes, for example, the position of the eyes, the nose base, the eyebrows, and so on. Identifying facial landmark points will be helpful in building applications such as virtual makeup rooms, face morphing, Augmented Reality (AR) filters, and so on.
Facial key point identification made with the dlib library looks something like the following:
