I’d like present solution regarding Deep Machine in case of Real-time to face detection based on Yolo Model (read more) and created on Darknet(Darknet). Most of the existing face detection solutions detect the face only when the face is face-to-face directed to the camera or camcorder. In the case when the person turns the head, most systems have problems with face detection, the same applies to different angles of face positions. Other problems with detection occur – when only part of the face is exposed.
The solution below works in all the cases discussed, as shown in the pictures. The solution detect face when head is turned (different tilt angle of face) in relation to camcorder or camera, or is showed only part of face