Is it possible to get more than 5 (possibly all 68, like in dlib) facial landmark units using Facenet?
I don't think so. The second vector returned by this detect_face has only 10 points.
There are five landmarks detected by MTCNN and these are left eye, right eye, nose, left mouth corner, and right mouth corner. It would not be straight forward to detect a larger number of landmarks.
Thanks for the response @davidsandberg and @achbogga
Any idea if it will be possible to detect emotions from only 5 facial landmarks?
@mayank26saxena you could train on a picture of the whole face instead of the landmarks
Most helpful comment
There are five landmarks detected by MTCNN and these are left eye, right eye, nose, left mouth corner, and right mouth corner. It would not be straight forward to detect a larger number of landmarks.