Face detection technology and quadric surfaces

[Yourmother]

Hello INTP'ers, how would you answer the following question?

Face detection technology employs curvature analysis and uses derivatives of surfaces to evaluate determining features. Pick the part of your face that has the largest curvature, and then decide which quadric surface best approximates it. Justify your answers.

http://cgit.nutn.edu.tw:8080/cgit/PaperDL/KYC_100330172044.PDF is the article linked to this. I would probably think that the chin is the steepest part of the body and could be represented by a hyperbolic paraboloid, but I could be wrong.

 

[wanderingSpark]

The chin would be a hyperbolic paraboloid. BUT - the chin is not the highest area of contrast (making it difficult to track) on the face. A face is not all chin; a mapped feature needs to be put into context with other features to make a face. Chins also vary too much from person to person.. However, the chin could be used as a secondary qualifier after a face has been detected to determine whose face it is.

The standard, including the method in the link, is to look for the eyes and nose.
When the eyes and nose are looked at as a set they have a larger curvature than the chin.

If you want to explore the training models to find chins and have any programming experience I can point you to openCV and http://docs.opencv.org/2.4/doc/user_guide/ug_traincascade.html

edit: Also, think about how consistently can the feature be tracked over time. From the perspective of the camera, transformations on the eyes and nose are less than changes the chin can go through.