As promised I’m posting a related video. In this video a vector model of a human face has been done, the characteristics of the face and (to some extent) the expression may be changed by using this model. It isn’t shown if it can handle the jaw openness, and (again) still there is a lot of work to do.

The point here is that once the vector model of the face has been extracted it is possible to compare that face with others according to those dimensions. Similarly to the way people describe faces, not as a matrix of pixels but as a set of characteristics like these, usually (IMHO) mentioning only the ones that are further from the mean.

On a side note I need to say that research related with images enables cool videos, and even if graphs are meant to be represented graphically the video resulting from reasoning over RDF graphs would probably be less amusing.