To practice sculpting, I'd like to try and sculpt celebrity faces. So I thought I could use some screengrabs from movies as reference. But these are rarely ideal front or side images, but I wondered if one could still use them to synthesize some key coordinates of a face, like eye nose and ear position/distance/etc. to have an anchor during sculpting. A program that allows to mark the eyes and nose tip on various photos and that computes vertices at the proper 3d coordinates.

There is quite some theory out there (epipolar geometry, fundamental matrix, lots of funny things), for example on wikipedia, to reconstruct 3d coordinates based on photos but it seemed that one needs to know some key parameters of the camera (focal distance, or relative position of the camera used for different photos). I didnt find any approach that accepts just a set of projected 2d coordinates. I thought it should be possible in theory with enough sample points available, no?

Anyone knows a good and simple program/script that can help with deriving key 3d points from 2d photo sets?

Thanks :)