0

What's the easiest/cheapest way of measuring the depth of facial features to the millimeter?

I was thinking two photos taken by smartphones at a static distance and angle from person.

E.g.: to infer exact protrusion of nose at each point of arch, eyes &etc.

A T
  • 13,008
  • 21
  • 97
  • 158
  • Not sure if StackOverflow is the right StackExchange site to ask, or if Robotics or Physics is. Asking here because: https://stackoverflow.com/q/22764579 – A T Aug 20 '17 at 00:56
  • 1
    define easiest / cheapest. to achieve sub-mm resolution will be tricky. just taking a few smartphone snapshots won't do. you can get down to a few millimeters with Microsoft's Kinects. But sub-mm will cost you several thousands and will only work under very controlled conditions – Piglet Aug 20 '17 at 11:08
  • Not sure what's possible for under $100, I did get side-tracked into [3D Face Reconstruction from 2D Pictures](https://www.ncbi.nlm.nih.gov/pubmed/23319167) or even [InverseFaceNet](https://arxiv.org/pdf/1703.10956.pdf) and similar which infer geometry from single images. Happy to do multiple images or even video, and can have the smartphone camera up very close to the human. You say sub-mm is unlikely to be possible at this cost, what's the finest accuracy that would be possible? - Thanks again – A T Aug 20 '17 at 15:41
  • well I guess you can get down to a few millimeters of resolution if you have enough images and know where your smartphone is. for your budget I'd go with a Kinect I guess. just google Kinect facial recognition. and do more research online. You'll find such data in any scientific paper on that topic. – Piglet Aug 20 '17 at 15:46

0 Answers0