notes

2016-12-9: validated with python2 on f25. not bothering with python3 at this stage.
need to run vncserver as user on robot (no automation of running this up).
waitforkey requires the input focus to be the picture, or nothing happens.

the pip install of bob.ip.flandmark did fail with missing stdlib.h, but code appears to mostly work

after last image get a segfault.

Some really good examples of how poor AI is.
---
this is a python interface to flandmark (I think that I may need a clandmark i/f). I worked with it on robot
spin up a vnc session and then run python y.py and it identifies some of the landmarks on a picture of winona rider (?)

I also cloned a more recent version of opencv to make it work. not tested whether it works with the rpm or not