Open GoogleCodeExporter opened 9 years ago
Thanks for the experiments. The ASM fitting algorithm is an iterative
procedure. So you're absolutely right the initial guess is important.
I guess a possible solution might be:
1. Locate the position of mouth and eyes first (which is probably doable with the Android library), then we can start with a better guess.
2. Or, start in several random positions, run the fitting and choose the best fitting result.
Do you like to try that?
Original comment by cxcxcxcx@gmail.com
on 23 Oct 2012 at 5:19
I can locate the corners of the mouth and eyes and tip of the nose using the
Flandmark facial landmark detector library, which seems very robust. How would
I go about using this information to initialise the iterative procedure in ASM
to get improved results?
Original comment by willx...@gmail.com
on 13 Mar 2013 at 7:13
There is no easy way, I will think about it.
Original comment by cxcxcxcx@gmail.com
on 3 Apr 2013 at 6:38
Have you found any solution for this matter ?
Original comment by yasser.a...@rizze.com
on 26 May 2014 at 4:00
Original issue reported on code.google.com by
av...@pixtr.me
on 22 Oct 2012 at 2:27Attachments: