So first of all my previous post about Robin Hewitt's pdf articles and website, seemed to be a dead end when further researched, she works directly and only with C++ which can not be converted into Max MSP as I discovered recently they cannot even externally be connected...
Secondly, I have been looking through Open CV libraries because turns out I can load multiple files for recognition. The particular files are called "Haar Cascades" and they have particular algorithms which have many possibilities for each kind of shape of feature - they contain eyes, nose, mouth, profile and frontal face. These are good except when they are tracked they are done as a square when I want the nice eye cut out and mouth cut out - so that would be the next step. (example below)
There were a couple of others that were briefly viewed like softVNS, Eigenface algorithms etc.. but these were only briefly touched on and not in depth - I'm going to have to work harder on the Haar Cascades and Open CV sources and just get the eye field narrowed down - it has been done I'm just unsure on how...
Other Links:
http://www.quasimondo.com/archives/000688.php
http://www.computer-vision-software.com/blog/2009/11/faq-opencv-haartraining/
http://note.sonots.com/SciSoftware/haartraining.html
http://www.cycling74.com/docs/max5/refpages/max-ref/lcd.html
http://note.sonots.com/?cmd=htmlinsert&page=haartraining
No comments:
Post a Comment