NICHOLAS KEY

Quote of the day:

Sunday, October 19, 2008

Facial features extraction

Now you see me ... now you don't ... and now you see me again :P



If I understand technically how this actually works, the image recognition process involves facial features extraction. The minimum number of feature points is 2 - the reference points at your eyeballs. Well, of course, the more reference points you provide, the higher the accuracy you'll get.

I have this image with 19 reference points, just to substantiate my theory.


Coincidentally, I said coincidentally, all the 19 reference points matches. This includes the corners of my mouth, eyebrows, nose and chin. Depending on what pictures MyHeritage has in their database, the algorithm will find a match to my picture. So I was matched to Chow Yun-Fat - pictorially, not genetically.

Anyway, both of us are living in two separate worlds. I'm a software development engineer with MS in Computer Science from University of San Francisco working to make ends meet while Chow Yun-Fat is a famous actor. Not forgetting the huge difference of our gross salaries :)

To my readers, you may try this out at MyHeritage Celebrity Morph. I particularly like how they manipulate our pictures. For laymen, it's called morphing your pictures into your favorite artists, but for a curious computer science geek like me, it's called AI techniques put into use.

2 Comments:

  • Actually one of my undergraduate friends know the owner of that site - MyHeritage. It is pretty cool.

    By Blogger Unknown, at Monday, October 20, 2008 12:51:00 AM  

  • @shuchun: yupe, it is cool. otherwise i wouldn't write about this app :D

    i'm not sure if the platform engineer(s) at MyHeritage have the same thought as mine about extracting the facial features.

    Probably you can introduce me to him so that we can geek about what happens behind the scene.

    but anyway, thank you for dropping by!

    By Blogger Nicholas Key, at Monday, October 20, 2008 12:57:00 AM  

Post a Comment

<< Home