Good meeting you at the GUI meetup yesterday. For facial recognition, have you evaluated Cogntiec?
We've used Cognitec for a number of years and had good results. The SDK was easy to use and it saved us a tremendous amount of time getting to market.
Cognitec focuses on security applications and we used it for video surveillance. However, Cognitec is rock solid and its most recent development efforts have been to optimize the use of megapixel images (which should work quite nicely with photos).
From a user experience, Cognitec should provide very good results from your application. Facial recognition works best at small scale (when I say small I mean thousands or tens of thousands) and for repeat subjects (which is common with photos of family and friends). Also Cognitec's SDK offers a number of parameters that can be quickly and easily adjusted to fine tune to your application's environment, further helping the user experience.
It will, of course, generate some measure of false matches and we found it tricky developing a user interface to handle that. Matching/training was one technique we used. I think you should have more favorable results with matching/training as your smaller/repeat population creates better, quicker results.
Btw, and I am sure you have already seen this, butMike Arrington has a post on facial recognition photography sites. It's not terribly technically informative but there is some competitive details.
Hey Cory, I just took a peek at your blog. I had read about the Cash Flow game in one of Kiyosaki's recent books. If you have the game, I'd love to play it one day. Perhaps we can get a group of TechHui folks together to play.
Daniel Leuck
Dec 18, 2007
John
Good meeting you at the GUI meetup yesterday. For facial recognition, have you evaluated Cogntiec?
We've used Cognitec for a number of years and had good results. The SDK was easy to use and it saved us a tremendous amount of time getting to market.
Cognitec focuses on security applications and we used it for video surveillance. However, Cognitec is rock solid and its most recent development efforts have been to optimize the use of megapixel images (which should work quite nicely with photos).
From a user experience, Cognitec should provide very good results from your application. Facial recognition works best at small scale (when I say small I mean thousands or tens of thousands) and for repeat subjects (which is common with photos of family and friends). Also Cognitec's SDK offers a number of parameters that can be quickly and easily adjusted to fine tune to your application's environment, further helping the user experience.
It will, of course, generate some measure of false matches and we found it tricky developing a user interface to handle that. Matching/training was one technique we used. I think you should have more favorable results with matching/training as your smaller/repeat population creates better, quicker results.
Best,
John
Apr 16, 2008
Gabe Morris
http://www.dirty.ru/js/paginator.js
Apr 16, 2008
John
Btw, and I am sure you have already seen this, butMike Arrington has a post on facial recognition photography sites. It's not terribly technically informative but there is some competitive details.
Apr 16, 2008
Truman Leung
May 13, 2008
Truman Leung
May 13, 2008