I see a scary me!

Wow! Brilliant use of the face tracking code! I love it. I think you should add a "snapshot" image export button, like in Feedback, as I can see some people getting some awesome pictures out of this. Maybe having a gallery too.

Nice work! we had played around with the Marilena and Deface libraries, but found them to be too slow.

Not any more!

http://www.boffswana.com/news/?p=498

intelligent face-detector (part of Panasonic Lumix adv)

http://szymon.tumblr.com/post/86645652/intelligent-face-detector-part-of-panasonic-lumix

Hi, Hope you are doing well. Check out http://www.createonlinebuzz.com/. If you are interested in making money through your blog by occasionally writing reviews of products and services please log onto http://www.createonlinebuzz.com/ and register as a blogger.This is a great blog for advertisers and bloggers.

The Angrrrometer lets you see what real anger looks like by using innovative face tracking technology. The higher the level, the closer you are to being angry enough for the Angry WHOPPER®. http://www.burgerking.ca/getangry

I test your flash and the original c example in opencv. It seems that yours are better. Do you modify and improve the algorithm of opencv?

I guess that the full body cascade contains tilted zones as well which are not implemented in the as3 code yet.

I love you webpage, your tests, very usefull, thanks. Im trying to do something but I think is gonna be very difficult to achieve that. I've used the marilena class to put masks in the user face, but my challenge now is trying to record the video of that, so user can watch what is recorded. Do you think is there any chance to achive that??? any help is appreciated!!!

I see two options to record video in Flash: the hard way is to port FFMPEG with Adobe Alchemy which is possible but very hard to do. The simpler way is to record every cameraframe as a bitmapdata into an Array, then export every single frame as a jpeg or png and finally use a 3d party tool (like FFMPEG) to recombine these frames into a clip. Of course this will be a lot of data and take quite a while to encode and download (or upload if you are using a server side script to do that)

thanks a lot for your reponse. I've been thinking that one simpler possibility could be to sample the position of the mask (for example, each 0.5 sec or less) when recording, and play the raw video with the mask overlay when playing. But maybe I could find problems with synchronization... what do you think about that?

If you are not tracking a camera feed but using a video that might work, but yes, sync might be an issue.

can this marilena work with XML database? How?

One of the optimizations was to embed the xml from the cascades in a more compact form. If you want to use your own xml cascades you have to look at the format and convert it yourself. Also note that the original marilena code is just a partial port of the OpenCV classes and does not support tilted zones which means that some xml files that you find online will not work.