You have several options here, it depends on which target language you want to use for implementation. One option would be Flash, where you can use the Ohtsuka Masakazu face-detection library ported to AS3 , originally implemented in OpenCV .
http://www.libspark.org/wiki/mash/Marilena
This has been clarified by Mario Klingemann. You can find the project here: http://www.quasimondo.com/archives/000687.php , which includes the source code.
If the desired platform is a native web browser, without any proprietary plug-ins, the new WebRTC W3 initiative is one of the alternatives, there is one more that we will talk about later. You can find the implementation here: http://neave.com/webcam/html5/face/ .
Another solution would be to use web interfaces to communicate with the destination server, where heavy image detection analysis takes place. The whole process can be summarized in several stages:
- We access the webcam using the
getUserMedia function, which is an HTML5 function. - Then we send the webcam data to the server using websocket.
- Here we analyze the data obtained, using, for example, JavaCV / OpenCV to detect and mark any recognized face.
- We send data to the client using websockets again.
There are several options that we can use on the website:
Jetty : Provides Web sites and other integration.
OpenCV : a library that has all kinds of algorithms for manipulating images. We use their face recognition support.
JavaCV : Provides wrappers for widely used libraries by computer vision researchers. Using this java shell, we can analyze images directly with data received from Jetty.
And here's a cool implementation of the above description.
UPDATE
Currently the most impressive JS face recognition library: clmtrackr , which is a library for fitting facial models to faces in videos or images. A detailed explanation of the technology is here: http://auduno.tumblr.com/post/61888277175/fitting-faces
source share