I tried to find contact information for the project, but this form is the only thing I've found.

I just tried out your project, and I'm very excited to see it! after experimenting with the face detection part, and your samples, I had the following experiences / thoughs:

• if one moves his had too fast, because the web camera blurs the image, face tracking is confused for the duration of the movement. after one stops the head movement, it re-captures the face.

I wonder - one might be able to predict the vector of movement for the uncertain period, and use that info in the meantime?

• if one moves his head out of the camera's view, basically the face detection is gone forever. even if one moves back his head, the reference points are gone

in such cases, auto re-initialization might help, basically entering a state of actively looking for a face?

• looking the application working, basically I found that the number of reference points is always decreasing - some are lost with time, but no new ones are picked up.

maybe looking for new reference points while running would be of help?

• if one turns his head left or right (not facing the camera, but looking to the side), head recognition is lost. maybe a prediction could be used in these cases as well? or maybe adaptive tracking could be used - basically re-calibrating the features for the same region, and sort of 'accepting the fact' that the visuals of the face have changed?

basically I'm trying to think of improving the tracking nature of the solution.

(I'm doing all this with a Linux laptop and a Logitech webcam using the UVC driver)

Hi Maroy,

I agree with you in lots of points. Sorry for not replying earlier (I just didn't receive the issue tracking e-mail for some reason). Most of the ideas you've given are being implemented in the SVN tree but they are still not finished, so, probably in the next release a lot of what you said might be already working.

Thanks for sharing your points

Hello,

First of all congratulation for all the work you've done. I've built the demo and the above issues are still there. How could we make it more robust?

Thanks and regards,

Hi Agaric Perdereau, we are working on a version based on Active Appearance Models. I believe it will be way more robust, but it is still going to take a while.

Best regards, Daniel

Hi, i am a final year BE student. i am working on the ehci-0.6 library for tracking the the face in fedora 13 in vmware. i am able to compile the code but couldnot run the sample programs.i am facing the same program as posted by u.The error s as follows After I compile the program cd ehci-0.6 ./configure make I am able to do the above with out any errors. Later when try running samples I am getting the below errors cd samples make ./simple2d

(:5940): GStreamer-CRITICAL **: Trying to dispose element appsink1, but it is in READY instead of the NULL state. You need to explicitly set elements to the NULL state before dropping the final reference, to allow them to clean up. This problem may also be caused by a refcounting bug in the application or some element.

(:5940): GStreamer-CRITICAL **: Trying to dispose element ffmpegcsp1, but it is in READY instead of the NULL state. You need to explicitly set elements to the NULL state before dropping the final reference, to allow them to clean up. This problem may also be caused by a refcounting bug in the application or some element.

OpenCV Error: Unspecified error (unicap: failed to get info for device ) in CvCapture_Unicap::initDevice, file /builddir/build/BUILD/OpenCV-2.0.0/src/highgui/cvcap_unicap.cpp, line 139 terminate called after throwing an instance of 'cv::Exception' Aborted (core dumped)

and but ven tried using
./samples/simple2d the error i am facing is "couldnot initialise tha capturing" can u please suggest me to overcome this kind of error. blank window opens and goes on showing the error mentioned above and ends ven given ctrl + c. the expected output to open a window and display the capture the person present in front of the webcam and print d coordinates of the face captured

If you need further more details I can provide you regarding the project and error. Thanking you niveditha

Hi niveditha, thank you for your report. Sounds like your camera isn't working inside VMWare. First thing I would do is download OpenCV (yum install opencv, or something like that) go to the samples folder and run some facedetect or camshift... If it doesn't acquire frames, the problem is probably related to VMWare camera... If it goes like this, I'd ask for some help in OpenCV forums. Please, tell us what happened =) Best regards, Daniel