if you don't get an error message, then everything should be working OK. If not, here's a few things you may want to check: Have you tested it with the example image I provided? Are the faces in your images large enough (about 25x25 pixels)? Are they somewhat evenly illuminated and frontal (not rotated, no side views)? Have you tried cranking up the detection threshold?
bhupala: the library cannot detect eyes and nose directly. but the face detection result should be a good starting point for where to look for eyes, nose, etc.
james: thanks! the algorithm uses a cascaded classifier, however, unlike viola/jones, it does not use integral images or haar wavelets. i hope i will manage to write up something soon. when that happens, i'll definitely post a link in here. :)
Using the values stored in 'r', i have cut out the detected faces individually from some rgb images, and i am getting all faces in grayscale which is obvious because of the rgb2gray function i have used..
But is there any way to get those cut out faces in rgb ?? How to retain its original rgb composition after running the fdlibmex ??