New Member

[Marc .]

Hi Kourosh,
Be welcome to the team! (sorry for the delay).
As I could see in your message, you have a good knowledge of Face
Recognition system, can you tell us more about it?.. (btw, do you code
in Java?)

BR
Marc Rosenfeld


>>>>>>>>>>>>>

Kourosh Meshgi MSc Student @ Amir Kabir University of Technology
http://www.facebook.com/kourosh.meshgi
http://ceit.aut.ac.ir/~87131055/resume.html Thesis: Brain Inspired
Face Detectio

[Kourosh Meshgi]

Hi dear Marc and all

Thank you for your warm welcome, to be honest I didn't expect that and it was encouraging.
I'm currently studying on face detection as my MSc thesis. I'll be very active in this field for at least the upcoming year. (I'm gonna further my education in PhD hopefully on this subject again)
I don't have enough experience in Java(I have some skills on c++), but I'll gonna start it soon (my job requirement).
Currently I'm reviewing literature of face detection and brain inspired models to accommodate the detection task with it. My thesis is titled "Brain Inspired Face Detection" and I proposed to use VisNet model over this problem. This brain ispired model don't promise to have the best accuracy, but can achieve invariance over illiumination, rotation and translation. I hope I can add more invariance to it (like pose, facial expression, background clutter, self-occlussion or even occlusion)

As I read the terms of your group, I noticed that you count on Facebook as a repository for face images. But as I myself is very active there, I see tons of images with irrelevant tags (for example I tagged a flower to a friend of mine!) and tags on the body of people instead of their faces. The face RECOGNITION trust on images tags to have a face inside, so the photos need massive work to correct tags or to choose a fine gallery of pics with correct tags. Face DETECTION -that I'm working on- can help to verify whether that tag is correct or not. I have this idea in mind: to run a face detection algorithm over the image and extract face candidates, and then match them with the tags, if both voted an area as face, so it will be a face with higher certainity (a kind of information fusion mechanism like mixture of experts or so)

I'm always ready to face challenges in this domain at theoretical and experimental level. I like being here and hope we all have fruitful discussions in the group.
Wish you luck
Kourosh