How to train opencv with xm2vts database

asked 2014-04-08 02:21:24 -0500

nqthiep gravatar image

updated 2014-04-08 03:56:13 -0500

Hi all,

I had a xm2vts face database, that is used for facial point detect. But i don't know to use it for detect facial point with opencv.

Anyone can help me, please.

I found this article (http://ibug.doc.ic.ac.uk/resources/facial-point-annotations/) about xm2vts, which used for detect feature point of the face. But this article does not just how to use xm2vts with opencv. If anyone know xm2vts with opencv, plese help me.

edit retag flag offensive close merge delete

Comments

could you give some more details , what you have and what you're trying to do ?

also, opencv does not have any any 'readymade' facial landmark functions

berak gravatar imageberak ( 2014-04-08 02:37:44 -0500 )edit

like berak said, openCV doesn't provide this functionality yet and looking at the database, it seems their aren't any markers on the face images themselves, so you will need to detect the features yourself. Start with face detection, then upscale the region and do eye and mouth detection (all using the provided cascade classifiers). Then use the location of the detection to define facial features and landmark points!

StevenPuttemans gravatar imageStevenPuttemans ( 2014-04-08 03:02:31 -0500 )edit
1

there's this but afaik, it trains muct76 models(though there's a asm68 one in the data)

also this but that's using the arcane c-api (and never compiled for me)

and ofc, flandmark (again, different featureset)

berak gravatar imageberak ( 2014-04-08 04:15:40 -0500 )edit

Thanks berak & StevenPuttemans very much. I tried flandmark, but not ok. I want to do same http://ibug.doc.ic.ac.uk/resources/facial-point-annotations, have you any way to do same that?

nqthiep gravatar imagenqthiep ( 2014-04-08 04:32:19 -0500 )edit