What is the correct way to train a FaceRecognizer from croped images?

asked 2013-05-15 20:23:59 -0500

obl gravatar image

I wrote a function that detects the mouths of the images and now I want to train the eigenface FaceRecognizer with that results. The problem is that the images have different width x height, if I resize the images then they will stretched, so the FaceRecognizer training will have wrong results (am I right?). How can I resize/adjust the croped region for the best possible training?

edit retag flag offensive close merge delete