c++ - Face recognition using neural networks -

i doing project on face recognition, have used different methods eigenface, fisherface, lbp histograms , surf. these methods not giving me accurate result. surf gives matches exact same images, need match 1 image it's own different poses(wearing glasses,side pose,if covering face) etc. lbp compares histogram of images, i.e., color informations. when there high variation on lighting condition not showing results. heard neural networks, don't know that. possible train system accurately using neural networks. if possible how can that?

according this opencv page, there seem support machine learning. being said, support seem bit limited.

what do, to:

1. user opencv extract face of person.
2. change image grey scale.
3. try manipulate face same size.

all above should doable opencv (could wrong, haven't messed opencv in while) should save time.

next, take image, bitmap maybe, , feed bitmap vector neural network. alternatively, @matthiasb recommended, feed features instead of individual pixels. simplify data being passed, making network easier train.

as training, manipulate these images above, , feed them network. if person uses glasses occasionally, have cases of same person , without glasses, etc.