Multiple SVM for Face Recognition

edit

In my system, I want to show known faces are "known" otherwise it is "Unknown". I don't know whether it is possible by using one multi-class svm. That's why am using multiple svm.

Using multiclass svm is simple, just train it using more classes and more labels (1, 2, 3, ...). By the way, you need to have a negative class too (so just put some faces that are not known there). Test it, see what you'll get.

ok..but what labels are given for negative images? because +ve images are labeled as 1,2,3...

0 or 4 would do

If we move away from camera it shows some false prediction. But if we close to camera it give correct predictions. So it Is the problem of my training sets or it is not possible to grab some sort of distance using svm?

what are you using for 'features' here ?

SURF features..

Is it possible to achieve some sort of distance using SURF features?

@arya, sure. but since one image will deliver 20 surf features, and another 31, you'd need some clustering, like BOW to achieve same count for all images. then use the BOW vocabulary with the multi-class svm

@berak How to use bow with SVM? The vocabulary contains the same number of descriptors and/or features for each image?

But in my case when a person nearer to camera it identify as known but when distance increases the same person identify as unknown. Could you tell me why?

Yes. You have used too many features for recognition, so if the person is too far, the features found in the face area are not enough to realise the recognition; so the person is unknown. But be careful, there is another problem: too few features lead to confusions.

So what I do to overcome this? Do you have any suggestions or solution?

Set a minimum size of the face that will be recognized. And I think that someone who has done this can help you with some values. Otherwise you can test it yourself and detect the values.

I already set this minimum face size and maximum face size. double min_face_size=30; double max_face_size=400; these values are used at the time of detection. But I don't know where I set minimum face size at the time of recognition.

The detection is used as input of recognition, am I right? Then the minimum detection size should be larger. Just make some tests on some faces of different sizes (30, 40, 50, ...) and search for features in it. See how many are, compare with the number of features used for recognition. More: you can also match, for seeing how many features are not taken in consideration for recognition.

ok.Thank you for your reply..

Is it possible to grab some sort of distance to the decision plane using opencv-svm?

The docs are saying that if the last parameter of predict is set to true, then you will have the decision function value (the distance to the separating line)