Face recognition using haartraining

asked 2013-02-21 06:16:58 -0500

Ashwin gravatar image

As the title suggests.. I have a doubt that whether we can do face recognition using Haartraining. By taking the required number of positive and negative images.. And by marking The person that we want to be detected from the positive images.. Will it work? Any suggestions Please?

edit retag flag offensive close merge delete



please clarify, if you're talking about face DETECTION (a binary, yes or no, face or not question) or RECOGNITION (which is the most similar person in the db)

haartraing is quite successful at the detection task, but if you want to abuse it for recognition, you'd need to train one cascade per person, also you'd need all the cascades in memory in the test phase. so, not a good idea.

if you want face recognition, the contrib module has 3 FaceRecognizer methods, eigen, fisher, and lbp. they're much better suited for that .

and ofc, there's a sample: samples/cpp/facerec_demo.cpp

berak gravatar imageberak ( 2013-02-21 06:30:01 -0500 )edit

Thanks man.. Now I understand the difficulty of using face recognition with haartraining.

Ashwin gravatar imageAshwin ( 2013-02-22 00:56:41 -0500 )edit