About face prediction

asked 2015-06-03 14:37:50 -0600

lechautu1993 gravatar image

Hi, i'm a senior student, researching about OpenCV to create face recognition program. The program is running perfectly but i want to understand more about it. When i call Predict() to recognize a given face using its calculated LBP histogram, i wish to know how it works. i think that it calculates distances from histogram of given face to each histogram in database, compares the smallest distance with a threshold to return -1 or id of the closest histogram in database. Does it really calculate all distances linearly and greedily? In case i have thousands or millions faces in database, it would take a long time to return result if it uses that method. Thanks.

edit retag flag offensive close merge delete

Comments

I know nothing about face prediction but in opencv when you watch source file or doc you can find some bibliography. For example In opencv_contrib\modules\face\src\lbph_faces.cpp at line 30:

namespace cv { namespace face {

// Face Recognition based on Local Binary Patterns.
//
//  Ahonen T, Hadid A. and Pietikäinen M. "Face description with local binary
//  patterns: Application to face recognition." IEEE Transactions on Pattern
//  Analysis and Machine Intelligence, 28(12):2037-2041.

I hope it helps you.

LBerger gravatar imageLBerger ( 2015-06-03 14:55:32 -0600 )edit
2

"Does it really calculate all distances linearly " - yes.

berak gravatar imageberak ( 2015-06-04 00:40:37 -0600 )edit
add a comment