opencv - What are proper SVMParams to use for texture analysis? -

opencv - What are proper SVMParams to use for texture analysis? -

i'm working on face spoof detector using research in article called: "face spoofing detection single images using micro-texture analysis"

it's based on training svm using uniform 59-bin lbp histogram of face, training database http://parnec.nuaa.edu.cn/xtan/nuaaimposterdb_download.html. made simpler implementation using 3x3 neighborhood of entire face, because other research shows simpler analysis works improve when used in conjunction other classifiers, doing.

the problem have predictor not working @ all; homecoming 1 (i.e., real face), false photos training database itself!

my suspicion maybe svmparams not correct. proper params in implementation? have now:

svmparams.svm_type = cv::svm::c_svc; svmparams.kernel_type = cv::svm::rbf;

there no such thing proper parameter, without analyzing data. if info linearly separable can utilize linear kernel, if not rbf kernel better, or can seek others. in case need find optimum parameters. c linear kernel, c , gamma rbf kernel. looks didn't set these parameters. problem perchance parameter not kernel type, kernel parameters. please have @ this.

edit

also this , this might help you.

opencv