Question

I couldnt find proper references about how to pass the SVM parameters to the opencv cvSvm class. I got the parameter list, but not any tutorial on how to convert the parameters to the CV data structures.

Can I have some help please?

Answer 1

Construct a CvSVMParams object.

Its members are:

• svm_type: SVM algorithm to use
• kernel_type: SVM kernel form
• degree: Degree for polynomial kernel
• gamma: Scale for polynomial, RBF or sigmoid kernel
• coef0: Offset for polynomial or sigmoid kernel
• C: C ratio for selecting support vectors
• nu: nu value for nu-SVR algorithm
• p: p value for eps-SVR
• class_weights: Class weights for C-SVM
• term_crit: Termination criteria in CvTermCriteria

Change appropriate fields and pass it to CvSVM constructor. An SVM will be trained with your parameters.

CvSVMParams params ;
params.svm_type = CvSVM::C_SVC ;
params.kernel_type = CvSVM::RBF ;
params.gamma = 0.5 ;
CvSVM svm(&samples, &labels, 0, 0, params) ;

See documentation of CvSVM for more information:
http://opencv.willowgarage.com/documentation/cpp/support_vector_machines.html