Jack L.
Jack L.

Reputation: 1335

Opencv 3 SVM training

As you may know, many things changed in OpenCV 3 (in comparision to the openCV2 or the old first version).

In the old days, to train SVM one would use:

CvSVMParams params;
params.svm_type = CvSVM::C_SVC;
params.kernel_type = CvSVM::POLY;
params.gamma = 3;

CvSVM svm;
svm.train(training_mat, labels, Mat(), Mat(), params);

In the third version of API, there is no CvSVMParams nor CvSVM. Surprisingly, there is a documentation page about SVM, but it tells everything, but not how to really use it (at least I cannot make it out). Moreover, it looks like no one in the Internet uses SVM from OpenCV's 3.0.

Currently, I only managed to get the following:

ml::SVM.Params params;
params.svmType = ml::SVM::C_SVC;
params.kernelType = ml::SVM::POLY;
params.gamma = 3;

Can you please provide me with information, how to rewrite the actual training to openCV 3?

Upvotes: 11

Views: 23383

Answers (3)

Jdban101
Jdban101

Reputation: 377

I know this is an old post, but i came across it looking for the same solution. This tutorial is extremely helpful: http://docs.opencv.org/3.0-beta/doc/tutorials/ml/introduction_to_svm/introduction_to_svm.html

Upvotes: 2

berak
berak

Reputation: 39796

with opencv3.0, it's definitely different , but not difficult:

Ptr<ml::SVM> svm = ml::SVM::create();
// edit: the params struct got removed,
// we use setter/getter now:
svm->setType(ml::SVM::C_SVC);
svm->setKernel(ml::SVM::POLY);
svm->setGamma(3); 

Mat trainData; // one row per feature
Mat labels;    
svm->train( trainData , ml::ROW_SAMPLE , labels );
// ...
Mat query; // input, 1channel, 1 row (apply reshape(1,1) if nessecary)
Mat res;   // output
svm->predict(query, res);

Upvotes: 32

DiKorsch
DiKorsch

Reputation: 1270

I was porting my code from OpenCV 2.4.9 to 3.0.0-rc1 and had the same issue. Unfortunately the API has changes since the answer was posted, so I would like to update it accordingly:

Ptr<ml::SVM> svm = ml::SVM::create();
svm->setType(ml::SVM::C_SVC);
svm->setKernel(ml::SVM::POLY);
svm->setGamma(3);

Mat trainData; // one row per feature
Mat labels;    
Ptr<ml::TrainData> tData = ml::TrainData::create(trainData, ml::SampleTypes::ROW_SAMPLE, labels);
svm->train(tData);
// ...
Mat query; // input, 1channel, 1 row (apply reshape(1,1) if nessecary)
Mat res;   // output
svm->predict(query, res);

Upvotes: 9

Related Questions