svm multiclass classifier predicts same class for all objects

edit

I have 14 classes of images - these are sampled from a kinect video of 14 objects. Using HOG for feature extraction and SVM for classification. There are 20 training samples for each class. I have also tried KAZE feature extractor and Random Forest, but the same class is predicted for all. what am I doing wrong?

These are the parameters used for SVM
////create and train the svm 
//Ptr<ml::SVM> svm = SVM::create();
//svm->setKernel(SVM::RBF);
//svm->setType(SVM::C_SVC);
////Ptr<ml::ParamGrid> nogrid = ml::ParamGrid::create(0, 0, 0);//no need nu, coeff0 or p
//auto td = TrainData::create(mat, ROW_SAMPLE, labels);
//ParamGrid Cgrid = SVM::getDefaultGrid(SVM::C);
//ParamGrid pGrid = SVM::getDefaultGrid(SVM::P);
//ParamGrid gammaGrid = SVM::getDefaultGrid(SVM::GAMMA);
//ParamGrid nuGrid = SVM::getDefaultGrid(0);
//ParamGrid coeffGrid = SVM::getDefaultGrid(SVM::COEF);
//ParamGrid degreeGrid = SVM::getDefaultGrid(0);

//svm->trainAuto(td,30,Cgrid,gammaGrid,pGrid,nuGrid,coeffGrid,degreeGrid,true);

Thanks for any help