Updated SVM to allow for setting degree for POLY classification.\n Wrapped some …

…parameters to trainSVM in a CvSVMParams object to cut down on clutter.

Updated SVM to allow for setting degree for POLY classification.\n Wrapped some …
…parameters to trainSVM in a CvSVMParams object to cut down on clutter.
dgcrouse
1 parent 48ca4edd
Showing 1 changed file with 26 additions and 13 deletions
openbr/plugins/classification/svm.cpp
@@ -26,19 +26,12 @@ using namespace cv;
 namespace br
 {
-static void trainSVM(SVM &svm, Mat data, Mat lab, int kernel, int type, float C, float gamma, int folds, bool balanceFolds, int termCriteria)
+static void trainSVM(SVM &svm, Mat data, Mat lab, CvSVMParams params, float C, float gamma, int folds, bool balanceFolds)
 {
     if (data.type() != CV_32FC1)
         qFatal("Expected single channel floating point training data.");
-    CvSVMParams params;
-    params.kernel_type = kernel;
-    params.svm_type = type;
-    params.p = 0.1;
-    params.nu = 0.5;
-    params.term_crit = cvTermCriteria(CV_TERMCRIT_ITER+CV_TERMCRIT_EPS, termCriteria, FLT_EPSILON);
-
-    if ((C == -1) || ((gamma == -1) && (kernel == CvSVM::RBF))) {
+    if ((C == -1) || ((gamma == -1) && (params.kernel_type == CvSVM::RBF))) {
         try {
             svm.train_auto(data, lab, Mat(), Mat(), params, folds,
                            CvSVM::get_default_grid(CvSVM::C),
@@ -83,6 +76,7 @@ class SVMTransform : public Transform
     Q_PROPERTY(int termCriteria READ get_termCriteria WRITE set_termCriteria RESET reset_termCriteria STORED false)
     Q_PROPERTY(int folds READ get_folds WRITE set_folds RESET reset_folds STORED false)
     Q_PROPERTY(bool balanceFolds READ get_balanceFolds WRITE set_balanceFolds RESET reset_balanceFolds STORED false)
+    Q_PROPERTY(int degree READ get_degree WRITE set_degree RESET reset_degree STORED false)
 public:
     enum Kernel { Linear = CvSVM::LINEAR,
@@ -107,6 +101,7 @@ private:
     BR_PROPERTY(int, termCriteria, 1000)
     BR_PROPERTY(int, folds, 5)
     BR_PROPERTY(bool, balanceFolds, false)
+    BR_PROPERTY(int, degree, 2)
     SVM svm;
     QHash<QString, int> labelMap;
@@ -127,8 +122,16 @@ private:
             QList<int> dataLabels = _data.indexProperty(inputVariable, labelMap, reverseLookup);
             lab = OpenCVUtils::toMat(dataLabels);
         }
-
-        trainSVM(svm, data, lab, kernel, type, C, gamma, folds, balanceFolds, termCriteria);
+        
+        CvSVMParams params;
+        params.kernel_type = kernel;
+        params.svm_type = type;
+        params.p = 0.1;
+        params.nu = 0.5;
+        params.term_crit = cvTermCriteria(CV_TERMCRIT_ITER+CV_TERMCRIT_EPS, termCriteria, FLT_EPSILON);
+        params.degree = degree;
+        
+        trainSVM(svm, data, lab, params, C, gamma, folds, balanceFolds);
     }
     void project(const Template &src, Template &dst) const
@@ -192,6 +195,7 @@ class SVMDistance : public Distance
     Q_PROPERTY(int termCriteria READ get_termCriteria WRITE set_termCriteria RESET reset_termCriteria STORED false)
     Q_PROPERTY(int folds READ get_folds WRITE set_folds RESET reset_folds STORED false)
     Q_PROPERTY(bool balanceFolds READ get_balanceFolds WRITE set_balanceFolds RESET reset_balanceFolds STORED false)
+    Q_PROPERTY(int degree READ get_degree WRITE set_degree RESET reset_degree STORED false)
 public:
     enum Kernel { Linear = CvSVM::LINEAR,
@@ -212,6 +216,7 @@ private:
     BR_PROPERTY(int, termCriteria, 1000)
     BR_PROPERTY(int, folds, 5)
     BR_PROPERTY(bool, balanceFolds, false)
+    BR_PROPERTY(int, degree, 2)
     SVM svm;
@@ -235,8 +240,16 @@ private:
             }
         deltaData = deltaData.rowRange(0, index);
         deltaLab = deltaLab.rowRange(0, index);
-
-        trainSVM(svm, deltaData, deltaLab, kernel, type, -1, -1, folds, balanceFolds, termCriteria);
+        
+        CvSVMParams params;
+        params.kernel_type = kernel;
+        params.svm_type = type;
+        params.p = 0.1;
+        params.nu = 0.5;
+        params.term_crit = cvTermCriteria(CV_TERMCRIT_ITER+CV_TERMCRIT_EPS, termCriteria, FLT_EPSILON);
+        params.degree = degree;
+
+        trainSVM(svm, deltaData, deltaLab, params, -1, -1, folds, balanceFolds);
     }
     float compare(const Mat &a, const Mat &b) const