OpenCV altında SVM kullanmak. (OpenCV SVM Sample Code)

Bilindiği gibi OpenCV kütüphanesinin machine learning modülünde libsvm wraplenmiştir. Böylece biz OpenCV kullanıcıları harici bir kütüphane kullanmadan SVM(Support Vector Machines) i uygulamalarımızda kullanabiliriz. Aşağıda bu konuda bir örnek hazırlamaya çalıştım. Yorumlarda açıklamalar mevcuttur.

	#include<iostream>
	#include <cv.h>
	#include <cxcore.h>
	#include <ml.h>
	using namespace std;
	int main()
	{
	CvSVM SVM;
	CvMat trainData;
	CvMat labels;
	CvTermCriteria criteria = cvTermCriteria(CV_TERMCRIT_EPS,100,0.001); //eğitim için term kriterimiz.
	int array_labels[5]= {1,1,1,-1,-1}; //5 eğitim verisinin bildiğimiz etiketleri.(labels, responses).
	//ilk 3 veri 1 sınıfına, diğer iki veri ise -1 sınıfına ait.
	//eğitim verilerini diziye doldurdum. Gerçek uygulamalarda filedan okuma veya hesaplama şeklinde olur.
	//kodu basit tutmak için bu şekilde gösterdim.
	#pragma region data
	//5 egitim verisi olsun herbiri 40 tane feature içersin.
	float array_trainData[5][40] = { 1,0.443137,0.333333,0.392157,0.0313725,
	0.270588,0.870588,0.964706,1,0.439216,
	0.337255,0.380392,0.0235294,0.266667,0.862745,
	0.964706,1,0.431373,0.337255,0.368627,
	0.0156863, 0.262745, 0.854902, 0.964706, 1,
	0.423529, 0.341176, 0.360784, 0.0156863, 0.258824,
	0.847059, 0.964706, 1, 0.419608, 0.341176,
	0.34902, 0.0117647, 0.254902, 0.839216, 0.968627,
	0.572549,0.4,0.113725,0.0784314,0.137255,
	0.160784,0.00784314,0.337255,0.580392,0.407843,
	0.121569, 0.0745098, 0.133333, 0.156863, 0.00392157,
	0.341176, 0.584314, 0.411765, 0.12549, 0.0666667 ,
	0.129412, 0.152941, 0, 0.34902, 0.592157,
	0.419608, 0.129412, 0.0588235, 0.12549, 0.14902,
	0.00392157, 0.352941, 0.596078, 0.423529, 0.137255,
	0.054902, 0.121569, 0.145098, 0.00784314, 0.356863,
	0.882353, 0.741176, 0.305882, 0.305882, 0.278431,
	0.141176, 0.156863, 0.509804, 0.882353, 0.741176,
	0.305882, 0.305882, 0.27451, 0.141176, 0.14902,
	0.501961, 0.886275, 0.741176, 0.305882, 0.305882,
	0.270588, 0.141176, 0.141176, 0.498039, 0.886275,
	0.741176, 0.305882, 0.305882, 0.266667, 0.141176,
	0.137255, 0.490196, 0.886275, 0.737255, 0.305882,
	0.305882, 0.266667, 0.137255, 0.129412, 0.486275,
	0.929412, 0.917647, 0.717647, 0.356863, 0.133333,
	0.156863, 0.231373, 0.709804, 0.929412, 0.913725,
	0.721569, 0.356863, 0.137255, 0.164706, 0.219608,
	0.694118, 0.92549, 0.905882, 0.729412, 0.356863,
	0.141176, 0.168627, 0.207843,0.67451, 0.917647,
	0.901961, 0.733333, 0.360784, 0.141176, 0.172549,
	0.196078, 0.658824, 0.909804, 0.894118, 0.737255,
	0.360784, 0.141176, 0.180392,0.180392, 0.639216,
	0.760784, 0.372549, 0.407843, 0.298039, 0.160784,
	0.72549, 0.988235, 0.960784, 0.772549, 0.372549,
	0.411765, 0.305882, 0.160784, 0.709804, 0.992157,
	0.960784, 0.780392, 0.376471, 0.419608, 0.309804,
	0.160784, 0.698039, 0.992157, 0.960784, 0.784314,
	0.376471, 0.419608, 0.317647, 0.164706, 0.686275,
	0.996078, 0.964706, 0.792157, 0.376471, 0.423529,
	0.32549, 0.160784, 0.670588, 1, 0.964706
	};
	#pragma endregion data
	cvInitMatHeader(&trainData,5,40,CV_32FC1,array_trainData); //eğitim verimizi tutacak matris.
	cvInitMatHeader(&labels,1,5,CV_32SC1,array_labels); //etiketlerimizi tutacak matris.
	//classification olduğu için CV_32SC1 kullandık.
	SVM.train(&trainData,&labels,0,0,CvSVMParams(CvSVM::C_SVC,CvSVM::RBF,3,0.025,0,1,0.5,0.1,NULL,
	criteria)); //train fonksiyonumuz (libsvm ile kontrol etmek için libsvm ile
	//gelen svm-train.exe nin default parametreleri kullanıldı.)
	//Parametrelerin detaylı açıklaması için opencv docs.
	//SVM.save("test.volkan"); //eğittiğimiz modeli kaydedip daha sonra kullanabiliriz.
	for(int i = 0; i < 5; i++ )
	{
	double r;
	CvMat sample;
	cvGetRow(&trainData, &sample, i );
	r = SVM.predict( &sample ); //predict edip sonuçları kontrol edelim.
	}
	}

view raw opencvSVM.cpp hosted with ❤ by GitHub

Kod