4.Opencv svm生成是否戴眼镜的模型

使用svm生成是否戴眼镜的模型，感觉效果还不错。

具体实现的代码如下：

#include <iostream> #include <opencv2/core.hpp> #include <opencv2/imgproc.hpp> #include "opencv2/imgcodecs.hpp" #include <opencv2/highgui.hpp> #include <opencv2/ml.hpp>   using namespace cv; using namespace cv::ml; using namespace std;     int main(int argc ,char ** argv) {     /////////////////     string filename = argv[1];////the name of cvs file     string testFileName = argv[2];       //--------------------- 1. Set up training data randomly ---------------------------------------     /////////get trainData and labels from cvs file     Ptr<TrainData> Samples;     Samples = TrainData::loadFromCSV(filename.c_str(), 0);     Mat trainDataGet  = Samples->getTrainSamples();     Mat labelsGet   = Samples->getTrainResponses();     Mat trainData(trainDataGet.rows, trainDataGet.cols, CV_32FC1);     Mat labels(labelsGet.rows, labelsGet.cols, CV_32SC1);          trainDataGet.convertTo(trainData, CV_32FC1);     labelsGet.convertTo(labels, CV_32SC1);     printf("\n%d,%d,%d\n", labels.rows, labels.cols, labels.channels());     printf("\n%d,%d,%d\n", trainData.rows, trainData.cols, trainData.channels());   //    cout<<labelsGet<<endl;   //  cout<<labels<<endl;         //------------------------ 2. Set up the support vector machines parameters --------------------     //------------------------ 3. Train the svm ----------------------------------------------------     cout << "Starting training process" << endl;     //! [init]     Ptr<SVM> svm = SVM::create();     svm->setType(SVM::C_SVC);     svm->setGamma(3);     svm->setC(0.1);     svm->setKernel(SVM::LINEAR); svm->setTermCriteria(TermCriteria(TermCriteria::MAX_ITER, (int)1e7, 1e-6));     //! [init]     //! [train]     svm->train(trainData, ROW_SAMPLE, labels);     //! [train]     cout << "Finished training process" << endl;       //--------------------- 4. Set up predict  data randomly ---------------------------------------     /////////get trainData and labels from cvs file     Ptr<TrainData> SamplesTest;     SamplesTest = TrainData::loadFromCSV(testFileName.c_str(), 0);     Mat TestDataGet  = SamplesTest->getTrainSamples();     Mat TestData(TestDataGet.rows, TestDataGet.cols, CV_32FC1);       TestDataGet.convertTo(TestData, CV_32FC1);     printf("\n%d,%d,%d\n", TestData.rows, TestData.cols, TestData.channels());       //------------------------ 5. predict and print results  ----------------------------------------     for(int i = 0; i < TestData.rows; i++)     {         float response = svm->predict(TestData.row(i));           printf("*****%d********%f   \n", i,  response);     }     //------------------------ 6. save model  ----------------------------------------     FileStorage  modelFile("./svm.yml", FileStorage::WRITE);     svm->write(modelFile);       waitKey(0); }

 

其中，第一个入参是训练集的文件名，第二个入参是测试集的文件名。