libsvm库基于c++的简单使用

最新推荐文章于 2024-04-22 08:00:00 发布

SCH0

最新推荐文章于 2024-04-22 08:00:00 发布

阅读量2.4k

点赞数 1

文章标签： c++ 支持向量机机器学习

本文链接：https://blog.csdn.net/qq_41246375/article/details/121710945

版权

libsvm库是一个开源的svm多语言支持库，官网地址https://www.csie.ntu.edu.tw/~cjlin/libsvm/，github地址https://github.com/cjlin1/libsvm

1.在c++中使用Libsvm库，第一步是将github中的svm.h,svm.cpp拷贝到CMakeLists调用的第三方库位置，然后就可以在自己的代码中使用，关于CMakeLists的编写和svm原理可以自行了解。

2.训练

只是关键一部分，主要记录下相关语法调用

//SVM  基于libsvm
std::string model_path= "../MODEL/svm.xml";
void trainSVM()
{
    //定义训练集和测试集的样本数
    const int  sampleSum=18*3;

    //读取训练数据,先以模板为训练集
    boost::format fmt_train_SVM("../train/temp%d.png"); 
    std::vector<cv::Mat> train_data;
    for (size_t i = 1; i < sampleSum+1; i++)    //读取
    {
        cv::Mat data=cv::imread(  (fmt_train_SVM % i).str() ,0);
        cv::Mat dataShape=data.reshape(1,1);
        train_data.emplace_back(dataShape);
    }

    //设置标签      训练样本所属类别 
    int labels[sampleSum] = { 
        1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,
        1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,
        1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,
        // 1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,
    };
    svm_problem problem;
    problem.l=sampleSum;   //训练样本数
    problem.x = new svm_node*[sampleSum];   //特征矩阵
    problem.y = new double[sampleSum];     //标签矩阵
    for (int i=0; i<sampleSum; ++i) 
    {
        problem.x[i] = new svm_node[420*240+1]; //
        for(int j=0; j<420*240; ++j)
        {
            problem.x[i][j].index = j + 1;
            problem.x[i][j].value = train_data[i].at<uchar>(0,j);
        }
        problem.x[i][420*240].index = -1;
        problem.y[i] = labels[i];
    }
    //参数
    svm_parameter param;
    param.svm_type=C_SVC;
    param.kernel_type=LINEAR;
    param.degree=3;
    param.gamma=0.01;
    param.coef0=0;
    param.cache_size=20;
    param.eps=1e-6;
    param.C=1;
    param.nr_weight=0;
    param.weight_label=nullptr;
    param.weight=nullptr;
    param.nu=0.5;
    param.p=0.1;
    param.shrinking=0;
    param.probability=0;        //改成1后预测结果差得很大

    //train
    auto model = svm_train(&problem, &param);
    //save
    if (svm_save_model(model_path.c_str(), model))
    {
        std::cerr << "Save SVM to [" << model_path << "] FAILED" << std::endl;
    } 
    else
    {
        std::cout << "Save SVM to [" << model_path << "] SUCCEED." << std::endl;
    }


}

3.测试

//返回svm,基于libsvm
int testSVM(cv::Mat singleLabelMat)
{
    //1.get model
    svm_model *model=svm_load_model(model_path.c_str());

    //mat to svm_node
    cv::Mat sam=singleLabelMat.clone();
    cv::Mat samShape= sam.reshape(1,1);
    svm_node *sample=new svm_node[420*240+1];
    for (size_t i = 0; i < 420*240; i++)
    {
        sample[i].index=i+1;
        sample[i].value=samShape.at<uchar>(0,i);
    }
    sample[420*240+1].index=-1;
    double probresut[18];
    double resultLabel=svm_predict_probability(model,sample,probresut);
    //double resultLabel=svm_predict(model,sample);

    std::cout<<"labels: "<<resultLabel<<" "<<std::endl;
    
    
    return int(resultLabel);
}