Question

我需要一些基于svm的分类器的帮助。我正在尝试从图像计算HOG功能并使用它们来训练svm。现在我有一个矢量＆lt;矢量＆gt;包含每个图像的要素和行的列。为了训练CvSVM，我需要一个具有这些功能的Mat矩阵。如何将矢量矢量转换为具有相同形状的Mat？

vector<vector<float>> totFeaturesVector;
for all images:
    vector<float> featuresVector;
    //populate featuresVector with 3780 floats...
    totFeaturesVector.push_back(featuresVector);
end for.
//numCols = 3780 numRows = 6.   6 images with 3780 features each. 

//Need to convert totFeaturesVector to 
//Mat training_mat(overallSamples,numCols,CV_32FC1); Something like this.

Answer 1

假设final_output是6x3780 Mat

for(int i = 0; i < height; i++)
{
    for(int j = 0; j < width; j++)
    {
        final_output.at<float>(i,j) = vector[i][j];
    }
}

Answer 2

vector<vector<float>> totFeaturesVector;
Mat_<float> M;
for (const auto & v: totFeaturesVector)
{
    Mat_<float> r(v), t=r.t(); // you need to do this
    M.push_back(t); // because push_back(Mat_<float>(v).t()) does not work
}

opencv ::从向量传染媒介的席子

2 个答案: