read some images randomly using a percentage in OpenCv 3.0.0

Question

i would to devllop a SVM classifier,for that i have images, and i would to read 80% for training, and 20% for test randomly, using OpenCv 3.0.0 and C++. can you help me please. my images are in the same folder.

Answer 1

You can:

• read all files in the specified folder with OpenCV glob function. You'll save all filenames into a std::vector<std::string>,

• std::random_shuffle the vector of file names

• Keep only the first XX% names of the shuffled vector

• Do something with your images

Code:

#include <opencv2/opencv.hpp>
#include <vector>
#include <string>
#include <algorithm>

int main()
{
    // Read all .png files from the specified folder
    std::string folder = "D:/SO/img/*.png";
    std::vector<std::string> filenames;
    cv::glob(folder, filenames);

    // Random shuffle
    std::random_shuffle(filenames.begin(), filenames.end());

    // Keep 80%
    float how_many = 0.80;
    size_t N = filenames.size() * how_many;

    // Do something with N random images
    for (size_t i = 0; i < N; ++i)
    {
        // Load and show image
        cv::Mat3b img = cv::imread(filenames[i]);
        cv::imshow("Img", img);
        cv::waitKey(1);
    }

    return 0;
}