Java wrappings for vpImageConvert

[failiz]

Hi, I am planning to use Visp for Apriltag detection in java and I saw that there are no wrappings for vpImageConvert. My intention is to use this class to convert a frame obtained through opencv (Mat) to VpImageUChar.

Is there an easy way to make them or documentation about how to use the generator? I added this class to the file list and added some types in the gen_dict.jason, but every time that I tried to define the Mat type I got this error:

Severity    Code    Description Project File    Line    Suppression State
Error   MSB8066 Custom build for 'C:\visp-ws\visp-build-vc16\CMakeFiles\cad67428fd38b45df881ac291d5c0db4\visp_java_jar.rule;C:\visp-ws\visp-build-vc16\CMakeFiles\316d1088996a8f8c33372d6144645757\visp_java_jar.rule' exited with code 1.  visp_java_jar   C:\Program Files (x86)\Microsoft Visual Studio\2019\Community\MSBuild\Microsoft\VC\v160\Microsoft.CppCommon.targets 238 

And without this type, the generated jar states:

    //
    // C++: static void convert(Mat src, vpImage_vpRGBa dest, bool flip = false)
    //

    // Unknown type 'Mat' (I), skipping the function

Or is there another way to convert from Mat to VpImageUChar? Andres

[fspindle]

I didn't try but from you message I suggest to use cv::Mat instead of Mat if not already done

[failiz]

I already tried that and then the message does not appear, but the jar still states "Unknown type 'Mat' (I), skipping the function"

[fspindle]

From what I see, Mat to vpImage_vpRGBa should be introduced in generator/src/java/org/visp/utils/Converters.java which should ideally import org.opencv.core.Mat. But a quick test shows that org.opencv.core.Mat package is not found.

import org.visp.core.VpColVector;
import org.visp.core.VpHomogeneousMatrix;
import org.visp.core.VpImagePoint;
import org.visp.core.VpImageUChar;
import org.visp.core.VpImageRGBa;
import org.visp.core.VpCameraParameters;

import org.opencv.core.Mat;

public class Converters {

  public static VpImageRGBa Mat_to_VpImageRGBa(org.opencv.core.Mat mat) {
     org.visp.core.VpImageRGBa I;
     return I;
  }
  ...

The only solution I see for now is to implement the converter outside ViSP, typically in your app.

Maybe @s-trinh could give some hints.

[s-trinh]

Unfortunately, I have not used the ViSP Java interface since a long time.

After a quick look, I would try something like this:

• use this constructor: public VpImageUChar(byte[] array, int height, int width, boolean copyData)
• convert from OpenCV Mat to byte[] with something like this:

  byte[] imgData = new byte[(int) (img.total() * img.channels())];
  img.get(0, 0, imgData);

I don't remember how the Java binding works, so the easiest solution would be to do the conversion in Java I think.

[ViokingTung]

hi,I have the same issue,did you solve it?

[failiz]

I think I did not manage to fix it. I replace the filelist and the gen_dict.json files for the ones attached her to get a few extra functions (I forgot which ones), but not that one: visp.zip

I implemented the conversion following @s-trinh suggestion (most likely not a very efficient conversion):

    public static VpImageUChar Mat2VpImageUChar(Mat matrix) {
        int length = (int) (matrix.total() * matrix.elemSize());
        //System.out.println("Length mat1: " + length + " Size mat1: " + matrix.size());
        // Create the black and white image

        Mat mat1;
        if (matrix.type() != CvType.CV_8UC1) {
            mat1 = new Mat(matrix.size(), CvType.CV_8UC1);
            Imgproc.cvtColor(matrix, mat1, Imgproc.COLOR_RGB2GRAY);
        } else {
            mat1 = matrix;
        }

        // Creating buffer from the matrix
        length = (int) (mat1.total() * mat1.elemSize());
        //System.out.println("Length mat1: " + length + " size: " + mat1.size());
        byte buffer[] = new byte[length];
        mat1.get(0, 0, buffer);

        BufferedImage bImage = new BufferedImage(mat1.width(), mat1.height(), BufferedImage.TYPE_BYTE_GRAY);
        bImage.getRaster().setDataElements(0, 0, mat1.width(), mat1.height(), buffer);

        // Create VpImageUChar
        return new VpImageUChar(buffer, mat1.height(), mat1.width(), true);
    }

[ViokingTung]

I think I did not manage to fix it. I replace the filelist and the gen_dict.json files for the ones attached her to get a few extra functions (I forgot which ones), but not that one: visp.zip

I implemented the conversion following @s-trinh suggestion (most likely not a very efficient conversion):

    public static VpImageUChar Mat2VpImageUChar(Mat matrix) {
        int length = (int) (matrix.total() * matrix.elemSize());
        //System.out.println("Length mat1: " + length + " Size mat1: " + matrix.size());
        // Create the black and white image

        Mat mat1;
        if (matrix.type() != CvType.CV_8UC1) {
            mat1 = new Mat(matrix.size(), CvType.CV_8UC1);
            Imgproc.cvtColor(matrix, mat1, Imgproc.COLOR_RGB2GRAY);
        } else {
            mat1 = matrix;
        }

        // Creating buffer from the matrix
        length = (int) (mat1.total() * mat1.elemSize());
        //System.out.println("Length mat1: " + length + " size: " + mat1.size());
        byte buffer[] = new byte[length];
        mat1.get(0, 0, buffer);

        BufferedImage bImage = new BufferedImage(mat1.width(), mat1.height(), BufferedImage.TYPE_BYTE_GRAY);
        bImage.getRaster().setDataElements(0, 0, mat1.width(), mat1.height(), buffer);

        // Create VpImageUChar
        return new VpImageUChar(buffer, mat1.height(), mat1.width(), true);
    }

Thank you for your valuable advice, I rewrote the code with the same idea as yours, compare yours ,I know something where is wrong of my code,thank you so much,have a nice day!