Aruco pose estimation in Python differs from C++

[matialmar]

System information (version)

• OpenCV => 4.5.5
• Operating System / Platform => Ubuntu 20.02
• Compiler => G++

Detailed description

I performed pose estimation with Arucos in Python3 with OpenCV 4.5.5 (installed via pip) and works as expected.
Rewriting the code in C++ with OpenCV 4.5.5 (built from source) the axes are flipped and translated from the center.
The corners are detected as expected in both Python and C++

I cant provide the full image (IP reasons) but i can provide a sample showcasing the error.

The camera matrix and distortion coefficients are the same in both.
Camera matrix:
[1256.90057913919, 0, 879.4298086122573;
0, 1254.48702971061, 504.1243816908799;
0, 0, 1]

Dist coeffs:
[-0.0552371547544854, -0.09931921582386366, 0.0007686659523849842, -0.002406439165687887, 0.02099621803945325]

Expected output (from Python):

Resulting output (C++)

Steps to reproduce

Python code:

# Ommiting the code for reading the image and camera calibration

dictionary = cv2.aruco.Dictionary_get(ARUCO_DICT["DICT_6X6_250"])

parameters = cv2.aruco.DetectorParameters_create()
parameters.cornerRefinementMethod = cv2.aruco.CORNER_REFINE_SUBPIX

(corners, ids, _) = cv2.aruco.detectMarkers(image, dictionary, parameters)
rvec, tvec, _ = cv2.aruco.estimatePoseSingleMarkers(corners, 0.1, mtx, dist)

cv2.aruco.drawDetectedMarkers(image,corners,ids)
for i in range(0, len(ids)):
    cv2.aruco.drawAxis(image, mtx, dist, rvec[i], tvec[i], 0.05) 

# Ommiting the code for displaying the image

C++ Code

C++ code

std::vector< int > ids;
std::vector< std::vector<cv::Point2f> > corners;
std::vector< cv::Vec3d > tvecs;
std::vector< cv::Vec3d > rvecs;

cv::Ptr<cv::aruco::Dictionary> dictionary = cv::aruco::getPredefinedDictionary(cv::aruco::DICT_6X6_250);

cv::Ptr<cv::aruco::DetectorParameters> parameters  = cv::aruco::DetectorParameters::create();
parameters->cornerRefinementMethod = cv::aruco::CORNER_REFINE_SUBPIX;

// Ommiting the code for reading images and camera calibration

cv::aruco::detectMarkers(image, dictionary, corners, ids, parameters);
cv::aruco::estimatePoseSingleMarkers(corners, 0.1, cameraMatrix, distCoeffs, rvecs, tvecs);
    
cv::aruco::drawDetectedMarkers(image, corners, ids);
for(int i=0; i< ids.size(); i++)
    cv::drawFrameAxes(image, cameraMatrix, distCoeffs, rvecs[i], detection.tvec[i], 0.05);

// Ommiting the code for displaying the image

Issue submission checklist

• I report the issue, it's not a question

• I checked the problem with documentation, FAQ, open issues,
  forum.opencv.org, Stack Overflow, etc and have not found any solution

• I updated to the latest OpenCV version and the issue is still there

• There is reproducer code and related data files: videos, images, onnx, etc

[alalek]

Original unmodified image is necessary for reproducer.

[matialmar]

Original unmodified image is necessary for reproducer.

Sorry for the delay. Here is the unmodified image.

[asmorkalov]

Most probably it's duplicate for #3252. Alexander Panov already prepared fix for that in #3256