refineDetectedMarkers return different arrays depending on if new markers were found #2720
Description
System information (version)
- OpenCV => 4.5.0-1 (via Python 3)
- Operating System / Platform => Arch Linux x64
- Compiler => Installed via Arch pacman
Detailed description
The function refineDetectedMarkers returns different lists depending on whether it finds new markers or not. The following list are the detectedIds as they are before calling refineDetectedMarkers:
[array([[[2189.961 , 4129.8223],
[2423.625 , 4104.6997],
[2496.3433, 4220.5117],
[2247.3022, 4247.493 ]]], dtype=float32), array([[[1983.1108, 4056.9077],
[2162.0933, 4126.974 ],
[2000.5653, 4216.961 ],
[1821.8195, 4140.437 ]]], dtype=float32), array([[[1980.4174, 3944.8262],
[1972.7639, 4044.8574],
[1745.5054, 4043.5725],
[1766.1394, 3943.3354]]], dtype=float32), array([[[2447.0383, 4095.2173],
[2535.8245, 3999.946 ],
[2756.3477, 4044.2703],
[2676.597 , 4145.6904]]], dtype=float32)]
And the following is the list returned:
[array([[[2189.961 , 4129.8223]],
[[2423.625 , 4104.6997]],
[[2496.3433, 4220.5117]],
[[2247.3022, 4247.493 ]]], dtype=float32), array([[[1983.1108, 4056.9077]],
[[2162.0933, 4126.974 ]],
[[2000.5653, 4216.961 ]],
[[1821.8195, 4140.437 ]]], dtype=float32), array([[[1980.4174, 3944.8262]],
[[1972.7639, 4044.8574]],
[[1745.5054, 4043.5725]],
[[1766.1394, 3943.3354]]], dtype=float32), array([[[2447.0383, 4095.2173]],
[[2535.8245, 3999.946 ]],
[[2756.3477, 4044.2703]],
[[2676.597 , 4145.6904]]], dtype=float32), array([[[2173.7258, 3877.364 ]],
[[1994.933 , 3934.5142]],
[[1874.378 , 3861.4165]],
[[2037.4222, 3808.2634]]], dtype=float32)]
As refineDetectedMarkers does not modify the list if no new markers are found, this means that the return value of refineDetectedMakers can be either [array([[4*[x, y]]])] or [array([4*[[x, y]]])].
Steps to reproduce
para = aruco.DetectorParameters_create()
para.cornerRefinementMethod = aruco.CORNER_REFINE_SUBPIX
det_corners, det_ids, rejected = aruco.detectMarkers(image, someMarkers, parameters=para)
coords = np.array([[np.append(y, [1]) for y in x] for x in markers.values()], 'float32')
ids = np.array([[x] for x in markers.keys()])
board = aruco.Board_create(coords, lightTriggerMarkers, ids)
new_det_corners, new_det_ids, new_rejected, recovered = aruco.refineDetectedMarkers(gray, board, det_corners[:-1], det_ids[:-1], rejected + det_corners[-1:], parameters=para, errorCorrectionRate=3., checkAllOrders=False)
print(f"Corners changed from {det_corners} to {new_det_corners}")
Issue submission checklist
- I report the issue, it's not a question
- I checked the problem with
documentation, FAQ,open issues,
answers.opencv.org, Stack Overflow, etc and have not found solutiononly since a problem with inconsistent return types is obviously a bug - I updated to latest OpenCV version and the issue is still there
- There is reproducer code and related data files: videos, images, onnx, etc Look, I'm sorry. I've worked around the issue in my code, I'm just here to help you and just finding how python libraries are generated, to find the correct repository to report to, and write this report, has taken way too long already. The code above is usable if you provide an image with aruco markers and load it into the code.