Problems with drawChessboardCorners

[uwesterh]

When calling drawChessboardCorners via cv.drawChessboardCorners{image=infileIR, patternSize=patternGeom, corners=resultsIR, patternWasFound=foundPointsIR}, I get the following error:

/home/user/torch/install/bin/luajit: /home/user/torch/install/share/lua/5.1/cv/init.lua:328: attempt to call method 'cdata' (a nil value)
stack traceback:
    /home/user/torch/install/share/lua/5.1/cv/init.lua:328: in function 'prepare_for_wrapping'
    /home/user/torch/install/share/lua/5.1/cv/init.lua:337: in function 'wrap_tensor'
    /home/user/torch/install/share/lua/5.1/cv/calib3d/init.lua:447: in function 'drawChessboardCorners'

Since other fuction like findCirclesGrid or stereoCalibrate work fine with the same parameters, I suspect a bug in the wrapper code. Can you check this, please?

[ainur699]

drawChessboardCorners is working, you can check in demo,
i think ptoblem is that image or corners are nil value

[shrubb]

@uwesterh
What are argument types? Could you give us the context or a short code sample reproducing the problem? Looks like image or corners aren't nil but some of them isn't a Tensor

[uwesterh]

OK, here is a minimal code example:

cv = require 'cv'
require 'cv.highgui'
require 'cv.calib3d'
require 'cv.imgcodecs'
require 'cv.features2d'


local irImg=cv.imread{"a_image_with_asymmetric_circles_pattern.png"}

 cv.imshow{"A", irImg}
 cv.waitKey{-1}

local patternType = cv.CALIB_CB_ASYMMETRIC_GRID
local patternGeom = {height=11, width=4}    -- size of the pattern: x, y

local foundPointsIR, resultsIR=cv.findCirclesGrid{image=irImg, patternSize=patternGeom, flags=patternType}

print("Types: "..resultsIR:type()..", "..irImg:type()..", pattern found: "..tostring(foundPointsIR))
cv.drawChessboardCorners{image=irImg, patternSize=patternGeom, corners=resultsIR, patternWasFound=foundPointsIR}

The image is shown nicely and it looks like the pattern is found, at least foundPointsIR is true. The error message is slightly changed but the problem remains:

>$ th test.lua 
Types: torch.FloatTensor, torch.ByteTensor, pattern found: true 
/home/user/torch/install/bin/luajit: /home/user/torch/install/share/lua/5.1/cv/init.lua:328: 'struct SizeWrapper' has no member named 'cdata'
stack traceback:
    /home/user/torch/install/share/lua/5.1/cv/init.lua:328: in function 'prepare_for_wrapping'
    /home/user/torch/install/share/lua/5.1/cv/init.lua:337: in function 'wrap_tensor'
    /home/user/torch/install/share/lua/5.1/cv/calib3d/init.lua:447: in function 'drawChessboardCorners'
    test.lua:20: in main chunk
    [C]: in function 'dofile'
    ...user/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:145: in main chunk
    [C]: at 0x00406670

[shrubb]

Very strange. I get correct results with this code even when foundPointsIR is false, with different shots of this pattern. Can you post your image here or email it to me?

[uwesterh]

I was using torch-opencv b445dd7 from Feb 17 15:12:30 2016 +0300. After pulling the two most recent commits, drawChessboardCerners works as expected (with the same input image and code from my side).
So the issue is resolved. Thanks.