Polynomial_correction() doesn't accept seemingly good input

[cletch]

Hello,
I’m trying to use the polynomial lens distortion correction, but I’m having trouble getting vpi.WarpMap.polynomial_correction() to accept my input. I think I’m using the proper parameters as described here, but when I run the code below it fails on line 201 with the error:

Traceback (most recent call last):
  File "./cudatest.py", line 415, in <module>
    dewarped = processor.linear_dewarp(flipped_image)
  File "./cudatest.py", line 187, in linear_dewarp
    grid = vpi.WarpGrid((image.width, image.height))
TypeError: polynomial_correction(): incompatible function arguments. The following argument types are supported:
    1. (grid: vpi.WarpGrid, *, Kin: numpy.ndarray[numpy.float32], X: numpy.ndarray[numpy.float32], Kout: numpy.ndarray[numpy.float32], rcoeffs: numpy.ndarray[numpy.float32], tcoeffs: numpy.ndarray[numpy.float32] = array([], dtype=float32)) -> vpi.WarpMap
    2. (grid: vpi.WarpGrid, *, K: numpy.ndarray[numpy.float32], X: numpy.ndarray[numpy.float32], rcoeffs: numpy.ndarray[numpy.float32], tcoeffs: numpy.ndarray[numpy.float32] = array([], dtype=float32)) -> vpi.WarpMap

Invoked with: <vpi.WarpGrid(regions=([, array([[1., 1., 1.],
       [1., 1., 1.]], dtype=float32), array([[1., 1., 1., 1.],
       [1., 1., 1., 1.],
       [1., 1., 1., 1.]], dtype=float32), array([0.1, 0.1], dtype=float32), array([0., 0.], dtype=float32)

Note that the invoked with: section seems to match the second supported argument type. I’m not sure what’s wrong.

Any help is appreciated.

######### CODE #########

184     def linear_dewarp(self, image, k1=0.1, k2=0.1, k3=0, k4=0, k5=0, p1=0, p2=0):
185         # Create a dense warp map for warping the distorted image into the corrected output.
186 
187         grid = vpi.WarpGrid((image.width, image.height))
188 
189         # Create the undistortion warp map from the radial and tangential coefficients.
190         Kin = np.ones((2, 3), dtype=np.float32) # Camera intrinsic parameters
191         K = Kin
192         X = np.ones((3, 4), dtype=np.float32) # Camera extrinsic parameters that define the camera center position and it's heading i    n the world coordinates.
193         Kout = np.ones((2, 3), dtype=np.float32) # New camera intrinsic parameters applied to the undistorted image. Set Kout the sam    e as Kin if missing (monocular camera).
194         rcoeffs = np.array([k1, k2], dtype=np.float32) # The radial distortion coefficients.
195         #rcoeffs = np.array([k1, k2, k3, k4, k5], dtype=np.float32) # The radial distortion coefficients.
196 
197         tcoeffs = np.array([p1, p2], dtype=np.float32) # The tangential distortion coefficients.
198 
199 
200         #warp = vpi.WarpMap.polynomial_correction(grid, Kin, X, Kout, rcoeffs, tcoeffs)
201         warp = vpi.WarpMap.polynomial_correction(grid, K, X, rcoeffs, tcoeffs)
202 
203         with vpi.Backend.CUDA:
204             output = image.remap(warp, interp=vpi.Interp.CATMULL_ROM, border=vpi.Border.ZERO)
205             return output

[AastaLLL]

Hi

Please find the K in the below document (Pinhole Camera Model).

https://docs.nvidia.com/vpi/appendix_pinhole_camera.html

The intrinsic camera parameter K has a fixed format (only 4 degrees of freedom).
Please define it with fx, fy, cx, and cy rather than a 3x3 identity matrix.

Thanks.

[cletch]

Thanks for the suggestion. I’ve modified the K and X matrices to be of the correct format, rather than just full of ones, but I’m still getting the same error.

Traceback (most recent call last):
  File "./cudatest.py", line 425, in <module>
    dewarped = processor.linear_dewarp(flipped_image)
  File "./cudatest.py", line 204, in linear_dewarp
    rcoeffs = np.array([k1, k2], dtype=np.float32) # The radial distortion coefficients.
TypeError: polynomial_correction(): incompatible function arguments. The following argument types are supported:
    1. (grid: vpi.WarpGrid, *, Kin: numpy.ndarray[numpy.float32], X: numpy.ndarray[numpy.float32], Kout: numpy.ndarray[numpy.float32], rcoeffs: numpy.ndarray[numpy.float32], tcoeffs: numpy.ndarray[numpy.float32] = array([], dtype=float32)) -> vpi.WarpMap
    2. (grid: vpi.WarpGrid, *, K: numpy.ndarray[numpy.float32], X: numpy.ndarray[numpy.float32], rcoeffs: numpy.ndarray[numpy.float32], tcoeffs: numpy.ndarray[numpy.float32] = array([], dtype=float32)) -> vpi.WarpMap

Invoked with: <vpi.WarpGrid(regions=([, array([[1., 0., 1.],
       [0., 1., 1.],
       [0., 0., 1.]], dtype=float32), array([[1., 1., 0., 1.],
       [1., 1., 0., 1.],
       [0., 0., 1., 1.]], dtype=float32), array([0.1, 0.1], dtype=float32), array([0., 0.], dtype=float32)

How should the rcoeffs and tcoeffs matrices be formatted?

[AastaLLL]

Hi,

It looks like the error is more related to data type.
Would you mind sharing the complete Python source with us?
So we can check it directly.

Thanks.