OpenCV error: Expected cv::UMat for argument 'M'

Question

I'm new to Python, therefore having troubles with debugging a script. I'm trying to create an ' average face ' using an opencv script to recreate with my own images. Here's the Github repo I'm using, but same goes for this one and this one .

The land mark detection part works but the average.py script throws the error I don't understand how to solve.

The first two errors I solved by replacing the xrange() function with range() . Than the estimateRigidTransform() seemed depricated and there for I swapped it with estimateAffinePartial2D() so far so good.

Now the console throws me the following error:

TypeError: Expected cv::UMat for argument 'M'

This is the scripts' code snippet:

# Apply affine transform calculated using src_tri and dst_tri to src and
# output an image of size.
def apply_affine_transform(src, src_tri, dst_tri, size):

    # Given a pair of triangles, find the affine transform.
    warp_mat = cv2.getAffineTransform(np.float32(src_tri), np.float32(dst_tri))

    # Apply the Affine Transform just found to the src image
    dst = cv2.warpAffine(src, warp_mat, (size[0], size[1]), None,
        flags=cv2.INTER_LINEAR, borderMode=cv2.BORDER_REFLECT_101)

    return dst

For more reproducible code see repos in introduction of this question.

All help much appreciated!

Answer 1

Instead of using tform = cv2.estimateAffinePartial2D(np.array([inPts]), np.array([outPts])) and returning tform , return tform[0] .

Refer to the documentation for more details.

You will notice that estimateAffinePartial2D returns retVal and inliers . That's why when you are returning tform , you are getting a TypeError .

I have also created a PR to fix the code on our LearnOpenCV GitHub repository.

Vishwesh

Edit : You can check the PR here .