Error with valis-hest

[Kexin-xu-01]

Hi,

I am trying to use valis for Xenium H&E alignment. However, I ran into a problem installing valis-hest. I am running on python3.11. The error is
module 'cv2' has no attribute 'xfeatures2d' (cv2 version: 4.7.0). Do you know how to work around it?

Also, can I bypass 10x Genomics alignment and just run align_with_valis without the alignment matrix? In that case, can I set alignment matrix to an identity matrix like following?

alignment_matrix = read_xenium_alignment(alignment_file_path) if alignment_file_path else None

# fallback to identity 3x3 if alignment is missing
if alignment_matrix is None:
    # 3x3 homogeneous identity transform so matmul works as expected
    print('alignment file missing. Using identity matrix as alignment matrix. Remember to align image later')
    alignment_matrix = np.eye(3, dtype=float)

dict['pixel_size_um_estimated'] = self.__xenium_estimate_pixel_size(pixel_size_morph, alignment_matrix)

Thanks!

[pauldoucet]

Hello @Kexin-xu-01 ,

1. The cv2 error could be related to having multiple versions of cv2 installed. Could you try the following?

pip uninstall opencv-contrib-python
pip uninstall opencv-python
pip install opencv-contrib-python

2. If you don't have access to 10x genomics alignment files and you only want to run Valis non-rigid alignment, I would advise to create an identity alignment file inside your sample directory like this one:

imagealignment.csv

Once you've read your sample as a HESTData sample, you can then run align_with_valis