Thank you for sharing the scoring metrics code for the AIIB competition.

However, while using the evaluation_branch_metrics function in AIIB/scoring metrics/task1.py. This function seems to consume a significant amount of time when it processes the input pred and label numpy arrays of size 512512d using numpy calculations.

I was wondering if you could suggest any methods or techniques for speeding up the evaluation process within this function.
Thank you for your attention to this matter.

Issue Description:
The recent update to scikit-image (version 0.23.1) has introduced a change in the skeletonize_3d method. This method now returns a skeleton with boolean values, which causes incorrect functionality in the get_parsing method.

Current code:
skeleton = skeletonize_3d(mask)

Updated Code:
To ensure compatibility with the new version, you should convert the boolean output to np.uint8:
skeleton = skeletonize_3d(mask).astype(np.uint8)

Please be aware that the skeletonize_3d method will be renamed to skeletonize in future releases of scikit-image.

Hello,

I'm grateful for the AIIB23 dataset and wish to access it. However, the provided link in the repository (https://codalab.lisn.upsaclay.fr/competitions/13076#participate) seems to be inaccessible.

Could you assist with an updated link or guidance on how to obtain the dataset?

Thank you.