to_multiscale with Methods.ITK_LABEL_GAUSSIAN throws error for certain data types

[ahnsws]

Hello, we would like to use to_multiscale to generate downscaled pyramids of label arrays. Unfortunately, it looks like the ITK_LABEL_GAUSSIAN method only works for certain data types. Please see the snippet below:

import dask.array as da
import numpy as np
from multiscale_spatial_image import to_multiscale, Methods
from spatial_image import to_spatial_image, SpatialImage


def try_multiscale_label():
    """
    types:
    - int8   failed -> itk.support.extras.TemplateTypeError: itk.Image is not wrapped for input type `itk.SC, int`.
    - int16  worked, but max val not big enough
    - int32  failed -> itk.support.extras.TemplateTypeError: itk.BinShrinkImageFilter is not wrapped for input type `None`.
    - int64  failed -> itk.support.extras.TemplateTypeError: itk.Image is not wrapped for input type `itk.SL, int`.

    - uint8  worked, but max val not big enough
    - uint16 worked, but max val not big enough
    - uint32 failed -> itk.support.extras.TemplateTypeError: itk.BinShrinkImageFilter is not wrapped for input type `None`.
    - uint64 failed -> itk.support.extras.TemplateTypeError: itk.BinShrinkImageFilter is not wrapped for input type `None`.
    """
    labels_arr: da.Array = da.zeros(shape=(100, 100), dtype=np.uint64)
    labels_arr[:50, :50] = 1

    labels: SpatialImage = to_spatial_image(labels_arr, dims=("y", "x"))
    to_multiscale(
        labels,
        scale_factors=[dict(y=2, x=2)],
        method=Methods.ITK_LABEL_GAUSSIAN,
    )


if __name__ == "__main__":
    try_multiscale_label()

For int8, int32, int64, uint32, and uint64, there is an error thrown, and it seems like this will take me into the weeds a bit. If you have any suggestions on how to approach fixing this I am all ears.

Edit: the dependencies for the above snippet are:

python==3.10.12
multiscale-spatial-image==0.11.2
dask==2023.8.0
numpy==1.25.2
zarr==2.16.0
itk==5.3.0

Thank you,
Seb

[dzenanz]

Considering wrapping specification, enabling ITK_WRAP_signed_long during package creation should fix your problem. Read more about ITKPythonPackage, and its docs.