ValueError: shape mismatch

[jo-mueller]

trafficstars

System and Software

• aicsimageio Version: 4.6.4
• Python Version: 3.8.12
• Operating System: Windows 10

Description

Hi all, I am receiving an error when I try to open a common .czi-image, such as this one (Credit to @Theresa-S for the data). I open it with the following code:

from aicsimageio import AICSImage
img = AICSImage(r"D:\P2E_GFAP-Nestin__0001.czi")  # selects the first scene found
img.shape

and receive the following error:

  File "C:\Users\johamuel\AppData\Local\Temp\ipykernel_1184\4260751373.py", line 1, in <cell line: 1>
    img.shape

  File "C:\Users\johamuel\Anaconda3\envs\devbio-napari-env\lib\site-packages\aicsimageio\aics_image.py", line 533, in shape
    return self.xarray_dask_data.shape

  File "C:\Users\johamuel\Anaconda3\envs\devbio-napari-env\lib\site-packages\aicsimageio\aics_image.py", line 414, in xarray_dask_data
    self.reader.mosaic_xarray_dask_data

  File "C:\Users\johamuel\Anaconda3\envs\devbio-napari-env\lib\site-packages\aicsimageio\readers\reader.py", line 410, in mosaic_xarray_dask_data
    self._mosaic_xarray_dask_data = self._get_stitched_dask_mosaic()

  File "C:\Users\johamuel\Anaconda3\envs\devbio-napari-env\lib\site-packages\aicsimageio\readers\czi_reader.py", line 777, in _get_stitched_dask_mosaic
    return self._construct_mosaic_xarray(self.dask_data)

  File "C:\Users\johamuel\Anaconda3\envs\devbio-napari-env\lib\site-packages\aicsimageio\readers\czi_reader.py", line 730, in _construct_mosaic_xarray
    stitched = self._stitch_tiles(

  File "C:\Users\johamuel\Anaconda3\envs\devbio-napari-env\lib\site-packages\aicsimageio\readers\czi_reader.py", line 710, in _stitch_tiles
    ans[tuple(ans_indexes)] = data[tuple(data_indexes)]

  File "C:\Users\johamuel\Anaconda3\envs\devbio-napari-env\lib\site-packages\dask\array\core.py", line 1828, in __setitem__
    dsk = setitem_array(out, self, key, value)

  File "C:\Users\johamuel\Anaconda3\envs\devbio-napari-env\lib\site-packages\dask\array\slicing.py", line 1868, in setitem_array
    raise ValueError(

ValueError: shape mismatch: value array of shape (2236, 2236) could not be broadcast to indexing result of shape (2236, 586)

Expected Behavior

Data cannot be provided by the loaded img object (commands like img.data, img.get_dask_data() all fail with the above error.

I assume that aicsimageio tries to chunk up the image somehow but fails to reproduce the original image dimensions in case the original image size is some odd number (here: 18495 x 18495).

[evamaxfield]

Ugh. More mosaic bugs. So sorry about this @jo-mueller.

We know there is a Zeiss created tiling API coming soon so hopefully when that comes out we can resolve a whole host of these issues :shrug:

---

Do you need it prestitched or can you handle the stitching yourself? If so I would turn off reconstruct_mosaic

img = AICSImage("...", reconstruct_mosaic=False)

I wonder how m2stitch deals with this... https://github.com/AllenCellModeling/aicsimageio/issues/383 :thinking:

[jo-mueller]

Hey @JacksonMaxfield ,

thanks for the fast reply. The spatial relation of single mosaic tiles to the whole image is quite important on this one, but the proplem is not super-urgent. If Zeiss brings out some major updates I think it's feasible to wait for the new API to fix these issues rather than self-engineering a fix that may not be worthwhile.

Do you mind if I keep this issue open or could you ping me when there's any updates on this?

[evamaxfield]

Not at all! Please keep this open until fixed!

[BrianWhitneyAI]

We have added a new maintenance feature to clean up stale issues and PRs. Adding this comment to set a baseline for ‘Stale’

[BrianWhitneyAI]

I beleive that the example file's link has expired and I cant seem to replicate this bug with an alternative file. Feel free to reopen the issue!