AssertionError when using autocontext with h5 files in headless mode


I am using an automated pipeline combining cellprofiler .cppipe and ilastik .ilp files. The output from cellprofiler is a 6-channel 16bit composite image in h5 format, and I put this into an ilastik autocontext workflow that I have pretrained. Batch processing using this works smoothly on the local GUI version of ilastik1.3.3b1-OSX, but when I try to run this using headless mode using ilastik docker container (1.3.3post2) it fails and shoots up an AssertionError.

See above dropbox link to a folder where I have included the command used, error message, ilastik .ilp file and the .h5 input file to reproduce this error.

I would really appreciate any advice with this!

Thank you,

Hi @Karishma,

thanks a lot for providing these files. I can reproduce the problem and will look into it. If you can live with using 1.3.3b1 for now, this would be the quickest solution.


Hi @k-dominik,

It’s not really feasable for us to use the 1.3.3b1 locally as we’re using the ilastik docker containers as a way to run a whole automated image analysis pipeline on the cluster (the dataset we want to use is very large so this doesn’t run locally on them).

Thank you for looking into this! :slight_smile:

would it help you (for now) if we provided a 1.3.3b1 container?

I’m not sure, if the issue is a bug with the 1.3.3post2 containers then yes a 1.3.3b1 container might help, but at the moment im not sure exactly where the issue is

File “/ilastik-release/lib/python3.7/site-packages/ndstructs/”, line 84, in init
assert len(arr.shape) == len(axiskeys)

If it’s easy for you to provide a 1.3.3b1 container then I would love to try and see if that solves the issue!

Thanks so much for the help!

The issue is definitely with some code changes we did for 1.3.3postX …

So, I just pushed the 1.3.3b1 container. Hope that helps for now

Hi @k-dominik,

Thank you so much for the 1.3.3b1 container, using this our pipeline runs to completion now yay! :smiley:

1 Like