Read ilastik ouput on knime

Hi everyone,

I am beginner in Ilastik and Knime, so maybe this issue has been solved somewhere.
My current “non-Knime” worflow is : generate images (tif) of cells from my experiment, segment the cells with Ilastik (pixel classification), process the hdf5 files from Ilastik with a homemade jupyter script to fit the shape of the cells with an ellipse.

I would like to automatize image processing with Ilastik. So, Knime seems adapted to me (actually I already tried the Ilastik plugin for Fiji but it was too long).
Thanks to example workflow available on Ilastik integration page (https://www.knime.com/book/ilastik-integration), I have been able to run somes images of my experiment.
The result can sucessfully be seen in Image normalizer node.
However, in Knime, when I put the “Jupyter” node after the “Ilastik” node, my Jupyter script is not working.
I have the feeling that is because, in Knime, “Ilastik” outputs are dataframes and not hdf5 files.

Is that so ? And if it is, how to deal with this issue ?
Thanks you very much for your help !

Kindly yours !

Hi @chalbiophysics,

i vaguely recall that knime will export tiff files, but I don’t think I can give you any more help on this subject.

As an alternative approach, you could convert your notebook into a Python script that takes the ilastik output as an input. (You could probably just copy all your cells to a *.py file and use sys.argv to get the name of the input file.

Hi @k-dominik,

Thanks for your advices. I gave up with Knime and moved on ilastik headless mode with python (this topic : How to call and run ilastik from python?). I have new questions but I thinks it’s better to open a new topic.

Kindly.

Even though you’ve already moved on from KNIME, one thing that may have been causing the issue with your Jupyter node is that the knime ilastik node will output the probability maps and not the simple segmentation. You likely noticed this already but just in case you didn’t :grinning:

1 Like