Ilastik Plugins for Fiji


I am trying to implement a Fiji Macro for batch processing images using Ilastik plugins as a segmentation step.
The training of the Ilastik classifier in the main app (ilastik.exe) works fine.
I ran into a problem when trying to apply the trained classifier on images using following Fiji plugins:

“Plugins/ilastik/Run Object Classification Prediction” &
“Plugins/ilastik/Run Pixel Classification Prediction”

I did not get any classification result (segmentation or prediction) and there were some
strange errors in the Log like:

“…ilastik error:
The operation completed successfully. ilastik finished successfully!..
…something went wrong during processing ilastik pixel classification
Cleaning up”

The ilastik executable configuration is correct. And the ilastik main app works fine on its own.
I tried the Fiji plugins on Windows 10 and Ubuntu 18.04 machines.

Thanks for your Help!!

Links to Log files:

Pixel Classification Segmentation Log

Object Classification Segmentation Log


Hi @Tony,

thanks for providing us with all the log files. It all boils down to the following exception:

Exception: Your dataset's provided axistags (yx) do not have the correct dimensionality for your dataset, which has 5 dimensions.

The most likely explanation for this error is that you’ve trained your classifier with raw 2D data loaded directly into ilastik, which means that the classifier expects 2D data only (it says ‘axistags (yx)’). “Plugins/ilastik/Run Pixel Classification Prediction” always exports the data to an ilastik-compatible HDF5 dataset with 5D first before running the classifier. Since your classifier expects 2D data you get an error.

That’s why it’s really important to also train your classifier with 5D data. The easiest way to export your raw data to 5D HDF5 format supported by the plugin is to open your stack in Fiji and then export it with “Plugins/ilastik/Export HDF5”. Then use the exported HDF5 file for training.

This process is described in more details in the README for the ilastik’s Fiji plugin:

Please let us know if it solves the issue.


Hi @ilastik_team,

I had a similar issue of that described by @Tony and, lucky me, I found your reply:

Indeed the issue was due to the use of 2D raw data in training. I just wanted you to know in case it’s helpful. Thank you very much for pointing it :smiley: