Open tischi opened 4 years ago
this looks like a genuine bug...
The problem (set aside the code there) is that you are trying to predict with an image in the autocontext workflow, that is most likely "flatter" than the training image, meaning it has less z-slices, so it cannot accommodate filters with larger sigmas. Where the code fails, it would usually try to resolve it. This code, however, was written with the assumption of a single feature selection applet (like in pixel classification).
To kind of get it working, you could select larger sigma filters to operate in 2D only.
@wolny @k-dominik I tried to run ilastik headless (from Java) and got below error. Any ideas regarding what is wrong? Maybe some version mismatch issue (of ilastik and the pipeline file)?