Ilastik Random forest hyperparameters

Hello everyone,

In my bachelor thesis i am using the “Object Classification” workflow from ilastik. Is there any documentation on the default parameters like number of decision trees, depth of the trees, maximum nodes when the training stops?
The ilastik paper answers some of these questions for the “pixel classification” workflow, but is it the same for OC?
I am quite bad at finding my way through big python code, so any help would be great!

Cheers Justus

Hi @Justusschl,

the ilastik source code is a bit hard to navigate if you are not used to that (we are working on it…).

So in Object Classification it should be a single Forest with 100 trees.
We use the Random Forest of the vigra library. For the other parameters we use the defaults (see, e.g. here).

Cheers!

P.S.: thx @etadobson for tagging the question with ilastik!

1 Like