Suggest Features Ilastik training applet

Hey folks,

I am trying to understand how to interpret the ‘out of bag error’ - which is displayed to me after setting in training applet. I can choose between three different Methods to select the features. As recommend, I used the ‘Filter Method’. After running the feature selection I got a result like this:

with an ‘obb_error=0,770’. How can I interpret the out of bag error score? My method has an error probability of 0.7%?

How should I use the suggestion features?
The ‘out of bag error’ is applied using my annotations from all the images in the training applet. Is it advisable to zoom into the image, if so how much? (To increase the speed of the process)

Thanks for any further advices,
Paul

Meh, when I wrote my reply to this question in your other thread I almost made it a new topic :D. Maybe I shouldn’t answer posts chronologically, then I’d have picked up this thread first and could answer here. In any case, to what I’ve written there, I would add:

The cutout of your zoom only influences the final prediction with the various selected feature sets. The feature search takes into account all your training data. So if you had multiple images in your training, all these annotations are taken into account.

Is the prediction with ilastik in your current analysis workflow one of your performance bottlenecks?

I have tried to understand the obb error better in correlation to my annotations. Because when I look visually at the prediction of my annotations in training applet, I see a good segmentation (sometimes better than a histogram based segmentation workflow) :slight_smile: