Trainable Weka segmentation: How to UPDATE a classifier

Hi all,

I am a happy user of the Traininable Weka Segmentation plugin in Fiji.

I have to train a classifier on a discrete set of images. And I was wondering how to do that.

I open the first image, launch the plugin, and annotate the classes. From this I get a classifier model that I can save.

I need to update the classifier with the annotations from the 2nd image. But when I load the classifier, add annotations on the 2nd image and click Train classifier, it appears that the first classifier has been cleared.

Is there a way to incrementally train a classifier from several images?

2 Likes

Hello @tinevez and sorry for the late replay!

If the images are of the same size, you can load them as a stack and train on all of them at the same time.

Most classifiers in Weka are not re-trainable, so the solution we came up with was to save the annotations’ information into an ARFF file using “Save data”, and then loading the file in the second image using “Load data”. That way both the current annotation derived information and the previous one will be used during the training of the algorithm.

Cheers!

1 Like