Classify material objects for segmentation with MIB

The classifying function for membrane detection is very nice. Could we extend that to segment with more than 2 classes of object materials and using the customized deep neural network from Matlab to train/predict?
Thanks!

Hi @nix,
so far the classifiers and the graphcut segmentation are all limited to the single object, i.e. 2 classes. If you need to classify the images into multiple classes, I suggest to use Ilastik. You can export images and import results back to MIB using the Ilastik compatible HDF5 format (HDF5 with XML header).

Use of DNNs are in plans; the recent releases of Matlab allow to implement those fairly easily.
The general problem that in MIB we can only have image+model+mask+selection layers and generation of models with more than 2 classes would require the second model layer.
Alternatively, I can re-use the mask layer as the second model or send the results to another MIB container. I need to think about the best way.

Ilya