Trainable Weka Segmentation

[al.2001]

Hello all,

Is Trainable Weka Segmentation purely based on pixel intensity? Can it use other factors, mainly shape, to classify the image?

I am trying to use the pixel intensity and shape of the input training data be a factor when classifying the image. So for example, if one highlighted area is black and circular, then all the other black circular shapes will be detected, but the other colored circular shapes won't be, and the black non circular shapes will not be either. This is simply a simplified version of my problem and I would really like to know if TWS can use shape as a factor when classifying.

Here is a diagram:



Thanks,

Al