Edge features between the super-pixels (nodes)

[Mario Measic]

Hello Andreas,

How did you obtain the edge features for the edges between the super-pixels in the Image segmentation example for the VOC dataset (http://pystruct.github.io/auto_examples/image_segmentation.html#sphx-glr-auto-examples-image-segmentation-py)?

I understand that there are 3 features per edge, with the first two being symmetric and the last one being asymmetric.

Thank you in advance.

[Mario Measic]

For those of you that are experiencing the same issue. Here is the nice description of what can possibly serve as an edge potential: https://books.google.es/books?id=ITvZ9HDPizcC&pg=PA4&lpg=PA4&dq=color+sensitive+potts+model&source=bl&ots=9gS25JaJTL&sig=tovsSfBRnmjeeu-obP9RdE49wWk&hl=en&sa=X&ved=0ahUKEwjO8veAm8nYAhXDORQKHbQqC7oQ6AEIOzAC#v=onepage&q=color%20sensitive%20potts%20model&f=false

I have used colour potentials (differences between the superpixels) and the location potentials.

Thank you.