In case the user has the ground-truth information, he/she can decide to assign more than one
node to any class taking into account the ground-truth, can merge some segmented regions,
and can give classification performances on the remote- sensing images. Otherwise, each node is allowed to represent a different candidate class by the proposed network.