Region without any ground truth to build-up models
Currently, if a region doesn't contain any sample (from the input ground truth) the region is not taken into account in the model's list. The consequence of this is that no model is produced for this region and, therefore, the region can't be classified.
However several strategies are conceivable :
- Ask users to define their own strategy: by the use of a script during classification stages ...
- Use a metric to define the most similar region (similarity to be defined) and use it to classify the region
- Use a iota2 existing method for the fusion of the results of other models: majority voting, dempster-shafer ...
- ...
Edited by Jordi Inglada