Methods to merge regression from multiple seeds
Currently, when multiple runs are required, we set aside some polygons to validate the resulting fusion of predictions. To perform the fusion, 2 methods are available: majority voting and the Dempster-Shafer one (directly from otb FusionOfClassifications).
However, if the predictions come from a regression model, these methods cannot be used.
We need to replace the FusionOfClassifications call with something else. We can simply use the average value of all regression predictions. @mfauvel , any other suggestions on how to merge regressions?