At the end of the supervised learning page we are trying to show a side by side of the predicted label vs RGB image.

Since this change went in the RGB image is actually a (128, 128) array and matplotlib is styiling it with viridis.

This is an antipattern and a misuse of rf_rgb_composite, as i understand it.

We should re-think the way we do this visualization to make a correct RGB and possibly to enable the visual to cover an expanded geographic scope.