Supervised classification — the setup. Requires a priori knowledge about the scene:

• Which land-use types exist in the study area.
• The geographic location of reliable samples (training sites) for each class.

Without good training data, supervised classification fails. Garbage in, garbage out.