Precision vs. recall curves for the final classifiers on each test tile. In each panel, the colored curves represent a classifier trained with a tile, while the black curve is the ensemble classifier. Red dots correspond to a precision and recall fixed at a threshold equivalent to the recall ∼0.5 when training and testing on the same tile with 10 k-folds. The blue stars represent the working point selected by averaging the thresholds corresponding to the red points.