The Dice Coefficient is a similarity metric for masks, often used in segmentation tasks. Given a ground-truth mask G and a prediction mask P, it is defined as 2*|intersection(G, P)|/(|G|+|P|), where |X| denotes the number of elements in X.