Figure 1

Consensus precision and recall. The left figure shows the two consensus features of a ground truth for the alignment of five feature maps. The features of the feature maps are distinguished by the five types of marker. Corresponding features in the different maps are illustrated by the same colour. The right figure shows three consensus features of a consensus map determined by an alignment algorithm. Note that the red features were assigned to separate consensus features, and the blue ones as well. The consensus feature in the middle even contains features from the same map. Thereby, the alignment results in a low recall value of (1/2)·(5/(2·5) + 4/(2·4)) = 0.5. Since most of the determined consensus features are "relevant" the method achieved a precision of (1/2)·(5/7 + 4/5) ≐ 0.76.