Figure 15.

Confusion matrices for our classifier (a), CNN-1 [25] (b) and CNN-2 [35] (c) based on the patch-based medium RAFT-MIM${}^{Q_{12}}$ dataset.