Figure 6.
Results of ablation studies on the BreakHis dataset for binary and multi-class type classification with maximum voting criteria. Using the four different types of evaluation metrics, we select a different combination of MFNet maps and fuse them, i.e., DB-(1–4), such as quadruplets, triplets, pairs and individual concatenated feature maps. For example: DB-× or DB-, represents that we used only DB-1 (first depth block) or DB-12 (first and second depth blocks), respectively.