Fig. 1. Overview of the CAM pipeline. We investigate three phenomena of the feature maps (F). P1. The areas of the activated regions largely differ by channel. P2. The activated regions corresponding to the negative weights (wc < 0) often cover large parts of the target object (e.g. monkey). P3. The most activated regions of each channel largely overlap at small regions. The three modules of CAM in gray boxes (M1–M3) do not take these phenomena into account correctly. It results in localization being limited to small discriminative regions.