Figure 3
From: Module of Axis-based Nexus Attention for weakly supervised object localization
Illustration of conventional context fusion attention approaches and MoANA. Our MoANA calibrates the feature employing an fine-grained attention map generated with axis-based complementary information.
