Figure 4 | Scientific Reports

Figure 4

From: Module of Axis-based Nexus Attention for weakly supervised object localization

Figure 4

Qualitative comparison between our proposed method (MoANA) and CAM4 for WSOL task on the (a) CUB-200-2011 and (b) ILSVRC datasets. The red box is the GT bounding box, the green box is the predicted bounding box, and the green area is the segmented region to extract the bounding box after the threshold is applied. MoANA can generate more exact localization maps by tightly bounding the entire region of the object in an image. These maps were generated using Python 3.6.0, available at https://www.python.org.

Back to article page