CAM Component IoU
It evaluates how much the class activation map focuses on multiple components of the same category. First, the class activation map foreground area is divided into connected components, i.e., groups of pixels connected to each other. The IoU value is calculated between each ground truth bounding box and the connected components that intersect it. Then, the average IoU across all the components is taken.
Example
from odin.classes import Metrics
my_metric = Metrics.CAM_COMPONENT_IOU
# use my_metric as 'metric' parameter in the analyses
Tasks supported
Binary Classification | Single-label Classification | Multi-label Classification | Object Detection | Instance Segmentation |
---|---|---|---|---|
yes | yes | yes | no | no |