Monai Multiclass Confusion Metric

[mcleod-matthew-gene]

Hello All,

For the MONAI ConfusionMatrixMetric, how does one have it calculate the multi-class metric rather than multilabel metric. I don't see a parameter to pass in to tell it do multi-class metrics and none of the reduction methods seem to yield multiclass results, rather they are multilabel.

For example, see below. y and y_preds are of shapes (4,3), so batch size of 4 and 3 classes. This should be a simple multiclass problem, but as the code snippet highlights, the results returned by MONAI are correct if this is a multilabel problem.

The documentation of the ConfusionMatrix states: It can support both multi-classes and multi-labels classification and segmentation tasks. How do I get multi-class metrics?

Thanks!