We currently have the option to generate row-, and column-normalized confusion matrices, or non-normalized confusion matrices.
Having an option to generate normalized CM that also shows the counts are very helpful to better understand class-imbalances.
Example below:
