Matthews Correlation Coefficient not showing value per class in multi-class classification after update

[pedrogalher]

System information

• Windows 10
• TensorFlow version and how it was installed (source or binary): TF 2.5.0
• TensorFlow-Addons version and how it was installed (source or binary): 0.13.0
• Python version: 3.6
• Is GPU used? (yes/no): yes

Describe the bug

Regarding the Matthews Correlation Coefficient, in the multi-class classification, using tfa 0.12.1 generates the MCC for each of the classes (as I expected and similar to the result provided by the F1 Score). However, tfa 0.13.0 generates a single MCC value for all the classes. It would be nice if you could choose how you want your output to be.

Just as a clarification: I'm working in a semantic segmentation project. The multiclass scenario (3 classes) is reduced to 3 binary classifications (A vs. not A, B vs. not B and C vs. not C). So it would be great to extract the MCC for each class so that I can compare how the MCC performs when directly compared to the F1-Score for each class as the MCC considers all categories (TP, TN, FP and FN) unlike F1-Score.