What is best practice of augmentation on Imbalance dataset?

Hello, I just learned deep learning for my college needs.

After I browse articles on the internet, many say that data augmentation should be applied only to training data after data splitting (training:validation:testing).

But in my case, I have imbalanced data. If I only apply augmentation to the training data, the validation and testing data will still be unbalanced.
(Is this a problem?)

What is the best practice for this? I hope you can help me.

Thank you!

Note: This …

apply articles augmentation case college data dataset deep learning deeplearning hello internet practice testing training training data validation