Has anyone tried creating synthetic data to balance their ML training?

towardsdatascience.com