How do different machine learning models perform on classifying handwritten Vietnamese from 3 datasets?