image processing - How Real time augmentation balance the all the classes?

Question

In ImageDataGenerator it gives different transformation of each images for every epochs. For example We have 2 classes with 1200 images belongs to 0 class and 300 images belongs to 1 class. ImageDataGeneratorin keras transforms all the 1500 images but not an imbalance class images only for each epochs. I don't know how this transformation balances the imbalanced classes in our data and gives better results ??.

question from:https://stackoverflow.com/questions/65880259/how-real-time-augmentation-balance-the-all-the-classes

Answer 1

Waitting for answers