Data Augmentation is widely used to increase train data where there is not much of training data.

In vision, it is done by,

• Random Cropping
• Horizontal and Vertical Flipping
• Changing brightness, saturation, contrast
• Translation
• Affine Transformation
• Padding

In language, it is done by,

• Synonym
• Stemming
• Back Translation
• LLM Rephrasing

Acts as Regularization to reduce Overfitting when data is limited; related to Oversampling.