Data Augmentation

The creation of modified training examples, such as image transformations, to increase useful variation. Changes must preserve task-relevant meaning; more examples do not automatically improve model performance.