Text this: Enhancing self-supervised visual representation learning through adversarially generated examples.