ROTATIONAL DATA AUGMENTATION
studying it from all angles... even tries different paths to understand the route to the actual treasure!
Technical definition
A method for artificially increasing training dataset size by applying rotations to the data, in order to enhance model robustness to variations in orientation.