Enhancing Data Augmentation with Partial Content Masking: The YONA Method
YONA, a novel data augmentation method, simplifies the augmentation process by bisecting an image, substituting one half with noise, and applying data augmentation techniques to the remaining half. This approach reduces redundant information, encourages neural networks to recognize objects from incomplete views, and significantly enhances their robustness.
