When annotating entity label regions on an image Lucid AI will also create several different augmented versions of the image data to enhance the AI training process. Different augmentation methods are applied depending on the image data. Some simple examples of augmentations include view rotations, and color adjustments.
Augmented image data is extremely useful when only a small number of image regions have been defined for an entity label, though more importantly it can boost the Lucid AI recognition success anywhere from 15-50% when compared with no augmentation. Augmentation is also why you may see your AI trained on a much larger image dataset that your overall entity label regions.