RandomErasing preprocessing layer about keras-cv HOT 4 CLOSED

https://github.com/tensorflow/models/blob/master/official/vision/beta/ops/augment.py#L2021

from keras-cv.

RadomErasing and Cutout augmentation are almost alike, except in RandomErasing, the cutout happens with random sizes.

cutout: tfa.image.cutout

from keras-cv.

A comment from @sayakpaul:

Random Erasing [1] is another essential augmentation transform for controlling the amount of regularization when training models on ImageNet. It's also used by the recent line of papers [2, 3]. Short explanation: https://paperswithcode.com/method/random-erasing.
[1] Random Erasing: https://arxiv.org/abs/1708.04896
[2] ResNet Strikes Back: https://arxiv.org/abs/2110.00476
[3] A ConvNet for the 2020s: https://arxiv.org/abs/2201.03545

from keras-cv.

Thanks @chjort

from keras-cv.