Closed ternaus closed 4 years ago
We apply the corruption on each image directly after loading it with:
img = corrupt(img, severity=corruption_severity, corruption_name=corruption_ name)
See our implementation in mmdetection for details. Here is also the link to the correct line in our original pull request. The new pipeline based dataloader can be a bit hard to understand if you look at it for the first time.
How exactly do you apply the library to get exactly the same COCO-C dataset as was used in https://arxiv.org/abs/1907.07484 ?