Question about corruption image generation

[baek85]

Hi,

Thanks for sharing your great work! It’s a very instresting study on visual in-context learning.

I have a question about corruption-image generation strategy.

1. Did you use the exact same code for generating the ImageNet-C versions, with only the image resolution adjusted for your datasets?

2. Could you please provide the specific script or code used to generate these corrupted images?

Provide the generation code would be helpful for me.

Best regards,