What data preprocessing is used for the input to the Vision Transformer?

[24anisha]

Hi @rwightman, and thank you for such a useful repo! I have a quick question about the pretrained Vision Transformer:

What is the expected status of the data before passing it into the trained vision transformer? Is there specific data preprocessing that is done within the model itself, or does the model expect certain preprocessing to be done to the images already? If so, what is that preprocessing?

Thank you!