For normal PyTorch models we'd usually perform some sort of image preprocessing before inference (https://pytorch.org/docs/stable/torchvision/models.html), with a mean and a standard deviation. Do we do the same for the robust pre-trained models in this library?
For normal PyTorch models we'd usually perform some sort of image preprocessing before inference (https://pytorch.org/docs/stable/torchvision/models.html), with a mean and a standard deviation. Do we do the same for the robust pre-trained models in this library?