Hello,
I am wondering whether the current ResNet family of models from the Torch Vision have been trained with data augmentation or not. In the reference script here, there is no mention of data augmentation for the ResNet family, but how can I make sure there was no augmentation (without training a new model myself)?
Based on the accuracy gap between the two weight versions of the ResNet-50 model (ResNet50_Weights.IMAGENET1K_V1 and ResNet50_Weights.IMAGENET1K_V2), I assume the V1 version has no augmentation, but the V2 version has several augmentation policies. Am I right?
Thank you.