Question Regarding Tranform

[lab176344]

The implementation is simple and easy to use. Thank you for that. I have one doubt,

Given a mini batch with input x of size BxCxHxW

we apply transformations to get
y1 = self.transform(x)
y2 = self.transform(x)

So is this a batch transformation or image wise transformation

Because as per the paper "More specifically, it
produces two distorted views for all images of a batch X
sampled from a dataset" there are two distorted views only i interpret it as for one distorted view we apply the same transformation for the images in a batch