Inconsistent input video tensor shapes across different Transforms

[MichaelRipa]

Hi,

I was looking through the Transforms API docs and noticed an inconsistency with video input shapes across different transforms. Here are a few examples to illustrate my confusion:

• AugMix takes an input video tensor of shape (T, C, H, W)
• CutMix takes in a batch of videos of shape (B, C, T, H, W)
• Div255 takes in an input video tensor of shape (C, T, H, W)

Is there any reason as to why the 'channels' and 'temporal' dimensions are sometimes transposed? I tried looking for an answer to this question but couldn't find anything, so I hope this is the right place.

Thanks! 🙂