Document the subtleties of to/from tensor conversions in a gallery example

[NicolasHug]

Our utils for converting to and from tensors (to_tensor, pil_to_tensor, to_pil_image, etc.) are fairly subtle and can sometimes create confusion (#2483 (comment), #3782).

It'd be nice to have a gallery example that exhaustively illustrates the stuff one should be aware of when using them, for example in terms of contiguity, range conversion, dtype, etc. and also explain when to use one over the other; In particular we could try to illustrate the differences between to_tensor and pil_to_tensor on the same set of inputs

[puhuk]

@NicolasHug I'd like to make an example and description for this. Do you have any idea of which documents such contents should be in? Or is making separate page would better?

[vilktor370]

@NicolasHug

I know this is pretty old. Hopefully, it's still open for contribution!
I sent a PR. #7985. Could you review it whenever you are available?

I do have a couple more questions if you don't mind:

• I am not very sure how to show the differences in terms of contiguity between to_tensor and pil_to_tensor. Are there any performance differences between those two functions?
• I only made examples of to_tensor, pil_to_tensor and to_pil_image . Are there more functions that should be included in the example? Feel free to let me know if there are more functions you would like to include.

[NicolasHug]

Thanks @vilktor370 , I left a few comments on the PR, LMK what you think. A lot has changed since this issue was opened and it is less relevant now, so I'll close it