Vision: VideoClips object needs documentation

[TengdaHan]

The recent 0.4.2 update on VideoClips seems to bring lots of new functionality with video_reader backend without many explanations.

I think VideoClips need documentation, it's a very useful tool. Thanks!

[fmassa]

Hey,

A few of those things are still a bit experimental. In particular, I would like to remove if possible all the arguments starting with an _ from VideoClips, https://github.com/pytorch/vision/blob/8aec85deb447d415abc0788d8dc301a727cc69fb/torchvision/datasets/video_utils.py#L73-L74
They are currently there for performance reasons, and are private (start with an underscore) specifically because of that.

I think we do have code documentation for the VideoClips, but it's not exposed in the website. Is that what you meant about the documentation?

[TengdaHan]

Hi, yes the code is great. What I mean is the website documentation plus some examples. They would be helpful.

[fmassa]

I see, yeah, we can expose it in the website documentation.
We would need to modify one of the files in https://github.com/pytorch/vision/tree/master/docs/source (maybe datasets?) to add this entry.

Contributions are welcome!

[bjuncek]

@TengdaHan I'll sync up with fmassa and try to write these up in a next few weeks. Please let me know of particular examples for things you'd like to see in there (I'll stop by O.U. on Thu as well so that can happen offline as well).