[docs] torchvision.io.read_video docs do not (but should) specify image scale format

[vadimkantorov]

Is it int8, uint8, float32 from 0 to 1, -1/1? Is the format always same or input-file dependent? What happens for grayscale single-channel input videos? Will it still produce 3 channels?

cc @bjuncek

[bjuncek]

Hi @vadimkantorov
Should be uint8 in 0-255 by default; RGB24 pixel space regardless of the input.

feel free to send a PR with the doc modification. I'm technically not working until next week so might not get to it before then.