How can I run video inference.

[tobymuller233]

It seems that in this repo, only demos for picture inference are implemented. How can I inference videos using this fantastic algorithm? I'm looking for some suggestions on how to handle videos. The first thing comes out from me is to just clip the video into frames for inference. However, I cannot find any scripts in tools/ that do such things. The only script that I can find related to video converting is convert_video.py, which seems do nothing but iterate all frames in the video. plz let me know if there's already written script in this repo to do this thing, otherwise I could just implement it by myself. :)