[feature request] SRT improve video inference

[emmalacaille]

[feature request]

Which project are you using?

SRT

Issue description

Hi, Xuanyi. It's really an impressive work!

I would like to predict facial landmarks on videos, do you know if it is possible to input correct facial landmarks for the first frames of the video when processing it, so that the predictions are even more accurate for the rest of the video ?
If it is possible, can you please give some pointers on how to do it ?

Thank you very much !