yeungchenwa/OCR-SAM

video pipeline

darkzbaron opened this issue · 3 comments

darkzbaron commented

That looks great! Any idea as to have a fast pipeline for video?

Mountchicken commented

Hi @darkzbaron
Consider that the inference speed is slow since we are using SAM, a text detector, and a text recognizer. But we can implement an offline version for video inference.

  • 👀1
gelsas commented

Really intersting! I would love to use this for video aswell to automatically detect and remove any text.

  • 👍1
darkzbaron commented

I am using paddle now and there is a whole pipeline here : https://github.com/YaoFANGUK/video-subtitle-remover