Using Mediapipe

[lakpa-tamang9]

I have some queries:

• Can i run the same action recognition, but by using different pose estimation tool?
• Do i need to build new graph based on that tool?
• Since google's mediapipe pose provides simple api for pose estimation, wouldn't it be better to use that for action recognition?

I havent found a single implementation related to this. If anyone has any references, then please list out in this issue thread.