MediaPipe Pose Detection

[UUoocl]

I have a project where the idea is to use hand and body positions to control OBS.

Google MediaPipe Pose Landmarks are sent to OBS and trigger Advanced Scene Switcher conditions.

https://github.com/UUoocl/MediaPipe_for_OBS

Electron is used to run the MediaPipe model locally.

graph TD;
subgraph OBS
A(Sources)-->B(Video Projector Window);
A<-->D[Advanced Scene Switcher];
end

B(Projector Window)--> C(MediaPipe for OBS);

C <---> |WebSocket|A;

Loading

The body landmarks are stored in an OBS Text Source as JSON.

I'd appreciate any feedback on this project.

[WarmUpTill]

Seems like a cool idea! :)

Just one question:
What exactly is the purpose of the text source?
I am not familiar if / how text sources can be used by Browser sources, so I apologize if the answer is obvious.

If it is only intended to be a proxy to receive the landmark information from and pass it into advanced scene switcher variables, you might want to consider sending that information directly via websocket messages to the advanced scene switcher.

So something like this:

[UUoocl]

Thanks for commenting. I wasn't sure how to get the landmark information directly into Advanced Scene Switcher. That is a helpful tip.

The Text Source is just being used to store the landmark information. The Text Source is being read by both Advanced Scene Switcher and a Browser Source with a local HTML/javascript file.

The page in the Browser source is using the OBS WebSocket Js to read the content of the Text Source. The java Script in the browser reacts to the landmark information.