An isolated proof of concept for getting the FreeMoCap backend running in realtime.
Currently, it mocks realtime video capture by sending MultiFramePayload
objects through a queue to the realtime process (happens in mock_multiframe_payload.py
). The realtime pipeline pulls the payloads out of the queue, and sends each frame to a separate tracking process in separate payload
queues, which then sends the resulting array back to the realtime pipeline in output
queues. The arrays are then combined and triangulated, giving 3d data.
So far, this has been optimized from ~144 ms per frame for a naive version to ~70 ms per frame.
Install with pip install -e .
(or uv pip install -e .
).
To run, run __main__.py
.