Question about get_body_joints_pose data source in py_bindings.cpp and handling of missing joints

[goldenstasteworriors]

I have a question while reading the project code, specifically in py_bindings.cpp:
The line m.def("get_body_joints_pose", &getBodyJointsPose, "Get the body joints pose data (24 joints, 7 values each: x,y,z,qx,qy,qz,qw.");states that it reads data for 24 joints.

How does the code obtain data for all 24 joints when only using a headset, controllers, and a few Motion Trackers? If the hardware setup cannot directly provide all 24 joints, how is the full skeleton data generated?

If data for some joints is not directly obtained from the hardware, how are these missing joints handled? How are their values filled or inferred?

[AlfaKeNTAvR]

@goldenstasteworriors I might be wrong, but I saw somewhere online that PICO 4U is using a learning approach (neural network). If you know human skeleton and it's proportions based on the height, then by providing pose of some key points on that skeleton (in case of PICO hands, ankles, head and waist, optionally) you can use optimization or NN solution to predict other key points like knees, chest, elbows, but of course it will be an estimation, not the real data, but might be close enough.