Related, I have the same interests:
[UE5] Automated rigging and animating from video input. Deeplabcut? OmniTrax? BlendArMocap?
these keywords are too broad. trackers are for tracking visible parts, “pose estimation” is what you need for invisible parts. Such thing requires an AI model and can be used to estimate where all the bones in a skeleton of for example a cat are if you only see one side. Plenty of research has been done, DeepLabCut looks promising but there seems to be no Blender or UE integration yet.