Experimental: make use of 3D pose data in post-annotator
@christian.hansen: as discussed in the meeting you could try to implement a 3D viewer, load and visualise the 3d pose estimation data from the backend. does not have to be pretty at the moment but would be interesting to "see" the data (also for identifying possible miscalculations in other ML features that make use of the data).
Anton suggested to keep a small video reference somewhere.