Simple bounding box annotations should not target the Timeline
The text annotations we were using so far did target the parent Timeline with an absolute timestamp so that a text annotation can correlate with one or more videos simultaneously.
I think this is no longer useful for bounding boxes, as they result from analysing one specific video. Layered on top of a second camera perspective, the annotation is useless.
This is different for tracking as people can be seen from different angles, but automatically identifying someone is impossible with our current tracking solution (id by order of appearance, no unique identifier)... We'll get to that later.