After testing the behaviour for a bit, the hand and foot position is a bit wonky at times., especially if e.g. the hand is in front of the body. In these cases having additional cameras at different angles would help with the precision.
How to determine which camera to believe?
For the body, prefer he camera with better frontal view
for the limbs, prefer the camera that sees that the limb is not front of the body
also prefer the limb coordinates that are in the vicinity of the external controllers (tracking of controllers could be lost)
Phase 1: selectable camera source
Phase 2: get data from multiple sources and average results
Phase 3: instead of average, use a better method to mash the results together
After testing the behaviour for a bit, the hand and foot position is a bit wonky at times., especially if e.g. the hand is in front of the body. In these cases having additional cameras at different angles would help with the precision.
How to determine which camera to believe?