Closed Belberus closed 3 years ago
I'm going to adapt AIK's backend method and @Belberus will have to adapt the frontend.
The idea is to copy the mechanism that considers every special case for complete and incomplete skeletons.
Note: It is not technically possible right now to interpolate incomplete poses and draw points from different poses. Hence, interpolating between two incomplete poses will only be done for the annotated labels and not for the missing ones (in either of the skeletons).
Fixed for now
Since the coordinates for a person's keypoint are [x,y,visibility] we have to make sure that the interpolate function ignores the third coordinate.