We applied our methods to a human skeleton under three scenarios: interactive posing for keyframe animation; inferring missing data from motion capture; and constraint-based animation in which constraints, not joint angles, are interpolated.
