Animatable Virtual Humans: Learning Pose-Dependent Human Representations in UV Space for Interactive Performance Synthesis | Synapse