Looks really sophisticated! How do you differ between different body poses? Do you just try to match the covered areas with a template or do you evaluate specific body motions and edge data or something like that?