It seems possible, since Apple’s Vision framework can read body pose too. Maybe I can try it for the next update.