The commercial opportunity is not the skeleton overlay itself. It is the label, score or command generated after the body is understood.
Clinics can use pose recognition to document range of motion, flag compensatory movement and support remote follow-up with more objective data than self-reporting alone.
For training apps, the recognizer turns camera frames into repetitions, tempo, stability, joint-angle thresholds and form warnings that can be explained to the user instantly.
Games need low-latency recognition for scoring and control. Enterprise interfaces need deliberate gesture commands. Both benefit from clear movement semantics and reliable classification.