Surprised this isn’t in coreML. Seems useful for the Vision Pro or something
Might see it at WWDC this year?