The pin gets rendered in a 3D scene, and positioned via finding the difference between your current coordinates and the pins. The building stuff is so that people cannot see a pin when there should be a building in the way. I do this by using computer vision that analyses the map imagery.
Oh, that makes sense, but how do you render the pin in 3D when you see the pin in Augmented Reality. I see your screenshots where you see the pin on google maps and the pin when you view it I augmented reality.
It's a mimic - there isn't a method in apple maps that render it. I made the 3D versions myself. I use scene libraries that overlay the camera view (that is basically what AR is).