1. All in one app.
2.Serveral object targets( 5 or 10 )
3.There would be only 1 target in camera at the same time.
Such as there are 2 targets in scene. A cube and a sphere. When the cube in real world run into camera, the app could recognize it's cube target; When the sphere in real world run into camera, the app could recognize it's sphere target ,not cube.
I don't know whether I make it clear.