I'm trying to track a 3D object with the Android native sdk (the beta version). I create the model with the sample app and then use it into the ClientTracking3DActivity. I see that, when the model is created for an object close to me (like a telephone or a box) the rectangle is placed very close to the target. When the object is bigger (and the model is created from a wider distance) the rectangle is placed in another place, it still correctly follows camera movements but is placed in another place rather than near to the object. When the object to recognize is a building, finally, I cannot see the rectangle at all, and I can't see where it is actually placed.
What I see is, dependent on what I have to recognize, I have to apply different transformations to the rectangle matrix in order to place it on the recognized target.
Is it a problem of the beta? Can I do something to do so the rectangle is always placed on the target?