Can you clarify "swapping real life images for virtual images"?

Do you mean you have an OpenGL rectangle that covers the trackable on which you paint another image as a texture?

And on the iPad3 the calculations you make for the size of that rectangle come out smaller than the visible target? Or a rectangle of the same computed size comes out smaller?

In the simplest terms lets say you have a map in real life and the same map as an image.

You have the map as an image as an AR target and have a plane mapped to exactly the
same size as the target (don't forget this in in Unity so you can see the plane) that has
the map texture also.

The expected result when you look at the map through the device camera is for the map
to be recognised and the same map be super imposed over the top in exactly the same place.

On the iPad 3 this seems substantially off where as on the iPhone 4 it doesn't seem as off.
I appreciate you will never get 1:1 mapping but there is a very large difference on the iPad 3.

I have uploaded some screen shots for iPhone 4S, iPad 2 and iPad 3.
These were using video using Mobile Movie Texture plug in for Unity but the results
are the same if just using an image plane.

If you look at the surrounds of the image you will see there is a little of the under image showing
on the iPhone 4S, none on the iPad 2 and lots on the iPad 3.

iPhone 4S

iPad 2

iPad 3

Hi Gary, I've just tried this same test on a native implementation and I see the same effect. This may be due to our iPad3 camera calibration - we will look further into it.

Thanks for raising the issue. :-)

No worries in terms of any updates from you guys will they be in this thread, a more general manner or both?

Thanks.

More general, but if I remember I'll update it here as well.

Lovely, don't forget that the issue also appears on iPhone 4 and 4S, albeit to a lesser extent.