I've been working with the object/toy recognition stuff and it seems pretty spotty, In particular, figures with shiny paint jobs don't do very well. I suspect this is because the specular reflections on reflective paint create trackable detail points which are inaccurate since the reflections change with the lighitng environment. But then again, wouldn't shadows do that too?
I scanned my toy in a white product photography box--and it recognizes it there. But if I put the object on a table or something elsewhere in the room, it rarely recognies it.
Are there guideilnes for what makes a highly trackable 3D object? (Like how you have guidelines for Image Target creation best practices?)