We started to demonstrate our prototype to potential furniture buyers and what returned was quite surprising. Because of the lack of realism in the scene rendering ( as mentioned above ) our tool did not have the impact we were expecting.

Most of these people mentioned that although it was a “cool app”:

- camera calibration process should be automatic ( marker based is the best solution for now [ see mateio ] - marker less is a must but this is hard )
- the room has to obviously be empty or almost !
- if the rendered furniture is not visually close from reality this is pretty useless unless you consider this tool as a room planner and not a product showcase app
- people want customization features ( color / textures / size variations ….) when possible

So this is maybe a reason why many people prefer to use room planning tools like floorplanner.com (pretty successful) and guys from companies like squareclock focus on rendering quality.

My 2 cents.

This is not really new as there are already many companies like metaio.com working on this king of augmented reality apps.

Now room calibration is a thing but there are many other issues that need to be addressed like the creation of the 3D scaled/textured models database (huge work on the furniture vendor side) and as importantly being able to create realistic renderings (meaning taking into account the room lighting conditions / companies like squareclock.com are working on this aspect even if their solution is not AR based).

I wrote an article on this topic further discussing your post.
Augmented Reality for Home Furniture Placement. Welcome your comment too.