Furniture recognition with deep learning

I am wondering if good results could be archieved in image recognition for furniture. You make a picture of a piece of furniture (eg a chair) and then this will be compared with a database of furniture images and it will show you the result that is most similar to it. What accuracy could be expected here? Google reverse image search is performing pretty bad at this task. However image classification APIs like clarifai are producing quite good results for general classification. Could similiar results be achieved for classification between different types of chairs, tables,... if you trained the network on these data? I could imagine that it might be difficult as the differences between two chairs can be subtle.

I would just like to have some expert opinions before I spend a lot of time on a task which might not be possible to solve with our current techniques.