In the last decade, object recognition placed on tabletop has been advertised as a near future ability on next generation tabletops by several major information technology companies. Markers-based recognition seems the straightforward solution, but one can easily understand that this may not be a lasting solution for real word scenarios. In this paper, we study the marker-free object recognition ability of a commonly used tabletop. This tabletop does not rely on external cameras and thus ts perfectly a future connected house scenario. A proof-of-concept system has been developed and experimented on various objects. Based on this study, we raise some conclusions and perspectives on both software and hardware levels.