This work introduces a way to build up and use an extensive sensor-independent object model database. In a first step, a cost-effective and computationally cheap way to create colored point cloud models from common household objects by using a Microsoft Kinect camera is presented. Those object models are stored in a world-wide accessible, distributed database called RoboEarth. Finally, the models are used for recognizing the corresponding objects with any kind of camera. In the presented implementation the demonstration was done with both a Kinect and common RGB cameras.