@visionlib The gif looks very nice. Can the solution locate and recognize the object completely without prior frames of information - or does it need a good initial calibration to the object? i.e. can I turn on the camera, look into a scene and get 6DOF tracking for anything there?