The project SLAM4AR is focused on developing SLAM methods that enable applications in the area of augmented reality. Such augmented reality systems have various constraints: First, the precision of tracking and reconstruction should be high - also when moving over larger distances. Second, the methods need to be fast so that they can be run in realtime. And third, methods should handle setups with one or two cameras, with and without IMU information. We have therefore developed suitable methods for visual SLAM and visual odometry that address these requirements. Moreover, we have demonstrated virtual-object insertion in realtime demos to show the potential for augmented reality.
