In this research project by Taehee Lee and Tobias Hollerer from the UCSB "Four Eyes Lab", you see a desktop augmented by computer graphics in real time using a markerless camera tracking algorithm. The method uses landmark (SIFT) features and sparse optical flow. Even though SIFT detection is slow (~0.5s on our laptop), we achieve realtime performance thanks to a special multi-threaded software design. We use our HandyAR hand tracking methodology for the Augmented Reality interaction. See http://ilab.cs.ucsb.edu/projects/taehee/HandyAR/ (IEEE VR 2008) (less info)