In 2006, the best full technical paper at the prestigious ACM Multimedia conference was entitled "Tiling Slideshow". The system described in this paper automatically generates a photo slideshow that is driven by music beats and in which similar photos are displayed in the same frame. A sophisticated photo tiling mechanism greatly improves the viewing experience as images are not displayed one-by-one. The slideshow is not only nice to watch, it is indeed a real pleasure as music plays a central role.
The goal of this 4 to 6 months internship is to implement the algorithms described in this paper. The internship must, however, not be seen as a pure implementation experience. I will be much more rewarding because the intern will get hands on many key elements of advanced computer science such as image analysis, music analysis (for extracting relevant information from images and for dealing with music beats), high-dimensional clustering (to group together similar images), face detection, smart image cropping, ... Furthermore, it will eventually be a very enjoyable project once all the pieces fit together -- your friends will love it! Of course, any original enhancement will further improve the result.
The implementation will be done on top of an existing photo browser we are developing in cooperation with Reykjavík University; that is, the photo browser will supply the photos that must be viewed together along with attributes necessary for clustering. Excellent programming skills in C++ and/or Python are mandatory. Note that some of the building blocks needed to implement the full-fledged system are available off-the-shelf.