My undergraduate dissertation was an exercise in paper modelling using the GPU and ultimately the geometry shader.
The abstract is included below and the full version can be downloaded.
This research sought to create a model of paper through the use of GPU hardware and shader programs with a minimal number of polygons. The main aim of this research was to investigate the possibility of rendering a paper model that did not adhere to traditional techniques. This research proposed a method of using ray intersections to limit the size of the paper when curved and to therefore mathematically constrain the paper in real time. The other solution proposed by this software was to mathematically constrain the dimensions of the paper without the use of ray intersections. The solutions proposed by this research had a simple curvature of applied to them.
This research found that the use of ray intersections was very difficult to implement with a polygon based model and is perhaps better suited to the creation of a model that is built in a similar manner to volume rendering . It was also discovered that the second method proposed by this research was easy to implement and therefore would provide a good starting point for future research.