This pipeline implements LINE-MOD. For more information, see http://ar.in.tum.de/Main/StefanHinterstoisser. It is one of the best methods out there for generic rigid object recognition and it proceeds using very fast template matching. The version used in that package is the same as the original paper (and it is in OpenCV) but the tricks to get it work properly are in the pre/post processing steps.
To generate all the templates, it uses an automatic view generator. Those vary the in-plane rotations of the cameras and deals with the different scales and view points. Thousands of images with depth+mask are generated and fed to the OpenCV trainer.
The OpenCV detector is simply called
TODO: implement an ICP step like the ACCV paper, probably: Linear Least-Squares Optimization for Point-to-Plane ICP Surface Registration from Kok-Lim Low.