Performs object segmentation and simple object recognition for constrained scenes.
Given a point cloud, will attempt to identify and localize known objects.
Works for detecting objects from the IKEA dataset, assuming they are sitting on a table, which is the dominant plane in a scan. It will first find the dominant plane, then cluster objects based on the projections of the points on the plane. For each cluster, it will then attempt to fit the models of the IKEA objects. Models are assumed to be upright on the table and rotationally symmetric.
Functionality is broken down into two main parts: clustering and model fitting.