Most shape-based texture-less object detection, with an edge map as an input, are slow and unsuitable for real-time performance. This research aims at detecting multiple 3D objects (given 2D viewpoints) using shape information.
Initial results are available here.
New results are available here.