eb86d510361fc23b59f18c1bc9802cc6-Reviews.html

This paper proposes an algorithm to compute k-prototypes for a set of 3D rigid structures. Two structures are called rigid, if one can align them via translation, reflection and rotation. The authors first propose a 1-prototype algorithm to find the best representative 3D structure (the one minimizing the distance to all other structure subject to optimal rigid alignement). For k-prototype clustering, the algorithm then consists of building a correlation graph and using [11] to perform k-median clustering. For each obtained cluster, the 1-prototype algorithm is applied.