Conjugate Product Graphs for Globally Optimal 2D-3D Shape Matching

Ours → Results of Lähner et al. [28] (top) and ours (bottom) on the TOSCA dataset. (i) Matching with our approach (ii) 2D to 3D deformation transfer Figure 1 . We propose a novel formalism for globally optimal 2D contour to 3D shape matching based on shortest paths in the conjugate product graph. For the first time we make it possible to incorporate higher-order costs within a shortest path-based matching formalism, which in turn enables to integrate powerful priors, e.g. favouring locally rigid deformations. Left: Our method produces compelling 2D-3D matchings that significantly outperform the previous state of the art [28] . Right: Sketch-based 2D to 3D deformation transfer by (i) computing a 2D-3D matching using our approach, (ii) manipulating the 2D sketch, and then transferring 2D deformations to the 3D shape.