Introduction

The contest is finished and all results can be found in the final publication [3].

Matching deformable 3D shapes is an active area of research, which has attracted the interest of many researchers during the years. A wide variety of approaches have been proposed to tackle the problem of (nearly-)isometric shape matching with different levels of robustness against topological changes and geometric noise. Standard data sets typically concentrate on the type of non-rigid deformation (i.e. change in pose and in shape class) and only include minor factors of nuisance, such as small holes and numerical noise. By contrast, the topological change of triangular meshes due to the coalescence of spatially close surface regions have been much less investigated during the years; few datasets include such instances, and these are limited to the coalescence of small areas. From a real world perspective, during a typical 3D acquisition process parts that touch each other are hidden from the sensors and will create a topologically different surface. For example, arms hanging down and along the body of a human shape may touch the torso and the legs, and there the shape is coalesced although they are not physically connected. These topological changes are very hard to match and, to the best of our knowledge, there exists no benchmark focusing on this class of changes.

Our data set includes shapes from the KIDS dataset [1] and new shapes produced with a free version of DAZ Studio 4.9, but merges parts that are self-intersecting so as to create a non-intersecting manifold surface using the method described in [2]. This changes the topology of each shape in several regions and to various degrees, creating a diverse but realistic data set.


In the original shape, arm and fingers intersect with the body and leg respectively, penetrating the shape. In the new shape the intersecting parts are removed, and the visible parts are stitched to the body.

References

  1. Dense Non-Rigid Shape Correspondence Using Random Forests. E. Rodola, S. Rota Bulo, T. Windheuser, M. Vestner, and D. Cremers, In IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2014
  2. Exact and Robust (Self-)Intersections for Polygonal Meshes. Marcel Campen and Leif Kobbelt, Eurographics, 2010
  3. SHREC’16: Matching of Deformable Shapes with Topological Noise. Z. Lähner and E. Rodolà and M. M. Bronstein and D. Cremers and O. Burghard and L. Cosmo and A. Dieckmann and R. Klein and Y. Sahillioglu, In Proc. of 3DOR, 2016