3D Segmentation Benchmark-Repository
Genious Web

3D Segmentation Benchmark

Repository of static 3D-meshes

The models of the corpus come from the GAMMA database from INRIA, from the Princeton Shape Benchmark , from AIM@SHAPE, and from The Stanford 3D Scanning Repository.

Nowadays, the static 3D-mesh repository includes twenty-eight 3D-static meshes categorized in five classes (animal, furniture, hand, human, statuette). Each mesh is provided with four manually edited ground-truth segmentations.

Last Update: 2009-06-15

Each class of models is stored in a particular page:

  • animal: animals as horse, eagle, etc.
  • furniture: furnitures as chair, table, etc.
  • hand: human hands in different postures.
  • human: humanoids as baby, robot, etc.
  • bust: statuettes as bimba, vaselion, etc.
  • Each directory contains the following information:

  • modelname.wrl, modelname.ply and modelname.off: the model modelname in VRML, PLY and OFF formats without any color;
  • modelname.n.wrl: the n-th ground-truth segmentation of the model modelname in VRML format colored with a color per-segment;
  • modelname.n.offc: the n-th ground-truth segmentation of the model modelname in OFFc format ;
  • modelname.n.lab: the label file of the n-th ground-truth segmentation of the model modelname.
  • downlaod Download the whole corpus with ground-truth segmentations.