Vector Graph Representation for Deformation Transfer Using Poisson Interpolation

Abstract

Given an action sequence of a source object, the deformation transfer (DT) generates similar action sequence for a target object. Usually, the 3D shape model of an object is stored as a mesh. In this work, we propose a vector graph (VG) representation for deformation transfer framework. Furthermore, the VG representation is suitable even for problems involving skeleton of the shape, for example human gait. We pose the deformation transfer as the Poisson interpolation problem. The cue about the intermediate poses of the target object is captured by the temporal gradient of the source pose sequence. With an assumption that the reference poses of the target and the source are similar, and given the initial and final pose of the target, Poisson interpolation generates intermediate deformed poses of the target corresponding to the source deformation sequence. The proposed deformation transfer scheme preserves target shape. Moreover, the source sequence gradient information guiding the process of deformation ensures that the target deformation sequence imitates the source action. The qualitative and quantitative comparisons of the proposed method with three state-of-the-art methods also show the effectiveness of the proposed method.

Cite

Text

Domadiya et al. "Vector Graph Representation for Deformation Transfer Using Poisson Interpolation." IEEE/CVF Winter Conference on Applications of Computer Vision, 2018. doi:10.1109/WACV.2018.00101

Markdown

[Domadiya et al. "Vector Graph Representation for Deformation Transfer Using Poisson Interpolation." IEEE/CVF Winter Conference on Applications of Computer Vision, 2018.](https://mlanthology.org/wacv/2018/domadiya2018wacv-vector/) doi:10.1109/WACV.2018.00101

BibTeX

@inproceedings{domadiya2018wacv-vector,
  title     = {{Vector Graph Representation for Deformation Transfer Using Poisson Interpolation}},
  author    = {Domadiya, Prashant and Shah, Pratik and Mitra, Suman K.},
  booktitle = {IEEE/CVF Winter Conference on Applications of Computer Vision},
  year      = {2018},
  pages     = {876-884},
  doi       = {10.1109/WACV.2018.00101},
  url       = {https://mlanthology.org/wacv/2018/domadiya2018wacv-vector/}
}