English

NASA: Neural Articulated Shape Approximation

Computer Vision and Pattern Recognition 2022-07-25 v5 Graphics Machine Learning

Abstract

Efficient representation of articulated objects such as human bodies is an important problem in computer vision and graphics. To efficiently simulate deformation, existing approaches represent 3D objects using polygonal meshes and deform them using skinning techniques. This paper introduces neural articulated shape approximation (NASA), an alternative framework that enables efficient representation of articulated deformable objects using neural indicator functions that are conditioned on pose. Occupancy testing using NASA is straightforward, circumventing the complexity of meshes and the issue of water-tightness. We demonstrate the effectiveness of NASA for 3D tracking applications, and discuss other potential extensions.

Keywords

Cite

@article{arxiv.1912.03207,
  title  = {NASA: Neural Articulated Shape Approximation},
  author = {Boyang Deng and JP Lewis and Timothy Jeruzalski and Gerard Pons-Moll and Geoffrey Hinton and Mohammad Norouzi and Andrea Tagliasacchi},
  journal= {arXiv preprint arXiv:1912.03207},
  year   = {2022}
}

Comments

ECCV 2020; Project Page: https://nasa-eccv20.github.io/

R2 v1 2026-06-23T12:38:14.655Z