Neural Articulated Shape Approximation

Boyang Deng1, JP Lewis1, Timothy Jeruzalski1, Gerard Pons-Moll2, Geoffrey Hinton1, Mohammad Norouzi1, Andrea Tagliasacchi1,3

1Google Research
2Max Planck Institute for Informatics, Saarland Informatics Campus, Germany
3University of Toronto, Canada 

ECCV 2020

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.

Reconstruction

image image


Tracking

image image


Citation

@inproceedings{deng2019neural,
    title = {Neural Articulated Shape Approximation},
    author = {Deng, Boyang and Lewis, JP and Jeruzalski, Timothy and Pons-Moll, Gerard and Hinton, Geoffrey and Norouzi, Mohammad and Tagliasacchi, Andrea},
    booktitle = {The European Conference on Computer Vision (ECCV)},
    month = {August},
    organization = {{Springer}},
    year = {2020},
}