English

Geometry-Consistent Neural Shape Representation with Implicit Displacement Fields

Computer Vision and Pattern Recognition 2022-02-03 v3 Graphics Machine Learning

Abstract

We present implicit displacement fields, a novel representation for detailed 3D geometry. Inspired by a classic surface deformation technique, displacement mapping, our method represents a complex surface as a smooth base surface plus a displacement along the base's normal directions, resulting in a frequency-based shape decomposition, where the high frequency signal is constrained geometrically by the low frequency signal. Importantly, this disentanglement is unsupervised thanks to a tailored architectural design that has an innate frequency hierarchy by construction. We explore implicit displacement field surface reconstruction and detail transfer and demonstrate superior representational power, training stability and generalizability.

Keywords

Cite

@article{arxiv.2106.05187,
  title  = {Geometry-Consistent Neural Shape Representation with Implicit Displacement Fields},
  author = {Wang Yifan and Lukas Rahmann and Olga Sorkine-Hornung},
  journal= {arXiv preprint arXiv:2106.05187},
  year   = {2022}
}

Comments

Accepted to ICLR 2022, 20 pages including appendix. Code available at https://github.com/yifita/idf

R2 v1 2026-06-24T03:01:06.038Z