CNNs on surfaces using rotation-equivariant features

@article{Wiersma2020CNNsOS,
  title={CNNs on surfaces using rotation-equivariant features},
  author={Ruben Wiersma and E. Eisemann and K. Hildebrandt},
  journal={ACM Transactions on Graphics (TOG)},
  year={2020},
  volume={39},
  pages={92:1 - 92:12}
}
  • Ruben Wiersma, E. Eisemann, K. Hildebrandt
  • Published 2020
  • Computer Science
  • ACM Transactions on Graphics (TOG)
  • This paper is concerned with a fundamental problem in geometric deep learning that arises in the construction of convolutional neural networks on surfaces. Due to curvature, the transport of filter kernels on surfaces results in a rotational ambiguity, which prevents a uniform alignment of these kernels on the surface. We propose a network architecture for surfaces that consists of vector-valued, rotation-equivariant features. The equivariance property makes it possible to locally align… CONTINUE READING

    References

    Publications referenced by this paper.
    SHOWING 1-10 OF 13 REFERENCES
    Convolutional neural networks on surfaces via seamless toric covers
    97
    Learning shape correspondence with anisotropic convolutional neural networks
    283
    ZerNet: Convolutional Neural Networks on Arbitrary Surfaces Via Zernike Local Tangent Space Estimation
    4
    PFCNN: Convolutional Neural Networks on 3D Surfaces Using Parallel Frames
    18
    Geodesic Convolutional Neural Networks on Riemannian Manifolds
    412
    Surface Networks via General Covers
    8
    Harmonic Networks: Deep Translation and Rotation Equivariance
    269
    Geometric Deep Learning on Graphs and Manifolds Using Mixture Model CNNs
    651
    Spectral Networks and Locally Connected Networks on Graphs
    1522
    The Vector Heat Method
    17