Effective Rotation-invariant Point CNN with Spherical Harmonics kernels - Inria - Institut national de recherche en sciences et technologies du numérique Access content directly
Conference Papers Year :

Effective Rotation-invariant Point CNN with Spherical Harmonics kernels

Abstract

We present a novel rotation invariant architecture operating directly on point cloud data. We demonstrate how rotation invariance can be injected into a recently proposed point-based PCNN architecture, on all layers of the network. This leads to invariance to both global shape transformations, and to local rotations on the level of patches or parts, useful when dealing with non-rigid objects. We achieve this by employing a spherical harmonics-based kernel at different layers of the network, which is guaranteed to be invariant to rigid motions. We also introduce a more efficient pooling operation for PCNN using space-partitioning data-structures. This results in a flexible, simple and efficient architecture that achieves accurate results on challenging shape analysis tasks, including classification and segmentation, without requiring data-augmentation typically employed by non-invariant approaches.
Fichier principal
Vignette du fichier
Spherical_harmonic_cnn_Pro_version.pdf (4.33 Mo) Télécharger le fichier
Origin : Files produced by the author(s)
Loading...

Dates and versions

hal-02167454 , version 1 (27-06-2019)
hal-02167454 , version 2 (16-09-2019)

Identifiers

  • HAL Id : hal-02167454 , version 2

Cite

Adrien Poulenard, Marie-Julie Rakotosaona, Yann Ponty, Maks Ovsjanikov. Effective Rotation-invariant Point CNN with Spherical Harmonics kernels. 3DV - International Conference on 3D Vision - 2019, Sep 2019, Québec City, Canada. ⟨hal-02167454v2⟩
238 View
936 Download

Share

Gmail Facebook Twitter LinkedIn More