Shape-Pose Disentanglement Using SE(3)-Equivariant Vector Neurons

Oren Katzir; Dani Lischinski; Daniel Cohen-Or

doi:10.1007/978-3-031-20062-5_27

Shape-Pose Disentanglement Using SE(3)-Equivariant Vector Neurons

Oren Katzir^*
, Dani Lischinski
, Daniel Cohen-Or

^*Corresponding author for this work

The Rachel and Selim Benin School of Engineering and Computer Science

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

12 Scopus citations

Abstract

We introduce an unsupervised technique for encoding point clouds into a canonical shape representation, by disentangling shape and pose. Our encoder is stable and consistent, meaning that the shape encoding is purely pose-invariant, while the extracted rotation and translation are able to semantically align different input shapes of the same class to a common canonical pose. Specifically, we design an auto-encoder based on Vector Neuron Networks, a rotation-equivariant neural network, whose layers we extend to provide translation-equivariance in addition to rotation-equivariance only. The resulting encoder produces pose-invariant shape encoding by construction, enabling our approach to focus on learning a consistent canonical pose for a class of objects. Quantitative and qualitative experiments validate the superior stability and consistency of our approach.

Original language	English
Title of host publication	Computer Vision – ECCV 2022 - 17th European Conference, Proceedings
Editors	Shai Avidan, Gabriel Brostow, Moustapha Cissé, Giovanni Maria Farinella, Tal Hassner
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	468-484
Number of pages	17
ISBN (Print)	9783031200618
DOIs	https://doi.org/10.1007/978-3-031-20062-5_27
State	Published - 2022
Event	17th European Conference on Computer Vision, ECCV 2022 - Tel Aviv, Israel Duration: 23 Oct 2022 → 27 Oct 2022

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	13663 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	17th European Conference on Computer Vision, ECCV 2022
Country/Territory	Israel
City	Tel Aviv
Period	23/10/22 → 27/10/22

Bibliographical note

Publisher Copyright:
© 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.

Keywords

Canonical pose
Equivariance
Point clouds
Shape-pose disentanglement

Access to Document

10.1007/978-3-031-20062-5_27

Cite this

Katzir, O., Lischinski, D., & Cohen-Or, D. (2022). Shape-Pose Disentanglement Using SE(3)-Equivariant Vector Neurons. In S. Avidan, G. Brostow, M. Cissé, G. M. Farinella, & T. Hassner (Eds.), Computer Vision – ECCV 2022 - 17th European Conference, Proceedings (pp. 468-484). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 13663 LNCS). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-031-20062-5_27

Katzir, Oren ; Lischinski, Dani ; Cohen-Or, Daniel. / Shape-Pose Disentanglement Using SE(3)-Equivariant Vector Neurons. Computer Vision – ECCV 2022 - 17th European Conference, Proceedings. editor / Shai Avidan ; Gabriel Brostow ; Moustapha Cissé ; Giovanni Maria Farinella ; Tal Hassner. Springer Science and Business Media Deutschland GmbH, 2022. pp. 468-484 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{b8d291f1776c4ed2bcef305ed6bd6fd5,

title = "Shape-Pose Disentanglement Using SE(3)-Equivariant Vector Neurons",

abstract = "We introduce an unsupervised technique for encoding point clouds into a canonical shape representation, by disentangling shape and pose. Our encoder is stable and consistent, meaning that the shape encoding is purely pose-invariant, while the extracted rotation and translation are able to semantically align different input shapes of the same class to a common canonical pose. Specifically, we design an auto-encoder based on Vector Neuron Networks, a rotation-equivariant neural network, whose layers we extend to provide translation-equivariance in addition to rotation-equivariance only. The resulting encoder produces pose-invariant shape encoding by construction, enabling our approach to focus on learning a consistent canonical pose for a class of objects. Quantitative and qualitative experiments validate the superior stability and consistency of our approach.",

keywords = "Canonical pose, Equivariance, Point clouds, Shape-pose disentanglement",

author = "Oren Katzir and Dani Lischinski and Daniel Cohen-Or",

note = "Publisher Copyright: {\textcopyright} 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.; 17th European Conference on Computer Vision, ECCV 2022 ; Conference date: 23-10-2022 Through 27-10-2022",

year = "2022",

doi = "10.1007/978-3-031-20062-5\_27",

language = "אנגלית",

isbn = "9783031200618",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Science and Business Media Deutschland GmbH",

pages = "468--484",

editor = "Shai Avidan and Gabriel Brostow and Moustapha Ciss{\'e} and Farinella, \{Giovanni Maria\} and Tal Hassner",

booktitle = "Computer Vision – ECCV 2022 - 17th European Conference, Proceedings",

address = "גרמניה",

}

Katzir, O, Lischinski, D & Cohen-Or, D 2022, Shape-Pose Disentanglement Using SE(3)-Equivariant Vector Neurons. in S Avidan, G Brostow, M Cissé, GM Farinella & T Hassner (eds), Computer Vision – ECCV 2022 - 17th European Conference, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 13663 LNCS, Springer Science and Business Media Deutschland GmbH, pp. 468-484, 17th European Conference on Computer Vision, ECCV 2022, Tel Aviv, Israel, 23/10/22. https://doi.org/10.1007/978-3-031-20062-5_27

Shape-Pose Disentanglement Using SE(3)-Equivariant Vector Neurons. / Katzir, Oren; Lischinski, Dani; Cohen-Or, Daniel.
Computer Vision – ECCV 2022 - 17th European Conference, Proceedings. ed. / Shai Avidan; Gabriel Brostow; Moustapha Cissé; Giovanni Maria Farinella; Tal Hassner. Springer Science and Business Media Deutschland GmbH, 2022. p. 468-484 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 13663 LNCS).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Shape-Pose Disentanglement Using SE(3)-Equivariant Vector Neurons

AU - Katzir, Oren

AU - Lischinski, Dani

AU - Cohen-Or, Daniel

PY - 2022

Y1 - 2022

N2 - We introduce an unsupervised technique for encoding point clouds into a canonical shape representation, by disentangling shape and pose. Our encoder is stable and consistent, meaning that the shape encoding is purely pose-invariant, while the extracted rotation and translation are able to semantically align different input shapes of the same class to a common canonical pose. Specifically, we design an auto-encoder based on Vector Neuron Networks, a rotation-equivariant neural network, whose layers we extend to provide translation-equivariance in addition to rotation-equivariance only. The resulting encoder produces pose-invariant shape encoding by construction, enabling our approach to focus on learning a consistent canonical pose for a class of objects. Quantitative and qualitative experiments validate the superior stability and consistency of our approach.

AB - We introduce an unsupervised technique for encoding point clouds into a canonical shape representation, by disentangling shape and pose. Our encoder is stable and consistent, meaning that the shape encoding is purely pose-invariant, while the extracted rotation and translation are able to semantically align different input shapes of the same class to a common canonical pose. Specifically, we design an auto-encoder based on Vector Neuron Networks, a rotation-equivariant neural network, whose layers we extend to provide translation-equivariance in addition to rotation-equivariance only. The resulting encoder produces pose-invariant shape encoding by construction, enabling our approach to focus on learning a consistent canonical pose for a class of objects. Quantitative and qualitative experiments validate the superior stability and consistency of our approach.

KW - Canonical pose

KW - Equivariance

KW - Point clouds

KW - Shape-pose disentanglement

UR - https://www.scopus.com/pages/publications/85144562058

U2 - 10.1007/978-3-031-20062-5_27

DO - 10.1007/978-3-031-20062-5_27

M3 - ???researchoutput.researchoutputtypes.contributiontobookanthology.conference???

AN - SCOPUS:85144562058

SN - 9783031200618

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 468

EP - 484

BT - Computer Vision – ECCV 2022 - 17th European Conference, Proceedings

A2 - Avidan, Shai

A2 - Brostow, Gabriel

A2 - Cissé, Moustapha

A2 - Farinella, Giovanni Maria

A2 - Hassner, Tal

PB - Springer Science and Business Media Deutschland GmbH

T2 - 17th European Conference on Computer Vision, ECCV 2022

Y2 - 23 October 2022 through 27 October 2022

ER -

Katzir O, Lischinski D, Cohen-Or D. Shape-Pose Disentanglement Using SE(3)-Equivariant Vector Neurons. In Avidan S, Brostow G, Cissé M, Farinella GM, Hassner T, editors, Computer Vision – ECCV 2022 - 17th European Conference, Proceedings. Springer Science and Business Media Deutschland GmbH. 2022. p. 468-484. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-031-20062-5_27

Shape-Pose Disentanglement Using SE(3)-Equivariant Vector Neurons

Abstract

Publication series

Conference

Bibliographical note

Keywords

Access to Document

Other files and links

Fingerprint

Cite this