Discrete Point Flow Networks for Efficient Point Cloud Generation

Roman Klokov; Edmond Boyer; Jakob Verbeek

doi:10.1007/978-3-030-58592-1_41

Discrete Point Flow Networks for Efficient Point Cloud Generation

Roman Klokov
, Edmond Boyer
, Jakob Verbeek

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

Abstract

Generative models have proven effective at modeling 3D shapes and their statistical variations. In this paper we investigate their application to point clouds, a 3D shape representation widely used in computer vision for which, however, only few generative models have yet been proposed. We introduce a latent variable model that builds on normalizing flows with affine coupling layers to generate 3D point clouds of an arbitrary size given a latent shape representation. To evaluate its benefits for shape modeling we apply this model for generation, autoencoding, and single-view shape reconstruction tasks. We improve over recent GAN-based models in terms of most metrics that assess generation and autoencoding. Compared to recent work based on continuous flows, our model offers a significant speedup in both training and inference times for similar or better performance. For single-view shape reconstruction we also obtain results on par with state-of-the-art voxel, point cloud, and mesh-based methods.

Original language	English
Title of host publication	Computer Vision – ECCV 2020 - 16th European Conference, Glasgow, 2020, Proceedings
Editors	Andrea Vedaldi, Horst Bischof, Thomas Brox, Jan-Michael Frahm
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	694-710
Number of pages	17
ISBN (Print)	9783030585914
DOIs	https://doi.org/10.1007/978-3-030-58592-1_41
Publication status	Published - 1 Jan 2020
Externally published	Yes
Event	16th European Conference on Computer Vision, ECCV 2020 - Glasgow, United Kingdom Duration: 23 Aug 2020 → 28 Aug 2020

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	12368 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	16th European Conference on Computer Vision, ECCV 2020
Country/Territory	United Kingdom
City	Glasgow
Period	23/08/20 → 28/08/20

Keywords

3D shape modeling
Generative modeling
Normalizing flows
Point cloud generation
Single view reconstruction

Access to Document

10.1007/978-3-030-58592-1_41

Cite this

Klokov, R., Boyer, E., & Verbeek, J. (2020). Discrete Point Flow Networks for Efficient Point Cloud Generation. In A. Vedaldi, H. Bischof, T. Brox, & J.-M. Frahm (Eds.), Computer Vision – ECCV 2020 - 16th European Conference, Glasgow, 2020, Proceedings (pp. 694-710). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 12368 LNCS). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-030-58592-1_41

Klokov, Roman ; Boyer, Edmond ; Verbeek, Jakob. / Discrete Point Flow Networks for Efficient Point Cloud Generation. Computer Vision – ECCV 2020 - 16th European Conference, Glasgow, 2020, Proceedings. editor / Andrea Vedaldi ; Horst Bischof ; Thomas Brox ; Jan-Michael Frahm. Springer Science and Business Media Deutschland GmbH, 2020. pp. 694-710 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{4fe92ed9f6bd430fbbdc8418edfa78f9,

title = "Discrete Point Flow Networks for Efficient Point Cloud Generation",

abstract = "Generative models have proven effective at modeling 3D shapes and their statistical variations. In this paper we investigate their application to point clouds, a 3D shape representation widely used in computer vision for which, however, only few generative models have yet been proposed. We introduce a latent variable model that builds on normalizing flows with affine coupling layers to generate 3D point clouds of an arbitrary size given a latent shape representation. To evaluate its benefits for shape modeling we apply this model for generation, autoencoding, and single-view shape reconstruction tasks. We improve over recent GAN-based models in terms of most metrics that assess generation and autoencoding. Compared to recent work based on continuous flows, our model offers a significant speedup in both training and inference times for similar or better performance. For single-view shape reconstruction we also obtain results on par with state-of-the-art voxel, point cloud, and mesh-based methods.",

keywords = "3D shape modeling, Generative modeling, Normalizing flows, Point cloud generation, Single view reconstruction",

author = "Roman Klokov and Edmond Boyer and Jakob Verbeek",

note = "Publisher Copyright: {\textcopyright} 2020, Springer Nature Switzerland AG.; 16th European Conference on Computer Vision, ECCV 2020 ; Conference date: 23-08-2020 Through 28-08-2020",

year = "2020",

month = jan,

day = "1",

doi = "10.1007/978-3-030-58592-1\_41",

language = "English",

isbn = "9783030585914",

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 = "694--710",

editor = "Andrea Vedaldi and Horst Bischof and Thomas Brox and Jan-Michael Frahm",

booktitle = "Computer Vision – ECCV 2020 - 16th European Conference, Glasgow, 2020, Proceedings",

}

Klokov, R, Boyer, E & Verbeek, J 2020, Discrete Point Flow Networks for Efficient Point Cloud Generation. in A Vedaldi, H Bischof, T Brox & J-M Frahm (eds), Computer Vision – ECCV 2020 - 16th European Conference, Glasgow, 2020, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 12368 LNCS, Springer Science and Business Media Deutschland GmbH, pp. 694-710, 16th European Conference on Computer Vision, ECCV 2020, Glasgow, United Kingdom, 23/08/20. https://doi.org/10.1007/978-3-030-58592-1_41

Discrete Point Flow Networks for Efficient Point Cloud Generation. / Klokov, Roman; Boyer, Edmond; Verbeek, Jakob.
Computer Vision – ECCV 2020 - 16th European Conference, Glasgow, 2020, Proceedings. ed. / Andrea Vedaldi; Horst Bischof; Thomas Brox; Jan-Michael Frahm. Springer Science and Business Media Deutschland GmbH, 2020. p. 694-710 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 12368 LNCS).

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

TY - GEN

T1 - Discrete Point Flow Networks for Efficient Point Cloud Generation

AU - Klokov, Roman

AU - Boyer, Edmond

AU - Verbeek, Jakob

PY - 2020/1/1

Y1 - 2020/1/1

N2 - Generative models have proven effective at modeling 3D shapes and their statistical variations. In this paper we investigate their application to point clouds, a 3D shape representation widely used in computer vision for which, however, only few generative models have yet been proposed. We introduce a latent variable model that builds on normalizing flows with affine coupling layers to generate 3D point clouds of an arbitrary size given a latent shape representation. To evaluate its benefits for shape modeling we apply this model for generation, autoencoding, and single-view shape reconstruction tasks. We improve over recent GAN-based models in terms of most metrics that assess generation and autoencoding. Compared to recent work based on continuous flows, our model offers a significant speedup in both training and inference times for similar or better performance. For single-view shape reconstruction we also obtain results on par with state-of-the-art voxel, point cloud, and mesh-based methods.

AB - Generative models have proven effective at modeling 3D shapes and their statistical variations. In this paper we investigate their application to point clouds, a 3D shape representation widely used in computer vision for which, however, only few generative models have yet been proposed. We introduce a latent variable model that builds on normalizing flows with affine coupling layers to generate 3D point clouds of an arbitrary size given a latent shape representation. To evaluate its benefits for shape modeling we apply this model for generation, autoencoding, and single-view shape reconstruction tasks. We improve over recent GAN-based models in terms of most metrics that assess generation and autoencoding. Compared to recent work based on continuous flows, our model offers a significant speedup in both training and inference times for similar or better performance. For single-view shape reconstruction we also obtain results on par with state-of-the-art voxel, point cloud, and mesh-based methods.

KW - 3D shape modeling

KW - Generative modeling

KW - Normalizing flows

KW - Point cloud generation

KW - Single view reconstruction

U2 - 10.1007/978-3-030-58592-1_41

DO - 10.1007/978-3-030-58592-1_41

M3 - Conference contribution

AN - SCOPUS:85097409856

SN - 9783030585914

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

SP - 694

EP - 710

BT - Computer Vision – ECCV 2020 - 16th European Conference, Glasgow, 2020, Proceedings

A2 - Vedaldi, Andrea

A2 - Bischof, Horst

A2 - Brox, Thomas

A2 - Frahm, Jan-Michael

PB - Springer Science and Business Media Deutschland GmbH

T2 - 16th European Conference on Computer Vision, ECCV 2020

Y2 - 23 August 2020 through 28 August 2020

ER -

Klokov R, Boyer E, Verbeek J. Discrete Point Flow Networks for Efficient Point Cloud Generation. In Vedaldi A, Bischof H, Brox T, Frahm JM, editors, Computer Vision – ECCV 2020 - 16th European Conference, Glasgow, 2020, Proceedings. Springer Science and Business Media Deutschland GmbH. 2020. p. 694-710. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-030-58592-1_41

Discrete Point Flow Networks for Efficient Point Cloud Generation

Abstract

Publication series

Conference

Keywords

Access to Document

Fingerprint

Cite this