Synthetic Ground Truth for Presegmentation of Known Objects for Effortless Pose Estimation

Frederik Haarslev; William Kristian Juel; Norbert Krüger; Leon Bodenhagen

doi:10.5220/0009163904820489

Synthetic Ground Truth for Presegmentation of Known Objects for Effortless Pose Estimation

Frederik Haarslev, William Kristian Juel, Norbert Krüger, Leon Bodenhagen

Publikation: Kapitel i bog/rapport/konference-proceeding › Konferencebidrag i proceedings › Forskning › peer review

299 Downloads (Pure)

Abstract

We present a method for generating synthetic ground truth for training segmentation networks for presegmenting point clouds in pose estimation problems. Our method replaces global pose estimation algorithms such as RANSAC which requires manual fine-tuning with a robust CNN, without having to hand-label segmentation masks for the given object. The data is generated by blending cropped images of the objects with arbitrary backgrounds. We test the method in two scenarios, and show that networks trained on the generated data segments the objects with high accuracy, allowing them to be used in a pose estimation pipeline.

Originalsprog	Engelsk
Titel	Proceedings of the 15th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications - Volume 4: VISAPP
Redaktører	Giovanni Maria Farinella, Petia Radeva, Jose Braz
Vol/bind	4
Forlag	SCITEPRESS Digital Library
Publikationsdato	2020
Sider	482-489
ISBN (Elektronisk)	978-989-758-402-2
DOI	https://doi.org/10.5220/0009163904820489
Status	Udgivet - 2020
Begivenhed	15th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications - Valletta, Malta Varighed: 27. feb. 2020 → 29. feb. 2020 Konferencens nummer: 15

Konference

Konference	15th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications
Nummer	15
Land/Område	Malta
By	Valletta
Periode	27/02/2020 → 29/02/2020

Dokumenter og links

10.5220/0009163904820489Licens: CC BY-NC-ND

Open Access VersionLicens: CC BY-NC-ND

SDU link

Tjek adgangen til materialet i Syddansk Universitetsbiblioteks søgemaskine

1 Ph.d.-afhandling

Spatio-Temporal Vision for Human-Robot Interaction using Deep Learning
Bidragets oversatte titel: Spatio-Temporalt Vision for Menneske-Robot Interaktion via Deep LearningHaarslev, F., 19. nov. 2021, Syddansk Universitet. Det Tekniske Fakultet. 153 s.
Publikation: Afhandling › Ph.d.-afhandling

1 Afsluttet

SMOOTH: Seamless huMan-robot interactiOn fOr THe support of elderly people
Bodenhagen, L. (Projektdeltager), Palinko, O. (Projektdeltager), Manoonpong, P. (Projektdeltager), Haarslev, F. (Projektdeltager), Krüger, N. (Koordinator), Juel, W. K. (Projektdeltager), Kollakidou, A. (Projektdeltager) & Naik, L. (Projektdeltager)
01/04/2017 → 31/01/2021
Projekter: Projekt › Forskning

Citationsformater

Haarslev, F., Juel, W. K., Krüger, N., & Bodenhagen, L. (2020). Synthetic Ground Truth for Presegmentation of Known Objects for Effortless Pose Estimation. I G. M. Farinella, P. Radeva, & J. Braz (red.), Proceedings of the 15th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications - Volume 4: VISAPP (Bind 4, s. 482-489). SCITEPRESS Digital Library. https://doi.org/10.5220/0009163904820489

Haarslev, Frederik ; Juel, William Kristian ; Krüger, Norbert et al. / Synthetic Ground Truth for Presegmentation of Known Objects for Effortless Pose Estimation. Proceedings of the 15th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications - Volume 4: VISAPP. red. / Giovanni Maria Farinella ; Petia Radeva ; Jose Braz. Bind 4 SCITEPRESS Digital Library, 2020. s. 482-489

@inproceedings{d76b69dbb79f4f76b414f5a33c9f80b5,

title = "Synthetic Ground Truth for Presegmentation of Known Objects for Effortless Pose Estimation",

abstract = "We present a method for generating synthetic ground truth for training segmentation networks for presegmenting point clouds in pose estimation problems. Our method replaces global pose estimation algorithms such as RANSAC which requires manual fine-tuning with a robust CNN, without having to hand-label segmentation masks for the given object. The data is generated by blending cropped images of the objects with arbitrary backgrounds. We test the method in two scenarios, and show that networks trained on the generated data segments the objects with high accuracy, allowing them to be used in a pose estimation pipeline.",

keywords = "Object Segmentation, Pose Estimation, Synthetic Ground Truth Generation",

author = "Frederik Haarslev and Juel, {William Kristian} and Norbert Kr{\"u}ger and Leon Bodenhagen",

year = "2020",

doi = "10.5220/0009163904820489",

language = "English",

volume = "4",

pages = "482--489",

editor = "Farinella, {Giovanni Maria} and Petia Radeva and Jose Braz",

booktitle = "Proceedings of the 15th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications - Volume 4: VISAPP",

publisher = "SCITEPRESS Digital Library",

note = "15th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications , VISAPP 2020 ; Conference date: 27-02-2020 Through 29-02-2020",

}

Haarslev, F, Juel, WK, Krüger, N & Bodenhagen, L 2020, Synthetic Ground Truth for Presegmentation of Known Objects for Effortless Pose Estimation. i GM Farinella, P Radeva & J Braz (red), Proceedings of the 15th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications - Volume 4: VISAPP. bind 4, SCITEPRESS Digital Library, s. 482-489, 15th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications , Valletta, Malta, 27/02/2020. https://doi.org/10.5220/0009163904820489

Synthetic Ground Truth for Presegmentation of Known Objects for Effortless Pose Estimation. / Haarslev, Frederik; Juel, William Kristian; Krüger, Norbert et al.
Proceedings of the 15th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications - Volume 4: VISAPP. red. / Giovanni Maria Farinella; Petia Radeva; Jose Braz. Bind 4 SCITEPRESS Digital Library, 2020. s. 482-489.

Publikation: Kapitel i bog/rapport/konference-proceeding › Konferencebidrag i proceedings › Forskning › peer review

TY - GEN

T1 - Synthetic Ground Truth for Presegmentation of Known Objects for Effortless Pose Estimation

AU - Haarslev, Frederik

AU - Juel, William Kristian

AU - Krüger, Norbert

AU - Bodenhagen, Leon

N1 - Conference code: 15

PY - 2020

Y1 - 2020

N2 - We present a method for generating synthetic ground truth for training segmentation networks for presegmenting point clouds in pose estimation problems. Our method replaces global pose estimation algorithms such as RANSAC which requires manual fine-tuning with a robust CNN, without having to hand-label segmentation masks for the given object. The data is generated by blending cropped images of the objects with arbitrary backgrounds. We test the method in two scenarios, and show that networks trained on the generated data segments the objects with high accuracy, allowing them to be used in a pose estimation pipeline.

AB - We present a method for generating synthetic ground truth for training segmentation networks for presegmenting point clouds in pose estimation problems. Our method replaces global pose estimation algorithms such as RANSAC which requires manual fine-tuning with a robust CNN, without having to hand-label segmentation masks for the given object. The data is generated by blending cropped images of the objects with arbitrary backgrounds. We test the method in two scenarios, and show that networks trained on the generated data segments the objects with high accuracy, allowing them to be used in a pose estimation pipeline.

KW - Object Segmentation

KW - Pose Estimation

KW - Synthetic Ground Truth Generation

U2 - 10.5220/0009163904820489

DO - 10.5220/0009163904820489

M3 - Article in proceedings

AN - SCOPUS:85083547970

VL - 4

SP - 482

EP - 489

BT - Proceedings of the 15th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications - Volume 4: VISAPP

A2 - Farinella, Giovanni Maria

A2 - Radeva, Petia

A2 - Braz, Jose

PB - SCITEPRESS Digital Library

T2 - 15th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications

Y2 - 27 February 2020 through 29 February 2020

ER -

Haarslev F, Juel WK, Krüger N , Bodenhagen L. Synthetic Ground Truth for Presegmentation of Known Objects for Effortless Pose Estimation. I Farinella GM, Radeva P, Braz J, red., Proceedings of the 15th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications - Volume 4: VISAPP. Bind 4. SCITEPRESS Digital Library. 2020. s. 482-489 doi: 10.5220/0009163904820489

Synthetic Ground Truth for Presegmentation of Known Objects for Effortless Pose Estimation

Abstract

Konference

Dokumenter og links

SDU link

Fingeraftryk

Relaterede publikationer

Spatio-Temporal Vision for Human-Robot Interaction using Deep Learning

Relaterede projekter

SMOOTH: Seamless huMan-robot interactiOn fOr THe support of elderly people

Citationsformater