Object proposals estimation in depth image using compact 3D shape manifolds

Shuai Zheng; Victor Adrian Prisacariu; Melinos Averkiou; Ming Ming Cheng; Niloy J. Mitra; Jamie Shotton; Philip H.S. Torr; Carsten Rother

doi:10.1007/978-3-319-24947-6_16

Object proposals estimation in depth image using compact 3D shape manifolds

Shuai Zheng^*, Victor Adrian Prisacariu, Melinos Averkiou, Ming Ming Cheng, Niloy J. Mitra, Jamie Shotton, Philip H.S. Torr, Carsten Rother

^*Corresponding author for this work

Computer, Electrical and Mathematical Sciences and Engineering

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

11 Scopus citations

Abstract

Man-made objects, such as chairs, often have very large shape variations, making it challenging to detect them. In this work we investigate the task of finding particular object shapes from a single depth image. We tackle this task by exploiting the inherently low dimensionality in the object shape variations, which we discover and encode as a compact shape space. Starting from any collection of 3D models, we first train a low dimensional Gaussian Process Latent Variable Shape Space. We then sample this space, effectively producing infinite amounts of shape variations, which are used for training. Additionally, to support fast and accurate inference, we improve the standard 3D object category proposal generation pipeline by applying a shallow convolutional neural network-based filtering stage. This combination leads to considerable improvements for proposal generation, in both speed and accuracy. We compare our full system to previous state-of-the-art approaches, on four different shape classes, and show a clear improvement.

Original language	English (US)
Title of host publication	Pattern Recognition - 37th German Conference, GCPR 2015, Proceedings
Editors	Bastian Leibe, Juergen Gall, Peter Gehler
Publisher	Springer Verlag
Pages	196-208
Number of pages	13
ISBN (Print)	9783319249469
DOIs	https://doi.org/10.1007/978-3-319-24947-6_16
State	Published - 2015
Event	37th German Conference on Pattern Recognition, GCPR 2015 - Aachen, Germany Duration: Oct 7 2015 → Oct 10 2015

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	9358
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Other

Other	37th German Conference on Pattern Recognition, GCPR 2015
Country/Territory	Germany
City	Aachen
Period	10/7/15 → 10/10/15

Bibliographical note

Publisher Copyright:
© Springer International Publishing Switzerland 2015.

ASJC Scopus subject areas

Theoretical Computer Science
General Computer Science

Access to Document

10.1007/978-3-319-24947-6_16

Cite this

Zheng, S., Prisacariu, V. A., Averkiou, M., Cheng, M. M., Mitra, N. J., Shotton, J., Torr, P. H. S., & Rother, C. (2015). Object proposals estimation in depth image using compact 3D shape manifolds. In B. Leibe, J. Gall, & P. Gehler (Eds.), Pattern Recognition - 37th German Conference, GCPR 2015, Proceedings (pp. 196-208). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9358). Springer Verlag. https://doi.org/10.1007/978-3-319-24947-6_16

Zheng, Shuai ; Prisacariu, Victor Adrian ; Averkiou, Melinos et al. / Object proposals estimation in depth image using compact 3D shape manifolds. Pattern Recognition - 37th German Conference, GCPR 2015, Proceedings. editor / Bastian Leibe ; Juergen Gall ; Peter Gehler. Springer Verlag, 2015. pp. 196-208 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{bc02964a01c34011b29e566a1cc0d315,

title = "Object proposals estimation in depth image using compact 3D shape manifolds",

abstract = "Man-made objects, such as chairs, often have very large shape variations, making it challenging to detect them. In this work we investigate the task of finding particular object shapes from a single depth image. We tackle this task by exploiting the inherently low dimensionality in the object shape variations, which we discover and encode as a compact shape space. Starting from any collection of 3D models, we first train a low dimensional Gaussian Process Latent Variable Shape Space. We then sample this space, effectively producing infinite amounts of shape variations, which are used for training. Additionally, to support fast and accurate inference, we improve the standard 3D object category proposal generation pipeline by applying a shallow convolutional neural network-based filtering stage. This combination leads to considerable improvements for proposal generation, in both speed and accuracy. We compare our full system to previous state-of-the-art approaches, on four different shape classes, and show a clear improvement.",

author = "Shuai Zheng and Prisacariu, {Victor Adrian} and Melinos Averkiou and Cheng, {Ming Ming} and Mitra, {Niloy J.} and Jamie Shotton and Torr, {Philip H.S.} and Carsten Rother",

note = "Publisher Copyright: {\textcopyright} Springer International Publishing Switzerland 2015.; 37th German Conference on Pattern Recognition, GCPR 2015 ; Conference date: 07-10-2015 Through 10-10-2015",

year = "2015",

doi = "10.1007/978-3-319-24947-6_16",

language = "English (US)",

isbn = "9783319249469",

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

publisher = "Springer Verlag",

pages = "196--208",

editor = "Bastian Leibe and Juergen Gall and Peter Gehler",

booktitle = "Pattern Recognition - 37th German Conference, GCPR 2015, Proceedings",

address = "Germany",

}

Zheng, S, Prisacariu, VA, Averkiou, M, Cheng, MM, Mitra, NJ, Shotton, J, Torr, PHS & Rother, C 2015, Object proposals estimation in depth image using compact 3D shape manifolds. in B Leibe, J Gall & P Gehler (eds), Pattern Recognition - 37th German Conference, GCPR 2015, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 9358, Springer Verlag, pp. 196-208, 37th German Conference on Pattern Recognition, GCPR 2015, Aachen, Germany, 10/7/15. https://doi.org/10.1007/978-3-319-24947-6_16

Object proposals estimation in depth image using compact 3D shape manifolds. / Zheng, Shuai; Prisacariu, Victor Adrian; Averkiou, Melinos et al.
Pattern Recognition - 37th German Conference, GCPR 2015, Proceedings. ed. / Bastian Leibe; Juergen Gall; Peter Gehler. Springer Verlag, 2015. p. 196-208 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9358).

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

TY - GEN

T1 - Object proposals estimation in depth image using compact 3D shape manifolds

AU - Zheng, Shuai

AU - Prisacariu, Victor Adrian

AU - Averkiou, Melinos

AU - Cheng, Ming Ming

AU - Mitra, Niloy J.

AU - Shotton, Jamie

AU - Torr, Philip H.S.

AU - Rother, Carsten

PY - 2015

Y1 - 2015

N2 - Man-made objects, such as chairs, often have very large shape variations, making it challenging to detect them. In this work we investigate the task of finding particular object shapes from a single depth image. We tackle this task by exploiting the inherently low dimensionality in the object shape variations, which we discover and encode as a compact shape space. Starting from any collection of 3D models, we first train a low dimensional Gaussian Process Latent Variable Shape Space. We then sample this space, effectively producing infinite amounts of shape variations, which are used for training. Additionally, to support fast and accurate inference, we improve the standard 3D object category proposal generation pipeline by applying a shallow convolutional neural network-based filtering stage. This combination leads to considerable improvements for proposal generation, in both speed and accuracy. We compare our full system to previous state-of-the-art approaches, on four different shape classes, and show a clear improvement.

AB - Man-made objects, such as chairs, often have very large shape variations, making it challenging to detect them. In this work we investigate the task of finding particular object shapes from a single depth image. We tackle this task by exploiting the inherently low dimensionality in the object shape variations, which we discover and encode as a compact shape space. Starting from any collection of 3D models, we first train a low dimensional Gaussian Process Latent Variable Shape Space. We then sample this space, effectively producing infinite amounts of shape variations, which are used for training. Additionally, to support fast and accurate inference, we improve the standard 3D object category proposal generation pipeline by applying a shallow convolutional neural network-based filtering stage. This combination leads to considerable improvements for proposal generation, in both speed and accuracy. We compare our full system to previous state-of-the-art approaches, on four different shape classes, and show a clear improvement.

UR - http://www.scopus.com/inward/record.url?scp=84952306663&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-24947-6_16

DO - 10.1007/978-3-319-24947-6_16

M3 - Conference contribution

AN - SCOPUS:84952306663

SN - 9783319249469

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

SP - 196

EP - 208

BT - Pattern Recognition - 37th German Conference, GCPR 2015, Proceedings

A2 - Leibe, Bastian

A2 - Gall, Juergen

A2 - Gehler, Peter

PB - Springer Verlag

T2 - 37th German Conference on Pattern Recognition, GCPR 2015

Y2 - 7 October 2015 through 10 October 2015

ER -

Zheng S, Prisacariu VA, Averkiou M, Cheng MM, Mitra NJ, Shotton J et al. Object proposals estimation in depth image using compact 3D shape manifolds. In Leibe B, Gall J, Gehler P, editors, Pattern Recognition - 37th German Conference, GCPR 2015, Proceedings. Springer Verlag. 2015. p. 196-208. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-319-24947-6_16

Object proposals estimation in depth image using compact 3D shape manifolds

Abstract

Publication series

Other

Bibliographical note

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this