Instance segmentation of indoor scenes using a coverage loss

Nathan Silberman; David Sontag; Rob Fergus

doi:10.1007/978-3-319-10590-1_40

Instance segmentation of indoor scenes using a coverage loss

Nathan Silberman, David Sontag, Rob Fergus

Computer Science

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

Abstract

A major limitation of existing models for semantic segmentation is the inability to identify individual instances of the same class: when labeling pixels with only semantic classes, a set of pixels with the same label could represent a single object or ten. In this work, we introduce a model to perform both semantic and instance segmentation simultaneously. We introduce a new higher-order loss function that directly minimizes the coverage metric and evaluate a variety of region features, including those from a convolutional network. We apply our model to the NYU Depth V2 dataset, obtaining state of the art results.

Original language	English (US)
Title of host publication	Computer Vision, ECCV 2014 - 13th European Conference, Proceedings
Publisher	Springer Verlag
Pages	616-631
Number of pages	16
Edition	PART 1
ISBN (Print)	9783319105895
DOIs	https://doi.org/10.1007/978-3-319-10590-1_40
State	Published - 2014
Event	13th European Conference on Computer Vision, ECCV 2014 - Zurich, Switzerland Duration: Sep 6 2014 → Sep 12 2014

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Number	PART 1
Volume	8689 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Other

Other	13th European Conference on Computer Vision, ECCV 2014
Country/Territory	Switzerland
City	Zurich
Period	9/6/14 → 9/12/14

Keywords

Deep Learning
Semantic Segmentation

ASJC Scopus subject areas

Theoretical Computer Science
General Computer Science

Access to Document

10.1007/978-3-319-10590-1_40

Cite this

Silberman, N., Sontag, D., & Fergus, R. (2014). Instance segmentation of indoor scenes using a coverage loss. In Computer Vision, ECCV 2014 - 13th European Conference, Proceedings (PART 1 ed., pp. 616-631). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 8689 LNCS, No. PART 1). Springer Verlag. https://doi.org/10.1007/978-3-319-10590-1_40

Instance segmentation of indoor scenes using a coverage loss. / Silberman, Nathan; Sontag, David; Fergus, Rob.
Computer Vision, ECCV 2014 - 13th European Conference, Proceedings. PART 1. ed. Springer Verlag, 2014. p. 616-631 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 8689 LNCS, No. PART 1).

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

Silberman, N, Sontag, D & Fergus, R 2014, Instance segmentation of indoor scenes using a coverage loss. in Computer Vision, ECCV 2014 - 13th European Conference, Proceedings. PART 1 edn, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), no. PART 1, vol. 8689 LNCS, Springer Verlag, pp. 616-631, 13th European Conference on Computer Vision, ECCV 2014, Zurich, Switzerland, 9/6/14. https://doi.org/10.1007/978-3-319-10590-1_40

Silberman N, Sontag D, Fergus R. Instance segmentation of indoor scenes using a coverage loss. In Computer Vision, ECCV 2014 - 13th European Conference, Proceedings. PART 1 ed. Springer Verlag. 2014. p. 616-631. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 1). doi: 10.1007/978-3-319-10590-1_40

Silberman, Nathan ; Sontag, David ; Fergus, Rob. / Instance segmentation of indoor scenes using a coverage loss. Computer Vision, ECCV 2014 - 13th European Conference, Proceedings. PART 1. ed. Springer Verlag, 2014. pp. 616-631 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 1).

@inproceedings{32d20fe5694a4ebc8bed17819615697c,

title = "Instance segmentation of indoor scenes using a coverage loss",

abstract = "A major limitation of existing models for semantic segmentation is the inability to identify individual instances of the same class: when labeling pixels with only semantic classes, a set of pixels with the same label could represent a single object or ten. In this work, we introduce a model to perform both semantic and instance segmentation simultaneously. We introduce a new higher-order loss function that directly minimizes the coverage metric and evaluate a variety of region features, including those from a convolutional network. We apply our model to the NYU Depth V2 dataset, obtaining state of the art results.",

keywords = "Deep Learning, Semantic Segmentation",

author = "Nathan Silberman and David Sontag and Rob Fergus",

year = "2014",

doi = "10.1007/978-3-319-10590-1_40",

language = "English (US)",

isbn = "9783319105895",

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

publisher = "Springer Verlag",

number = "PART 1",

pages = "616--631",

booktitle = "Computer Vision, ECCV 2014 - 13th European Conference, Proceedings",

edition = "PART 1",

note = "13th European Conference on Computer Vision, ECCV 2014 ; Conference date: 06-09-2014 Through 12-09-2014",

}

TY - GEN

T1 - Instance segmentation of indoor scenes using a coverage loss

AU - Silberman, Nathan

AU - Sontag, David

AU - Fergus, Rob

PY - 2014

Y1 - 2014

N2 - A major limitation of existing models for semantic segmentation is the inability to identify individual instances of the same class: when labeling pixels with only semantic classes, a set of pixels with the same label could represent a single object or ten. In this work, we introduce a model to perform both semantic and instance segmentation simultaneously. We introduce a new higher-order loss function that directly minimizes the coverage metric and evaluate a variety of region features, including those from a convolutional network. We apply our model to the NYU Depth V2 dataset, obtaining state of the art results.

AB - A major limitation of existing models for semantic segmentation is the inability to identify individual instances of the same class: when labeling pixels with only semantic classes, a set of pixels with the same label could represent a single object or ten. In this work, we introduce a model to perform both semantic and instance segmentation simultaneously. We introduce a new higher-order loss function that directly minimizes the coverage metric and evaluate a variety of region features, including those from a convolutional network. We apply our model to the NYU Depth V2 dataset, obtaining state of the art results.

KW - Deep Learning

KW - Semantic Segmentation

UR - http://www.scopus.com/inward/record.url?scp=84906507967&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84906507967&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-10590-1_40

DO - 10.1007/978-3-319-10590-1_40

M3 - Conference contribution

AN - SCOPUS:84906507967

SN - 9783319105895

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

SP - 616

EP - 631

BT - Computer Vision, ECCV 2014 - 13th European Conference, Proceedings

PB - Springer Verlag

T2 - 13th European Conference on Computer Vision, ECCV 2014

Y2 - 6 September 2014 through 12 September 2014

ER -

Instance segmentation of indoor scenes using a coverage loss

Abstract

Publication series

Other

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this