Aggregating crowdsourced image segmentations

Doris Jung Lin Lee, Akash Das Sarma, Aditya G Parameswaran

Research output: Contribution to journalConference article

Abstract

Instance-level image segmentation provides rich information crucial for scene understanding in a variety of real-world applications. In this paper, we evaluate multiple crowdsourced algorithms for the image segmentation problem, including novel worker-aggregation-based methods and retrieval-based methods from prior work. We characterize the different types of worker errors observed in crowdsourced segmentation, and present a clustering algorithm as a preprocessing step that is able to capture and eliminate errors arising due to workers having different semantic perspectives. We demonstrate that aggregation-based algorithms attain higher accuracies than existing retrieval-based approaches, while scaling better with increasing numbers of worker segmentations.

Original languageEnglish (US)
JournalCEUR Workshop Proceedings
Volume2173
StatePublished - Jan 1 2018
Event2018 HCOMP Works in Progress and Demonstration Papers, HCOMP WIP and DEMO 2018 - Zurich, Switzerland
Duration: Jul 5 2018Jul 8 2018

Fingerprint

Image segmentation
Agglomeration
Clustering algorithms
Semantics

ASJC Scopus subject areas

  • Computer Science(all)

Cite this

Lee, D. J. L., Sarma, A. D., & Parameswaran, A. G. (2018). Aggregating crowdsourced image segmentations. CEUR Workshop Proceedings, 2173.

Aggregating crowdsourced image segmentations. / Lee, Doris Jung Lin; Sarma, Akash Das; Parameswaran, Aditya G.

In: CEUR Workshop Proceedings, Vol. 2173, 01.01.2018.

Research output: Contribution to journalConference article

Lee, DJL, Sarma, AD & Parameswaran, AG 2018, 'Aggregating crowdsourced image segmentations', CEUR Workshop Proceedings, vol. 2173.
Lee DJL, Sarma AD, Parameswaran AG. Aggregating crowdsourced image segmentations. CEUR Workshop Proceedings. 2018 Jan 1;2173.
Lee, Doris Jung Lin ; Sarma, Akash Das ; Parameswaran, Aditya G. / Aggregating crowdsourced image segmentations. In: CEUR Workshop Proceedings. 2018 ; Vol. 2173.
@article{4bc8b3e10f0a4abcb7b71f5004abe41a,
title = "Aggregating crowdsourced image segmentations",
abstract = "Instance-level image segmentation provides rich information crucial for scene understanding in a variety of real-world applications. In this paper, we evaluate multiple crowdsourced algorithms for the image segmentation problem, including novel worker-aggregation-based methods and retrieval-based methods from prior work. We characterize the different types of worker errors observed in crowdsourced segmentation, and present a clustering algorithm as a preprocessing step that is able to capture and eliminate errors arising due to workers having different semantic perspectives. We demonstrate that aggregation-based algorithms attain higher accuracies than existing retrieval-based approaches, while scaling better with increasing numbers of worker segmentations.",
author = "Lee, {Doris Jung Lin} and Sarma, {Akash Das} and Parameswaran, {Aditya G}",
year = "2018",
month = "1",
day = "1",
language = "English (US)",
volume = "2173",
journal = "CEUR Workshop Proceedings",
issn = "1613-0073",
publisher = "CEUR-WS",

}

TY - JOUR

T1 - Aggregating crowdsourced image segmentations

AU - Lee, Doris Jung Lin

AU - Sarma, Akash Das

AU - Parameswaran, Aditya G

PY - 2018/1/1

Y1 - 2018/1/1

N2 - Instance-level image segmentation provides rich information crucial for scene understanding in a variety of real-world applications. In this paper, we evaluate multiple crowdsourced algorithms for the image segmentation problem, including novel worker-aggregation-based methods and retrieval-based methods from prior work. We characterize the different types of worker errors observed in crowdsourced segmentation, and present a clustering algorithm as a preprocessing step that is able to capture and eliminate errors arising due to workers having different semantic perspectives. We demonstrate that aggregation-based algorithms attain higher accuracies than existing retrieval-based approaches, while scaling better with increasing numbers of worker segmentations.

AB - Instance-level image segmentation provides rich information crucial for scene understanding in a variety of real-world applications. In this paper, we evaluate multiple crowdsourced algorithms for the image segmentation problem, including novel worker-aggregation-based methods and retrieval-based methods from prior work. We characterize the different types of worker errors observed in crowdsourced segmentation, and present a clustering algorithm as a preprocessing step that is able to capture and eliminate errors arising due to workers having different semantic perspectives. We demonstrate that aggregation-based algorithms attain higher accuracies than existing retrieval-based approaches, while scaling better with increasing numbers of worker segmentations.

UR - http://www.scopus.com/inward/record.url?scp=85052405350&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85052405350&partnerID=8YFLogxK

M3 - Conference article

AN - SCOPUS:85052405350

VL - 2173

JO - CEUR Workshop Proceedings

JF - CEUR Workshop Proceedings

SN - 1613-0073

ER -