Layered embeddings for amodal instance segmentation

Yanfeng Liu, Eric T. Psota, Lance C. Pérez

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

The proposed method extends upon the representational output of semantic instance segmentation by explicitly including both visible and occluded parts. A fully convolutional network is trained to produce consistent pixel-level embedding across two layers such that, when clustered, the results convey the full spatial extent and depth ordering of each instance. Results demonstrate that the network can accurately estimate complete masks in the presence of occlusion and outperform leading top-down bounding-box approaches. Source code available at https://github.com/yanfengliu/layered_embeddings.

Original languageEnglish (US)
Title of host publicationImage Analysis and Recognition - 16th International Conference, ICIAR 2019, Proceedings
EditorsFakhri Karray, Alfred Yu, Aurélio Campilho
PublisherSpringer Verlag
Pages102-111
Number of pages10
ISBN (Print)9783030272012
DOIs
StatePublished - Jan 1 2019
Event16th International Conference on Image Analysis and Recognition, ICIAR 2019 - Waterloo, Canada
Duration: Aug 27 2019Aug 29 2019

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume11662 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference16th International Conference on Image Analysis and Recognition, ICIAR 2019
CountryCanada
CityWaterloo
Period8/27/198/29/19

Fingerprint

Masks
Segmentation
Pixels
Semantics
Occlusion
Mask
Pixel
Output
Estimate
Demonstrate

Keywords

  • Amodal segmentation
  • Occlusion recovery
  • Pixel embedding
  • Semantic instance segmentation

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Liu, Y., Psota, E. T., & Pérez, L. C. (2019). Layered embeddings for amodal instance segmentation. In F. Karray, A. Yu, & A. Campilho (Eds.), Image Analysis and Recognition - 16th International Conference, ICIAR 2019, Proceedings (pp. 102-111). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11662 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-030-27202-9_9

Layered embeddings for amodal instance segmentation. / Liu, Yanfeng; Psota, Eric T.; Pérez, Lance C.

Image Analysis and Recognition - 16th International Conference, ICIAR 2019, Proceedings. ed. / Fakhri Karray; Alfred Yu; Aurélio Campilho. Springer Verlag, 2019. p. 102-111 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11662 LNCS).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Liu, Y, Psota, ET & Pérez, LC 2019, Layered embeddings for amodal instance segmentation. in F Karray, A Yu & A Campilho (eds), Image Analysis and Recognition - 16th International Conference, ICIAR 2019, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 11662 LNCS, Springer Verlag, pp. 102-111, 16th International Conference on Image Analysis and Recognition, ICIAR 2019, Waterloo, Canada, 8/27/19. https://doi.org/10.1007/978-3-030-27202-9_9
Liu Y, Psota ET, Pérez LC. Layered embeddings for amodal instance segmentation. In Karray F, Yu A, Campilho A, editors, Image Analysis and Recognition - 16th International Conference, ICIAR 2019, Proceedings. Springer Verlag. 2019. p. 102-111. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-030-27202-9_9
Liu, Yanfeng ; Psota, Eric T. ; Pérez, Lance C. / Layered embeddings for amodal instance segmentation. Image Analysis and Recognition - 16th International Conference, ICIAR 2019, Proceedings. editor / Fakhri Karray ; Alfred Yu ; Aurélio Campilho. Springer Verlag, 2019. pp. 102-111 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{ce68e4f642ba4aa2bd7583409b0ad434,
title = "Layered embeddings for amodal instance segmentation",
abstract = "The proposed method extends upon the representational output of semantic instance segmentation by explicitly including both visible and occluded parts. A fully convolutional network is trained to produce consistent pixel-level embedding across two layers such that, when clustered, the results convey the full spatial extent and depth ordering of each instance. Results demonstrate that the network can accurately estimate complete masks in the presence of occlusion and outperform leading top-down bounding-box approaches. Source code available at https://github.com/yanfengliu/layered_embeddings.",
keywords = "Amodal segmentation, Occlusion recovery, Pixel embedding, Semantic instance segmentation",
author = "Yanfeng Liu and Psota, {Eric T.} and P{\'e}rez, {Lance C.}",
year = "2019",
month = "1",
day = "1",
doi = "10.1007/978-3-030-27202-9_9",
language = "English (US)",
isbn = "9783030272012",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "102--111",
editor = "Fakhri Karray and Alfred Yu and Aur{\'e}lio Campilho",
booktitle = "Image Analysis and Recognition - 16th International Conference, ICIAR 2019, Proceedings",

}

TY - GEN

T1 - Layered embeddings for amodal instance segmentation

AU - Liu, Yanfeng

AU - Psota, Eric T.

AU - Pérez, Lance C.

PY - 2019/1/1

Y1 - 2019/1/1

N2 - The proposed method extends upon the representational output of semantic instance segmentation by explicitly including both visible and occluded parts. A fully convolutional network is trained to produce consistent pixel-level embedding across two layers such that, when clustered, the results convey the full spatial extent and depth ordering of each instance. Results demonstrate that the network can accurately estimate complete masks in the presence of occlusion and outperform leading top-down bounding-box approaches. Source code available at https://github.com/yanfengliu/layered_embeddings.

AB - The proposed method extends upon the representational output of semantic instance segmentation by explicitly including both visible and occluded parts. A fully convolutional network is trained to produce consistent pixel-level embedding across two layers such that, when clustered, the results convey the full spatial extent and depth ordering of each instance. Results demonstrate that the network can accurately estimate complete masks in the presence of occlusion and outperform leading top-down bounding-box approaches. Source code available at https://github.com/yanfengliu/layered_embeddings.

KW - Amodal segmentation

KW - Occlusion recovery

KW - Pixel embedding

KW - Semantic instance segmentation

UR - http://www.scopus.com/inward/record.url?scp=85071465763&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85071465763&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-27202-9_9

DO - 10.1007/978-3-030-27202-9_9

M3 - Conference contribution

AN - SCOPUS:85071465763

SN - 9783030272012

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

SP - 102

EP - 111

BT - Image Analysis and Recognition - 16th International Conference, ICIAR 2019, Proceedings

A2 - Karray, Fakhri

A2 - Yu, Alfred

A2 - Campilho, Aurélio

PB - Springer Verlag

ER -