Simultaneous Edge Alignment and Learning

Zhiding Yu, Weiyang Liu, Yang Zou, Chen Feng, Srikumar Ramalingam, B. V.K.Vijaya Kumar, Jan Kautz

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

Abstract

Edge detection is among the most fundamental vision problems for its role in perceptual grouping and its wide applications. Recent advances in representation learning have led to considerable improvements in this area. Many state of the art edge detection models are learned with fully convolutional networks (FCNs). However, FCN-based edge learning tends to be vulnerable to misaligned labels due to the delicate structure of edges. While such problem was considered in evaluation benchmarks, similar issue has not been explicitly addressed in general edge learning. In this paper, we show that label misalignment can cause considerably degraded edge learning quality, and address this issue by proposing a simultaneous edge alignment and learning framework. To this end, we formulate a probabilistic model where edge alignment is treated as latent variable optimization, and is learned end-to-end during network training. Experiments show several applications of this work, including improved edge detection with state of the art performance, and automatic refinement of noisy annotations.

Original languageEnglish (US)
Title of host publicationComputer Vision – ECCV 2018 - 15th European Conference, 2018, Proceedings
EditorsVittorio Ferrari, Cristian Sminchisescu, Martial Hebert, Yair Weiss
PublisherSpringer-Verlag
Pages400-417
Number of pages18
ISBN (Print)9783030012182
DOIs
StatePublished - Jan 1 2018
Event15th European Conference on Computer Vision, ECCV 2018 - Munich, Germany
Duration: Sep 8 2018Sep 14 2018

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume11207 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other15th European Conference on Computer Vision, ECCV 2018
CountryGermany
CityMunich
Period9/8/189/14/18

Fingerprint

Edge detection
Alignment
Edge Detection
Labels
Perceptual Grouping
Misalignment
Latent Variables
Probabilistic Model
Annotation
Refinement
Learning
Tend
Benchmark
Experiments
Optimization
Evaluation
Experiment

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Yu, Z., Liu, W., Zou, Y., Feng, C., Ramalingam, S., Kumar, B. V. K. V., & Kautz, J. (2018). Simultaneous Edge Alignment and Learning. In V. Ferrari, C. Sminchisescu, M. Hebert, & Y. Weiss (Eds.), Computer Vision – ECCV 2018 - 15th European Conference, 2018, Proceedings (pp. 400-417). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11207 LNCS). Springer-Verlag. https://doi.org/10.1007/978-3-030-01219-9_24

Simultaneous Edge Alignment and Learning. / Yu, Zhiding; Liu, Weiyang; Zou, Yang; Feng, Chen; Ramalingam, Srikumar; Kumar, B. V.K.Vijaya; Kautz, Jan.

Computer Vision – ECCV 2018 - 15th European Conference, 2018, Proceedings. ed. / Vittorio Ferrari; Cristian Sminchisescu; Martial Hebert; Yair Weiss. Springer-Verlag, 2018. p. 400-417 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11207 LNCS).

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

Yu, Z, Liu, W, Zou, Y, Feng, C, Ramalingam, S, Kumar, BVKV & Kautz, J 2018, Simultaneous Edge Alignment and Learning. in V Ferrari, C Sminchisescu, M Hebert & Y Weiss (eds), Computer Vision – ECCV 2018 - 15th European Conference, 2018, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 11207 LNCS, Springer-Verlag, pp. 400-417, 15th European Conference on Computer Vision, ECCV 2018, Munich, Germany, 9/8/18. https://doi.org/10.1007/978-3-030-01219-9_24
Yu Z, Liu W, Zou Y, Feng C, Ramalingam S, Kumar BVKV et al. Simultaneous Edge Alignment and Learning. In Ferrari V, Sminchisescu C, Hebert M, Weiss Y, editors, Computer Vision – ECCV 2018 - 15th European Conference, 2018, Proceedings. Springer-Verlag. 2018. p. 400-417. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-030-01219-9_24
Yu, Zhiding ; Liu, Weiyang ; Zou, Yang ; Feng, Chen ; Ramalingam, Srikumar ; Kumar, B. V.K.Vijaya ; Kautz, Jan. / Simultaneous Edge Alignment and Learning. Computer Vision – ECCV 2018 - 15th European Conference, 2018, Proceedings. editor / Vittorio Ferrari ; Cristian Sminchisescu ; Martial Hebert ; Yair Weiss. Springer-Verlag, 2018. pp. 400-417 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{fa30c50dac384b018d6d2aac546db855,
title = "Simultaneous Edge Alignment and Learning",
abstract = "Edge detection is among the most fundamental vision problems for its role in perceptual grouping and its wide applications. Recent advances in representation learning have led to considerable improvements in this area. Many state of the art edge detection models are learned with fully convolutional networks (FCNs). However, FCN-based edge learning tends to be vulnerable to misaligned labels due to the delicate structure of edges. While such problem was considered in evaluation benchmarks, similar issue has not been explicitly addressed in general edge learning. In this paper, we show that label misalignment can cause considerably degraded edge learning quality, and address this issue by proposing a simultaneous edge alignment and learning framework. To this end, we formulate a probabilistic model where edge alignment is treated as latent variable optimization, and is learned end-to-end during network training. Experiments show several applications of this work, including improved edge detection with state of the art performance, and automatic refinement of noisy annotations.",
author = "Zhiding Yu and Weiyang Liu and Yang Zou and Chen Feng and Srikumar Ramalingam and Kumar, {B. V.K.Vijaya} and Jan Kautz",
year = "2018",
month = "1",
day = "1",
doi = "10.1007/978-3-030-01219-9_24",
language = "English (US)",
isbn = "9783030012182",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer-Verlag",
pages = "400--417",
editor = "Vittorio Ferrari and Cristian Sminchisescu and Martial Hebert and Yair Weiss",
booktitle = "Computer Vision – ECCV 2018 - 15th European Conference, 2018, Proceedings",

}

TY - GEN

T1 - Simultaneous Edge Alignment and Learning

AU - Yu, Zhiding

AU - Liu, Weiyang

AU - Zou, Yang

AU - Feng, Chen

AU - Ramalingam, Srikumar

AU - Kumar, B. V.K.Vijaya

AU - Kautz, Jan

PY - 2018/1/1

Y1 - 2018/1/1

N2 - Edge detection is among the most fundamental vision problems for its role in perceptual grouping and its wide applications. Recent advances in representation learning have led to considerable improvements in this area. Many state of the art edge detection models are learned with fully convolutional networks (FCNs). However, FCN-based edge learning tends to be vulnerable to misaligned labels due to the delicate structure of edges. While such problem was considered in evaluation benchmarks, similar issue has not been explicitly addressed in general edge learning. In this paper, we show that label misalignment can cause considerably degraded edge learning quality, and address this issue by proposing a simultaneous edge alignment and learning framework. To this end, we formulate a probabilistic model where edge alignment is treated as latent variable optimization, and is learned end-to-end during network training. Experiments show several applications of this work, including improved edge detection with state of the art performance, and automatic refinement of noisy annotations.

AB - Edge detection is among the most fundamental vision problems for its role in perceptual grouping and its wide applications. Recent advances in representation learning have led to considerable improvements in this area. Many state of the art edge detection models are learned with fully convolutional networks (FCNs). However, FCN-based edge learning tends to be vulnerable to misaligned labels due to the delicate structure of edges. While such problem was considered in evaluation benchmarks, similar issue has not been explicitly addressed in general edge learning. In this paper, we show that label misalignment can cause considerably degraded edge learning quality, and address this issue by proposing a simultaneous edge alignment and learning framework. To this end, we formulate a probabilistic model where edge alignment is treated as latent variable optimization, and is learned end-to-end during network training. Experiments show several applications of this work, including improved edge detection with state of the art performance, and automatic refinement of noisy annotations.

UR - http://www.scopus.com/inward/record.url?scp=85055090697&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85055090697&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-01219-9_24

DO - 10.1007/978-3-030-01219-9_24

M3 - Conference contribution

AN - SCOPUS:85055090697

SN - 9783030012182

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

SP - 400

EP - 417

BT - Computer Vision – ECCV 2018 - 15th European Conference, 2018, Proceedings

A2 - Ferrari, Vittorio

A2 - Sminchisescu, Cristian

A2 - Hebert, Martial

A2 - Weiss, Yair

PB - Springer-Verlag

ER -