Learning Context Flexible Attention Model for Long-Term Visual Place Recognition

Zetao Chen, Lingqiao Liu, Inkyu Sa, Zongyuan Ge, Margarita Chli

Research output: Contribution to journalArticleResearchpeer-review

Abstract

Identifying regions of interest in an image has long been of great importance in a wide range of tasks, including place recognition. In this letter, we propose a novel attention mechanism with flexible context, which can be incorporated into existing feedforward network architecture to learn image representations for long-term place recognition. In particular, in order to focus on regions that contribute positively to place recognition, we introduce a multiscale context-flexible network to estimate the importance of each spatial region in the feature map. Our model is trained end-to-end for place recognition and can detect regions of interest of arbitrary shape. Extensive experiments have been conducted to verify the effectiveness of our approach and the results demonstrate that our model can achieve consistently better performance than the state of the art on standard benchmark datasets. Finally, we visualize the learned attention maps to generate insights into what attention the network has learned.

Original languageEnglish
Article number8421024
Pages (from-to)4015-4022
Number of pages8
JournalIEEE Robotics and Automation Letters
Volume3
Issue number4
DOIs
Publication statusPublished - 1 Oct 2018

Keywords

  • deep learning in robotics and automation
  • Localization
  • visual-based navigation

Cite this

Chen, Zetao ; Liu, Lingqiao ; Sa, Inkyu ; Ge, Zongyuan ; Chli, Margarita . / Learning Context Flexible Attention Model for Long-Term Visual Place Recognition. In: IEEE Robotics and Automation Letters. 2018 ; Vol. 3, No. 4. pp. 4015-4022.
@article{558dd257ac05458fb5fb36fe4913faea,
title = "Learning Context Flexible Attention Model for Long-Term Visual Place Recognition",
abstract = "Identifying regions of interest in an image has long been of great importance in a wide range of tasks, including place recognition. In this letter, we propose a novel attention mechanism with flexible context, which can be incorporated into existing feedforward network architecture to learn image representations for long-term place recognition. In particular, in order to focus on regions that contribute positively to place recognition, we introduce a multiscale context-flexible network to estimate the importance of each spatial region in the feature map. Our model is trained end-to-end for place recognition and can detect regions of interest of arbitrary shape. Extensive experiments have been conducted to verify the effectiveness of our approach and the results demonstrate that our model can achieve consistently better performance than the state of the art on standard benchmark datasets. Finally, we visualize the learned attention maps to generate insights into what attention the network has learned.",
keywords = "deep learning in robotics and automation, Localization, visual-based navigation",
author = "Zetao Chen and Lingqiao Liu and Inkyu Sa and Zongyuan Ge and Margarita Chli",
year = "2018",
month = "10",
day = "1",
doi = "10.1109/LRA.2018.2859916",
language = "English",
volume = "3",
pages = "4015--4022",
journal = "IEEE Robotics and Automation Letters",
issn = "2377-3766",
publisher = "IEEE, Institute of Electrical and Electronics Engineers",
number = "4",

}

Learning Context Flexible Attention Model for Long-Term Visual Place Recognition. / Chen, Zetao; Liu, Lingqiao; Sa, Inkyu; Ge, Zongyuan; Chli, Margarita .

In: IEEE Robotics and Automation Letters, Vol. 3, No. 4, 8421024, 01.10.2018, p. 4015-4022.

Research output: Contribution to journalArticleResearchpeer-review

TY - JOUR

T1 - Learning Context Flexible Attention Model for Long-Term Visual Place Recognition

AU - Chen, Zetao

AU - Liu, Lingqiao

AU - Sa, Inkyu

AU - Ge, Zongyuan

AU - Chli, Margarita

PY - 2018/10/1

Y1 - 2018/10/1

N2 - Identifying regions of interest in an image has long been of great importance in a wide range of tasks, including place recognition. In this letter, we propose a novel attention mechanism with flexible context, which can be incorporated into existing feedforward network architecture to learn image representations for long-term place recognition. In particular, in order to focus on regions that contribute positively to place recognition, we introduce a multiscale context-flexible network to estimate the importance of each spatial region in the feature map. Our model is trained end-to-end for place recognition and can detect regions of interest of arbitrary shape. Extensive experiments have been conducted to verify the effectiveness of our approach and the results demonstrate that our model can achieve consistently better performance than the state of the art on standard benchmark datasets. Finally, we visualize the learned attention maps to generate insights into what attention the network has learned.

AB - Identifying regions of interest in an image has long been of great importance in a wide range of tasks, including place recognition. In this letter, we propose a novel attention mechanism with flexible context, which can be incorporated into existing feedforward network architecture to learn image representations for long-term place recognition. In particular, in order to focus on regions that contribute positively to place recognition, we introduce a multiscale context-flexible network to estimate the importance of each spatial region in the feature map. Our model is trained end-to-end for place recognition and can detect regions of interest of arbitrary shape. Extensive experiments have been conducted to verify the effectiveness of our approach and the results demonstrate that our model can achieve consistently better performance than the state of the art on standard benchmark datasets. Finally, we visualize the learned attention maps to generate insights into what attention the network has learned.

KW - deep learning in robotics and automation

KW - Localization

KW - visual-based navigation

UR - http://www.scopus.com/inward/record.url?scp=85060562544&partnerID=8YFLogxK

U2 - 10.1109/LRA.2018.2859916

DO - 10.1109/LRA.2018.2859916

M3 - Article

VL - 3

SP - 4015

EP - 4022

JO - IEEE Robotics and Automation Letters

JF - IEEE Robotics and Automation Letters

SN - 2377-3766

IS - 4

M1 - 8421024

ER -