Filter early, match late: Improving network-based visual place recognition
CNNs have excelled at performing place recognition over time, particularly when the neural network is optimized for localization in the current environmental conditions. In this paper we investigate the concept of feature map filtering, where, rather than using all the activations within a convoluti...
Published in: | 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) |
---|---|
Main Authors: | , , |
Format: | Book Part |
Language: | unknown |
Published: |
Institute of Electrical and Electronics Engineers Inc.
2019
|
Subjects: | |
Online Access: | https://eprints.qut.edu.au/198106/ |
id |
ftqueensland:oai:eprints.qut.edu.au:198106 |
---|---|
record_format |
openpolar |
spelling |
ftqueensland:oai:eprints.qut.edu.au:198106 2024-05-19T07:44:23+00:00 Filter early, match late: Improving network-based visual place recognition Hausler, Stephen Jacobson, Adam Milford, Michael 2019-11 application/pdf https://eprints.qut.edu.au/198106/ unknown Institute of Electrical and Electronics Engineers Inc. https://eprints.qut.edu.au/198106/1/56922609.pdf doi:10.1109/IROS40897.2019.8967783 Hausler, Stephen, Jacobson, Adam, & Milford, Michael (2019) Filter early, match late: Improving network-based visual place recognition. In Proceedings of the 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). Institute of Electrical and Electronics Engineers Inc., United States of America, pp. 3268-3275. http://purl.org/au-research/grants/arc/FT140101229 https://eprints.qut.edu.au/198106/ Institute for Future Environments; Science & Engineering Faculty free_to_read http://creativecommons.org/licenses/by-nc/4.0/ IEEE © 2019 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. Proceedings of the 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) Chapter in Book, Report or Conference volume 2019 ftqueensland https://doi.org/10.1109/IROS40897.2019.8967783 2024-04-23T23:54:23Z CNNs have excelled at performing place recognition over time, particularly when the neural network is optimized for localization in the current environmental conditions. In this paper we investigate the concept of feature map filtering, where, rather than using all the activations within a convolutional tensor, only the most useful activations are used. Since specific feature maps encode different visual features, the objective is to remove feature maps that detract from the ability to recognize a location across appearance changes. Our key innovation is to filter the feature maps in an early convolutional layer, but then continue to run the network and extract a feature vector using a later, more viewpoint invariant layer in the same network. Our approach improves the condition and viewpoint invariance of a pre-trained network, using as little as a single training image pair from the deployment environment. An exhaustive experimental analysis is performed to determine the full scope of causality between early layer filtering and late layer extraction. For validity, we use three datasets: Oxford RobotCar, Nordland, and Gardens Point, achieving overall superior performance to NetVLAD. The work provides a number of new avenues for exploring CNN optimizations, without requiring any re-training of the network weights. Book Part Nordland Nordland Nordland Queensland University of Technology: QUT ePrints 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) 3268 3275 |
institution |
Open Polar |
collection |
Queensland University of Technology: QUT ePrints |
op_collection_id |
ftqueensland |
language |
unknown |
description |
CNNs have excelled at performing place recognition over time, particularly when the neural network is optimized for localization in the current environmental conditions. In this paper we investigate the concept of feature map filtering, where, rather than using all the activations within a convolutional tensor, only the most useful activations are used. Since specific feature maps encode different visual features, the objective is to remove feature maps that detract from the ability to recognize a location across appearance changes. Our key innovation is to filter the feature maps in an early convolutional layer, but then continue to run the network and extract a feature vector using a later, more viewpoint invariant layer in the same network. Our approach improves the condition and viewpoint invariance of a pre-trained network, using as little as a single training image pair from the deployment environment. An exhaustive experimental analysis is performed to determine the full scope of causality between early layer filtering and late layer extraction. For validity, we use three datasets: Oxford RobotCar, Nordland, and Gardens Point, achieving overall superior performance to NetVLAD. The work provides a number of new avenues for exploring CNN optimizations, without requiring any re-training of the network weights. |
format |
Book Part |
author |
Hausler, Stephen Jacobson, Adam Milford, Michael |
spellingShingle |
Hausler, Stephen Jacobson, Adam Milford, Michael Filter early, match late: Improving network-based visual place recognition |
author_facet |
Hausler, Stephen Jacobson, Adam Milford, Michael |
author_sort |
Hausler, Stephen |
title |
Filter early, match late: Improving network-based visual place recognition |
title_short |
Filter early, match late: Improving network-based visual place recognition |
title_full |
Filter early, match late: Improving network-based visual place recognition |
title_fullStr |
Filter early, match late: Improving network-based visual place recognition |
title_full_unstemmed |
Filter early, match late: Improving network-based visual place recognition |
title_sort |
filter early, match late: improving network-based visual place recognition |
publisher |
Institute of Electrical and Electronics Engineers Inc. |
publishDate |
2019 |
url |
https://eprints.qut.edu.au/198106/ |
genre |
Nordland Nordland Nordland |
genre_facet |
Nordland Nordland Nordland |
op_source |
Proceedings of the 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) |
op_relation |
https://eprints.qut.edu.au/198106/1/56922609.pdf doi:10.1109/IROS40897.2019.8967783 Hausler, Stephen, Jacobson, Adam, & Milford, Michael (2019) Filter early, match late: Improving network-based visual place recognition. In Proceedings of the 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). Institute of Electrical and Electronics Engineers Inc., United States of America, pp. 3268-3275. http://purl.org/au-research/grants/arc/FT140101229 https://eprints.qut.edu.au/198106/ Institute for Future Environments; Science & Engineering Faculty |
op_rights |
free_to_read http://creativecommons.org/licenses/by-nc/4.0/ IEEE © 2019 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. |
op_doi |
https://doi.org/10.1109/IROS40897.2019.8967783 |
container_title |
2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) |
container_start_page |
3268 |
op_container_end_page |
3275 |
_version_ |
1799484166095503360 |