Learning-Based Dimensionality Reduction for Computing Compact and Effective Local Feature Descriptors
(2023) 2023 IEEE International Conference on Robotics and Automation, ICRA 2023 In Proceedings - IEEE International Conference on Robotics and Automation 2023-May. p.6189-6195- Abstract
A distinctive representation of image patches in form of features is a key component of many computer vision and robotics tasks, such as image matching, image retrieval, and visual localization. State-of-the-art descriptors, from hand-crafted descriptors such as SIFT to learned ones such as HardNet, are usually high-dimensional; 128 dimensions or even more. The higher the dimensionality, the larger the memory consumption and computational time for approaches using such descriptors. In this paper, we investigate multi-layer perceptrons (MLPs) to extract low-dimensional but high-quality descriptors. We thoroughly analyze our method in unsuper-vised, self-supervised, and supervised settings, and evaluate the dimensionality reduction... (More)
A distinctive representation of image patches in form of features is a key component of many computer vision and robotics tasks, such as image matching, image retrieval, and visual localization. State-of-the-art descriptors, from hand-crafted descriptors such as SIFT to learned ones such as HardNet, are usually high-dimensional; 128 dimensions or even more. The higher the dimensionality, the larger the memory consumption and computational time for approaches using such descriptors. In this paper, we investigate multi-layer perceptrons (MLPs) to extract low-dimensional but high-quality descriptors. We thoroughly analyze our method in unsuper-vised, self-supervised, and supervised settings, and evaluate the dimensionality reduction results on four representative descriptors. We consider different applications, including visual localization, patch verification, image matching and retrieval. The experiments show that our lightweight MLPs trained using supervised method achieve better dimensionality reduction than PCA. The lower-dimensional descriptors generated by our approach outperform the original higher-dimensional descriptors in downstream tasks, especially for the hand-crafted ones. The code is available at https://github.com/PRBonn/descriptor-dr.
(Less)
- author
- Dong, Hao ; Chen, Xieyuanli ; Dusmanu, Mihai ; Larsson, Viktor LU ; Pollefeys, Marc and Stachniss, Cyrill
- organization
- publishing date
- 2023
- type
- Chapter in Book/Report/Conference proceeding
- publication status
- published
- subject
- host publication
- Proceedings - ICRA 2023 : IEEE International Conference on Robotics and Automation - IEEE International Conference on Robotics and Automation
- series title
- Proceedings - IEEE International Conference on Robotics and Automation
- volume
- 2023-May
- pages
- 7 pages
- publisher
- IEEE - Institute of Electrical and Electronics Engineers Inc.
- conference name
- 2023 IEEE International Conference on Robotics and Automation, ICRA 2023
- conference location
- London, United Kingdom
- conference dates
- 2023-05-29 - 2023-06-02
- external identifiers
-
- scopus:85168662000
- ISSN
- 1050-4729
- ISBN
- 9798350323658
- DOI
- 10.1109/ICRA48891.2023.10161381
- language
- English
- LU publication?
- yes
- id
- 5c3cd743-ebde-4449-9df4-a7dd2efe19ad
- date added to LUP
- 2023-12-01 11:19:51
- date last changed
- 2023-12-01 11:19:51
@inproceedings{5c3cd743-ebde-4449-9df4-a7dd2efe19ad, abstract = {{<p>A distinctive representation of image patches in form of features is a key component of many computer vision and robotics tasks, such as image matching, image retrieval, and visual localization. State-of-the-art descriptors, from hand-crafted descriptors such as SIFT to learned ones such as HardNet, are usually high-dimensional; 128 dimensions or even more. The higher the dimensionality, the larger the memory consumption and computational time for approaches using such descriptors. In this paper, we investigate multi-layer perceptrons (MLPs) to extract low-dimensional but high-quality descriptors. We thoroughly analyze our method in unsuper-vised, self-supervised, and supervised settings, and evaluate the dimensionality reduction results on four representative descriptors. We consider different applications, including visual localization, patch verification, image matching and retrieval. The experiments show that our lightweight MLPs trained using supervised method achieve better dimensionality reduction than PCA. The lower-dimensional descriptors generated by our approach outperform the original higher-dimensional descriptors in downstream tasks, especially for the hand-crafted ones. The code is available at https://github.com/PRBonn/descriptor-dr.</p>}}, author = {{Dong, Hao and Chen, Xieyuanli and Dusmanu, Mihai and Larsson, Viktor and Pollefeys, Marc and Stachniss, Cyrill}}, booktitle = {{Proceedings - ICRA 2023 : IEEE International Conference on Robotics and Automation}}, isbn = {{9798350323658}}, issn = {{1050-4729}}, language = {{eng}}, pages = {{6189--6195}}, publisher = {{IEEE - Institute of Electrical and Electronics Engineers Inc.}}, series = {{Proceedings - IEEE International Conference on Robotics and Automation}}, title = {{Learning-Based Dimensionality Reduction for Computing Compact and Effective Local Feature Descriptors}}, url = {{http://dx.doi.org/10.1109/ICRA48891.2023.10161381}}, doi = {{10.1109/ICRA48891.2023.10161381}}, volume = {{2023-May}}, year = {{2023}}, }