Hyperbolic Visual Embedding Learning for Zero-Shot Recognition

Shaoteng Liu; Jingjing Chen; Liangming Pan; Chong-Wah Ngo; Tat-Seng Chua; Yu-Gang Jiang

doi:10.1109/CVPR42600.2020.00929

Hyperbolic Visual Embedding Learning for Zero-Shot Recognition

Shaoteng Liu, Jingjing Chen^*, Liangming Pan, Chong-Wah Ngo, Tat-Seng Chua, Yu-Gang Jiang

^*Corresponding author for this work

Department of Computer Science

Research output: Chapters, Conference Papers, Creative and Literary Works › RGC 32 - Refereed conference paper (with host publication) › peer-review

137 Citations (Scopus)

Abstract

This paper proposes a Hyperbolic Visual Embedding Learning Network for zero-shot recognition. The network learns image embeddings in hyperbolic space, which is capable of preserving the hierarchical structure of semantic classes in low dimensions. Comparing with existing zeroshot learning approaches, the network is more robust because the embedding feature in hyperbolic space better represents class hierarchy and thereby avoid misleading resulted from unrelated siblings. Our network outperforms exiting baselines under hierarchical evaluation with an extremely challenging setting, i.e., learning only from 1,000
categories to recognize 20,841 unseen categories. While under flat evaluation, it has competitive performance as state-of-the-art methods but with five times lower embedding dimensions. Our code is publicly available ^∗.

Original language	English
Title of host publication	2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2020
Subtitle of host publication	Proceedings
Publisher	IEEE
Pages	9270-9278
ISBN (Electronic)	978-1-7281-7168-5
ISBN (Print)	978-1-7281-7169-2
DOIs	https://doi.org/10.1109/CVPR42600.2020.00929
Publication status	Published - Jun 2020
Event	2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2020) - Virtual, Seattle, United States Duration: 13 Jun 2020 → 19 Jun 2020 http://cvpr2020.thecvf.com/ http://openaccess.thecvf.com/content_CVPR_2020/html/Guo_Zero-Reference_Deep_Curve_Estimation_for_Low-Light_Image_Enhancement_CVPR_2020_paper.html https://ieeexplore.ieee.org/xpl/conhome/9142308/proceeding http://cvpr2021.thecvf.com/ https://ieeexplore.ieee.org/xpl/conhome/1000147/all-proceedings https://openaccess.thecvf.com/CVPR2021

Publication series

Name	IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR
Publisher	Institute of Electrical and Electronics Engineers
ISSN (Print)	1063-6919
ISSN (Electronic)	2575-7075

Conference

Conference	2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2020)
Abbreviated title	CVPR2020
Place	United States
City	Seattle
Period	13/06/20 → 19/06/20
Internet address	http://cvpr2020.thecvf.com/ http://openaccess.thecvf.com/content_CVPR_2020/html/Guo_Zero-Reference_Deep_Curve_Estimation_for_Low-Light_Image_Enhancement_CVPR_2020_paper.html https://ieeexplore.ieee.org/xpl/conhome/9142308/proceeding http://cvpr2021.thecvf.com/ https://ieeexplore.ieee.org/xpl/conhome/1000147/all-proceedings https://openaccess.thecvf.com/CVPR2021

Bibliographical note

Research Unit(s) information for this publication is provided by the author(s) concerned.

Access Output

10.1109/CVPR42600.2020.00929

Other Access Options

Check CityUHK Library

Permanent Link

Right click to copy link

Cite this

Liu, S., Chen, J., Pan, L., Ngo, C.-W., Chua, T.-S., & Jiang, Y.-G. (2020). Hyperbolic Visual Embedding Learning for Zero-Shot Recognition. In 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2020: Proceedings (pp. 9270-9278). (IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR). IEEE. https://doi.org/10.1109/CVPR42600.2020.00929

@inproceedings{a86e089bdfca457088b861261ee623a4,

title = "Hyperbolic Visual Embedding Learning for Zero-Shot Recognition",

abstract = "This paper proposes a Hyperbolic Visual Embedding Learning Network for zero-shot recognition. The network learns image embeddings in hyperbolic space, which is capable of preserving the hierarchical structure of semantic classes in low dimensions. Comparing with existing zeroshot learning approaches, the network is more robust because the embedding feature in hyperbolic space better represents class hierarchy and thereby avoid misleading resulted from unrelated siblings. Our network outperforms exiting baselines under hierarchical evaluation with an extremely challenging setting, i.e., learning only from 1,000categories to recognize 20,841 unseen categories. While under flat evaluation, it has competitive performance as state-of-the-art methods but with five times lower embedding dimensions. Our code is publicly available ∗.",

author = "Shaoteng Liu and Jingjing Chen and Liangming Pan and Chong-Wah Ngo and Tat-Seng Chua and Yu-Gang Jiang",

note = "Research Unit(s) information for this publication is provided by the author(s) concerned.; 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2020), CVPR2020 ; Conference date: 13-06-2020 Through 19-06-2020",

year = "2020",

month = jun,

doi = "10.1109/CVPR42600.2020.00929",

language = "English",

isbn = "978-1-7281-7169-2",

series = "IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR",

publisher = "IEEE",

pages = "9270--9278",

booktitle = "2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2020",

address = "United States",

url = "http://cvpr2020.thecvf.com/, http://openaccess.thecvf.com/content_CVPR_2020/html/Guo_Zero-Reference_Deep_Curve_Estimation_for_Low-Light_Image_Enhancement_CVPR_2020_paper.html, https://ieeexplore.ieee.org/xpl/conhome/9142308/proceeding, http://cvpr2021.thecvf.com/, https://ieeexplore.ieee.org/xpl/conhome/1000147/all-proceedings, https://openaccess.thecvf.com/CVPR2021",

}

Liu, S, Chen, J, Pan, L, Ngo, C-W, Chua, T-S & Jiang, Y-G 2020, Hyperbolic Visual Embedding Learning for Zero-Shot Recognition. in 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2020: Proceedings. IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR, IEEE, pp. 9270-9278, 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2020), Seattle, Washington, United States, 13/06/20. https://doi.org/10.1109/CVPR42600.2020.00929

Hyperbolic Visual Embedding Learning for Zero-Shot Recognition. / Liu, Shaoteng; Chen, Jingjing; Pan, Liangming et al.
2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2020: Proceedings. IEEE, 2020. p. 9270-9278 (IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR).

Research output: Chapters, Conference Papers, Creative and Literary Works › RGC 32 - Refereed conference paper (with host publication) › peer-review

TY - GEN

T1 - Hyperbolic Visual Embedding Learning for Zero-Shot Recognition

AU - Liu, Shaoteng

AU - Chen, Jingjing

AU - Pan, Liangming

AU - Ngo, Chong-Wah

AU - Chua, Tat-Seng

AU - Jiang, Yu-Gang

N1 - Research Unit(s) information for this publication is provided by the author(s) concerned.

PY - 2020/6

Y1 - 2020/6

N2 - This paper proposes a Hyperbolic Visual Embedding Learning Network for zero-shot recognition. The network learns image embeddings in hyperbolic space, which is capable of preserving the hierarchical structure of semantic classes in low dimensions. Comparing with existing zeroshot learning approaches, the network is more robust because the embedding feature in hyperbolic space better represents class hierarchy and thereby avoid misleading resulted from unrelated siblings. Our network outperforms exiting baselines under hierarchical evaluation with an extremely challenging setting, i.e., learning only from 1,000categories to recognize 20,841 unseen categories. While under flat evaluation, it has competitive performance as state-of-the-art methods but with five times lower embedding dimensions. Our code is publicly available ∗.

AB - This paper proposes a Hyperbolic Visual Embedding Learning Network for zero-shot recognition. The network learns image embeddings in hyperbolic space, which is capable of preserving the hierarchical structure of semantic classes in low dimensions. Comparing with existing zeroshot learning approaches, the network is more robust because the embedding feature in hyperbolic space better represents class hierarchy and thereby avoid misleading resulted from unrelated siblings. Our network outperforms exiting baselines under hierarchical evaluation with an extremely challenging setting, i.e., learning only from 1,000categories to recognize 20,841 unseen categories. While under flat evaluation, it has competitive performance as state-of-the-art methods but with five times lower embedding dimensions. Our code is publicly available ∗.

UR - http://www.scopus.com/inward/record.url?scp=85094638032&partnerID=8YFLogxK

UR - https://www.scopus.com/record/pubmetrics.uri?eid=2-s2.0-85094638032&origin=recordpage

U2 - 10.1109/CVPR42600.2020.00929

DO - 10.1109/CVPR42600.2020.00929

M3 - RGC 32 - Refereed conference paper (with host publication)

SN - 978-1-7281-7169-2

T3 - IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR

SP - 9270

EP - 9278

BT - 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2020

PB - IEEE

T2 - 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2020)

Y2 - 13 June 2020 through 19 June 2020

ER -

Hyperbolic Visual Embedding Learning for Zero-Shot Recognition

Abstract

Publication series

Conference

Bibliographical note

Access Output

Other Access Options

Permanent Link

Other Files and Links

Fingerprint

Cite this