AIA: Attention in Attention Within Collaborate Domains

Le Zhang; Qi Feng; Yao Lu; Chang Liu; Guangming Lu

doi:10.1007/978-3-031-18907-4_47

AIA: Attention in Attention Within Collaborate Domains

Le Zhang
, Qi Feng
, Yao Lu^*
, Chang Liu
, Guangming Lu

^*Corresponding author for this work

Harbin Institute of Technology Shenzhen
Industrial and Commercial Bank of China Limited
Guangdong Provincial Key Laboratory of Novel Security Intelligence Technologies

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

Attention mechanisms can effectively improve the performance of the mobile networks with a limited computational complexity cost. However, existing attention methods extract importance from only one domain of the networks, hindering further performance improvement. In this paper, we propose the Attention in Attention (AIA) mechanism integrating One Dimension Frequency Channel Attention (1D FCA) with Joint Coordinate Attention (JCA) to collaboratively adjust the channel and coordinate weights in frequency and spatial domains, respectively. Specifically, 1D FCA using 1D Discrete Cosine Transform (DCT) adaptively extract and enhance the necessary channel information in the frequency domain. The JCA using explicit and implicit coordinate information extract and embed position feature into frequency channel attention. Extensive experiments on different datasets demonstrate that the proposed AIA mechanism can effectively improve the accuracy with only a limited computation complexity cost.

Original language	English
Title of host publication	Pattern Recognition and Computer Vision - 5th Chinese Conference, PRCV 2022, Proceedings
Editors	Shiqi Yu, Jianguo Zhang, Zhaoxiang Zhang, Tieniu Tan, Pong C. Yuen, Yike Guo, Junwei Han, Jianhuang Lai
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	605-617
Number of pages	13
ISBN (Print)	9783031189067
DOIs	https://doi.org/10.1007/978-3-031-18907-4_47
State	Published - 2022
Externally published	Yes
Event	5th Chinese Conference on Pattern Recognition and Computer Vision, PRCV 2022 - Shenzhen, China Duration: 4 Nov 2022 → 7 Nov 2022

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	13534 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	5th Chinese Conference on Pattern Recognition and Computer Vision, PRCV 2022
Country/Territory	China
City	Shenzhen
Period	4/11/22 → 7/11/22

Keywords

Attention in Attention
Discrete Cosine Transform
Frequency Domain Attention
Spatial Domain Attention

Access to Document

10.1007/978-3-031-18907-4_47

Cite this

Zhang, L., Feng, Q., Lu, Y., Liu, C., & Lu, G. (2022). AIA: Attention in Attention Within Collaborate Domains. In S. Yu, J. Zhang, Z. Zhang, T. Tan, P. C. Yuen, Y. Guo, J. Han, & J. Lai (Eds.), Pattern Recognition and Computer Vision - 5th Chinese Conference, PRCV 2022, Proceedings (pp. 605-617). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 13534 LNCS). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-031-18907-4_47

Zhang, Le ; Feng, Qi ; Lu, Yao et al. / AIA : Attention in Attention Within Collaborate Domains. Pattern Recognition and Computer Vision - 5th Chinese Conference, PRCV 2022, Proceedings. editor / Shiqi Yu ; Jianguo Zhang ; Zhaoxiang Zhang ; Tieniu Tan ; Pong C. Yuen ; Yike Guo ; Junwei Han ; Jianhuang Lai. Springer Science and Business Media Deutschland GmbH, 2022. pp. 605-617 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{bb8c6d40ce734bc58ccd35459e455a50,

title = "AIA: Attention in Attention Within Collaborate Domains",

abstract = "Attention mechanisms can effectively improve the performance of the mobile networks with a limited computational complexity cost. However, existing attention methods extract importance from only one domain of the networks, hindering further performance improvement. In this paper, we propose the Attention in Attention (AIA) mechanism integrating One Dimension Frequency Channel Attention (1D FCA) with Joint Coordinate Attention (JCA) to collaboratively adjust the channel and coordinate weights in frequency and spatial domains, respectively. Specifically, 1D FCA using 1D Discrete Cosine Transform (DCT) adaptively extract and enhance the necessary channel information in the frequency domain. The JCA using explicit and implicit coordinate information extract and embed position feature into frequency channel attention. Extensive experiments on different datasets demonstrate that the proposed AIA mechanism can effectively improve the accuracy with only a limited computation complexity cost.",

keywords = "Attention in Attention, Discrete Cosine Transform, Frequency Domain Attention, Spatial Domain Attention",

author = "Le Zhang and Qi Feng and Yao Lu and Chang Liu and Guangming Lu",

note = "Publisher Copyright: {\textcopyright} 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.; 5th Chinese Conference on Pattern Recognition and Computer Vision, PRCV 2022 ; Conference date: 04-11-2022 Through 07-11-2022",

year = "2022",

doi = "10.1007/978-3-031-18907-4\_47",

language = "英语",

isbn = "9783031189067",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Science and Business Media Deutschland GmbH",

pages = "605--617",

editor = "Shiqi Yu and Jianguo Zhang and Zhaoxiang Zhang and Tieniu Tan and Yuen, \{Pong C.\} and Yike Guo and Junwei Han and Jianhuang Lai",

booktitle = "Pattern Recognition and Computer Vision - 5th Chinese Conference, PRCV 2022, Proceedings",

address = "德国",

}

Zhang, L, Feng, Q, Lu, Y, Liu, C & Lu, G 2022, AIA: Attention in Attention Within Collaborate Domains. in S Yu, J Zhang, Z Zhang, T Tan, PC Yuen, Y Guo, J Han & J Lai (eds), Pattern Recognition and Computer Vision - 5th Chinese Conference, PRCV 2022, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 13534 LNCS, Springer Science and Business Media Deutschland GmbH, pp. 605-617, 5th Chinese Conference on Pattern Recognition and Computer Vision, PRCV 2022, Shenzhen, China, 4/11/22. https://doi.org/10.1007/978-3-031-18907-4_47

AIA: Attention in Attention Within Collaborate Domains. / Zhang, Le; Feng, Qi; Lu, Yao et al.
Pattern Recognition and Computer Vision - 5th Chinese Conference, PRCV 2022, Proceedings. ed. / Shiqi Yu; Jianguo Zhang; Zhaoxiang Zhang; Tieniu Tan; Pong C. Yuen; Yike Guo; Junwei Han; Jianhuang Lai. Springer Science and Business Media Deutschland GmbH, 2022. p. 605-617 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 13534 LNCS).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - AIA

T2 - 5th Chinese Conference on Pattern Recognition and Computer Vision, PRCV 2022

AU - Zhang, Le

AU - Feng, Qi

AU - Lu, Yao

AU - Liu, Chang

AU - Lu, Guangming

PY - 2022

Y1 - 2022

N2 - Attention mechanisms can effectively improve the performance of the mobile networks with a limited computational complexity cost. However, existing attention methods extract importance from only one domain of the networks, hindering further performance improvement. In this paper, we propose the Attention in Attention (AIA) mechanism integrating One Dimension Frequency Channel Attention (1D FCA) with Joint Coordinate Attention (JCA) to collaboratively adjust the channel and coordinate weights in frequency and spatial domains, respectively. Specifically, 1D FCA using 1D Discrete Cosine Transform (DCT) adaptively extract and enhance the necessary channel information in the frequency domain. The JCA using explicit and implicit coordinate information extract and embed position feature into frequency channel attention. Extensive experiments on different datasets demonstrate that the proposed AIA mechanism can effectively improve the accuracy with only a limited computation complexity cost.

AB - Attention mechanisms can effectively improve the performance of the mobile networks with a limited computational complexity cost. However, existing attention methods extract importance from only one domain of the networks, hindering further performance improvement. In this paper, we propose the Attention in Attention (AIA) mechanism integrating One Dimension Frequency Channel Attention (1D FCA) with Joint Coordinate Attention (JCA) to collaboratively adjust the channel and coordinate weights in frequency and spatial domains, respectively. Specifically, 1D FCA using 1D Discrete Cosine Transform (DCT) adaptively extract and enhance the necessary channel information in the frequency domain. The JCA using explicit and implicit coordinate information extract and embed position feature into frequency channel attention. Extensive experiments on different datasets demonstrate that the proposed AIA mechanism can effectively improve the accuracy with only a limited computation complexity cost.

KW - Attention in Attention

KW - Discrete Cosine Transform

KW - Frequency Domain Attention

KW - Spatial Domain Attention

UR - https://www.scopus.com/pages/publications/85142668820

U2 - 10.1007/978-3-031-18907-4_47

DO - 10.1007/978-3-031-18907-4_47

M3 - 会议稿件

AN - SCOPUS:85142668820

SN - 9783031189067

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

SP - 605

EP - 617

BT - Pattern Recognition and Computer Vision - 5th Chinese Conference, PRCV 2022, Proceedings

A2 - Yu, Shiqi

A2 - Zhang, Jianguo

A2 - Zhang, Zhaoxiang

A2 - Tan, Tieniu

A2 - Yuen, Pong C.

A2 - Guo, Yike

A2 - Han, Junwei

A2 - Lai, Jianhuang

PB - Springer Science and Business Media Deutschland GmbH

Y2 - 4 November 2022 through 7 November 2022

ER -

Zhang L, Feng Q, Lu Y, Liu C, Lu G. AIA: Attention in Attention Within Collaborate Domains. In Yu S, Zhang J, Zhang Z, Tan T, Yuen PC, Guo Y, Han J, Lai J, editors, Pattern Recognition and Computer Vision - 5th Chinese Conference, PRCV 2022, Proceedings. Springer Science and Business Media Deutschland GmbH. 2022. p. 605-617. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-031-18907-4_47