Towards Explainable Computer Vision Methods via Uncertainty Activation Map

Seungyoun Shin; Wonho Bae; Junhyug Noh; Sungjoon Choi

doi:10.1007/978-3-031-47634-1_1

Towards Explainable Computer Vision Methods via Uncertainty Activation Map

Seungyoun Shin, Wonho Bae, Junhyug Noh, Sungjoon Choi

Department of Artificial Intelligence

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

Abstract

This paper focuses on the problem of highlighting the input image regions that result in increasing predictive uncertainty. In particular, we focus on two types of uncertainty, epistemic and aleatoric, and present an uncertainty activation mapping method that can incorporate both types of uncertainty. To this end, we first utilize a mixture-of-experts model combined with class-activation mapping (CAM). The proposed method is extensively evaluated in two different scenarios: multi-label and artificial noise injection scenarios, where we show that our proposed method can effectively capture uncertain regions.

Original language	English
Title of host publication	Pattern Recognition - 7th Asian Conference, ACPR 2023, Proceedings
Editors	Huimin Lu, Michael Blumenstein, Sung-Bae Cho, Cheng-Lin Liu, Yasushi Yagi, Tohru Kamiya
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	1-14
Number of pages	14
ISBN (Print)	9783031476334
DOIs	https://doi.org/10.1007/978-3-031-47634-1_1
State	Published - 2023
Event	7th Asian Conference on Pattern Recognition, ACPR 2023 - Kitakyushu, Japan Duration: 5 Nov 2023 → 8 Nov 2023

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	14406 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	7th Asian Conference on Pattern Recognition, ACPR 2023
Country/Territory	Japan
City	Kitakyushu
Period	5/11/23 → 8/11/23

Bibliographical note

Publisher Copyright:
© 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.

Keywords

Explainable machine learning
Uncertainty estimation
Visual explanation

Access to Document

10.1007/978-3-031-47634-1_1

Cite this

Shin, S., Bae, W., Noh, J., & Choi, S. (2023). Towards Explainable Computer Vision Methods via Uncertainty Activation Map. In H. Lu, M. Blumenstein, S.-B. Cho, C.-L. Liu, Y. Yagi, & T. Kamiya (Eds.), Pattern Recognition - 7th Asian Conference, ACPR 2023, Proceedings (pp. 1-14). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 14406 LNCS). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-031-47634-1_1

Shin, Seungyoun ; Bae, Wonho ; Noh, Junhyug et al. / Towards Explainable Computer Vision Methods via Uncertainty Activation Map. Pattern Recognition - 7th Asian Conference, ACPR 2023, Proceedings. editor / Huimin Lu ; Michael Blumenstein ; Sung-Bae Cho ; Cheng-Lin Liu ; Yasushi Yagi ; Tohru Kamiya. Springer Science and Business Media Deutschland GmbH, 2023. pp. 1-14 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{44f056fb945a43319937328e218eac2c,

title = "Towards Explainable Computer Vision Methods via Uncertainty Activation Map",

abstract = "This paper focuses on the problem of highlighting the input image regions that result in increasing predictive uncertainty. In particular, we focus on two types of uncertainty, epistemic and aleatoric, and present an uncertainty activation mapping method that can incorporate both types of uncertainty. To this end, we first utilize a mixture-of-experts model combined with class-activation mapping (CAM). The proposed method is extensively evaluated in two different scenarios: multi-label and artificial noise injection scenarios, where we show that our proposed method can effectively capture uncertain regions.",

keywords = "Explainable machine learning, Uncertainty estimation, Visual explanation",

author = "Seungyoun Shin and Wonho Bae and Junhyug Noh and Sungjoon Choi",

note = "Publisher Copyright: {\textcopyright} 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.; 7th Asian Conference on Pattern Recognition, ACPR 2023 ; Conference date: 05-11-2023 Through 08-11-2023",

year = "2023",

doi = "10.1007/978-3-031-47634-1_1",

language = "English",

isbn = "9783031476334",

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 = "1--14",

editor = "Huimin Lu and Michael Blumenstein and Sung-Bae Cho and Cheng-Lin Liu and Yasushi Yagi and Tohru Kamiya",

booktitle = "Pattern Recognition - 7th Asian Conference, ACPR 2023, Proceedings",

}

Shin, S, Bae, W, Noh, J & Choi, S 2023, Towards Explainable Computer Vision Methods via Uncertainty Activation Map. in H Lu, M Blumenstein, S-B Cho, C-L Liu, Y Yagi & T Kamiya (eds), Pattern Recognition - 7th Asian Conference, ACPR 2023, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 14406 LNCS, Springer Science and Business Media Deutschland GmbH, pp. 1-14, 7th Asian Conference on Pattern Recognition, ACPR 2023, Kitakyushu, Japan, 5/11/23. https://doi.org/10.1007/978-3-031-47634-1_1

Towards Explainable Computer Vision Methods via Uncertainty Activation Map. / Shin, Seungyoun; Bae, Wonho; Noh, Junhyug et al.
Pattern Recognition - 7th Asian Conference, ACPR 2023, Proceedings. ed. / Huimin Lu; Michael Blumenstein; Sung-Bae Cho; Cheng-Lin Liu; Yasushi Yagi; Tohru Kamiya. Springer Science and Business Media Deutschland GmbH, 2023. p. 1-14 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 14406 LNCS).

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

TY - GEN

T1 - Towards Explainable Computer Vision Methods via Uncertainty Activation Map

AU - Shin, Seungyoun

AU - Bae, Wonho

AU - Noh, Junhyug

AU - Choi, Sungjoon

PY - 2023

Y1 - 2023

N2 - This paper focuses on the problem of highlighting the input image regions that result in increasing predictive uncertainty. In particular, we focus on two types of uncertainty, epistemic and aleatoric, and present an uncertainty activation mapping method that can incorporate both types of uncertainty. To this end, we first utilize a mixture-of-experts model combined with class-activation mapping (CAM). The proposed method is extensively evaluated in two different scenarios: multi-label and artificial noise injection scenarios, where we show that our proposed method can effectively capture uncertain regions.

AB - This paper focuses on the problem of highlighting the input image regions that result in increasing predictive uncertainty. In particular, we focus on two types of uncertainty, epistemic and aleatoric, and present an uncertainty activation mapping method that can incorporate both types of uncertainty. To this end, we first utilize a mixture-of-experts model combined with class-activation mapping (CAM). The proposed method is extensively evaluated in two different scenarios: multi-label and artificial noise injection scenarios, where we show that our proposed method can effectively capture uncertain regions.

KW - Explainable machine learning

KW - Uncertainty estimation

KW - Visual explanation

UR - http://www.scopus.com/inward/record.url?scp=85177231081&partnerID=8YFLogxK

U2 - 10.1007/978-3-031-47634-1_1

DO - 10.1007/978-3-031-47634-1_1

M3 - Conference contribution

AN - SCOPUS:85177231081

SN - 9783031476334

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

SP - 1

EP - 14

BT - Pattern Recognition - 7th Asian Conference, ACPR 2023, Proceedings

A2 - Lu, Huimin

A2 - Blumenstein, Michael

A2 - Cho, Sung-Bae

A2 - Liu, Cheng-Lin

A2 - Yagi, Yasushi

A2 - Kamiya, Tohru

PB - Springer Science and Business Media Deutschland GmbH

T2 - 7th Asian Conference on Pattern Recognition, ACPR 2023

Y2 - 5 November 2023 through 8 November 2023

ER -

Shin S, Bae W, Noh J, Choi S. Towards Explainable Computer Vision Methods via Uncertainty Activation Map. In Lu H, Blumenstein M, Cho SB, Liu CL, Yagi Y, Kamiya T, editors, Pattern Recognition - 7th Asian Conference, ACPR 2023, Proceedings. Springer Science and Business Media Deutschland GmbH. 2023. p. 1-14. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-031-47634-1_1