Interpretability of Deep Network Structures for SAR Image Target Recognition Based on Generative Adversarial Mask

Yuhang Zhang; Jie Geng; Wen Jiang

doi:10.1109/ICUS58632.2023.10318287

Interpretability of Deep Network Structures for SAR Image Target Recognition Based on Generative Adversarial Mask

Yuhang Zhang, Jie Geng, Wen Jiang

School of Electronics and Information

Northwestern Polytechnical University Xian

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

1 Scopus citations

Abstract

Deep neural networks (DNN) have been widely used in SAR image Target recognition and have achieved great success. However, the black box attribute of DNN makes the mechanism of the network inexplicable and confusing. In this paper, a network layer analysis model based on generative adversarial mask (GAM) is proposed to enhance the structure interpretability of DNNs. Inspired by network adaptive search, we introduce a mask layer and construct a layer analysis model. After sparse training and generative adversarial training, the contribution of the network layer can be calculated based on the mask layer parameters. We conducted sufficient experiments on several typical DNNs to verify the effectiveness of GAM.

Original language	English
Title of host publication	Proceedings of 2023 IEEE International Conference on Unmanned Systems, ICUS 2023
Editors	Rong Song
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	786-791
Number of pages	6
ISBN (Electronic)	9798350316308
DOIs	https://doi.org/10.1109/ICUS58632.2023.10318287
State	Published - 2023
Event	2023 IEEE International Conference on Unmanned Systems, ICUS 2023 - Hefei, China Duration: 13 Oct 2023 → 15 Oct 2023

Publication series

Name	Proceedings of 2023 IEEE International Conference on Unmanned Systems, ICUS 2023

Conference

Conference	2023 IEEE International Conference on Unmanned Systems, ICUS 2023
Country/Territory	China
City	Hefei
Period	13/10/23 → 15/10/23

Keywords

Deep neural network
SAR image recognition
Structure interpretability

Access to Document

10.1109/ICUS58632.2023.10318287

Cite this

Zhang, Y., Geng, J., & Jiang, W. (2023). Interpretability of Deep Network Structures for SAR Image Target Recognition Based on Generative Adversarial Mask. In R. Song (Ed.), Proceedings of 2023 IEEE International Conference on Unmanned Systems, ICUS 2023 (pp. 786-791). (Proceedings of 2023 IEEE International Conference on Unmanned Systems, ICUS 2023). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ICUS58632.2023.10318287

Zhang, Yuhang ; Geng, Jie ; Jiang, Wen. / Interpretability of Deep Network Structures for SAR Image Target Recognition Based on Generative Adversarial Mask. Proceedings of 2023 IEEE International Conference on Unmanned Systems, ICUS 2023. editor / Rong Song. Institute of Electrical and Electronics Engineers Inc., 2023. pp. 786-791 (Proceedings of 2023 IEEE International Conference on Unmanned Systems, ICUS 2023).

@inproceedings{d7a5b575d8ed40e1b67fd520f06950fa,

title = "Interpretability of Deep Network Structures for SAR Image Target Recognition Based on Generative Adversarial Mask",

abstract = "Deep neural networks (DNN) have been widely used in SAR image Target recognition and have achieved great success. However, the black box attribute of DNN makes the mechanism of the network inexplicable and confusing. In this paper, a network layer analysis model based on generative adversarial mask (GAM) is proposed to enhance the structure interpretability of DNNs. Inspired by network adaptive search, we introduce a mask layer and construct a layer analysis model. After sparse training and generative adversarial training, the contribution of the network layer can be calculated based on the mask layer parameters. We conducted sufficient experiments on several typical DNNs to verify the effectiveness of GAM.",

keywords = "Deep neural network, SAR image recognition, Structure interpretability",

author = "Yuhang Zhang and Jie Geng and Wen Jiang",

note = "Publisher Copyright: {\textcopyright} 2023 IEEE.; 2023 IEEE International Conference on Unmanned Systems, ICUS 2023 ; Conference date: 13-10-2023 Through 15-10-2023",

year = "2023",

doi = "10.1109/ICUS58632.2023.10318287",

language = "英语",

series = "Proceedings of 2023 IEEE International Conference on Unmanned Systems, ICUS 2023",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "786--791",

editor = "Rong Song",

booktitle = "Proceedings of 2023 IEEE International Conference on Unmanned Systems, ICUS 2023",

}

Zhang, Y, Geng, J & Jiang, W 2023, Interpretability of Deep Network Structures for SAR Image Target Recognition Based on Generative Adversarial Mask. in R Song (ed.), Proceedings of 2023 IEEE International Conference on Unmanned Systems, ICUS 2023. Proceedings of 2023 IEEE International Conference on Unmanned Systems, ICUS 2023, Institute of Electrical and Electronics Engineers Inc., pp. 786-791, 2023 IEEE International Conference on Unmanned Systems, ICUS 2023, Hefei, China, 13/10/23. https://doi.org/10.1109/ICUS58632.2023.10318287

Interpretability of Deep Network Structures for SAR Image Target Recognition Based on Generative Adversarial Mask. / Zhang, Yuhang; Geng, Jie ; Jiang, Wen.
Proceedings of 2023 IEEE International Conference on Unmanned Systems, ICUS 2023. ed. / Rong Song. Institute of Electrical and Electronics Engineers Inc., 2023. p. 786-791 (Proceedings of 2023 IEEE International Conference on Unmanned Systems, ICUS 2023).

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

TY - GEN

T1 - Interpretability of Deep Network Structures for SAR Image Target Recognition Based on Generative Adversarial Mask

AU - Zhang, Yuhang

AU - Geng, Jie

AU - Jiang, Wen

PY - 2023

Y1 - 2023

N2 - Deep neural networks (DNN) have been widely used in SAR image Target recognition and have achieved great success. However, the black box attribute of DNN makes the mechanism of the network inexplicable and confusing. In this paper, a network layer analysis model based on generative adversarial mask (GAM) is proposed to enhance the structure interpretability of DNNs. Inspired by network adaptive search, we introduce a mask layer and construct a layer analysis model. After sparse training and generative adversarial training, the contribution of the network layer can be calculated based on the mask layer parameters. We conducted sufficient experiments on several typical DNNs to verify the effectiveness of GAM.

AB - Deep neural networks (DNN) have been widely used in SAR image Target recognition and have achieved great success. However, the black box attribute of DNN makes the mechanism of the network inexplicable and confusing. In this paper, a network layer analysis model based on generative adversarial mask (GAM) is proposed to enhance the structure interpretability of DNNs. Inspired by network adaptive search, we introduce a mask layer and construct a layer analysis model. After sparse training and generative adversarial training, the contribution of the network layer can be calculated based on the mask layer parameters. We conducted sufficient experiments on several typical DNNs to verify the effectiveness of GAM.

KW - Deep neural network

KW - SAR image recognition

KW - Structure interpretability

UR - http://www.scopus.com/inward/record.url?scp=85180127375&partnerID=8YFLogxK

U2 - 10.1109/ICUS58632.2023.10318287

DO - 10.1109/ICUS58632.2023.10318287

M3 - 会议稿件

AN - SCOPUS:85180127375

T3 - Proceedings of 2023 IEEE International Conference on Unmanned Systems, ICUS 2023

SP - 786

EP - 791

BT - Proceedings of 2023 IEEE International Conference on Unmanned Systems, ICUS 2023

A2 - Song, Rong

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 2023 IEEE International Conference on Unmanned Systems, ICUS 2023

Y2 - 13 October 2023 through 15 October 2023

ER -

Zhang Y, Geng J , Jiang W. Interpretability of Deep Network Structures for SAR Image Target Recognition Based on Generative Adversarial Mask. In Song R, editor, Proceedings of 2023 IEEE International Conference on Unmanned Systems, ICUS 2023. Institute of Electrical and Electronics Engineers Inc. 2023. p. 786-791. (Proceedings of 2023 IEEE International Conference on Unmanned Systems, ICUS 2023). doi: 10.1109/ICUS58632.2023.10318287

Interpretability of Deep Network Structures for SAR Image Target Recognition Based on Generative Adversarial Mask

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this