TSRFormer: Transformer Based Two-stage Refinement for Single Image Shadow Removal

Hua En Chang; Chia Hsuan Hsieh; Hao Hsiang Yang; I. Hsiang Chen; Yi Chung Chen; Yuan Chun Chiang; Zhi Kai Huang; Wei Ting Chen; Sy Yen Kuo

doi:10.1109/CVPRW59228.2023.00148

TSRFormer: Transformer Based Two-stage Refinement for Single Image Shadow Removal

Hua En Chang^*, Chia Hsuan Hsieh^*, Hao Hsiang Yang, I. Hsiang Chen, Yi Chung Chen, Yuan Chun Chiang, Zhi Kai Huang, Wei Ting Chen, Sy Yen Kuo

^*Corresponding author for this work

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

9 Scopus citations

Abstract

Single-image shadow removal aims to remove undesired shadow information from captured images. With the development of deep convolutional neural networks, several methods have been proposed to achieve promising performance in shadow removal. However, they still struggle with limited performance due to the non-homogeneous intensity distribution of the shadow. To address this issue, we propose a two-stage shadow removal architecture based on the transformer called TSRFormer. The proposed architecture is divided into shadow removal and content refinement networks. These two stages adopt different transformer architectures and remove the shadow based on different information to achieve effective shadow removal. Experiments performed on challenging benchmark show that the proposed model achieves the 2nd highest SSIM in the NTIRE 2023 Image Shadow Removal Challenge. The source code will be public after the acceptance of this paper.

Original language	English
Title of host publication	Proceedings - 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2023
Publisher	IEEE Computer Society
Pages	1436-1446
Number of pages	11
ISBN (Electronic)	9798350302493
DOIs	https://doi.org/10.1109/CVPRW59228.2023.00148
State	Published - 2023
Externally published	Yes
Event	2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2023 - Vancouver, Canada Duration: 18 06 2023 → 22 06 2023

Publication series

Name	IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops
Volume	2023-June
ISSN (Print)	2160-7508
ISSN (Electronic)	2160-7516

Conference

Conference	2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2023
Country/Territory	Canada
City	Vancouver
Period	18/06/23 → 22/06/23

Bibliographical note

Publisher Copyright:
© 2023 IEEE.

Access to Document

10.1109/CVPRW59228.2023.00148

Cite this

Chang, H. E., Hsieh, C. H., Yang, H. H., Chen, I. H., Chen, Y. C., Chiang, Y. C., Huang, Z. K., Chen, W. T., & Kuo, S. Y. (2023). TSRFormer: Transformer Based Two-stage Refinement for Single Image Shadow Removal. In Proceedings - 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2023 (pp. 1436-1446). (IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops; Vol. 2023-June). IEEE Computer Society. https://doi.org/10.1109/CVPRW59228.2023.00148

Chang, Hua En ; Hsieh, Chia Hsuan ; Yang, Hao Hsiang et al. / TSRFormer : Transformer Based Two-stage Refinement for Single Image Shadow Removal. Proceedings - 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2023. IEEE Computer Society, 2023. pp. 1436-1446 (IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops).

@inproceedings{38bf4ee6393f4a4aa8310d589eb4aec6,

title = "TSRFormer: Transformer Based Two-stage Refinement for Single Image Shadow Removal",

abstract = "Single-image shadow removal aims to remove undesired shadow information from captured images. With the development of deep convolutional neural networks, several methods have been proposed to achieve promising performance in shadow removal. However, they still struggle with limited performance due to the non-homogeneous intensity distribution of the shadow. To address this issue, we propose a two-stage shadow removal architecture based on the transformer called TSRFormer. The proposed architecture is divided into shadow removal and content refinement networks. These two stages adopt different transformer architectures and remove the shadow based on different information to achieve effective shadow removal. Experiments performed on challenging benchmark show that the proposed model achieves the 2nd highest SSIM in the NTIRE 2023 Image Shadow Removal Challenge. The source code will be public after the acceptance of this paper.",

author = "Chang, \{Hua En\} and Hsieh, \{Chia Hsuan\} and Yang, \{Hao Hsiang\} and Chen, \{I. Hsiang\} and Chen, \{Yi Chung\} and Chiang, \{Yuan Chun\} and Huang, \{Zhi Kai\} and Chen, \{Wei Ting\} and Kuo, \{Sy Yen\}",

note = "Publisher Copyright: {\textcopyright} 2023 IEEE.; 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2023 ; Conference date: 18-06-2023 Through 22-06-2023",

year = "2023",

doi = "10.1109/CVPRW59228.2023.00148",

language = "英语",

series = "IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops",

publisher = "IEEE Computer Society",

pages = "1436--1446",

booktitle = "Proceedings - 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2023",

address = "美国",

}

Chang, HE, Hsieh, CH, Yang, HH, Chen, IH, Chen, YC, Chiang, YC, Huang, ZK, Chen, WT & Kuo, SY 2023, TSRFormer: Transformer Based Two-stage Refinement for Single Image Shadow Removal. in Proceedings - 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2023. IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops, vol. 2023-June, IEEE Computer Society, pp. 1436-1446, 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2023, Vancouver, Canada, 18/06/23. https://doi.org/10.1109/CVPRW59228.2023.00148

TSRFormer: Transformer Based Two-stage Refinement for Single Image Shadow Removal. / Chang, Hua En; Hsieh, Chia Hsuan; Yang, Hao Hsiang et al.
Proceedings - 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2023. IEEE Computer Society, 2023. p. 1436-1446 (IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops; Vol. 2023-June).

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

TY - GEN

T1 - TSRFormer

T2 - 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2023

AU - Chang, Hua En

AU - Hsieh, Chia Hsuan

AU - Yang, Hao Hsiang

AU - Chen, I. Hsiang

AU - Chen, Yi Chung

AU - Chiang, Yuan Chun

AU - Huang, Zhi Kai

AU - Chen, Wei Ting

AU - Kuo, Sy Yen

PY - 2023

Y1 - 2023

N2 - Single-image shadow removal aims to remove undesired shadow information from captured images. With the development of deep convolutional neural networks, several methods have been proposed to achieve promising performance in shadow removal. However, they still struggle with limited performance due to the non-homogeneous intensity distribution of the shadow. To address this issue, we propose a two-stage shadow removal architecture based on the transformer called TSRFormer. The proposed architecture is divided into shadow removal and content refinement networks. These two stages adopt different transformer architectures and remove the shadow based on different information to achieve effective shadow removal. Experiments performed on challenging benchmark show that the proposed model achieves the 2nd highest SSIM in the NTIRE 2023 Image Shadow Removal Challenge. The source code will be public after the acceptance of this paper.

AB - Single-image shadow removal aims to remove undesired shadow information from captured images. With the development of deep convolutional neural networks, several methods have been proposed to achieve promising performance in shadow removal. However, they still struggle with limited performance due to the non-homogeneous intensity distribution of the shadow. To address this issue, we propose a two-stage shadow removal architecture based on the transformer called TSRFormer. The proposed architecture is divided into shadow removal and content refinement networks. These two stages adopt different transformer architectures and remove the shadow based on different information to achieve effective shadow removal. Experiments performed on challenging benchmark show that the proposed model achieves the 2nd highest SSIM in the NTIRE 2023 Image Shadow Removal Challenge. The source code will be public after the acceptance of this paper.

UR - https://www.scopus.com/pages/publications/85170826284

U2 - 10.1109/CVPRW59228.2023.00148

DO - 10.1109/CVPRW59228.2023.00148

M3 - 会议稿件

AN - SCOPUS:85170826284

T3 - IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops

SP - 1436

EP - 1446

BT - Proceedings - 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2023

PB - IEEE Computer Society

Y2 - 18 June 2023 through 22 June 2023

ER -

Chang HE, Hsieh CH, Yang HH, Chen IH, Chen YC, Chiang YC et al. TSRFormer: Transformer Based Two-stage Refinement for Single Image Shadow Removal. In Proceedings - 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2023. IEEE Computer Society. 2023. p. 1436-1446. (IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops). doi: 10.1109/CVPRW59228.2023.00148

TSRFormer: Transformer Based Two-stage Refinement for Single Image Shadow Removal

Abstract

Publication series

Conference

Bibliographical note

Access to Document

Other files and links

Fingerprint

Cite this