SS-GANs: Text-to-image via stage by stage generative adversarial networks

Ming Tian; Yuting Xue; Chunna Tian; Lei Wang; Donghu Deng; Wei Wei

doi:10.1007/978-3-030-31723-2_40

SS-GANs: Text-to-image via stage by stage generative adversarial networks

Ming Tian, Yuting Xue, Chunna Tian, Lei Wang, Donghu Deng, Wei Wei

School of Computer Science

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

Abstract

Realistic text-to-image synthesis has achieved great improvements in recent years. However, most work ignores the relationship between low and high resolution and prefers to adopt identical module in different stages. It is obviously inappropriate because the differences in various generation stages are huge. Therefore, we propose a novel structure of network named SS-GANs, in which specific modules are added in different stages to satisfy the unique requirements. In addition, we also explore an effective training way named coordinated train and a simple negative sample selection mechanism. Lastly, we train our model on Oxford-102 dataset, which outperforms the state-of-the-art models.

Original language	English
Title of host publication	Pattern Recognition and Computer Vision 2nd Chinese Conference, PRCV 2019, Proceedings, Part II
Editors	Zhouchen Lin, Liang Wang, Tieniu Tan, Jian Yang, Guangming Shi, Nanning Zheng, Xilin Chen, Yanning Zhang
Publisher	Springer
Pages	475-486
Number of pages	12
ISBN (Print)	9783030317225
DOIs	https://doi.org/10.1007/978-3-030-31723-2_40
State	Published - 2019
Event	2nd Chinese Conference on Pattern Recognition and Computer Vision, PRCV 2019 - Xi'an, China Duration: 8 Nov 2019 → 11 Nov 2019

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	11858 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	2nd Chinese Conference on Pattern Recognition and Computer Vision, PRCV 2019
Country/Territory	China
City	Xi'an
Period	8/11/19 → 11/11/19

Keywords

Coordinated train
Different stages
Negative samples
Text-to-image

Access to Document

10.1007/978-3-030-31723-2_40

Cite this

Tian, M., Xue, Y., Tian, C., Wang, L., Deng, D., & Wei, W. (2019). SS-GANs: Text-to-image via stage by stage generative adversarial networks. In Z. Lin, L. Wang, T. Tan, J. Yang, G. Shi, N. Zheng, X. Chen, & Y. Zhang (Eds.), Pattern Recognition and Computer Vision 2nd Chinese Conference, PRCV 2019, Proceedings, Part II (pp. 475-486). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11858 LNCS). Springer. https://doi.org/10.1007/978-3-030-31723-2_40

Tian, Ming ; Xue, Yuting ; Tian, Chunna et al. / SS-GANs : Text-to-image via stage by stage generative adversarial networks. Pattern Recognition and Computer Vision 2nd Chinese Conference, PRCV 2019, Proceedings, Part II. editor / Zhouchen Lin ; Liang Wang ; Tieniu Tan ; Jian Yang ; Guangming Shi ; Nanning Zheng ; Xilin Chen ; Yanning Zhang. Springer, 2019. pp. 475-486 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{9e09c3ea696d4c2f8b300ce0f04f7fcf,

title = "SS-GANs: Text-to-image via stage by stage generative adversarial networks",

abstract = "Realistic text-to-image synthesis has achieved great improvements in recent years. However, most work ignores the relationship between low and high resolution and prefers to adopt identical module in different stages. It is obviously inappropriate because the differences in various generation stages are huge. Therefore, we propose a novel structure of network named SS-GANs, in which specific modules are added in different stages to satisfy the unique requirements. In addition, we also explore an effective training way named coordinated train and a simple negative sample selection mechanism. Lastly, we train our model on Oxford-102 dataset, which outperforms the state-of-the-art models.",

keywords = "Coordinated train, Different stages, Negative samples, Text-to-image",

author = "Ming Tian and Yuting Xue and Chunna Tian and Lei Wang and Donghu Deng and Wei Wei",

note = "Publisher Copyright: {\textcopyright} Springer Nature Switzerland AG 2019.; 2nd Chinese Conference on Pattern Recognition and Computer Vision, PRCV 2019 ; Conference date: 08-11-2019 Through 11-11-2019",

year = "2019",

doi = "10.1007/978-3-030-31723-2_40",

language = "英语",

isbn = "9783030317225",

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

publisher = "Springer",

pages = "475--486",

editor = "Zhouchen Lin and Liang Wang and Tieniu Tan and Jian Yang and Guangming Shi and Nanning Zheng and Xilin Chen and Yanning Zhang",

booktitle = "Pattern Recognition and Computer Vision 2nd Chinese Conference, PRCV 2019, Proceedings, Part II",

}

Tian, M, Xue, Y, Tian, C, Wang, L, Deng, D & Wei, W 2019, SS-GANs: Text-to-image via stage by stage generative adversarial networks. in Z Lin, L Wang, T Tan, J Yang, G Shi, N Zheng, X Chen & Y Zhang (eds), Pattern Recognition and Computer Vision 2nd Chinese Conference, PRCV 2019, Proceedings, Part II. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 11858 LNCS, Springer, pp. 475-486, 2nd Chinese Conference on Pattern Recognition and Computer Vision, PRCV 2019, Xi'an, China, 8/11/19. https://doi.org/10.1007/978-3-030-31723-2_40

SS-GANs: Text-to-image via stage by stage generative adversarial networks. / Tian, Ming; Xue, Yuting; Tian, Chunna et al.
Pattern Recognition and Computer Vision 2nd Chinese Conference, PRCV 2019, Proceedings, Part II. ed. / Zhouchen Lin; Liang Wang; Tieniu Tan; Jian Yang; Guangming Shi; Nanning Zheng; Xilin Chen; Yanning Zhang. Springer, 2019. p. 475-486 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11858 LNCS).

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

TY - GEN

T1 - SS-GANs

T2 - 2nd Chinese Conference on Pattern Recognition and Computer Vision, PRCV 2019

AU - Tian, Ming

AU - Xue, Yuting

AU - Tian, Chunna

AU - Wang, Lei

AU - Deng, Donghu

AU - Wei, Wei

N1 - Publisher Copyright: © Springer Nature Switzerland AG 2019.

PY - 2019

Y1 - 2019

N2 - Realistic text-to-image synthesis has achieved great improvements in recent years. However, most work ignores the relationship between low and high resolution and prefers to adopt identical module in different stages. It is obviously inappropriate because the differences in various generation stages are huge. Therefore, we propose a novel structure of network named SS-GANs, in which specific modules are added in different stages to satisfy the unique requirements. In addition, we also explore an effective training way named coordinated train and a simple negative sample selection mechanism. Lastly, we train our model on Oxford-102 dataset, which outperforms the state-of-the-art models.

AB - Realistic text-to-image synthesis has achieved great improvements in recent years. However, most work ignores the relationship between low and high resolution and prefers to adopt identical module in different stages. It is obviously inappropriate because the differences in various generation stages are huge. Therefore, we propose a novel structure of network named SS-GANs, in which specific modules are added in different stages to satisfy the unique requirements. In addition, we also explore an effective training way named coordinated train and a simple negative sample selection mechanism. Lastly, we train our model on Oxford-102 dataset, which outperforms the state-of-the-art models.

KW - Coordinated train

KW - Different stages

KW - Negative samples

KW - Text-to-image

UR - http://www.scopus.com/inward/record.url?scp=85076999853&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-31723-2_40

DO - 10.1007/978-3-030-31723-2_40

M3 - 会议稿件

AN - SCOPUS:85076999853

SN - 9783030317225

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

SP - 475

EP - 486

BT - Pattern Recognition and Computer Vision 2nd Chinese Conference, PRCV 2019, Proceedings, Part II

A2 - Lin, Zhouchen

A2 - Wang, Liang

A2 - Tan, Tieniu

A2 - Yang, Jian

A2 - Shi, Guangming

A2 - Zheng, Nanning

A2 - Chen, Xilin

A2 - Zhang, Yanning

PB - Springer

Y2 - 8 November 2019 through 11 November 2019

ER -

Tian M, Xue Y, Tian C, Wang L, Deng D, Wei W. SS-GANs: Text-to-image via stage by stage generative adversarial networks. In Lin Z, Wang L, Tan T, Yang J, Shi G, Zheng N, Chen X, Zhang Y, editors, Pattern Recognition and Computer Vision 2nd Chinese Conference, PRCV 2019, Proceedings, Part II. Springer. 2019. p. 475-486. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-030-31723-2_40

SS-GANs: Text-to-image via stage by stage generative adversarial networks

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this