Expression synthesis based on progressive growing of generative adversarial network and StyleGAN2

Nengsheng Bao; Shengbao Guo; Jian Gao

doi:10.1117/12.2660973

2 February 2023 Expression synthesis based on progressive growing of generative adversarial network and StyleGAN2

Nengsheng Bao, Shengbao Guo, Jian Gao

Proceedings Volume 12462, Third International Symposium on Computer Engineering and Intelligent Communications (ISCEIC 2022); 1246224 (2023) https://doi.org/10.1117/12.2660973
Event: International Symposium on Computer Engineering and Intelligent Communications (ISCEIC 2022), 2022, Xi'an, China

Abstract

An expression synthesis method based on Progressive Generative Adversarial Network (PGGAN) and StyleGAN2 is proposed to create customized datasets in order to address the problems with traditional expression dataset creation methods, such as their high cost, length of time, and difficulty in avoiding the influence of subjective annotation factors. In order to separate identity features from expression features in expression images, this paper first uses PGGAN to generate images of various resolution layers. It then builds a feedback network for this image and obtains the feature latent codes corresponding to the generated images of each resolution layer. The target face image is then produced with specific expressions based on the image generation direction dictated by the fusion result, using the StyleGAN2 network model to fuse the latent codes of identity information in the target face image with the latent codes of expression information in the original expression image. Finally, the Fréchet onset distance (FID) and structural similarity (SSIM) of the synthesized image and the original image are compared. The mean values of FID and SSIM compared with the original image in this paper are 34.61 and 0.90, respectively. It is difficult for human visual perception to distinguish the real from the fake, proving the authenticity of the synthesized image and the efficacy of the synthesis method.

Citation Download Citation

Nengsheng Bao, Shengbao Guo, and Jian Gao "Expression synthesis based on progressive growing of generative adversarial network and StyleGAN2", Proc. SPIE 12462, Third International Symposium on Computer Engineering and Intelligent Communications (ISCEIC 2022), 1246224 (2 February 2023); https://doi.org/10.1117/12.2660973

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available

Members: $17.00

Non-members: $21.00 ADD TO CART

PROCEEDINGS
10 PAGES

DOWNLOAD PAPER SAVE TO MY LIBRARY

GET CITATION

RIGHTS & PERMISSIONS

Get copyright permission Get copyright permission on Copyright Marketplace

KEYWORDS

Image fusion

Feature extraction

Data modeling

Image resolution

Visualization

Gallium nitride

Neural networks

Show All Keywords

Keywords/Phrases

Search In:

Publication Years