As time goes by: dynamic face aging method using CycleGAN

Wanzhou Liu

doi:10.1117/12.2631420

18 March 2022 As time goes by: dynamic face aging method using CycleGAN

Wanzhou Liu

Proceedings Volume 12168, International Conference on Computer Graphics, Artificial Intelligence, and Data Processing (ICCAID 2021); 121681C (2022) https://doi.org/10.1117/12.2631420
Event: International Conference on Computer Graphics, Artificial Intelligence, and Data Processing (ICCAID 2021), 2021, Harbin, China

Abstract

Achieving image-to-image style transfer using various types of generative adversarial networks is a popular question in computer vision, with numerous works being proposed. However, the current typical research on the use of neural networks to achieve image-to-image style transfer is to improve the output performance through the improvement of the model, and the intermediate process of style transfer has not been paid much attention. In this paper, based on the implementation of CycleGAN style translation, the gradients of non-leaf nodes are calculated and extracted by using the autograd function in PyTorch; and finally, the aging images of faces at different ages are dynamically generated by using the middle layer gradient. The experimental results, based on the AGFW-v2 dataset, tell us that this is capable for the initial research purpose; and the presentation of the process of style transfer intermediate enriches the stage changes of age. The processing of intermediate layer features in this study can also be used in areas such as photo enhancement, image dehazing and face swapping to achieve controllability of model processing results.

Citation Download Citation

Wanzhou Liu "As time goes by: dynamic face aging method using CycleGAN", Proc. SPIE 12168, International Conference on Computer Graphics, Artificial Intelligence, and Data Processing (ICCAID 2021), 121681C (18 March 2022); https://doi.org/10.1117/12.2631420

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