Publication

Re-training StyleGAN-A first step towards building large, scalable synthetic facial datasets

Varkarakis, Viktor
; 
Bazrafkan, Shabab
; 
Corcoran, Peter
Thumbnail Image
Files
Thumbnail Image
09180189.pdf
Adobe PDF, 959.5 KB
Identifiers
http://hdl.handle.net/10379/16785
https://doi.org/10.13025/18638
Repository DOI
Publication Date
2020-08-31
Keywords
sunthetic face data, face recognition, generative adversarial networks, GANs, StyleGAN
Type
Conference Paper
Downloads
Show all metadata
Citation
Varkarakis, Viktor, Bazrafkan, Shabab, & Corcoran, Peter. (2020). Re-training StyleGAN-A first step towards building large, scalable synthetic facial datasets. Paper presented at the 31st Irish Signals and Systems Conference (ISSC), Letterkenny, Ireland, 11-12 June. doi:10.1109/ISSC49989.2020.9180189
Abstract
StyleGAN is a state-of-art generative adversarial network architecture that generates random 2D high-quality synthetic facial data samples. In this paper we recap the StyleGAN architecture and training methodology and present our experiences of retraining it on a number of alternative public datasets. Practical issues and challenges arising from the retraining process are discussed. Tests and validation results are presented and a comparative analysis of several different re-trained StyleGAN weightings is provided. The role of this tool in building large, scalable datasets of synthetic facial data is also discussed.
Funder
Science Foundation Ireland
FotoNation Ltd
Publisher
Institute of Electrical and Electronics Engineers (IEEE)
Publisher DOI
10.1109/ISSC49989.2020.9180189
Rights
Attribution-NonCommercial-NoDerivs 3.0 Ireland
cbn
Collections
Electrical and Electronic Engineering (Conference Papers)