GRB-Sty: Redesign of Generative Residual Block for StyleGAN

We have previously published a paper introducing a novel module, the Generative Residual Block (GRB), which successfully enhances GAN performance. However, the experiments in the earlier paper were conducted on baseline models using spectral normalization, a technique seldom used today. To address t...

Full description

Saved in:
Bibliographic Details
Main Authors: Seung Park, Yong-Goo Shin
Format: Article
Language:English
Published: Elsevier 2025-04-01
Series:ICT Express
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S2405959525000232
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:We have previously published a paper introducing a novel module, the Generative Residual Block (GRB), which successfully enhances GAN performance. However, the experiments in the earlier paper were conducted on baseline models using spectral normalization, a technique seldom used today. To address this problem, we investigate the effectiveness of GRB on contemporary StyleGAN-based models. This paper introduces an enhanced version of GRB, termed GRB-Sty, which consistently boosts the performance of StyleGAN-based models and demonstrates versatility across various aspects. The significant performance enhancements observed in extensive experiments on multiple benchmark datasets highlight the compatibility of GRB-Sty with state-of-the-art methods.
ISSN:2405-9595