LoopNet for fine-grained fashion attributes editing

• Published in: Expert systems with applications Vol. 259; p. 125182
• Main Authors: Zou, Xingxing, Zhu, Shumin, Wong, Wai Keung
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.01.2025
• Subjects: Attribute disentanglement; Fashion editing; GAN inversion; GAN inversion; Fashion editing; Attribute disentanglement
• ISSN: 0957-4174
• DOI: 10.1016/j.eswa.2024.125182

Generative Adversarial Networks (GANs) have revolutionized the field of image synthesis by transforming randomly sampled latent codes into high-fidelity synthesized images. However, current methods fall short in manipulating a wide range of fashion attributes due to semantic ambiguity and lack of disentanglement. This work focuses on fashion attribute editing by proposing an encoder-based GAN inversion method, namely LoopNet. To enable high-fidelity image inversion and fine-grained attribute editing, it refines edited images through two encoder–decoder stages, utilizing predefined directions from principal component analysis on latent codes and canny loss for detail enhancement. Experiments show LoopNet’s effectiveness in attribute disentanglement and manipulation, outperforming seven state-of-the-art image inversion methods. •A novel image inversion framework, LoopNet, designed for fashion attribute editing.•The use of semantic attribute directions to enhance model fidelity and editability.•The introduction of a Canny loss function to refine the details of inverted images.•Comprehensive experimental shows LoopNet’s ability in fashion attributes editing.