Skip to main navigation Skip to search Skip to main content

Coarse-to-Fine Attribute Editing for Fashion Images

Research output: Chapter in book / Conference proceedingConference article published in proceeding or bookAcademic researchpeer-review

Abstract

With the development of Generative Adversarial Networks, attribute editing has been more and more popular in computer vision. The previous works employed the multi-domain image-to-image translation framework to solve attribute editing. However, it is difficult to generate consistent texture with the original images. Meanwhile, the un-target attribute regions are changed by using these methods. To address these problems, this paper presents a coarse-to-fine attribute editing scheme (CFAE) for fashion images. CFAE is composed of coarse stage and refine stage. In the coarse stage, we design a landmark-based attention scheme in conjunction with StarGAN [5] to locate and edit the target attribute regions. Subsequently, DeepFillv2 [16] is employed in the refine stage to make the target attribute regions consistent with the original image. In experiment section, we compare our method with several state-of-the-art methods on OUTFIT Dataset and the results demonstrate the effectiveness of CFAE.

Original languageEnglish
Title of host publicationArtificial Intelligence - 1st CAAI International Conference, CICAI 2021, Proceedings
EditorsLu Fang, Yiran Chen, Guangtao Zhai, Jane Wang, Ruiping Wang, Weisheng Dong
PublisherSpringer Science and Business Media Deutschland GmbH
Pages396-407
Number of pages12
ISBN (Print)9783030930455
DOIs
Publication statusPublished - 2021
Event1st CAAI International Conference on Artificial Intelligence, CICAI 2021 - Hangzhou, China
Duration: 5 Jun 20216 Jun 2021

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume13069 LNAI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference1st CAAI International Conference on Artificial Intelligence, CICAI 2021
Country/TerritoryChina
CityHangzhou
Period5/06/216/06/21

Keywords

  • Attribute editing
  • Deep learning
  • Fashion
  • GAN

ASJC Scopus subject areas

  • Theoretical Computer Science
  • General Computer Science

Fingerprint

Dive into the research topics of 'Coarse-to-Fine Attribute Editing for Fashion Images'. Together they form a unique fingerprint.

Cite this