Skip to main navigation Skip to search Skip to main content

M-VTON: Multi-layer Virtual Try-on System

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

Abstract

Fashion recommendation is an issue of considerable importance in the fashion ecommerce industry. The essential part of the recommendation task is how to represent outfits. We propose a multi-layer try-on system (M-VTON) based on deep learning methods. M-VTON can generates vivid try-on images to represent an outfit using its separate product images. It includes a fashion keypoint detection model and a semantic segmentation model. The detection model aims to estimate the keypoints of garments in order to calculate an item’s scale and position. The segmentation model serves to separate garments into front and back pieces, which are used to generate the multi-layers of outfits. We conducted experiments on two mainstream fashion datasets and results support the effectiveness of our proposed approach.

Original languageEnglish
Title of host publicationInternational Conference on Design and Semantics of Form and Movement
EditorsMiguel Bruns, Lin-Lin Chen, Jun Hu, Sara Colombo, Yihyun Lim, Steven Kyffin, Ozcan Vieira, E. Jeroen Raijmakers, Lucia Rampino, Edgar Rodriguez Ramirez, Dagmar Johanna Steffen, Calvin Wong
PublisherThe Hong Kong Polytechnic University
Pages94-104
Number of pages11
ISBN (Print)9789623678704
Publication statusPublished - 2023
Event12th International Conference on Design and Semantics of Form and Movement, DeSForM 2023 - Hong Kong, Hong Kong
Duration: 5 Jul 20237 Jul 2023

Publication series

NameInternational Conference on Design and Semantics of Form and Movement
ISSN (Electronic)2706-6150

Conference

Conference12th International Conference on Design and Semantics of Form and Movement, DeSForM 2023
Country/TerritoryHong Kong
CityHong Kong
Period5/07/237/07/23

Keywords

  • fashion compatibility learning
  • Fashion recommendation
  • keypoint detection
  • semantic segmentation
  • virtual try-on

ASJC Scopus subject areas

  • Arts and Humanities (miscellaneous)
  • Human-Computer Interaction
  • Mechanics of Materials

Fingerprint

Dive into the research topics of 'M-VTON: Multi-layer Virtual Try-on System'. Together they form a unique fingerprint.

Cite this