Garment Style Creator: Using StarGAN for Image-to-Image Translation of Multidomain Garments

On the basis of StarGAN, this study developed a garment style-generation system named Garment Style Creator. This system first requires the user to import a human image to produce distinct parts, such as head, arms, and garment by applying the human parsing technique. Then, this system separates the...

Full description

Saved in:
Bibliographic Details
Published in:IEEE multimedia 2022-01, Vol.29 (1), p.85-93
Main Authors: Chou, Chien-Hsing, Han, Ping-Hsuan, Chang, Chia-Chun, Hsieh, Yi-Zeng
Format: Magazinearticle
Language:English
Subjects:
Garment
Garments
Generators
Human Parsing
Image processing
Solid modeling
StarGAN
Style Transfer
Textile composites
Three-dimensional displays
Training data
Virtual Try-On
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:On the basis of StarGAN, this study developed a garment style-generation system named Garment Style Creator. This system first requires the user to import a human image to produce distinct parts, such as head, arms, and garment by applying the human parsing technique. Then, this system separates the training images that contain only the garment part and removes the other semantic parts as background. Subsequently, images of the garment are imported into StarGAN to produce images of multiple garment styles and concurrently generate garment images of the human model wearing plain, lattice, stripe, and polka-dotted fabrics. Finally, we designed a style-mixing system that allows users to blend multiple garment images into one image. The system facilitates the generation of images of numerous garment styles simultaneously and thus assists designers and users in finding the ideal garment style for them.
ISSN:1070-986X
1941-0166
DOI:10.1109/MMUL.2021.3139760