ETRI-Knowledge Sharing Plaform

ENGLISH

성과물

논문 검색
구분 SCI
연도 ~ 키워드

상세정보

학술대회 Generating and Modifying High Resolution Fashion Model Image using StyleGAN
Cited 2 time in scopus Download 7 time Share share facebook twitter linkedin kakaostory
저자
최인문, 박순찬, 박지영
발행일
202210
출처
International Conference on Information and Communication Technology Convergence (ICTC) 2022, pp.1536-1538
DOI
https://dx.doi.org/10.1109/ICTC55196.2022.9952574
협약과제
22IH1900, 5G를 활용하는 차세대 1인 콘텐츠 기반의 문화상품 커뮤니케이션 마켓 플랫폼 개발, 박지영
초록
In this paper, a research of synthesizing fashion model images by utilizing a state-of-the-art generative adversarial network (i.e., GAN) is introduced. After training GAN with fashion model images, the network was able to generate realistic fashion model images having various characteristics such as pose and clothes. Moreover, two image modifications named Fashion Model Morphing and Fashion Transfer are also proposed by merging attributes of two generated fashion model images. The research investigates the effectiveness of using GAN for fashion to create a large number of images for exploring new design and styles. The generated images are even more beneficial for fashion industries because the generated images have no legal issues such as portrait right and copyright.
KSP 제안 키워드
High-resolution, generative adversarial network, legal issues, state-of-The-Art