Title:
SYSTEM AND METHOD FOR TRAINING HAIR IMAGE REMOVAL NEURAL NETWORK
Document Type and Number:
WIPO Patent Application WO/2023/163273
Kind Code:
A1
Abstract:
A system and a method for training a neural network for removing a hair image included in a skin image are disclosed. In a training system, according to one embodiment, for removing a hair image included in a skin image through unsupervised learning based on a generative adversarial neural network including a hair removal neural network and an authenticity determination neural network implemented through a synthesis neural network, the hair removal neural network is configured to receive a skin image including hair so as to generate a hair-removed image from which a hair image has been removed, and the authenticity determination neural network can be configured to determine, on the basis of the hair-removed image and a comparative image, whether the hair-removed image is a fake image generated by means of the hair removal neural network.
Inventors:
YOU WON SANG (KR)
LEE YOUNG CHAN (KR)
PARK BUM JIN (KR)
LEE YOUNG CHAN (KR)
PARK BUM JIN (KR)
Application Number:
PCT/KR2022/004666
Publication Date:
August 31, 2023
Filing Date:
March 31, 2022
Export Citation:
Assignee:
ONDESIGN INC (KR)
INDUSTRY UNIV COOPERATION FOUNDATION SUN MOON UNIV (KR)
INDUSTRY UNIV COOPERATION FOUNDATION SUN MOON UNIV (KR)
International Classes:
G06T5/00; G06N20/00; G06T7/00
Foreign References:
| KR102240885B1 | 2021-04-14 | |||
| KR20210089166A | 2021-07-15 | |||
| JP2021165962A | 2021-10-14 | |||
| KR102288645B1 | 2021-08-10 | |||
| US20220005227A1 | 2022-01-06 |
Attorney, Agent or Firm:
KIM, Dae Young et al. (KR)
Download PDF: