METHOD AND DEVICE FOR GENERATING TRAINING DATA AND COMPUTER PROGRAM STORED IN COMPUTER-READABLE RECORDING MEDIUM

WIPO Patent Application WO/2019/039757

A1

Disclosed is a method for generating flawed data of a target domain by using flawed data of a source domain according to an embodiment of the present disclosure for implementing such an objective as described above. The method for generating flawed data of a target domain by using flawed data of a source domain comprises the steps of: inputting flawed data of a source domain, to which a first mask has been applied, and flawed data of the source domain, to which the first mask has not been applied, to a restoration algorithm; firstly training the restoration algorithm so as to generate flawed data of the source domain from which the first mask has been restored; inputting normal data of the source domain, to which a second mask has been applied, and normal data of the source domain, to which the second mask has not been applied, to the restoration algorithm; secondly training the restoration algorithm so as to generate normal data of the source domain from which the second mask has been restored; inputting normal data of a target domain, to which the second mask has been applied, and normal data of the target domain, to which the second mask has not been applied, to the restoration algorithm; and thirdly training the restoration algorithm so as to generate normal data of the target domain from which the second mask has been restored.

KI HONGDO (KR)
KIM HYUNJUN (KR)

PCT/KR2018/008619

February 28, 2019

July 30, 2018

Click for automatic bibliography generation  

SUALAB CO LTD (KR)

G06T5/00; G06V10/764; G06V10/774

KR101688458B1	2016-12-23
JP2005156334A	2005-06-16
JPH1114553A	1999-01-22
JP2001202265A	2001-07-27
JP2008139074A	2008-06-19

LEE, Dae Ho et al. (KR)

View/Download PDF      PDF Help

Previous Patent: PUMPING TYPE COSMETIC CONTAINER HAVING DIFFUSING PLATE

Next Patent: METHOD FOR GENERATING AND LEARNING IMPROVED NEURAL NETWORK