Title:
METHOD FOR TRAINING AND TESTING DATA EMBEDDING NETWORK TO GENERATE MARKED DATA BY INTEGRATING ORIGINAL DATA WITH MARK DATA, AND TRAINING DEVICE AND TESTING DEVICE USING THE SAME
Document Type and Number:
WIPO Patent Application WO/2020/032420
Kind Code:
A1
Abstract:
A method for learning a data embedding network is provided. The method includes steps of: a learning device acquiring and inputting original training data and mark training data into the data embedding network which integrates them and generates marked training data; inputting the marked training data into a learning network which applies a network operation to them and generates 1-st characteristic information, and inputting the original training data into the learning network which applies a network operation to them and generates 2-nd characteristic information; learning the data embedding network such that a data error is minimized, by referring to part of errors referring to the 1-st and the 2-nd characteristic information and errors referring to task specific outputs and their ground truths, and a marked data score is maximized, and learning a discriminator such that a original data score is maximized and the marked data score is minimized.
More Like This:
Inventors:
KIM TAE HOON (KR)
Application Number:
PCT/KR2019/008944
Publication Date:
February 13, 2020
Filing Date:
July 19, 2019
Export Citation:
Assignee:
DEEPING SOURCE INC (KR)
International Classes:
G06F21/16; G06F21/30; G06N3/08
Domestic Patent References:
| WO2018131405A1 | 2018-07-19 |
Foreign References:
| US20030095683A1 | 2003-05-22 | |||
| US20060269098A1 | 2006-11-30 | |||
| KR20170092631A | 2017-08-11 | |||
| KR20180058116A | 2018-05-31 | |||
| KR101837939B1 | 2018-03-13 |
Other References:
A. A. SIROTA ET AL.: "Neural Network Functional Models and Algorithms for Information Conversion in Order to Create Digital Watermarks", RADIOELECTRONICS AND COMMUNICATIONS SYSTEMS, vol. 58, no. 1, January 2015 (2015-01-01), pages 1 - 10, XP035441176, ISSN: 0735-2727, Retrieved from the Internet [retrieved on 20191017], DOI: 10.3103/S073527271501001X
YUSUKE UCHIDA ET AL.: "Embedding Watermarks into Deep Neural Networks", ARXIV:1701.04082V2, 20 April 2017 (2017-04-20), pages 1 - 10, XP081275035, Retrieved from the Internet [retrieved on 20191017]
JIREN ZHU ET AL.: "HiDDeN: Hiding Data With Deep Networks", 26 July 2018, CORNELL UNIVERSITY LIBRARY
BAZRAFKAN SHABAB ET AL., VERSATILE AUXILIARY CLASSIFIER WITH GENERATIVE ADVERSARIAL NETWORK (VAC+GAN), MULTI CLASS SCENARIOS TRAINING CONDITIONAL GENERATORS, 19 June 2018 (2018-06-19)
LUO CHUNJIE ET AL.: "Cosine Normalization: Using Cosine Similarity Instead of Dot Product in Neural Networks", 20 February 2017, CORNELL UNIVERSITY LIBRARY
YUSUKE UCHIDA ET AL.: "Embedding Watermarks into Deep Neural Networks", ARXIV:1701.04082V2, 20 April 2017 (2017-04-20), pages 1 - 10, XP081275035, Retrieved from the Internet
JIREN ZHU ET AL.: "HiDDeN: Hiding Data With Deep Networks", 26 July 2018, CORNELL UNIVERSITY LIBRARY
BAZRAFKAN SHABAB ET AL., VERSATILE AUXILIARY CLASSIFIER WITH GENERATIVE ADVERSARIAL NETWORK (VAC+GAN), MULTI CLASS SCENARIOS TRAINING CONDITIONAL GENERATORS, 19 June 2018 (2018-06-19)
LUO CHUNJIE ET AL.: "Cosine Normalization: Using Cosine Similarity Instead of Dot Product in Neural Networks", 20 February 2017, CORNELL UNIVERSITY LIBRARY
Attorney, Agent or Firm:
SU INTELLECTUAL PROPERTY (KR)
Download PDF: