Title:
DEEP LEARNING MODEL GENERATION SYSTEM AND METHOD BASED ON HIERARCHICAL TRANSFER LEARNING FOR ENVIRONMENTAL INFORMATION RECOGNITION
Document Type and Number:
WIPO Patent Application WO/2024/043390
Kind Code:
A1
Abstract:
Provided is a deep learning model generation method based on hierarchical transfer learning for environmental information recognition. The method includes the steps of: primarily training a deep learning model on the basis of a non-target domain dataset; analyzing feature information for each dataset of a target domain and a similar target domain collected as having a similar purpose to the target domain; generating a class answer sheet for each dataset of the similar target domain on the basis of an analysis result of the feature information; performing secondary training with respect to the primarily trained deep learning model on the basis of the dataset of the similar target domain having the generated class answer sheet; and performing tertiary training with respect to the secondarily trained deep learning model on the basis of the dataset of the target domain.
More Like This:
Inventors:
SEO KYEONG EUN (KR)
YANG CHANG MO (KR)
KIM DONG CHIL (KR)
YANG CHANG MO (KR)
KIM DONG CHIL (KR)
Application Number:
PCT/KR2022/014926
Publication Date:
February 29, 2024
Filing Date:
October 04, 2022
Export Citation:
Assignee:
KOREA ELECTRONICS TECHNOLOGY (KR)
International Classes:
G06N3/08; G06N3/04
Foreign References:
| KR20210010505A | 2021-01-27 | |||
| KR20210136344A | 2021-11-17 | |||
| KR102261187B1 | 2021-06-07 | |||
| KR20200046173A | 2020-05-07 | |||
| US20190108640A1 | 2019-04-11 |
Attorney, Agent or Firm:
JIMYUNG PATENT FIRM (KR)
Download PDF: