Title:
IMAGE CHARACTER STRUCTURED OUTPUT METHOD AND APPARATUS, AND ELECTRONIC DEVICE AND STORAGE MEDIUM
Document Type and Number:
WIPO Patent Application WO/2023/246912
Kind Code:
A1
Abstract:
Disclosed in the present invention are an image character structured output method and apparatus, and an electronic device and a storage medium. The method comprises: performing character row detection on an image to be subjected to recognition, so as to obtain character row data; performing a semantic feature encoding/decoding operation on the character row data, so as to obtain a character row semantic feature, and recognizing character row text information on the basis of the character row semantic feature; performing a feature enhancement operation and an image feature extraction operation on the character row data, so as to obtain a character row image feature; fusing the character row semantic feature and the character row image feature, and performing feature adjacency matrix learning on a fused feature, so as to obtain a feature adjacency matrix; performing structural feature extraction on the character row data and the feature adjacency matrix, so as to recognize character row structural information; and outputting structured information on the basis of the character row text information and the character row structural information. In this way, by means of the present invention, an image feature and a semantic feature of a character row are enhanced, and the efficiency and performance of character structuring are improved.
Inventors:
CHENG LONG (CN)
WU QINLONG (CN)
WU QINLONG (CN)
Application Number:
PCT/CN2023/101885
Publication Date:
December 28, 2023
Filing Date:
June 21, 2023
Export Citation:
Assignee:
CHINA MOBILE SUZHOU SOFTWARE TECH CO LTD (CN)
CHINA MOBILE COMMUNICATIONS GROUP CO LTD (CN)
CHINA MOBILE COMMUNICATIONS GROUP CO LTD (CN)
International Classes:
G06F40/126; G06F40/30; G06N3/0442; G06N3/0464; G06N3/092; G06V30/148; G06V30/18; G06V30/19
Foreign References:
| CN116266259A | 2023-06-20 | |||
| CN112001368A | 2020-11-27 | |||
| CN113297975A | 2021-08-24 | |||
| CN110569846A | 2019-12-13 |
Attorney, Agent or Firm:
CHINA PAT INTELLECTUAL PROPERTY OFFICE (CN)
Download PDF:
Previous Patent: T CELL RECEPTOR-BASED BISPECIFIC POLYPEPTIDE MOLECULE AND USE THEREOF
Next Patent: POWER STATE DETERMINATION METHOD AND APPARATUS
Next Patent: POWER STATE DETERMINATION METHOD AND APPARATUS