Title:
SEMANTIC SEGMENTATION MODEL TRAINING METHOD AND APPARATUS, AND ELECTRONIC DEVICE AND STORAGE MEDIUM
Document Type and Number:
WIPO Patent Application WO/2024/012251
Kind Code:
A1
Abstract:
Provided in the embodiments of the present disclosure are a semantic segmentation model training method and apparatus, and an electronic device and a storage medium. The method comprises: acquiring a sample image, and extracting, by means of a semantic segmentation model to be trained, a visual image feature corresponding to the sample image; processing the sample image to obtain a text image feature corresponding to the sample image, wherein the text image feature is an image feature which is generated by language description text for the sample image; fusing the visual image feature with the text image feature, so as to obtain a multi-modal feature, and performing image segmentation prediction on the basis of the multi-modal feature, so as to obtain a target loss; and training said semantic segmentation model on the basis of the target loss, so as to obtain a target semantic segmentation model.
Inventors:
QIN JIE (CN)
WU JIE (CN)
XIAO XUEFENG (CN)
WU JIE (CN)
XIAO XUEFENG (CN)
Application Number:
PCT/CN2023/104527
Publication Date:
January 18, 2024
Filing Date:
June 30, 2023
Export Citation:
Assignee:
BEIJING ZITIAO NETWORK TECHNOLOGY CO LTD (CN)
International Classes:
G06V10/26; G06T7/11; G06V10/40; G06V10/80
Domestic Patent References:
| WO2020079704A1 | 2020-04-23 |
Foreign References:
| CN113657400A | 2021-11-16 | |||
| CN112990218A | 2021-06-18 | |||
| CN114723996A | 2022-07-08 | |||
| CN112184738A | 2021-01-05 | |||
| CN114283127A | 2022-04-05 | |||
| CN110245710A | 2019-09-17 | |||
| US20210049397A1 | 2021-02-18 |
Attorney, Agent or Firm:
LINKER IP LLC et al. (CN)
Download PDF: