Title:
TRAINING METHOD AND APPARATUS FOR SEMANTIC SEGMENTATION NETWORK, DEVICE, AND STORAGE MEDIUM
Document Type and Number:
WIPO Patent Application WO/2023/035535
Kind Code:
A1
Abstract:
The present application relates to the field of neural networks and smart cities in artificial intelligence technologies, and specifically discloses a training method and apparatus for a semantic segmentation network, a device, and a storage medium. The method comprises: obtaining a sample image and the point-of-interest data of the sample image; aligning the point-of-interest data with the sample image to obtain an aligned sample image; determining a seed point in the sample image; generating a seed clue according to the aligned sample image and the seed point; and inputting the seed clue and the sample image into a convolutional network, training the convolutional network by using a seed loss function, and using a trained convolutional network as a semantic segmentation network. The present application can improve the recognition accuracy of the semantic segmentation network.
Inventors:
ZHENG XIMIN (CN)
CHEN ZHENHONG (CN)
SHU CHANG (CN)
CHEN YOUXIN (CN)
CHEN ZHENHONG (CN)
SHU CHANG (CN)
CHEN YOUXIN (CN)
Application Number:
PCT/CN2022/072182
Publication Date:
March 16, 2023
Filing Date:
January 14, 2022
Export Citation:
Assignee:
PING AN TECH SHENZHEN CO LTD (CN)
International Classes:
G06N3/04
Domestic Patent References:
| WO2021122446A1 | 2021-06-24 |
Foreign References:
| CN113780532A | 2021-12-10 | |||
| CN109063723A | 2018-12-21 | |||
| CN109993173A | 2019-07-09 | |||
| CN111259936A | 2020-06-09 |
Attorney, Agent or Firm:
SHENZHEN LIDAO INTELLECTUAL PROPERTY AGENCY (GENERAL PARTNERSHIP) (CN)
Download PDF: