Title:
MULTI-VIEW SEMANTIC SEGMENTATION METHOD AND APPARATUS, ELECTRONIC DEVICE, AND STORAGE MEDIUM
Document Type and Number:
WIPO Patent Application WO/2023/216654
Kind Code:
A1
Abstract:
Disclosed in embodiments of the present disclosure are a multi-view semantic segmentation method and apparatus, an electronic device, and a storage medium. The method comprises: determining first image data respectively corresponding to at least two first-type views to obtain at least two pieces of first image data; determining first semantic segmentation features under second-type views respectively corresponding to the at least two pieces of first image data to obtain at least two first semantic segmentation features; fusing the at least two first semantic segmentation features to obtain a fused semantic segmentation feature; and obtaining a fused semantic segmentation result on the basis of the fused semantic segmentation feature. The embodiments of the present disclosure can achieve an end-to-end multi-view semantic segmentation result by means of middle fusion only using a camera, a radar, etc., without performing post-processing, such that a processing time is effectively shortened, thereby reducing an auxiliary delay, and solving the problem in the prior art of a large delay caused by the need of transmission to a post-processing module for post-processing.
More Like This:
Inventors:
WANG MENGYUAN (CN)
ZHU HONGMEI (CN)
ZHANG QIAN (CN)
ZHU HONGMEI (CN)
ZHANG QIAN (CN)
Application Number:
PCT/CN2023/074402
Publication Date:
November 16, 2023
Filing Date:
February 03, 2023
Export Citation:
Assignee:
BEIJING HORIZON ROBOTICS TECH RES & DEVELOPMENT CO LTD (CN)
International Classes:
G06V20/56
Foreign References:
CN114821506A | 2022-07-29 | |||
CN110348351A | 2019-10-18 | |||
CN114187311A | 2022-03-15 | |||
CN108345887A | 2018-07-31 | |||
CN112733919A | 2021-04-30 | |||
US20210334556A1 | 2021-10-28 |
Attorney, Agent or Firm:
CHINABLE IP (CN)
Download PDF: