Title:
TARGET SEGMENTATION METHOD AND APPARATUS, AND ELECTRONIC DEVICE
Document Type and Number:
WIPO Patent Application WO/2023/273173
Kind Code:
A1
Abstract:
A target segmentation method and apparatus, and an electronic device, which relate to the field of artificial intelligence, in particular, to computer vision and deep learning technology, and can be specifically applied to smart city and smart transportation scenarios. The specific implementation solution involves: according to a video to be subjected to recognition, generating a frame to be subjected to recognition, the previous frame of said frame, and a reference frame; inputting same into an encoding network, and generating a feature map of the frame to be subjected to recognition, a target object reference-frame feature map and a target object previous-frame feature map; generating a first correlation matrix and a second correlation matrix; generating a first correlation feature map and a second correlation feature map; and generating a current-frame target segmented image according to the feature map of the frame to be subjected to recognition. By means of the embodiments of the present disclosure, a target in a frame to be subjected to recognition can be detected. By means of the embodiments of the present disclosure, a correlation matrix of a feature map of a frame to be subjected to recognition is acquired according to reference-frame and previous-frame feature maps that only include a target object, such that attention is focused on the target object, thereby improving the accuracy of recognizing the target object.
Inventors:
HAN WENHUA (CN)
Application Number:
PCT/CN2021/136548
Publication Date:
January 05, 2023
Filing Date:
December 08, 2021
Export Citation:
Assignee:
BEIJING BAIDU NETCOM SCI & TECH CO LTD (CN)
International Classes:
G06T7/10
Foreign References:
| CN113570606A | 2021-10-29 | |||
| CN112950640A | 2021-06-11 | |||
| CN112070044A | 2020-12-11 | |||
| CN109685805A | 2019-04-26 |
Other References:
WANG ZIQIN; XU JUN; LIU LI; ZHU FAN; SHAO LING: "RANet: Ranking Attention Network for Fast Video Object Segmentation", 2019 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV), IEEE, 27 October 2019 (2019-10-27), pages 3977 - 3986, XP033723818, DOI: 10.1109/ICCV.2019.00408
Attorney, Agent or Firm:
TSINGYIHUA INTELLECTUAL PROPERTY LLC (CN)
Download PDF: