US 11,816,891 B2
Video recognition method and apparatus, electronic device and storage medium
Wenhao Wu, Beijing (CN); and Yuxiang Zhao, Beijing (CN)
Assigned to BEIJING BAIDU NETCOM SCIENCE AND TECHNOLOGY CO., LTD., Beijing (CN)
Filed by BEIJING BAIDU NETCOM SCIENCE AND TECHNOLOGY CO., LTD., Beijing (CN)
Filed on Aug. 25, 2021, as Appl. No. 17/411,254.
Claims priority of application No. 202110081599.6 (CN), filed on Jan. 21, 2021.
Prior Publication US 2021/0383128 A1, Dec. 9, 2021
Int. Cl. G06V 20/40 (2022.01)
CPC G06V 20/41 (2022.01) [G06V 20/46 (2022.01); G06V 20/49 (2022.01)] 9 Claims
OG exemplary drawing
 
1. A video recognition method, comprising:
dividing a to-be-recognized video into at least two video segments;
extracting video frames from the at least two video segments, and performing feature recognition on the video frames to obtain initial semantic feature blocks of the at least two video segments;
fusing each of the initial semantic feature blocks, and obtaining a fused target semantic feature block; and
determining, according to the fused target semantic feature block, a type of the to-be-recognized video;
wherein the fusing the obtained initial semantic feature blocks, and obtaining the fused target semantic feature block comprises: fusing, according to a dynamic movement strategy, feature information of different channels of the each of the initial semantic feature blocks, and obtaining the fused target semantic feature block;
wherein the fusing, according to the dynamic movement strategy, the feature information of the different channels of the each of the initial semantic feature blocks comprises:
moving feature information of a first channel of a current initial semantic feature block to an initial semantic feature block that is ranked before the current initial semantic feature block in time sequence and adjacent to the current initial semantic feature block; and
moving feature information of a second channel of the current initial semantic feature block to an initial semantic feature block that is ranked after the current initial semantic feature block in time sequence and adjacent to the current initial semantic feature block.