Title:
VIDEO RETRIEVAL METHOD AND APPARATUS, DEVICE, AND STORAGE MEDIUM
Document Type and Number:
WIPO Patent Application WO/2022/007827
Kind Code:
A1
Abstract:
A video retrieval method and apparatus, a device, and a storage medium. The method comprises: obtaining a comparison video clip from a video library according to the time length of a video to be tested (S110); determining the similarity between said video and the comparison video clip by means of a target spatial-temporal neural network, a spatial-temporal convolutional layer of the target spatial-temporal neural network being configured to be capable of respectively performing two-dimensional convolution and time dimension information processing (S120); and traversing the video library and outputting the retrieval result according to the similarity (S130). According to the method, in the process that the target spatial-temporal neural network determines the similarity between said video and the comparison video clip, the spatial-temporal convolutional layer of the target spatial-temporal neural network performs two-dimensional convolution and processes spatial features of said video and the comparison video clip, and then processes time dimension information of said video and the comparison video clip. By means of a mode in which the spatial features and the time dimension information are successively processed, the low-complexity space-time feature extraction is realized.
Inventors:
WU ZHENZHI (CN)
ZHU YAOLONG (CN)
ZHU YAOLONG (CN)
Application Number:
PCT/CN2021/104913
Publication Date:
January 13, 2022
Filing Date:
July 07, 2021
Export Citation:
Assignee:
LYNXI TECH CO LTD (CN)
International Classes:
G06F16/732; G06N3/04
Foreign References:
| CN111831852A | 2020-10-27 | |||
| CN111241338A | 2020-06-05 | |||
| CN111104555A | 2020-05-05 |
Other References:
LU, YAOYAO: "Research on Video Retrieval Technology Based on 3D Convolutional Neural NetworK", CHINESE MASTER’S THESES FULL-TEXT DATABASE, INFORMATION SCIENCE AND TECHNOLOGY, 1 January 2017 (2017-01-01), XP055887339, [retrieved on 20220203], DOI: 10.16565/j.cnki.1006-7744.2019.02.13
CAMUÑAS-MESA LUIS, LINARES-BARRANCO BERNABÉ, SERRANO-GOTARREDONA TERESA: "Neuromorphic Spiking Neural Networks and Their Memristor-CMOS Hardware Implementations", MATERIALS, vol. 12, no. 17, 27 August 2019 (2019-08-27), pages 2745, XP055887337, DOI: 10.3390/ma12172745
CAMUÑAS-MESA LUIS, LINARES-BARRANCO BERNABÉ, SERRANO-GOTARREDONA TERESA: "Neuromorphic Spiking Neural Networks and Their Memristor-CMOS Hardware Implementations", MATERIALS, vol. 12, no. 17, 27 August 2019 (2019-08-27), pages 2745, XP055887337, DOI: 10.3390/ma12172745
Attorney, Agent or Firm:
TEE&HOWE INTELLECTUAL PROPERTY ATTORNEYS (CN)
Download PDF: