Video retrieval method and device, electronic equipment and computer readable storage medium

Abstract

The invention provides a video retrieval method and device, electronic equipment and a computer readable storage medium, and relates to the technical field of data processing. The video retrieval method comprises the steps of when video retrieval is carried out, firstly obtaining retrieval content, wherein the type of the retrieval content comprises voice or text; converting the retrieval content into a query vector according to a pre-trained text feature encoder; determining a target video matched with the retrieval content according to the query vector and a visual feature vector of each video in a retrieval library; wherein the visual feature vector is obtained by performing feature extraction on a corresponding video in a retrieval library through a self-supervised pre-trained visual encoder. In this way, the user can quickly and accurately retrieve the matched video through voice or text description based on content guessing or memorizing of the video to be searched, and compared with an existing video retrieval mode based on manual text labeling, the accuracy and the working efficiency are improved, the cost is reduced, and the user experience is enhanced.

Description

technical field [0001] The present invention relates to the technical field of data processing, in particular to a video retrieval method, device, electronic equipment and computer-readable storage medium. Background technique [0002] In recent years, the domestic video industry has shown explosive growth, and video is gradually becoming an important way for people to entertain, learn and socialize. Education and training videos account for a considerable proportion of online education and video sharing platforms. For the platform side, how to quickly and accurately retrieve relevant videos from massive amounts of video data in a more convenient way according to users’ learning intentions And recommending it to users has become one of the important indicators for content monetization. [0003] In the traditional video retrieval method, artificial text annotation is performed on the video frame image, and the video data is described according to the content of the video fra...

AI Technical Summary

Problems solved by technology

However, this retrieval method has the following disadvantages: first, the annotation is mainly processed manually, so it is greatly affected by the subjective factors of the annotator, which may cause different descriptions of the same video; second, the text description is a A fixed abstraction of video scene content, so specific video tags are only suitable for specific retrieval; third, the amount of video data is large, and adding annotations manually is a lot of work, especially for today's growing video volume. High cost and low productivity

[0004] In summary, the existing video retrieval methods have the problems of low accuracy, high cost and low work efficiency, and the user experience is poor

Images