Multi-label labeling method and device, storage medium, electronic equipment and program product

By constructing a pre-defined tag library and a two-stage collaborative annotation method, and using a multimodal model to determine the candidate tag set and content parsing results, the problem of low accuracy of generative models in multi-tag annotation is solved, and high-precision multi-tag annotation of multimedia objects is achieved.

CN122392066APending Publication Date: 2026-07-14BEIJING XIAOMI MOBILE SOFTWARE CO LTD +1
0 Cites 0 Cited by

Patent Information

Application Number
CN202610629077.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-05-08
Publication Date
2026-07-14

AI Technical Summary

Technical Problem

Existing generative models have the problem of strong single-label image classification ability but low multi-label labeling accuracy in the field of multi-label annotation, especially when the image contains multiple objects of different categories, they cannot accurately generate multiple labels.

Method used

A pre-defined tag library is constructed. A first multimodal model is used to determine the candidate tag set based on semantic similarity. A second multimodal model is used to parse the content and determine the multi-tag annotation results of multimedia objects based on the degree of association. A two-stage collaborative annotation method is adopted.

Benefits of technology

It improves the recall accuracy and precision of multi-tags for multimedia objects, enhances the semantic alignment accuracy between multimedia objects and tags in the preset tag library, and strengthens the accuracy of multi-tag annotation.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention provides a multi-label labeling method and device, a storage medium, electronic equipment and a program product. The method comprises the following steps: acquiring a multimedia object to be labeled and a preset label library; determining a candidate label set from the preset label library by using a first multi-modal model based on the semantic similarity between the multimedia object and each label in the preset label library; and performing content analysis on the multimedia object by using a second multi-modal model to obtain a content analysis result, and determining at least one target label from the candidate label set as a labeling result of the multimedia object based on the association degree of the result and each label in the candidate label set. The label recall rate of the multimedia object is improved by introducing a candidate mechanism, and the multi-label recall precision and accuracy of the multimedia object are improved through a two-stage collaborative labeling mode.
Need to check novelty before this filing date? Find Prior Art