Label identification method, label identification model training method, device and equipment

Abstract

The invention provides a label identification method, a label identification model training method, devices and equipment, and relates to the field of video processing of network media, and the method comprises the steps of extracting a plurality of modal features of a target image or video; carrying out feature fusion on the plurality of modal features to obtain a fused feature; based on the fusion feature and the intermediate feature of the (i-1) th layer of classifier in the M layers of classifiers, obtaining the intermediate feature of the ith layer of classifier until the intermediate feature of the Mth layer of classifier is obtained; based on the middle features of the Mth-layer classifier, utilizing the Mth-layer classifier to output probability distribution features; and determining a label of the target image or video based on the probability distribution feature. According to the method provided by the invention, the features of the multiple modes are extracted and fused, and the fused features are subjected to feature multiplexing of different levels by using the M-layer classifier, so that the accuracy of the classifier is improved, that is, the accuracy of the probability distribution features output by the Mth-layer classifier is improved, and that is, the accuracy of tag identification is improved.

Description

technical field [0001] The embodiment of the present application relates to the field of video processing of network media, and more specifically, relates to a method for identifying tags, a method, device and equipment for training a tag identification model. Background technique [0002] With the arrival of the fifth-generation mobile communication technology (5-Generation, 5G) and the development of mobile Internet platforms, more and more videos have been accumulated on Internet platforms, and the consumption of short videos and images has shown a blowout explosion, which has led to Intelligent understanding of image or video content becomes essential in all aspects of visual content. The most basic intelligent image understanding task is to label images or videos with accurate and rich labels, so as to ensure that users or downstream tasks can quickly find images or videos, and improve the quality and efficiency of retrieval. [0003] So far, the methods for identifyin...

AI Technical Summary

Problems solved by technology

However, expressing the features of images or videos only through visual features will lead to insufficient feature expression, which in turn will lead to low recognition accuracy of labels; , which cannot meet the recognition accuracy requirements for quickly searching images or video scenes

Images