Multi-label image classification method, device and equipment based on image convolution

Abstract

The invention provides a multi-label image classification method, a multi-label image classification device and the multi-label image classification equipment based on image convolution, and relates to the technical field of intelligent identification. The method comprises the steps of obtaining a to-be-classified image; performing feature extraction on the to-be-classified image to obtain the image feature information; inputting the image feature information into a pre-trained multi-label classifier, wherein the multi-label classifier is a classifier which is obtained by performing multi-label relation modeling through a graph convolutional network and performing learning and comprises multi-label relation information; and determining a label corresponding to the to-be-classified image according to the at least one label score outputted by the multi-label classifier. According to the multi-label image classification method, device and equipment provided by the embodiment of the invention, the image classification precision can be improved.

Description

technical field [0001] The present invention relates to the field of intelligent recognition technology, in particular to a multi-label image classification method, device and equipment based on graph convolution. Background technique [0002] Since images always contain multiple labels in natural scenes, multi-label image classification is more practical than single-label. The purpose of multi-label image classification is to classify images in All objects are predicted. Since the image contains multiple labels, the number of classification results is exponentially higher than that of a single label. Compared with the single-label image classification problem, the multi-label image classification problem is more difficult and the accuracy is lower. [0003] Existing methods include using graphs to model the relationship between labels, so as to artificially impose constraints on the final predicted results in order to reduce the number of classification results. Because t...

AI Technical Summary

Problems solved by technology

At present, the use of deep learning to solve multi-label classification problems is mainly divided into two categories: 1. The method based on the graph model mainly uses the Recurrent Neural Network (Recurrent Neural Network, RNN) to model the graph, which is very dependent on the order of the input labels, and due to RNN length limitation leads to inflexibility; 2. Based on the attention mechanism, only the relationship between local image tags is modeled instead of the global relationship

Images