Image recognition method and device based on semi-supervised relation measurement network

Abstract

The invention relates to an image recognition method and device based on a semi-supervised relation measurement network, and the method comprises the steps of: carrying out data expansion of all labeled data and unlabeled data in an image data set to obtain an expanded image data set; carrying out clustering operation on labeled data in the expanded image data set to obtain category atomic image data of each category; carrying out random noise addition processing on the expanded image data set, and inputting the image data set subjected to noise addition and category atomic image data into a semi-supervised relation measurement network model to obtain different category template comparison scores of labeled data and unlabeled data; calculating cross entropy loss and mean square error loss according to the comparison scores of different types of templates; performing training according to cross entropy loss and mean square error loss to obtain a trained semi-supervised relation measurement network model; and identifying a to-be-identified image through the trained semi-supervised relation measurement network model to determine the category of the to-be-identified image.

Description

technical field [0001] The invention relates to the technical field of image recognition, in particular to an image recognition method and device based on a semi-supervised relational metric network. Background technique [0002] With the rapid development of machine learning technology, many scholars at home and abroad have used machine learning methods to solve problems in various industries, and it has also shown great value in the field of image recognition and classification. Since the end of the last century, many scholars have conducted extensive and in-depth research on image classification methods. Many image classification methods have been developed in this field, including common wavelet-based, neural networks, Bayesian networks, association rules, decision trees, rough The classification technology of a single mode such as sets, and then combined various classifiers and distributed systems. Among them, the deep learning method performed very well. [0003] Alt...

AI Technical Summary

Problems solved by technology

[0006] In order to overcome the problems existing in the related technologies, the present invention provides an image recognition method and device based on a semi-supervised relational metric network, thereby effectively solving the problems of small data volume, unbalanced categories, and few labels in the image data set, and realizing Better classification performance in the field of image classification

Images