Small sample image classification method and system based on semantic perception graph neural network

Abstract

The invention discloses a small sample image classification method and system based on a semantic perception graph neural network, and aims to solve the problem of category semantic fuzziness caused by a picture-level label. The method comprises the steps of: firstly, refining the expression of edges; refining an edge from a scalar representing the global similarity of adjacent nodes to a vector representing the similarity of each pixel position of the adjacent nodes; and then, on the basis of edge refining, further introducing a semantic calibration module to update edge features, calculating a relation matrix between adjacent nodes, and converting the relation matrix into an edge value of each pixel position to enable the edge to explicitly express semantic similarity between the nodes, and therefore, the final classification result is improved through updating and spreading semantic information of the multi-layer graph neural network. Experiments on a miniImageNet data set show that the method can improve the precision of small sample image classification.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and in particular relates to a small-sample image classification method and system based on a semantic perception graph neural network. Background technique [0002] In recent years, thanks to the rapid improvement of computer computing power, deep learning has become a hot topic in the field of artificial intelligence research. However, behind the prosperity and development of deep learning is the support of large-scale manually labeled data sets, and the more complex the network, the more large data sets are needed for training. However, data in some special fields is very scarce, such as the discrimination of rare cases in medicine, and the existing limited medical images are far from enough to train a good deep model. Not only that, the existing deep neural network models often have poor generalization. For example, a good cat and dog classifier is trained with a large number of cat ...

AI Technical Summary

Problems solved by technology

[0005] The technical problem to be solved by the present invention is to provide a small-sample image classification method and system based on a semantic-aware graph neural network in order to solve the problems caused by image-level labels by constructing a semantic-aware graph neural network. fuzzy question

Images