Zero sample image classification method and system

Abstract

The invention discloses a zero sample image classification method and system. The method includes the following steps of: S10, inputting training data belonging to a visible category and a category tag thereof for feature extraction; S20, inputting the semantic assistance information of all category tags to obtain the semantic embedded representations of respective tags and semantic difference measure between the tags; S30, establishing a zero sample classification model, performing semantic consistency regularization on the model based on the semantic differences between the categories; S40,iteratively updating model parameters until convergence; and S50, predicting the category tag of a test image. The method and system establish a semantic consistency regularized zero sample classification model so that the output of the model conforms to the semantic neighborhood relation between categories to adapt to the semantic structure of the target category so as to obtain high classification accuracy.

Description

technical field [0001] The invention relates to the field of image classification and zero-sample classification learning, in particular to a zero-sample image classification method and system based on category-based semantic consistency regularization. Background technique [0002] With the development and popularization of the Internet and shooting equipment, multimedia data such as images on the Internet are generated rapidly. Due to the high labeling cost of image data and the emergence of more image datasets with fine-grained categories, training data in some emerging domains is usually difficult to obtain. On the other hand, even in the same domain, training data and test data may have related but different feature and semantic distributions, which do not satisfy the traditional assumption of same distribution. Therefore, in order to effectively predict the newly generated target data, how to use labeled training data from related fields to learn knowledge that can ad...

AI Technical Summary

Problems solved by technology

Due to the high labeling cost of image data and the emergence of more image datasets with fine-grained categories, it is usually difficult to obtain training data in some emerging domains.

On the other hand, even in the same domain, training data and test data may have related but different feature and semantic distributions, which do not satisfy the traditional assumption of the same distribution

Images