Hash cross-modal information retrieval method based on dictionary pair learning

Abstract

The invention discloses a hash cross-modal information retrieval method based on dictionary pair learning. The hash cross-modal information retrieval method comprises the following steps of: (1) generating a training set; (2) constructing a hash objective function based on dictionary pair learning; (3) optimizing the hash objective function based on dictionary pair learning; (4) calculating a hashcode of a test set; and (5) obtaining a retrieval result. According to the invention, an objective function is constructed by using a dictionary pair thought, a coefficient embedding matrix with smaller modal difference is obtained by changing the data of different modals through a dictionary pair, so that the problem that the data of different modals is directly mapped to the Hamming space is solved, the problem of isomerism of the data of different modals can be better solved, and the method has higher average retrieval precision when solving the problem of cross-modal retrieval.

Description

technical field [0001] The invention relates to the field of computer technology, and further relates to a hash cross-modal information retrieval method based on dictionary pairs in the field of information retrieval technology. The present invention can be used in existing information retrieval applications, including data of multiple modalities such as text and images, and realizes fast retrieval between data of different modalities. Background technique [0002] Multimodal data is becoming more and more common in daily life, and the demand for cross-modal retrieval is also increasing. For example, articles on social media often use images or videos to supplement text descriptions, and information with pictures and texts has become more and more popular. Generally, people hope to retrieve related images or videos through text. These data usually have the characteristics of a large amount of data. In order to achieve fast retrieval of cross-modal data, many studies now foc...

AI Technical Summary

Problems solved by technology

This method directly maps data of different modalities to Hamming space, and introduces class label constraints and discrete constraints when learning hash codes. Although this method has good retrieval efficiency, there are still shortcomings in this method. The disadvantage is that this method directly maps data of different modalities to Hamming space, ignoring the heterogeneity of data of different modalities. Specifically, data of different modalities has different representation methods and different data dimensions. It is difficult to obtain a consistent hash code when it is directly mapped to the Hamming space, thus reducing the retrieval accuracy

However, the shortcomings of this method are: 1. This method directly maps data of different modalities to Hamming space, ignoring the heterogeneity of data of different modalities, and it is difficult to obtain a consistent hash code when calculating the hash code. 2. Although the introduction of graph constraints can maintain the similarity of the original data in the hash code, thereby improving the retrieval accuracy, it also greatly increases the computational complexity, which is not conducive to large-scale Applications of scale data

Images