A Cross-modal Hash Retrieval Method Guided by Class Semantics

Abstract

A cross -modal hash retrieval method based on semantic guidance belongs to the field of computer technology, including: 1) projection learning guided by semantic semantics; 2) hash code learning; 3) similarity between modes and modes between modalsKeep; 4) the construction and optimization of the total target function.The present invention mainly ignores the problem of class semantics in the supervision method. Considering the semantic correlation between the semantic semantic association through the semantic embedded space, and the semantic semantics as the middle layer, the encoder ‑ decoder paradigm is used for projection learning, further furtherGround generates a hash function with a hash code and a specific modal.In addition, because the semantic semantics has established visible and invisible classes, it also solves the problem of retrieval of the invisible domains at the same time.Experiments show that the present invention effectively capture the semantic correlation between categories, improves the quality of the hash code and the performance of cross -modular retrieval, and has the ability to deal with cross -modal retrieval tasks and zero sample cross -modal retrieval tasks.

Description

technical field [0001] The invention belongs to the technical field of computers and relates to a cross-modal hash retrieval method based on class semantic guidance. Background technique [0002] With the advent of the era of big data, multimedia data on the Internet, such as images, text, audio, etc., is growing exponentially. The diversification and multi-dimensional characteristics of these multimedia data make people's retrieval needs change from traditional single-modal data retrieval, such as image retrieval, text retrieval, etc., to mutual retrieval between multi-modal data, especially cross-modal data. search. Cross-modal retrieval is to use samples from one modality to obtain related results from another modality, such as using an image to retrieve text or video related to it. In recent years, many researchers have made unremitting efforts and produced a large number of research results. However, in large-scale retrieval tasks, cross-modal retrieval methods suffe...

AI Technical Summary

Problems solved by technology

Such a model not only enhances the discriminativeness of learned hash codes, but also enables the present invention to solve the retrieval problem of invisible domains

Images