Big data cross-modal retrieval method and system based on deep integration Hash

Abstract

The invention provides a big data cross-modal retrieval method based on deep integration Hash. The method comprises the following steps that: identifying the data type of data to be retrieved, if thedata type of the data to be retrieved is an image, inputting an image to be retrieved into a trained image Hash network, and obtaining an image binary system code corresponding to the image to be retrieved; calculating a first Hamming distance between the image binary system code corresponding to the image to be retrieved and the binary system code corresponding to each statement in a search library; and selecting a preset quantity of statements with a minimum Hamming distance from the search library as the data type of the data to be retrieved as an image retrieval result. By use of the big data cross-modal retrieval method and system based on the deep integration Hash, an internal cross corresponding relationship between visual data and natural language is captured, so that the compact Hash code of the image and the statement is generated in an end-to-end deep learning framework, and cross-modal retrieval accuracy is improved.

Description

technical field [0001] The present invention relates to the technical field of computer data management, and more specifically, to a large data cross-modal retrieval method and system based on deep fusion hashing. Background technique [0002] With the rapid development of information technology, high-dimensional multimedia data in search engines and social networks continues to increase. How to perform efficient cross-modal approximate nearest neighbor search in these massive high-dimensional big data has become an important and urgent problem. Cross-modal retrieval uses data from one modality as a query condition to return relevant results from another modality, such as retrieving images with sentences or retrieving sentences with images. A powerful solution for efficient cross-modal approximate nearest neighbor search is the hashing method (Hashing), which compresses high-dimensional data into compact binary codes, thus greatly improving its storage efficiency and retriev...

AI Technical Summary

Problems solved by technology

[0004] However, these techniques cannot effectively capture the spatial dependence of images and the temporal dynamics of sentences, so they cannot learn powerful feature representations and cross-modal embeddings, and cannot alleviate the heterogeneity of different modalities, so they cannot generate images for cross-modal High-quality, compact hash coding for modal retrieval, which does not perform well in cross-modal retrieval

Images