Cross-modal Hash retrieval method based on deep learning

Abstract

The invention discloses a cross-modal Hash retrieval method based on deep learning. Assuming that the pixel feature vector set of the image modality of n objects is shown in the specification, the method is characterized by comprising the following steps: (1) obtaining binary Hash codes B shared by an image mode and a text mode, deep neural network parameters theta gamma and theta i of the image mode and the text mode, and projection matrixes Pv and Pi of the image mode and the text mode by using an objective function designed based on a deep learning technology; (2) solving unknown variablesB, theta gamma, theta i, Pv and Pi in the objective function in an alternative updating mode; (3) based on the obtained deep neural network parameters theta gamma and theta i of the image modality andthe text modality, and the projection matrixes Pv and Pi, generating the binary Hash codes; (4) calculating a Hamming distance from the query sample to each sample in the retrieval sample set based on the generated binary hash code; and (5) finishing retrieval of the query sample by using a cross-mode retriever based on approximate nearest neighbor search. According to the method, the cross-modalHash retrieval performance is effectively improved.

Description

technical field [0001] The invention relates to a cross-modal hash retrieval method based on deep learning. Background technique [0002] With the rapid development of science and technology and social productivity, the era of big data has come quietly. The so-called big data refers to the collection of data that cannot be captured, managed and processed by conventional software tools within a certain period of time. IBM proposed that big data has 5V characteristics, namely: Volume (large amount of data), Variety (variety of types and sources), Value (data value density is relatively low, but sometimes it is precious), Velocity (data growth rate is fast) ), Veracity (quality of data). Big data can also be considered as an information asset that requires a new processing model to have stronger decision-making power, insight discovery power, and process optimization capabilities. [0003] Information retrieval is an important aspect of data processing, and in the face of bi...

AI Technical Summary

Problems solved by technology

[0007] The purpose of the present invention is to provide a cross-modal hash retrieval method based on deep learning, which solves the problem that the existing cross-modal hash retrieval method based on shallow learning structure cannot well mine the identification of original features. information problem

Images