Unlock instant, AI-driven research and patent intelligence for your innovation.

A cross-modal retrieval method and system based on discrete online hash learning

A cross-modal and hashing technology, which is applied in the field of cross-modal retrieval methods and systems based on discrete online hash learning, can solve the problems of large-scale errors between discrete values ​​and real values, insufficient directness and effectiveness, and information loss. Achieve the effect of avoiding quantization error, reducing time complexity and good performance

Active Publication Date: 2021-11-02
SHANDONG UNIV
View PDF7 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

The relaxation-based strategy and the auxiliary variable strategy relax the hash code from the Hamming space to the real-valued space to avoid complex optimization, these two strategies may lead to a large quantization error between the discrete value and the real value, thus result in loss of information and performance degradation
They preserve the similarity between incoming data and previously accumulated data in latent space (real-valued space), which is less direct and efficient than methods that directly measure similarity in Hamming space

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • A cross-modal retrieval method and system based on discrete online hash learning
  • A cross-modal retrieval method and system based on discrete online hash learning
  • A cross-modal retrieval method and system based on discrete online hash learning

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0038] In this embodiment, in order to deal with the cross-modal retrieval task of large-scale streaming data, the invention proposes an online cross-modal hash retrieval method based on supervised learning. This method improves the generalization ability of the model and can learn hash functions of multiple modalities, and the computational complexity of this method is linearly related to the size of the new data block, which improves the efficiency in large-scale cross-modal retrieval .

[0039] In the method designed by the present invention, when a new data block appears in the tth round, it is mainly carried out in two steps: (1) generate an r-bit hash code for the new incoming data, and ensure that the existing data The hash code of M remains unchanged; (2) Update the hash function of M-mode to adapt to the new incoming data and existing data. The following will divide the method into three parts in order to introduce the technical content in detail.

[0040] It should...

Embodiment 2

[0108] The purpose of this embodiment is to provide a computing device, including a memory, a processor, and a computer program stored in the memory and operable on the processor, and the processor implements the steps of the above method when executing the program.

Embodiment 3

[0110] The purpose of this embodiment is to provide a discrete online cross-modal hash retrieval system based on supervised learning, including:

[0111] The hash code learning module is configured to: acquire simulated stream data;

[0112] Based on the simulated stream data, the hash code of the second data block of the latest round is discretely updated, and the hash code of the first data block of the previous round of the latest round is kept unchanged, so as to learn a unified hash code from different modes. Greek code;

[0113] The hash function learning module is configured to: according to the hash code of the latest round of the second data block, obtain the projection matrix that maps the feature to the hash code for each modality, that is, the hash function, and perform the hash function Learn and update the hash function of each modality to handle the learned hash code;

[0114] The retrieval module is configured to: calculate the Hamming distance between the sa...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The present invention proposes a cross-modal retrieval method and system based on discrete online hash learning, including: obtaining simulated stream data; based on the simulated stream data, discretely updating the hash code of the latest round of the second data block, Keep the hash code of the first data block of the previous round of the latest round unchanged, and learn a unified hash code from different modalities; according to the hash code of the second data block of the latest round, obtain the hash code of each module separately. The state maps the feature to the projection matrix of the hash code, that is, the hash function, to learn the hash function, and update the hash function of each mode to process the learned hash code; calculate the simulation based on the updated hash function The Hamming distance between samples in the stream data, according to the Hamming distance, returns the retrieval sample of another modality that is close to the sample to be queried. The invention learns a unified hash code from different modalities, so that the hash code can integrate information of multiple modalities.

Description

technical field [0001] The invention belongs to the field of cross-media retrieval, and in particular relates to a cross-modal retrieval method and system based on discrete online hash learning. Background technique [0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art. [0003] With the popularity of multimedia content on the Internet, multimedia data from various search engines and social media has exploded. Generally speaking, the massive multimedia data generated by users on the Internet is incrementally generated, that is, it appears dynamically in the form of data streams. With the increasing demand of users for cross-modal retrieval, the traditional cross-modal hash retrieval method is time-consuming and computationally expensive, which is difficult to meet the requirements. Therefore, online cross-modal hash methods have attracted widespread interest in recent years....

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Patents(China)
IPC IPC(8): G06F16/483G06N20/00
CPCG06F16/483G06N20/00
Inventor 罗昕詹雨薇付婷许信顺
Owner SHANDONG UNIV