Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Image Retrieval Method and Device Based on Specific Class Prototype in Deep Hash Network

A network image and hashing technology is applied in the field of deep hashing network image retrieval methods and devices, and can solve problems such as restricting the improvement of hash retrieval performance.

Active Publication Date: 2022-05-31
WUHAN INSTITUTE OF TECHNOLOGY
View PDF8 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] The technical problem to be solved by the present invention is to provide a data retrieval method and device based on a specific class prototype deep hash network, which solves the semantic gap and domain gap between semantic tags and hash codes, which restricts the improvement of hash retrieval performance The problem

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
  • Image Retrieval Method and Device Based on Specific Class Prototype in Deep Hash Network
  • Image Retrieval Method and Device Based on Specific Class Prototype in Deep Hash Network
  • Image Retrieval Method and Device Based on Specific Class Prototype in Deep Hash Network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0059] This example provides figure 1 A data retrieval method based on a class-specific prototype deep hash network is shown, including:

[0060] S1, establish a deep hash network model; the deep hash network model learns discrete class-specific prototypes from the semantic labels of image data as the intermediate semantic representation of semantic labels, and establishes a rough semantic relationship between the hash code of image data and the class-specific prototypes , and construct the refined semantic relationship between all hash codes through paired supervisory information;

[0061] S2, acquiring target image information to be retrieved;

[0062] S3, the target image information is retrieved according to the deep hash network model, and a retrieval result is obtained.

[0063] Specifically, the retrieval process includes: preprocessing the image to be retrieved, extracting the feature vector, inputting the feature vector into the deep hash network to obtain the hash ...

Embodiment 2

[0125] like figure 2 As shown, the present embodiment provides an image data retrieval device based on a specific class prototype deep hash network, including:

[0126] memory 91 for storing computer programs;

[0127] The processor 92 is configured to execute a computer program to implement any one of the above-mentioned deep hash network image retrieval methods based on specific class prototypes.

[0128] Specifically, the processor 92 and the memory 91 are electrically connected. The processor 91 can access the memory 92, read the program and data in the memory 92, and be used to execute the image data method based on the depth hash network of the specific class prototype; The program and data of the image data method of the hash network are used to save the result obtained after the processor 92 executes the method.

Embodiment 3

[0130] This embodiment provides a non-volatile computer-readable storage medium on which computer program instructions are stored. When the computer program instructions are executed by a processor, any one of the above-mentioned deep hash network image retrieval based on a specific class prototype can be realized. method.

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 invention discloses a deep hash network image retrieval method and device based on a specific class prototype, and relates to the technical field of computer information retrieval. The data retrieval method based on a specific class prototype deep hash network specifically includes: establishing a deep hash network model; obtaining target image information to be retrieved; retrieving the target image information according to the deep hash network model, and obtaining a retrieval result . The present invention obtains discrete specific class prototypes through label information decomposition and learning, as the intermediate semantic representation of semantic labels, and establishes a rough semantic relationship between the hash code of the image data and the specific class prototypes, and through paired The supervisory information constructs a refined semantic relationship between all hash codes. This intermediate representation narrows the semantic gap between semantic labels and hash codes, and solves the semantic gap and domain gap between semantic labels and hash codes. Improved performance of hash retrieval.

Description

technical field [0001] The invention belongs to the technical field of computer information retrieval, in particular to a deep hash network image retrieval method and device based on a specific class prototype. Background technique [0002] Hashing methods, especially deep hashing, have recently become popular in large-scale multimedia retrieval, which can be used in various large-scale multimedia data search tasks. However, there is still a "semantic gap" between hash codes extracted from many multimedia data and semantic labels, that is, there is a difference between the similarity calculated from hash codes extracted from underlying visual features and the semantic similarity understood by humans. , and the "domain gap" problem, that is, there is a difference between the hash code in the Hamming space and the one-hot space in which the semantic label is located, which causes the insufficient semantic representation ability of the hash code and greatly weakens and restrict...

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
IPC IPC(8): G06F16/583G06F16/55G06F16/538G06F16/51G06V10/74G06V10/82G06K9/62G06N3/04G06N3/08
CPCG06F16/583G06F16/55G06F16/538G06F16/51G06N3/08G06N3/045G06F18/22Y02D10/00
Inventor 马雷罗心怡刘红李璇
Owner WUHAN INSTITUTE OF TECHNOLOGY
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products