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

A method of image convolution feature generation based on multi-region cross weights

A multi-area, weighted technology, applied in still image data retrieval, metadata still image retrieval, biological neural network model, etc., can solve the problem of low accuracy, achieve the effect of improving retrieval accuracy and suppressing background noise areas

Active Publication Date: 2020-06-16
GUILIN UNIV OF ELECTRONIC TECH
View PDF2 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] What the present invention aims to solve is the problem of low accuracy of CNN feature descriptors in the application scenario of "searching images with images", and provides a method for generating image convolution features based on multi-region cross weights, which makes the features more compact and more discriminative and robust

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 method of image convolution feature generation based on multi-region cross weights
  • A method of image convolution feature generation based on multi-region cross weights
  • A method of image convolution feature generation based on multi-region cross weights

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0040] In order to make the object, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below in combination with specific examples and with reference to the accompanying drawings.

[0041] A method for generating image convolution features based on multi-region cross weights, such as figure 1 As shown, it specifically includes the following steps:

[0042] Step 1: Select any convolutional neural network model for classification, and cut off the step of classification in the network (that is, the fully connected layer) to ensure that the network model does not have constraints on the size of the image.

[0043] For the selection of the CNN model, it can be any pre-trained model without any fine-tuning on the retrieved data set. For the CNN model itself, it is necessary to remove the relevant fully connected layers to ensure that the image is input into the CNN network at its original size.

[0044] In ...

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 an image convolution characteristic generating method based on multi-region intersection weight values. According to the method, target positions in K characteristic images arevaguely marked out, and obtained target positions and depth convolution characteristics are adopted to calculate spatial weight maps; the region sizes of the K characteristic images under different scales are calculated, and spatial weights, channel weights and region weights in different regions are calculated separately; regarding each region, intersection weights are adopted to carry out aggregation operation, characteristic vectors of multiple regions are added up, and K-dimensional characteristic representation of the images is obtained. When applied to task retrieving, the generated characteristic representation shows significant advantages, target regions of the images are well highlighted, at the same time, background noise regions are inhibited, descriptors of the images are applied to image retrieval, the accuracy of retrieval can be improved, and the purpose of accurate retrieval is achieved.

Description

technical field [0001] The invention relates to the technical field of image retrieval, in particular to a method for generating image convolution features based on multi-region cross weights. Background technique [0002] With the rapid development of the mobile Internet and the widespread popularity of smart devices, users upload and download massive images every day. On the one hand, more and more images have enriched the image resources on the Internet and brought various conveniences to people; on the other hand, with the explosive growth of image resources, it has also brought many problems to people. Effectively and accurately find the information you really need in the massive data. How to effectively organize, express and retrieve images so that users can find image data efficiently and quickly from a large amount of image data. [0003] Content-based image retrieval (Content-based Image Retrieval, CBIR) was proposed by Kato T in 1992. In the past few years, the ...

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): G06F16/58G06N3/04
Inventor 董荣胜程德强李凤英
Owner GUILIN UNIV OF ELECTRONIC TECH
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

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com