Depth hash image retrieval method based on cosine measurement

A technology of image retrieval and depth, applied in still image data retrieval, other database retrieval, still image data clustering/classification, etc.

Active Publication Date: 2019-10-08
BEIJING UNIV OF TECH
View PDF5 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0006] In order to effectively solve the problems existing in the existing deep hash method, the present invention provides a deep hash image retrieval method based on cosine metric, which proposes a deep hash framework under the

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
  • Depth hash image retrieval method based on cosine measurement
  • Depth hash image retrieval method based on cosine measurement
  • Depth hash image retrieval method based on cosine measurement

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0056] The invention proposes a cosine metric-based deep hash image retrieval method. The concrete realization steps of this invention are as follows:

[0057] Step 1: Select the image data set. For the images in it, randomly select a part as the test set, then select a part from the remaining data as the test set, and finally the remaining part as the database set.

[0058] Step 2: Construct a deep learning network for learning hash functions. The present invention adopts the CNN-F network structure as the basic part of image feature learning, wherein the last layer of CNN-F is replaced with a full connection with C neurons layer so that the output of the penultimate layer maps to Hamming space. Simultaneously, the present invention uses a twin network to learn a hash function, that is, uses two CNN-Fs to learn a hash function, which have shared weights and the same network structure. The paired sample images are used as input to these two networks. The specific model para...

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 provides a depth hash image retrieval method based on cosine measurement. For huge image data on the Internet, in order to meet the requirements of users, a rapid and accurate image retrieval method is found to become an urgent problem to be solved. Based on cosine measurement, diversity of vector length can be effectively reduced, and then retrieval performance is improved while category information is introduced into a loss function and the loss function is combined with cosine measurement constraint, so that similarity maintenance in the same network can be learned jointly, and classification information can be fully utilized. By the adoption of the scheme, large-scale image retrieval can be effectively achieved, and experiments prove that the performance of the scheme issuperior to that of an existing image retrieval method, and the scheme has very important application value.

Description

technical field [0001] The invention relates to the field of image retrieval, in particular to a depth hash image retrieval method based on cosine metric. Background technique [0002] In recent years, with the explosive growth of network multimedia data, hundreds of thousands of images are uploaded to the Internet every day. In the face of such large-scale multimedia data, it becomes difficult to retrieve relevant images from massive images according to different user needs. extremely difficult. Therefore, content-based image retrieval has received more and more attention in commercial applications and academic fields. Assuming that both the images in the database and the query image are represented by real-valued features, the easiest way to retrieve relevant images is to sort the images in the database according to the one-by-one distance metric between the database image and the query image in the feature space, and then return The image result with the smallest distan...

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): G06F16/55G06F16/901G06N3/04
CPCG06F16/55G06F16/9014G06N3/045
Inventor 毋立芳李丰简萌胡文进赵宽陈禹锟
Owner BEIJING UNIV OF TECH
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

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

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap