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

Zero-sample image classification method based on structure propagation

A sample image and classification method technology, which is applied to computer parts, instruments, character and pattern recognition, etc., can solve the problem of inaccurate zero-sample image classification method

Active Publication Date: 2018-11-20
XIAN UNIV OF TECH
View PDF3 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to provide a zero-sample image classification method based on structure propagation, which unifies the visual and semantic distribution structure and sorts out the relationship between the two, and solves the inaccurate problem of the existing zero-sample image classification method

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
  • Zero-sample image classification method based on structure propagation
  • Zero-sample image classification method based on structure propagation
  • Zero-sample image classification method based on structure propagation

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0103] In the image database of animals with attributes (AwA), there are 40 visible categories and 10 unseen categories, with a total of 30473 images. Each category has 85 attributes. According to whether a certain category contains 85 attributes, it is marked as 1 if it is included, and 0 if it is not included. An 85-dimensional semantic representation vector can be formed.

[0104] The sample representation of the image is a 1024-dimensional vector that outputs the final layer of the GoogleNet deep model trained on the ImageNet dataset by directly inputting the image.

[0105] The comparison methods in the following table are respectively: output structure joint embedding method (SJE), latent embedding model method (LatEm), comprehensive classifier method (SynC) and the zero-shot image classification method (SP) based on structure propagation of the present invention.

[0106] By comparison, the recognition rates of the four methods are shown in Table 1:

[0107] Table 1

...

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 zero-sample image classification method based on structure propagation. The method comprises the steps of (1) calculating semantic structures and image structures of all categories, (2) establishing a structural relationship model suitable for zero-sample image classification according to calculation formulas of the semantic structures and the image structures in step 1 and solving an optimization formula, (3) continuing to update the image structures according to the optimization formula in the step (2), and (4) carrying out loop iteration of the steps 2 and 3 and carrying out forward structure propagation until signs of stable visible categories are obtained. According to the zero-sample image classification method based on structure propagation, the average recognition rate is highest, the accuracy and efficiency of classification are improved, and the effect of better zero-sample image classification is obtained.

Description

technical field [0001] The invention belongs to the technical field of image classification methods, and in particular relates to a zero-sample image classification method based on structure propagation. Background technique [0002] In recent years, deep learning has made breakthroughs in image target recognition. Under certain experimental conditions, it has even surpassed human discrimination. However, most of deep learning is a supervised learning method, which requires a lot of Annotated training samples for the classes. In practical applications, most of the time, we cannot obtain a large number of labeled samples or require a lot of manpower and material resources to obtain such samples. Therefore, in order to solve the problem of how to identify the categories of training samples without visual images, zero Sample learning is proposed. [0003] In the zero-shot classification problem, the known information is the image samples of the seen categories and the semanti...

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 Applications(China)
IPC IPC(8): G06K9/62
CPCG06F18/24
Inventor 蔺广逢缪亚林范引娣陈万军张二虎朱虹
Owner XIAN UNIV OF TECH