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

Zero sample image identification method based on attributive learning of discriminative sample

A technology of sample attributes and sample images, which is applied in the field of image recognition to achieve the effect of de-dependence

Inactive Publication Date: 2017-07-25
NANJING UNIV OF POSTS & TELECOMM
View PDF2 Cites 11 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] The technical problem to be solved by the present invention is to provide a zero-shot image recognition method based on discriminative sample attribute learning to alleviate the impact of domain migration on image recognition accuracy in view of the defects involved in the background technology

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 identification method based on attributive learning of discriminative sample
  • Zero sample image identification method based on attributive learning of discriminative sample
  • Zero sample image identification method based on attributive learning of discriminative sample

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0050] Below in conjunction with accompanying drawing, technical scheme of the present invention is described in further detail:

[0051] The invention discloses a zero-sample image recognition method based on discriminative sample attribute learning: figure 2 As shown, first we assume that the source domain and the target domain share the same projection matrix, so that we can jointly learn the projection matrix from the feature space to the attribute space on the source domain and the target domain. Then, we can use the projection matrix to map the image feature data of the target domain to the attribute space. Finally, we adopt the simplest classification model - nearest neighbor to classify the target domain samples. A zero-shot image recognition method based on discriminative sample attribute learning according to the present invention comprises the following steps:

[0052] (1) Determination of the target domain projection matrix:

[0053] Let the source domain set S...

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 identification method based on the attributive learning of a discriminative sample. According to the method, firstly, a projection matrix is learned commonly in a source domain and a target domain, and a prototype of each type in the source domain and the target domain is adopted to adjust the learned projection matrix. Secondly, based on the learned projection matrix, the image features of the target domain are mapped in an attribute space to obtain the attribute representation thereof. Finally, in the attribute space, by using a nearest neighbor classifier, images are classified. For the target domain, the existing projection matrix learning method does not consider the distribution difference between the target domain and the source domain, so that the domain migration phenomenon occurs more easily. However, the above method mitigates the above effect by comprehensively utilizing the sample information of the source domain and the target domain, so that the higher image identification accuracy is realized.

Description

technical field [0001] The invention relates to the field of image recognition, in particular to a zero-sample image recognition method based on discriminative sample attribute learning. Background technique [0002] The so-called zero-shot image recognition (zero shotrecognition) is to learn a recognition model for samples without label data. Therefore, it is an important part of the field of pattern recognition and computer vision. It has received extensive attention from the research community and has achieved rapid development. But most zero-shot image recognition techniques are: first learn a model from the source domain, and then directly apply the model to the target domain to predict the attribute representation of the image. Such learning methods do not consider the domain transfer problem. Since the labels of source and target domains are different, this poses a new challenge for researchers—how to mitigate the impact of domain transfer problem on the final classi...

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): G06K9/62
CPCG06F18/24147G06F18/2415
Inventor 吴松松汪禄高广谓郁俊荆晓远岳东
Owner NANJING UNIV OF POSTS & TELECOMM
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