Image classification method based on deep neural network subspace coding

A technology of deep neural network and classification method, applied in the direction of neural learning method, biological neural network model, neural architecture, etc., can solve the problem of increasing output feature dimension, achieve model complexity and accuracy, reduce feature dimension, and realize feature Effect of Dimension Size and Classification Accuracy

Inactive Publication Date: 2019-12-03
西安宏规电子科技有限公司
View PDF5 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

These two methods are first-order pooling methods, which have strong limitations on the feature representation of images. Some researchers have also tried to make some improvements to the feature pooling layer of the deep neural network, but these improvements are to a certain extent. The classification accuracy is improved, but the output feature dimension is greatly increased.

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 classification method based on deep neural network subspace coding
  • Image classification method based on deep neural network subspace coding

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0024] The present invention will be further described in detail below in conjunction with the accompanying drawings, which are explanations rather than limitations of the present invention.

[0025] Such as figure 1 Shown is a flow chart of the present invention, comprising the following steps:

[0026] Step 1: Divide the image set to be classified into training sets {A i} dataset.

[0027] Step 2: Select a deep neural network model, you can choose a deep convolutional neural network model;

[0028] Step 3: Write the local feature output layer of the deep neural network model as a matrix Each row i∈[1,c] represents a feature map, each column j∈[1,hw] represents a spatial position, c is the number of channels of the feature map, h is the height of the feature map, and w is the width of the feature map ;

[0029] make is the singular value decomposition of matrix X, where u i is the left singular vector of matrix X, v i is the right singular vector of matrix X, σ i i...

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 classification method based on deep neural network subspace coding, and belongs to the technical field of artificial intelligence, computer vision and machine learning. According to the method, output features of a deep neural network are mapped into a low-dimensional manifold (Grassmann manifold) space of a column orthogonal matrix, and a Grassmann classifier hasthe same compact form, so that the parameter size of the classifier is remarkably reduced; according to the Grassmann projection method, the feature dimension can be reduced, and the classifier modelis further compressed; on the premise of ensuring strong feature discrimination, the feature dimension is reduced, the classifier model is compressed, and the balance between the size of the feature dimension and the classification precision is realized.

Description

technical field [0001] The invention belongs to the technical fields of artificial intelligence, computer vision and machine learning, and in particular relates to an image classification method based on deep neural network subspace coding. Background technique [0002] The current deep neural network has been widely used in various research directions of artificial intelligence, computer vision and machine learning, such as speech recognition, image classification, target detection and 3D scene reconstruction, etc. Generally, image classification methods based on supervised learning can be divided into two steps: the first step is to extract image features; the second step is to learn one or several image classifiers. The image classification method based on deep neural network effectively unifies these two steps into a whole for end-to-end training. The final key step of the image feature extraction method based on deep neural network is feature pooling. Feature pooling ...

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): G06K9/62G06N3/04G06N3/08
CPCG06N3/08G06N3/045G06F18/23G06F18/214
Inventor 魏星张玥龚怡宏
Owner 西安宏规电子科技有限公司
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