Residual convolutional neural network image classification method based on multi-path feature weighting

A convolutional neural network and feature weighting technology, applied in the field of computer vision, can solve problems such as poor performance, and achieve the effect of improving performance and improving feature weighting effects

Active Publication Date: 2018-11-06
ZHEJIANG UNIV OF TECH
View PDF3 Cites 17 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] In order to overcome the poor performance of the existing image classification methods when applied to complex images, 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
  • Residual convolutional neural network image classification method based on multi-path feature weighting
  • Residual convolutional neural network image classification method based on multi-path feature weighting

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0021] The present invention will be further described below in conjunction with the flowchart.

[0022] refer to figure 1 and figure 2 , a residual convolutional neural network image classification method based on multi-way feature weighting, comprising the following steps:

[0023] 1) First, the input image of the model is the preprocessed original image, and the preprocessed image must be all cropped to a fixed size. In order to facilitate model training, it is best to keep the length and width of the fixed size consistent. The specific size is determined by the model The specific application and model size are determined. Common input image sizes are: 512, 299, 224, etc.

[0024] 2) Perform large-scale convolution and pooling operations on the image, such as convolution with a convolution kernel size of 7×7 and a stride of 2 and maximum pooling with a size of 3×3 and a stride of 2. The significance of choosing a larger convolution kernel size is to extract the underly...

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 relates to a residual convolutional neural network image classification method based on multi-path feature weighting. The residual convolutional neural network image classification method based on multi-path feature weighting includes the following steps: 1) an input image of a model is an original image which is preprocessed, and the preprocessed image is cut into a fixed size; 2) alarge size of convolution operation and pooling operation are performed on the image; 3) the features output from the step 2) are transmitted to a first multi-path feature weighting residual module;and 4) the output of the multi-path feature weighting residual module in the step 3) is continuously transmitted to the next multi-path feature weighting residual module, and after passing a pluralityof multi-path feature weighting residual modules, the size of the output feature image is gradually reduced until to become a smaller size, and then the smaller size is reduced to be a feature pointthrough an average pooling layer; and the obtained feature point is directly sent into a classification layer for classification or for classification after passing a full connection layer. The residual convolutional neural network image classification method based on multi-path feature weighting is applied to a complicated image classification task, enriches feature expression, and avoids the problem of gradient loss caused by increase of neural network depth.

Description

technical field [0001] The invention relates to the field of computer vision, is a kind of deep learning technology, and is mainly used for training deep image classification models, especially an image classification method. Background technique [0002] In recent years, with the exponential growth of computer computing power and the emergence of new neural network architectures, deep learning technology has begun to shine in the fields of computer vision, speech recognition, and natural language processing. In the field of computer vision, the emergence of convolutional neural networks has greatly improved the performance of computers in image segmentation and image recognition tasks, and its recognition accuracy is much higher than traditional machine learning algorithms. At present, image recognition technology based on convolutional neural network has been widely used. [0003] A conventional convolutional neural network includes a convolutional layer, a pooling layer,...

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): G06K9/62G06N3/04
CPCG06N3/045G06F18/24
Inventor 刘义鹏李湛青陈朋蒋莉王海霞梁荣华
Owner ZHEJIANG 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