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

An Image Recognition Method Based on Variational Group Convolution

An image recognition and convolution technology, applied in the field of image recognition of variable group convolution, can solve the problem of overfitting without considering the input channel, reduce the possibility of overfitting, be easy to implement, and improve the degree of generalization Effect

Active Publication Date: 2018-11-13
BEIHANG UNIV
View PDF9 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] Although various methods have different advantages, the problem with the above methods is that they do not consider whether the input channels between different convolution kernels of the same layer will lead to overfitting

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
  • An Image Recognition Method Based on Variational Group Convolution
  • An Image Recognition Method Based on Variational Group Convolution
  • An Image Recognition Method Based on Variational Group Convolution

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0024] The present invention is described in detail below in conjunction with accompanying drawing and embodiment,

[0025] Such as figure 1 Shown, the concrete realization steps of the present invention are as follows:

[0026] (1) For each convolution kernel in m convolution kernels (the number of m can be adjusted manually, it is recommended that m The combination of , assuming a total of n' combined feature maps, that is,

[0027]

[0028] The feature maps of m combinations are randomly selected from these n' combinations, and each such combination corresponds to a convolution kernel, and there are m convolution kernels in total.

[0029] (2) For the m combined feature maps and corresponding m convolution kernels randomly extracted in the second step, it is assumed that the i-th convolution kernel among the m convolution kernels of the current l layer corresponds to k input features , the k feature maps are convolved,

[0030]

[0031] where W i Is the convolut...

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 present invention relates to an image recognition method based on variable group convolution. Aiming at the overfitting problem existing in the current image recognition algorithm based on deep convolutional neural network, the idea of ​​random channel combination is adopted. For each processing layer, first Channel splitting is performed on the input feature map, and then the channel combination is arranged, and different channel combinations are assigned to each convolution kernel, and finally the convolution activation feature map of this layer is calculated. The present invention can effectively improve the performance of the same layer in feature extraction. The degree of data randomization reduces the possibility of overfitting of model parameters, thereby improving the performance of convolutional neural networks in image retrieval, image matching and other problems.

Description

technical field [0001] The invention relates to an image recognition method of variable group convolution, which is used in the fields of video monitoring, image retrieval and the like. Background technique [0002] Image recognition is a typical application in computer vision. With the development of computer vision technology, more and more applications hope to improve the effect of image recognition models based on deep convolutional neural networks without increasing the amount of model parameters. Overfitting is an important problem. The product neural network is composed of multiple layers, so it is very important to improve the resistance of each layer to overfitting. [0003] There are many ways to improve the overfitting resistance of deep convolutional neural network models. One of the main methods is to add regularization terms to the model. For example, an L2 norm regularization term is added to the parameters of the model. The L2 norm refers to the sum of the...

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 Patents(China)
IPC IPC(8): G06K9/62G06N3/04G06N3/08
CPCG06N3/08G06N3/045G06F18/2111
Inventor 张弘辛淼张泽宇
Owner BEIHANG UNIV
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

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

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com