Robust convolution kernel number adaptation method based on corner radiation domain

A convolution kernel and radiation area technology, applied in the field of convolution kernel quantity adaptation, can solve problems such as poor interpretability, high computing overhead, convolution kernel size limit, etc., and achieve good stability and high interpretability

Active Publication Date: 2018-12-21
SHANDONG UNIV OF SCI & TECH
View PDF3 Cites 3 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, existing methods for adapting the number of convolution kernels are limited by the size of the convolution kernels, resulting in problems such as poor interpretability and high computational overhead.

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
  • Robust convolution kernel number adaptation method based on corner radiation domain
  • Robust convolution kernel number adaptation method based on corner radiation domain
  • Robust convolution kernel number adaptation method based on corner radiation domain

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0050] The specific embodiment of the present invention will be further described below in conjunction with accompanying drawing and specific embodiment:

[0051] In a convolutional neural network (CNN), the convolution kernel mainly performs convolution operations, and the convolution kernel is usually initialized in the form of a random matrix, which is learned through error backpropagation (BP) during network training. After the BP operation, the convolutional kernel continuously learns valuable features to increase its weights. To facilitate classification, CNN brings superimposed weights into the scoring function. The larger the activation value obtained by the scoring function, the better the convolution kernel meets the requirements, and the easier it is to distinguish the target image from other images.

[0052] The convolution kernel is like a filter, which is used to extract the local features of the image. When the convolution layer contains only one convolution k...

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 provides a robust convolution kernel number adaptation method based on corner radiation domain, the number of convolution cores can be adaptively determined, By sequentially performing denoising operation, corner detection operation and corner radiation domain operation, the denoising operation and corner detection operation are used to extract points with rich features in the imageand facilitate the extraction of corner radiation domain, and the corner radiation domain operation is used to find the corner expansion area of auxiliary functions. The method of the invention is notlimited by the size of the convolution kernel, and the output results are consistent and robust for different sizes. The convolution kernel adapted by the method well matches a specific data set, This method can improve the accuracy by 3% compared with the comparable method. More importantly, this method can reduce the computational cost by 15% compared with the similar method in the corner radiation domain.

Description

technical field [0001] The invention relates to the field of adaptation of the number of convolution kernels, in particular to a method for adapting the number of robust convolution kernels based on the corner radiation domain. Background technique [0002] In recent years, deep learning has attracted widespread attention from all walks of life and promoted the rapid development of a series of applied research in the field of artificial intelligence. As an important branch of deep learning, convolutional neural network (CNN) is closely related to computer vision research. With the CNN model put into the field of computer vision, computer vision has made remarkable progress. The feature learning strategy of convolutional neural network has been widely successful in a series of fields such as computer vision, and has gradually replaced the traditional research on artificially designed features and has become a new research hotspot. [0003] As one of the important hyperparam...

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): G06K9/62G06K9/40G06K9/46
CPCG06V10/30G06V10/44G06F18/214
Inventor 杜玉越梁煜王路张福新亓亮刘伟
Owner SHANDONG UNIV OF SCI & 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