Method for acquiring neural network training set and system thereof

A neural network training and acquisition method technology, applied in neural learning methods, biological neural network models, special data processing applications, etc., can solve the problem of image recognition training results that affect neural network models, the subjectivity of screening results, and the difficulty of screening results Errors and other problems, to achieve the effect of improving generalization ability, fast and efficient network training and parameter adjustment, and reducing data screening errors

Inactive Publication Date: 2018-09-28
SICHUAN FEIXUN INFORMATION TECH CO LTD
View PDF5 Cites 15 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

The problem is that the workload is extremely huge, the screening results are highly subjective, and the screening results are prone to errors. At the same time, using the wrong image sample data set to train the neural network will bring wrong classification results.
When traditional web crawler technology is used to obtain image data, the quality of the crawled images generally shows a downward trend as the number of crawls increases, resulting in large noise in the data crawled by traditional web crawlers, which affects subsequent neural network-based models. Training Results for Image Recognition

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
  • Method for acquiring neural network training set and system thereof
  • Method for acquiring neural network training set and system thereof
  • Method for acquiring neural network training set and system thereof

Examples

Experimental program
Comparison scheme
Effect test

no. 1 example

[0066] The first embodiment of the present invention, such as figure 1 Shown:

[0067] A method for obtaining a neural network training set, comprising steps:

[0068] S100 acquires category information of the image set to be screened;

[0069] S200 Screen the target image, and obtain the target image data set corresponding to the image set to be screened; the target image is image data whose similarity with the sample image reaches a preset similarity threshold; the sample image is the image set to be screened The template image corresponding to the category information of ;

[0070] S300 Perform adversarial training on the target image data to obtain a sample data set.

[0071] Specifically, in this embodiment, when processing image recognition or image classification or other machine learning tasks, how to improve the performance (recognition rate, classification accuracy) of the neural network model, because the larger the amount of data in the neural network model , t...

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 method for acquiring a neural network training set and a system thereof. The method comprises the steps of: S100, acquiring category information of an image set to be filtered; S200, filtering a target image, and obtaining a target image data set corresponding to the image set to be filtered, wherein the similarity of the target image and the sample image reaches the imagedata of a preset similarity threshold, and the sample image is a template image corresponding to the category information of the image set to be filtered; and S300, performing a confrontation training on the target image data to obtain a sample data set. The invention is capable of reducing manual screening of sample data sets, improving screening efficiency and screening reliability, and improving the accuracy of the neural network.

Description

technical field [0001] The invention relates to the field of data processing, in particular to a method for obtaining a training set of a neural network and a system thereof. Background technique [0002] In recent years, with the continuous development of computer vision technology, especially the rapid development of neural network models, people's demand for image data required for computer vision training, especially for image data with accurate label information, is increasing. [0003] Convolutional Neural Networks (CNN) is a kind of deep learning algorithm and an important processing and analysis tool in the field of image recognition. In recent years, it has become one of the research hotspots in many scientific fields. The advantage of the neural network model algorithm is that it does not need to use any manually labeled features when training the model, and can automatically explore the features hidden in the input variables. At the same time, the weight sharing f...

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): G06N3/08G06K9/62G06F17/30
CPCG06N3/08G06F18/214
Inventor 罗培元
Owner SICHUAN FEIXUN INFORMATION TECH CO LTD
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