A semi-supervised image classification method based on random area interpolation

A classification method, a semi-supervised technology, applied in neural learning methods, computer components, instruments, etc., to achieve good classification effect, easy classification, and good generalization performance

Active Publication Date: 2022-06-14
SOUTH CHINA UNIV OF TECH
View PDF6 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005]The purpose of the present invention is to overcome the problem that existing deep neural networks always tend to learn the most discriminative features to obtain high training accuracy, and may focus on not necessarily Important or unwanted areas, especially with limited supervision

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
  • A semi-supervised image classification method based on random area interpolation
  • A semi-supervised image classification method based on random area interpolation
  • A semi-supervised image classification method based on random area interpolation

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0043] like figure 1 The semi -supervised image classification method based on the random area interpolation provided by this embodiment, including the following steps:

[0044] S1. Divide the training data and divide the real label image collection No real tag image collection Test data set details as follows:

[0045] First, the image is flipped horizontally at a probability of 0.5, and then fills the width and height of the image. Then the image is randomly cut into a pixel of 32X32, then the image pixels are reduced, and finally the albinization process is processed; as needed as needed;Classify image data, first divide image data into training data and test data sets Two categories; then divide the training data into two categories: real label image collection Collection with no real tag image The ratio is 1:50, that is, the training data is equal to A real label image is recorded as Right now A real label image is recorded as Right now

[0046] S2, from no real la...

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 a semi-supervised image classification method based on random area interpolation. The method selects a small number of images with real labels from the training set, leaving other images as images without real labels; two types of images are simultaneously sent to random area interpolation module; the interpolation process will be different. Images with real labels can be directly interpolated to generate new augmented images, but images without real labels cannot be interpolated normally, so the teacher network will first obtain high-confidence label information as an The temporary label of the real label image is then interpolated; the network is trained with the new augmented image until the network model is trained to a preset number of times. The method of the invention performs random area interpolation on two types of images at the same time, generates new augmented images for training classification networks, and improves the generalization performance of the training model.

Description

Technical field [0001] The technology field of computer vision involves a semi -supervised image classification method based on random regional interpolation. Background technique [0002] As social media and network services increase, a large number of photos will be uploaded every day.Manually marked them for model training.However, there are not enough real label data in more and more applications, and no real label data is easy to obtain.In order to solve the problem of some data learning with real labels, the study of semi -supervision was studied to effectively reduce the dependence on real label data by using data without real labels. [0003] The semi -supervision learning is a key issue in the field of model recognition and the field of machine learning. It is also a way of learning a combination of supervision learning and unsupervised learning.The semi -supervisor learns to use a large number of real label data, and uses a small amount of real tag data to perform mode ...

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): G06V10/764G06V10/774G06V10/82G06K9/62G06N3/04G06N3/08
CPCG06N3/08G06N3/088G06N3/045G06F18/241G06F18/214
Inventor 曾祥平霍晓阳吴斯
Owner SOUTH CHINA 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