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

Multi-label image identification method and system based on DCGAN and GCN

A multi-label, recognition algorithm technology, applied in the field of image processing, which can solve problems such as target occlusion, inapplicability of multi-label images, and inconspicuous targets.

Active Publication Date: 2021-09-10
QILU UNIV OF TECH
View PDF5 Cites 4 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] The content in the multi-label image is relatively complex, and there may be problems such as occlusion of the target, complex background, and unobvious target.
It may not be applicable if the processing algorithm of a single-label image is directly applied to a multi-label image

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
  • Multi-label image identification method and system based on DCGAN and GCN
  • Multi-label image identification method and system based on DCGAN and GCN
  • Multi-label image identification method and system based on DCGAN and GCN

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0035] Such as figure 1 As shown, the present disclosure provides a multi-label recognition algorithm based on DCGAN and GCN, including:

[0036] Construct a DCGAN model based on the GAN model, and generate similar images based on the DCGAN model;

[0037] Generate similar images based on the DCGAN model, use the migration-based CNN algorithm to extract features, migrate the parameters of the neural network of the DCGAN model to the CNN algorithm for feature extraction of multi-label images, and use the GCN algorithm to generate categories through the relationship graph between training labels label classifier;

[0038] The class label classifier based on the GCN algorithm is used to classify and recognize multi-label images. After dot multiplying the features extracted by the CNN algorithm and the semantic feature vector matrix in the class classifier generated by the GCN algorithm, the multi-label classifier is used to process the image. identify.

[0039] Further, the ge...

Embodiment 2

[0089] A self-supervised learning multi-label recognition system based on DCGAN and GCN, implemented based on a server, the server includes:

[0090] The image generation module is configured to generate similar images based on the DCGAN model;

[0091] The feature extraction module is configured to extract features based on the migration-based CNN algorithm, migrate the parameters of the neural network of the DCGAN model to the CNN algorithm to perform feature extraction on multi-label images, and use the GCN algorithm to generate category labels through the relationship graph between training labels Classifier;

[0092] The image recognition module is configured to classify and recognize multi-label images based on the GCN algorithm. After dot multiplying the features extracted by the CNN algorithm and the semantic feature vector matrix in the category classifier generated by the GCN algorithm, the image is processed by the multi-label classifier. identify.

Embodiment 3

[0094] A computer-readable storage medium is used for storing computer instructions, and when the computer instructions are executed by a processor, a multi-label recognition algorithm based on DCGAN and GCN as described in the first aspect is completed.

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 multi-label identification algorithm based on DCGAN and GCN, and the algorithm comprises the steps: building a DCGAN model based on a GAN model, and generating a similar image based on the DCGAN model; extracting features based on a migrated CNN algorithm, migrating parameters of a neural network of the DCGAN model to the CNN algorithm to perform feature extraction on the multi-label image, and generating a category label classifier through a relation graph between training labels by using a GCN algorithm; generating a data pre-training model through a deep convolutional generative adversarial network, and migrating parameters of a convolutional neural network of the pre-training model to a target task to finely adjust the network so as to obtain a more accurate image identification effect. Meanwhile, when the image is generated, random noise is increased, so that the robustness of the pre-training model can be improved.

Description

technical field [0001] The present disclosure relates to the technical field of image processing, in particular to a multi-label image recognition method and system based on DCGAN (Deep Convolutional Adversarial Network) and GCN (Graph Neural Network). Background technique [0002] In the Internet era, multimedia data has become the mainstream of information, such as images and short videos, which have had an important impact on people's lives. Image recognition is a branch of computer vision. By labeling images with appropriate labels, the visual information conveyed by images is converted into semantic information that is easy for people to understand, so that people can better understand and analyze images. Single-label image classification algorithms have been studied for many years, such as support vector machines, follow-up forest algorithms, etc. Supervised deep learning algorithms such as convolutional neural networks have excellent performance in single-label image...

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): G06K9/62G06K9/46G06N3/04G06N3/08
CPCG06N3/084G06N3/045G06F18/22G06F18/241G06F18/214
Inventor 刘嵩来庆涵周梓涵
Owner QILU UNIV OF TECH
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