Unsupervised decoupling image generation method based on invariant information distillation

An image generation, unsupervised technology, applied in the field of computer vision, can solve problems such as poor image effect and poor image generation quality, and achieve the effect of improving image decoupling generation effect, decoupling ability, and standardized mutual information

Pending Publication Date: 2022-01-28
YANGTZE DELTA REGION INST (QUZHOU) UNIV OF ELECTRONIC SCI & TECH OF CHINA
View PDF0 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0008] The invention is a decoupling image generation method of unsupervised invariant information distillation clustering, which mainly solves the problems of poor image generation quality and poor image generation effect by class in the existing clustering generation confrontation network method

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
  • Unsupervised decoupling image generation method based on invariant information distillation
  • Unsupervised decoupling image generation method based on invariant information distillation
  • Unsupervised decoupling image generation method based on invariant information distillation

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0081] Step 1: Preprocessing the experimental data;

[0082] Get the CIFAR10 dataset from official sources. The CIFAR10 dataset is an image dataset composed of 60,000 color RGB images. The dataset contains 10 categories in total, and the number of images in each category is equal. In addition, the size of each image in the dataset is 32×32, and it is marked corresponding category information. In order to convert the image data into data that is easier for the deep learning model to learn, the image pixel values ​​are normalized to the interval [-1,1] and converted into tensors for storage.

[0083] Step 2: Perform random data enhancement operations on experimental data;

[0084] To perform random data enhancement processing on the image data processed in step 1, a total of four operations including random cropping, random horizontal flip, random brightness change and random grayscale are used. The specific random data enhancement process for each image is as follows: the fi...

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 an unsupervised decoupling image generation method based on invariant information distillation, and belongs to the field of computer vision. The method comprises the following steps: firstly, normalizing training images, respectively sampling noise from normal distribution and uniform distribution, and splicing to obtain mixed noise; through adjustment of parameters and details, the new network can be suitable for a decoupling image generation task. Besides, based on the idea of information invariant distillation, extra unsupervised information is introduced for the encoder to perform more sufficient clustering, so that the encoder has more accurate resolution capability for image categories under the unsupervised condition, and more correct category information is provided for the generator, thereby improving the decoupling generation effect of the generative adversarial network. According to the clustering scheme based on invariant information distillation, the decoupling image generation level of the generative adversarial network can be remarkably improved, and meanwhile the image generation quality of an existing method is improved.

Description

technical field [0001] The invention belongs to the field of computer vision, mainly relates to the generation of images, and is mainly applied to the film and television entertainment industry, product design, machine vision understanding and the like. Background technique [0002] Image generation refers to the use of computer vision technology, supplemented by deep learning methods to understand the representation of image content, and realize the technology of image generation. According to whether there is clear supervision information as a guide, it can be divided into two categories: supervised image generation and unsupervised image generation. Unsupervised image generation methods usually take images as input, learn the distribution of image data from random noise distributions, and establish a mapping between them through certain learning schemes and technical means. The supervised image generation method provides accurate guidance to the image generation process ...

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): G06K9/62G06N3/04G06N3/08
CPCG06N3/088G06N3/047G06N3/048G06N3/045G06F18/23
Inventor 陈志勇
Owner YANGTZE DELTA REGION INST (QUZHOU) UNIV OF ELECTRONIC SCI & TECH OF CHINA
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