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

Image conversion method and device based on analogy learning

An image and source image technology, applied in the field of computer vision, can solve the problem of lack of diversity of images, and achieve the effect of reducing model learning parameters, improving running speed, and improving diversity

Active Publication Date: 2020-11-27
INST OF INFORMATION ENG CAS
View PDF7 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, this method uses indirect constraints to ensure the consistency of image content, which usually results in a lack of diversity in the generated images.

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
  • Image conversion method and device based on analogy learning
  • Image conversion method and device based on analogy learning
  • Image conversion method and device based on analogy learning

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0040] The present invention will be described in further detail below through specific embodiments and accompanying drawings.

[0041] In this embodiment, a kind of image conversion method based on analogy learning proposed by the present invention is realized through an image conversion device based on analogy learning proposed by the present invention. This system includes the following modules:

[0042] The non-paired image data set construction module is responsible for collecting two types of images according to the difference of image categories and recording them as source image set and target image set respectively;

[0043] The image transformation network module is an image transformation network constructed based on a generative confrontation network. The image transformation network includes a generator and a discriminator for the mutual transformation between the source image and the target image. The loss function of the image transformation network includes the co...

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 an image conversion method and device based on analogy learning. According to the method, two types of non-paired images to be converted are sorted into a source image set and atarget image set, then an image conversion network is constructed, an analog loss function is used as a part of a loss function in the training process, and then image conversion is performed on a source image. According to the method, the analogy loss function is used, so that on one hand, the cross-class difference between the generated image and the source images is ensured, and on the other hand, the difference between any two source images can be ensured to be reserved in the generated image; based on a generative adversarial network structure sharing weight, intermediate hidden variables can be analogous in the same metric space. Meanwhile, model learning parameters can be reduced through weight sharing, and the operation speed is increased. According to the invention, the non-paired images can be used to train the image conversion network, and the real target image is obtained.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and specifically relates to a method and a device for realizing image conversion by using analogy learning and unmatched data. Background technique [0002] Image conversion is a relatively common problem in the field of image processing. Some basic computer vision tasks can also be transformed into image translation problems to solve. For example, image coloring can be understood as the conversion between grayscale images and color images; the conversion between photos and stick figures can be used in image retrieval; image segmentation, image super-resolution and image denoising can all be regarded as images A special case of conversion. [0003] Image conversion algorithms can be divided into two categories according to different algorithm theories: methods based on variational autoencoders and methods based on generative adversarial networks. [0004] Variational autoencoders are s...

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): G06K9/62G06N3/04G06N3/08
CPCG06N3/08G06N3/045G06F18/22G06F18/214
Inventor 王蕊梁栋操晓春
Owner INST OF INFORMATION ENG CAS
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

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com