Image multi-style conversion method based on latent variable feature generation

A latent variable, multi-style technology, applied in the field of image processing, can solve the problems of style representation and image quality that cannot meet the needs

Active Publication Date: 2020-04-10
HEFEI INNOVATION RES INST BEIHANG UNIV
View PDF3 Cites 16 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0009] An image multi-style conversion method based on latent variable feature generation proposed by the present invention can solve...

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 multi-style conversion method based on latent variable feature generation
  • Image multi-style conversion method based on latent variable feature generation
  • Image multi-style conversion method based on latent variable feature generation

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0096] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments It is a part of embodiments of the present invention, but not all embodiments.

[0097] Such as figure 1 As shown, the image multi-style conversion method generated based on latent variable features described in this embodiment includes:

[0098] S100, collecting image data;

[0099] S200. Preprocessing the image;

[0100] S300. Construct and optimize an image conversion model;

[0101] S400. Based on the image conversion model in step S300, the image data in step S200 is used as input to perform conversion processing on the image;

[0102] S500. Evaluate the converted image quality.

[0103] The following combination fig...

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 image multi-style conversion method based on latent variable feature generation can solve the technical problem that the existing image conversion method cannot meet the requirements in style representation and image quality. The method comprises the following steps: S100, collecting image data; s200, preprocessing the image; s300, an image conversion model is constructed and optimized; s400,based on the image conversion model in the step S300, taking the image data in the step S200 as input, and performing conversion processing on the image; and S500, evaluating the quality of the imageobtained by conversion. According to the method, the model is expanded and improved on the basis of MUNIT, and sufficient transmission of content feature information is provided for style conversion of the image by designing jump connection; the style code generator learns potential variables of image style codes, and rich image style conversion can be achieved; and meanwhile, the model can realize conversion of a specific style by taking an input style image as a reference, and has important guiding significance for development of a specific style conversion task.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to an image multi-style conversion method based on latent variable feature generation. Background technique [0002] Due to the advancement of deep learning, transfer learning under computer vision has developed rapidly, including image restoration, character transformation, super-resolution, attribute transformation, image segmentation, scene transformation, and style transformation. More and more researches are devoted to image translation tasks, but most of them are carried out under supervised learning, which requires a large amount of paired data as training support. While the challenging unsupervised image translation research has received more attention, the research in this paper is based on the mutual translation between unpaired multi-image domains. [0003] Inter-image translation builds on connections between image domains. Due to the variability of the relati...

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): G06T3/00G06T9/00
CPCG06T3/0012G06T9/001G06T9/002Y02T10/40
Inventor 张冀聪胡静斐王华武广
Owner HEFEI INNOVATION RES INST BEIHANG UNIV
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