Super-realistic drawing image style migration method based on deep learning

A realism, deep learning technology, applied in image enhancement, image analysis, image data processing and other directions, can solve problems such as affecting the quality of primitives, achieve good image quality, and avoid image generation failures.

Active Publication Date: 2020-09-29
TONGJI UNIV
View PDF11 Cites 9 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, the conditional generation confrontation network can no longer control the strength of the condition for the generation constraint. For the generation of surreal (mosaic) style images, the existence of the condition will greatly affect the quality of the primitive itself.

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
  • Super-realistic drawing image style migration method based on deep learning
  • Super-realistic drawing image style migration method based on deep learning
  • Super-realistic drawing image style migration method based on deep learning

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0115] In order to verify the performance of the present invention, this embodiment uses MS-COCO14 as the content map training set, generates a style map data set according to step 1, and iterates 200k times in the RTX2080Ti environment. In order to illustrate the advantages of the present invention, this embodiment uses the same data set under the same conditions to compare the traditional conditional generation network, the existing surrealism (mosaic) style migration method FAMOS (with reference) and the method of the present invention As can be seen from the comparison figure, the method of the present invention has a better performance in terms of surreal style transfer. Specific steps are as follows:

[0116] i. Generating style data sets according to step 1. In this embodiment, samples of two types of style map data are generated, such as Figure 14 with Figure 15 shown;

[0117] ii. Take α=0.4 and β=0.6 in the formula (6), and carry out training with a learning rat...

Embodiment 2

[0121] The method of the present invention can also be used for conventional style transfer tasks, and the specific implementation steps are as follows:

[0122] i. Prepare several images of the same style, and test four styles in the experiment, such as Figure 19 Shown:

[0123] a. A data set composed of several satellite images of Sydney obtained through Google Maps (the resolution of a single image is 1600*800);

[0124] b. Several paintings by Yayoi Kusama;

[0125] c. A Chinese landscape painting;

[0126] d. A green landscape painting style landscape painting;

[0127] ii. For styles (a), (c) and (d), this embodiment uses the (256, 256) slices of the original image; for Yayoi Kusama's painting (c), this embodiment only hopes to learn his strokes , and do not want to learn the content of the original painting, so it is necessary to properly enlarge the original image first, and then cut (256,256) slices from the enlarged image. The slices constitute the style map da...

Embodiment 3

[0132] This embodiment provides a computer-implemented system for migrating surrealist painting image styles based on deep learning, the system includes a processor and a memory, the memory stores a computer program, the processor is a GPU, and calls the computer program Steps 1 to 3 of the deep learning-based surrealist painting image style transfer method described in Example 1.

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 relates to a super-realistic drawing image style migration method based on deep learning. The method comprises the following steps: obtaining a style image training set and a content image training set; obtaining a convolutional neural network model based on the style image training set and the content image training set, the convolutional neural network model comprises a generator and a discriminator, performing down-sampling on an input picture by the generator, arranging a noise layer behind an encoder, and the discriminator is a Markov discriminator; and obtaining a real picture, calling the trained generator, and carrying out super-realistic subjective style conversion on the real picture. Compared with the prior art, the method has the advantages of high migration quality, good effect and the like.

Description

technical field [0001] The invention relates to an image style transfer method, in particular to a deep learning-based surreal painting image style transfer method. Background technique [0002] As a solid field in machine vision, style transfer has attracted extensive attention from industry and academia in recent years. Given an arbitrary content map and a specific style map, the task of style transfer is to generate an image with the style of the style map and the visual content of the content map. The mosaic primitive composition style is a surreal art form. How to use a certain number of image elements, such as a style atlas of flowers, fruits, etc., to recombine the input content map to generate a surreal (mosaic) style image is still an open problem. [0003] At present, the methods of style transfer mainly include three categories: methods based on Gram matrix, methods based on Markov random fields and methods based on generative confrontation networks. [0004] T...

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): G06T3/00G06T3/40G06T7/11G06N3/04G06K9/62
CPCG06T3/0012G06T3/4084G06T7/11G06T3/4007G06T2207/20081G06T2207/20084G06T2207/20016G06N3/045G06F18/214
Inventor 林澜杨怡汪澄
Owner TONGJI 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