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

Image fusion generation type face changing method based on face reconstruction

A face image and image fusion technology, which is applied in the field of face synthesis in computer vision, can solve the problems of face-changing texture mismatch, etc., to solve the mismatch of facial features, improve similarity and realism, and improve similarity and realism Effect

Active Publication Date: 2021-08-10
ZHEJIANG UNIV
View PDF8 Cites 8 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0006] In order to solve the problems existing in the technical background, the present invention proposes an image fusion generative face-changing method based on face reconstruction, which realizes the complementary advantages of the face-changing method based on 3D face reconstruction and generative confrontation network, and effectively solves the problem of The problem of face-changing texture mismatch in the field of face synthesis effectively improves the similarity and realism of face-changing

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 fusion generation type face changing method based on face reconstruction
  • Image fusion generation type face changing method based on face reconstruction
  • Image fusion generation type face changing method based on face reconstruction

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0033] The present invention is applicable to most face-changing scenarios, and the specific use of the present invention will be described in a clear, detailed and complete manner below.

[0034] The present invention uses 300W-LP as the main training data. The 300W-LP data set is obtained by distorting, deforming and flipping the faces of the 300W data set. CelebAMask-HQ is also used as the training data. CelebAMask-HQ contains 30,000 face images , and each image has an area attribute segmentation mask corresponding to the CelebA dataset. In the present invention, 1000 images are sampled from the CelebAMask-HQ data set for testing, and at the same time, in order to verify the versatility of the model, 1000 images are sampled from the Seeprettyface data set for testing.

[0035]For the 300W-LP data set containing 68 3D face key points, the 300W-LP data set is preprocessed, specifically the face alignment is carried out by the MTCNN method, and the face segmentation is carried...

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 image fusion generation type face changing method based on face reconstruction. The method comprises the following steps: 1) selecting a plurality of face images for preprocessing, obtaining a plurality of preprocessed face images, and forming a training set; 2) constructing a generative face replacement network; 3) inputting the training set into the generative face replacement network for training until the generative face replacement network converges, and obtaining a trained generative face replacement network; and 4) inputting a source face image to be replaced and a target face image into the trained generative face replacement network, and outputting to obtain a face replacement image. According to the method, the problem of texture mismatching in the face changing process is effectively solved, skin texture refined generation of three-dimensional face reconstruction is realized, so that the similarity and reality of face changing are effectively improved, and a high-quality face changing image is obtained.

Description

technical field [0001] The invention relates to a neural network-based face-changing method in the field of human face synthesis of computer vision, in particular to an image fusion generative face-changing method based on human face reconstruction. Background technique [0002] As a medium of identity information, face images are widely used in industries such as medical care, education, science, and culture. This also increases the risk of personal privacy leakage. The method of replacing the face in the image to protect the privacy and security of the people came into being. In computer vision, this method is known as face replacement. In addition, face replacement is often used to automate the replacement of actors in film and television dramas, liberating labor and quickly producing videos. To sum up, face replacement is of great significance to protecting the privacy and security of the public and promoting the development of the film and television industry. [00...

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): G06T17/00G06T15/00G06T5/50G06N3/04
CPCG06T17/00G06T15/005G06T5/50G06T2207/20081G06T2207/20084G06T2207/30201G06T2207/20221G06N3/045
Inventor 朱建科俞境心林利翔
Owner ZHEJIANG UNIV
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