Face image age conversion method based on gradient adversarial attack and generative adversarial model

A conversion method and image technology, applied in the field of image synthesis and translation, can solve the problems of face image age conversion technology that needs to be improved, face identity, posture background and illumination are not easy to maintain, and model training time is long.

Pending Publication Date: 2021-10-29
SOUTHEAST UNIV
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, the training time of the model is extremely long, and the features extracted during the training process designed solely for face age conversion are not representative enough. The images generated by large-span age conversion are not very realistic, and the face identity, posture, background and illumination in the age conversion process difficult to maintain
Therefore, the existing face image age conversion technology needs to be improved

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
  • Face image age conversion method based on gradient adversarial attack and generative adversarial model
  • Face image age conversion method based on gradient adversarial attack and generative adversarial model
  • Face image age conversion method based on gradient adversarial attack and generative adversarial model

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0044] Embodiment 1: see figure 1 , figure 2 , a face image age conversion method based on a gradient confrontation attack and a generation confrontation model, the method comprising the following steps:

[0045] Step 1: Use the face alignment method to cut out the aligned face image from the wild image;

[0046] Step 2: Use discriminator, decoder, face age estimation model and face recognition model;

[0047] Step 3: Use the average vector of the latent space of the decoder to initialize or use the updated latent space vector to decode the generated picture and then input it to the discriminator, face age estimation model, and face recognition model to obtain the relationship between the generated picture and the target age difference, the authenticity of generated pictures, and the face identity retention of generated pictures;

[0048] Step 4: The optimization target is obtained by weighting the target age difference, the authenticity of the generated image, and the fac...

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 a face image age conversion method based on gradient adversarial attack and a generative adversarial model, and the method mainly comprises the steps: firstly aligning face images, intercepting the face parts of the face images, obtaining an original image through the rotation of the images, adjusting the angle, and giving an expected target face age; initializing an implicit space vector, and inputting the implicit space vector into a decoder to obtain a generated picture; according to the algorithm, using an age estimator, a discriminator and a face recognizer to obtain the age and the trueness of a generated picture and the loss of the identity similarity between the generated picture and an original picture face; and with loss reduction as a target, carrying out gradient updating on the initialized hidden space vector, and then repeating the first two steps until the step length agreed by the algorithm is reached. By using the method, automatic conversion of face pictures can be realized, criminal investigation and cross-age face recognition are helped, and the invention can also be used for entertainment and movie and television.

Description

technical field [0001] The invention belongs to the field of image synthesis and translation, and in particular relates to an age conversion method of a human face image. Background technique [0002] As society enters the digital age, people's film and television entertainment activities are gradually enriched. The age conversion of face pictures by artificial intelligence methods can be used as video special effects and widely used in live broadcast platforms and social media. Many film and television works require the process of shooting characters from childhood to adulthood, and the traditional manual age conversion special effects production is expensive and time-consuming, and the use of artificial intelligence methods can greatly reduce costs. Criminal investigation often encounters situations where prisoners or missing persons have not appeared for decades. To face this problem, the traditional method uses experts to manually draw age-transformed portraits, which 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): G06K9/00G06K9/62G06N3/04G06N3/08
CPCG06N3/08G06N3/045G06F18/22
Inventor 杨绍枢永明
Owner SOUTHEAST 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