CN111612872B - Face age change image countermeasure generation method and system - Google Patents
Face age change image countermeasure generation method and system Download PDFInfo
- Publication number
- CN111612872B CN111612872B CN202010441089.0A CN202010441089A CN111612872B CN 111612872 B CN111612872 B CN 111612872B CN 202010441089 A CN202010441089 A CN 202010441089A CN 111612872 B CN111612872 B CN 111612872B
- Authority
- CN
- China
- Prior art keywords
- face
- image
- loss value
- real
- age
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T11/00—2D [Two Dimensional] image generation
- G06T11/60—Editing figures and text; Combining figures or text
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/168—Feature extraction; Face representation
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/178—Human faces, e.g. facial parts, sketches or expressions estimating age from face image; using age information for improving recognition
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computational Linguistics (AREA)
- Human Computer Interaction (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Multimedia (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Processing (AREA)
Abstract
The invention relates to a face age change image countermeasure generation method and a face age change image countermeasure generation system, wherein the generation method comprises the following steps: acquiring a plurality of pairs of real face images and target age feature vectors; a face generator based on an airspace attention mechanism obtains a composite image according to each pair of real face images and target age feature vectors; calculating a loss value of an image loss function according to each real face image and the corresponding composite image based on the face discriminator; iteratively adjusting weights of a face generator and a face discriminator by using a loss gradient back propagation algorithm according to the loss value until convergence; and obtaining the face image with the target age characteristic according to the face image to be processed and the corresponding target age characteristic vector based on the current face generator. The invention limits the change area of the input image through the generator through the airspace attention mechanism, and can reduce the possibility of pixel change in the image area irrelevant to age change, thereby reducing the probability of noise and distortion.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a face age-changing image countermeasure generation method and a face age-changing image countermeasure generation system.
Background
The face image after the age change is synthesized is an important branch of the image editing problem in the computer vision field, and the purpose of the face image is to generate a vivid face image with the appearance characteristics of the appointed age based on the given face image and the target age characteristic vector on the premise of keeping the face identity information unchanged.
With the wide application of deep learning theory, especially against the rapid development of the generation network (GENERATIVE ADVERSARIAL Networks, GANs), the existing face age-changing technology mostly uses a GANs-based model to synthesize realistic face images. Typically, age changes are focused only on local areas of the face (forehead, corners of eyes, mouth, etc.), and a reasonable age change model should only focus on pixel changes in these areas, while keeping the image information of other parts unchanged.
However, most existing models change the whole input image, so that the pixel values in the areas such as the image background and the like which are irrelevant to age change also change, thereby introducing noise and distortion and reducing the quality and fidelity of the generated image.
Disclosure of Invention
In order to solve the above problems in the prior art, that is, to improve the quality of the generated image, the present invention aims to provide a face age-varying image countermeasure generation method and system.
In order to solve the technical problems, the invention provides the following scheme:
A face age-changing image countermeasure generation method, the generation method comprising:
Acquiring a plurality of pairs of real face images and target age feature vectors;
A face generator based on an airspace attention mechanism obtains a composite image according to each pair of real face images and target age feature vectors;
Calculating a loss value of an image loss function according to each real face image and the corresponding composite image based on the face discriminator;
Iteratively adjusting weights of the face generator and the face discriminator by using a loss gradient back propagation algorithm according to the loss value until convergence to obtain a current face generator;
and obtaining the face image with the target age characteristic according to the face image to be processed and the corresponding target age characteristic vector based on the current face generator.
Optionally, the face generator based on the airspace attention mechanism obtains a composite image according to each pair of real face images and the target age feature vector, and specifically includes:
for each pair of real face images and target age feature vectors,
Splicing the real face image and the target age characteristic vector along the channel dimension to obtain a splicing result;
and a face generator based on an airspace attention mechanism obtains a corresponding synthetic image according to the splicing result.
Optionally, the real face image and the target age feature vector are divided into a real young face image and a target aging age feature vector, and a real old face image and a target aging age feature vector.
Optionally, the loss value of the image loss function includes:
the antagonism loss value, the age estimation loss value and the face reconstruction loss value of the real young face image, and the antagonism loss value, the age estimation loss value and the face reconstruction loss value of the real old face image.
Optionally, the loss value of the image loss function is calculated according to the following formula:
counter loss value of real-year old face image
Age estimation loss value of real-year old face image
Face reconstruction loss value of face image of real aged person
Counter loss value of true young face image
Age estimation loss value of real young face image
Face reconstruction loss value of true young face image
Wherein I y represents a true young face image, I o represents a true old face image, α y represents a target aging age feature vector, α o represents a target aging age feature vector, G r represents a face aging generator based on an airspace attention mechanism, G p represents a face aging generator based on an airspace attention mechanism, D r represents a young face discriminator, D p represents an old face discriminator, D r and D p superscript I represents a discrimination output for image authenticity, superscript α represents a regression output for a face age in an image, D represents a face discriminator, and G represents a face generator.
Optionally, iteratively adjusting weights of the face generator and the face discriminator by using a loss gradient back propagation algorithm according to the loss value until convergence, including:
the total loss value L is calculated according to the following formula:
Taking the total loss value L as an objective function according to Performing iterative optimization;
The weights of the generators G r and G p and the discriminators D r and D p are updated using a gradient back propagation algorithm until convergence.
In order to solve the technical problems, the invention also provides the following scheme:
a face age-changing image countermeasure generation system, the generation system comprising:
the acquisition unit is used for acquiring a plurality of pairs of real face images and target age characteristic vectors;
The synthesis unit is used for a face generator based on an airspace attention mechanism to obtain a synthesized image according to each pair of real face images and target age feature vectors;
The computing unit is used for computing a loss value of an image loss function according to each real face image and the corresponding composite image based on the face discriminator;
The iteration unit is used for iteratively adjusting the weights of the face generator and the face discriminator by using a loss gradient back propagation algorithm according to the loss value until convergence to obtain a current face generator;
The processing unit is used for obtaining the face image with the target age characteristic according to the face image to be processed and the corresponding target age characteristic vector based on the current face generator.
Optionally, the synthesizing unit includes:
The splicing module is used for splicing the real face image and the target age characteristic vector along the channel dimension aiming at each pair of real face image and the target age characteristic vector to obtain a splicing result;
And the synthesis module is used for obtaining a corresponding synthesized image according to the splicing result by a face generator based on an airspace attention mechanism.
In order to solve the technical problems, the invention also provides the following scheme:
a face age-changing image countermeasure generation system, comprising:
A processor; and
A memory arranged to store computer executable instructions that, when executed, cause the processor to:
Acquiring a plurality of pairs of real face images and target age feature vectors;
A face generator based on an airspace attention mechanism obtains a composite image according to each pair of real face images and target age feature vectors;
Calculating a loss value of an image loss function according to each real face image and the corresponding composite image based on the face discriminator;
Iteratively adjusting weights of the face generator and the face discriminator by using a loss gradient back propagation algorithm according to the loss value until convergence to obtain a current face generator;
and obtaining the face image with the target age characteristic according to the face image to be processed and the corresponding target age characteristic vector based on the current face generator.
In order to solve the technical problems, the invention also provides the following scheme:
A computer-readable storage medium storing one or more programs that, when executed by an electronic device comprising a plurality of application programs, cause the electronic device to:
Acquiring a plurality of pairs of real face images and target age feature vectors;
A face generator based on an airspace attention mechanism obtains a composite image according to each pair of real face images and target age feature vectors;
Calculating a loss value of an image loss function according to each real face image and the corresponding composite image based on the face discriminator;
Iteratively adjusting weights of the face generator and the face discriminator by using a loss gradient back propagation algorithm according to the loss value until convergence to obtain a current face generator;
and obtaining the face image with the target age characteristic according to the face image to be processed and the corresponding target age characteristic vector based on the current face generator.
According to the embodiment of the invention, the following technical effects are disclosed:
The invention limits the change area of the input image through the generator through the airspace attention mechanism, and can reduce the possibility of pixel change in the image area irrelevant to age change, thereby reducing the probability of noise and distortion introduction and improving the quality and reliability of the synthesized aging face image.
Drawings
FIG. 1 is a flow chart of a face age-variation image countermeasure generation method of the present invention;
FIG. 2 is a flowchart of an embodiment of the face age-variation image countermeasure generation method of the present invention;
FIG. 3 is a flow chart for the implementation of FIG. 2;
FIG. 4 is a block diagram of a generator based on airspace attention mechanisms;
fig. 5 is a schematic block diagram of the face age-varying image countermeasure generation system of the present invention.
Symbol description:
The system comprises an acquisition unit-1, a synthesis unit-2, a calculation unit-3, an iteration unit-4 and a processing unit-5.
Detailed Description
Preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are merely for explaining the technical principles of the present invention, and are not intended to limit the scope of the present invention.
The invention aims to provide a face age change image countermeasure generation method, which limits the change area of an input image through a generator through an airspace attention mechanism and can reduce the possibility of pixel change in an image area irrelevant to age change, thereby reducing the probability of noise and distortion introduction and improving the quality and reliability of a synthesized aging face image.
In order that the above-recited objects, features and advantages of the present invention will become more readily apparent, a more particular description of the invention will be rendered by reference to the appended drawings and appended detailed description.
As shown in fig. 1, the face age-varying image countermeasure generation method of the present invention includes:
step 100: acquiring a plurality of pairs of real face images and target age feature vectors;
Step 200: a face generator based on an airspace attention mechanism obtains a composite image according to each pair of real face images and target age feature vectors;
step 300: calculating a loss value of an image loss function according to each real face image and the corresponding composite image based on the face discriminator;
step 400: iteratively adjusting weights of the face generator and the face discriminator by using a loss gradient back propagation algorithm according to the loss value until convergence to obtain a current face generator;
Step 500: and obtaining the face image with the target age characteristic according to the face image to be processed and the corresponding target age characteristic vector based on the current face generator.
As shown in fig. 2 and fig. 3, the real face image and the target age feature vector are divided into a real young face image and a target aging age feature vector, and a real old face image and a target aging age feature vector.
Optionally, in step 200, the face generator based on the airspace attention mechanism obtains a composite image according to each pair of real face images and the target age feature vector, which specifically includes:
Step 201: aiming at each pair of real face images and target age feature vectors, splicing the real face images and the target age feature vectors along the channel dimension to obtain a splicing result;
Step 202: and a face generator based on an airspace attention mechanism obtains a corresponding synthetic image (shown in figure 4) according to the splicing result.
Specifically, the loss value of the image loss function includes: the antagonism loss value, the age estimation loss value and the face reconstruction loss value of the real young face image, and the antagonism loss value, the age estimation loss value and the face reconstruction loss value of the real old face image.
Further, a loss value of the image loss function is calculated according to the following formula:
counter loss value of real-year old face image
Age estimation loss value of real-year old face image
Face reconstruction loss value of face image of real aged person
Counter loss value of true young face image
Age estimation loss value of real young face image
Face reconstruction loss value of true young face image
Wherein I y represents a true young face image, I o represents a true old face image, α y represents a target aging age feature vector, α o represents a target aging age feature vector, G r represents a face aging generator based on an airspace attention mechanism, G p represents a face aging generator based on an airspace attention mechanism, D r represents a young face discriminator, D p represents an old face discriminator, D r and D p superscript I represents a discrimination output for image authenticity, superscript α represents a regression output for a face age in an image, D represents a face discriminator, and G represents a face generator.
In step 400, iteratively adjusting weights of the face generator and the face discriminator until convergence according to the loss value by using a loss gradient back propagation algorithm, including:
step 401: the total loss value L is calculated according to the following formula:
step 402: taking the total loss value L as an objective function according to Performing iterative optimization;
Step 403: the weights of the generators G r and G p and the discriminators D r and D p are updated using a gradient back propagation algorithm until convergence.
The invention uses GANs (GENERATIVE ADVERSARIAL Networks, anti-aging network) based model to generate the age-changing face image, the model mainly comprises three parts: generator, discriminator, loss function based on airspace attention mechanism. The generator based on the airspace attention mechanism comprises a face aging generator and a face rejuvenation generator, and a real young face image and a real old face image are taken as inputs respectively. The input image and the target age characteristic vector are spliced along the channel dimension and then input into a generator. The generator guides the image change region based on the airspace attention mechanism so that pixel values in the region irrelevant to the change can be maintained, and finally the age-changed face image is output.
In order to make the generated image as realistic as possible and to embody the appearance characteristics of the target age, we use two discriminators (an aged face discriminator and a young face discriminator) to distinguish the synthesized and true age-changed face images, and regress the ages of the face images input to the discriminators.
In order to supervise the training process of the model, the invention adopts the countermeasures, the age estimation loss and the face reconstruction loss to restrain the synthesized age-changing face image. Specifically, the contrast loss increases the fidelity of the synthesized face image by punishing the difference between the data distribution of the synthesized face and the real face; age estimation loss enables the generated face to have similar appearance characteristics to the real face of the same age; the face reconstruction errors reduce the variation of pixel values in image areas that are independent of age variation.
The invention is illustrated in more detail below by means of specific examples:
As shown in fig. 2 and 3, step S1, a real young face image (a young male face image is taken as an example in fig. 3) and a target aging age feature vector (the target age is 51-60 years old) are spliced and then input into a face aging generator based on an airspace attention mechanism, so as to obtain a combined adult face image; the face image of the real aged (taking an aged female face image as an example in fig. 3) and the target younger age feature vector (the target age 21-30 years) are spliced and then input into a face younger generator based on an airspace attention mechanism, so that a synthesized younger face image is obtained.
The step S1 specifically includes the following steps:
step S11, the real young face image and the target aging age feature vector are spliced along the channel dimension and then input into a face aging generator (the structure is shown in fig. 4) based on an airspace attention mechanism, and a combined adult face image is output.
Step S12, the face image of the real aged and the target younger age feature vector are spliced along the channel dimension and then input into a face younger generator (the structure is shown in fig. 4) based on an airspace attention mechanism, and the synthesized face image is output.
Step S2, sending the real and synthesized young face images into a young face discriminator, sending the real and synthesized old faces into an old face discriminator, respectively calculating the countermeasures, the age estimation loss and the face reconstruction loss, and iteratively adjusting the weights of the face aging generator, the face rejuvenation generator, the old face discriminator and the young face discriminator until convergence by using a loss gradient back propagation algorithm.
The step S2 specifically includes the following steps:
Step S21, based on the synthesized aged face image obtained in step S11, the synthesized aged face image and the real aged face image are input into an aged face discriminator, and the value of the loss function is calculated. The loss function is divided into three parts: counter loss value, age estimation loss value and face reconstruction loss value
And S22, calculating an antagonism loss value, an age estimation loss value and a face reconstruction loss value of the face image of the real old man.
Step S23, based on the synthesized young human face image obtained in the step S12, the synthesized young human face image and the true young human face image are input into a young human face discriminator, and the value of the loss function is calculated.
Step S24, calculating an antagonism loss value, an age estimation loss value and a face reconstruction loss value of the real young face image.
In addition, the invention also provides a face age-varying image countermeasure generating system which can improve the quality of the generated image.
As shown in fig. 5, the face age-variation image countermeasure generation system of the present invention includes an acquisition unit 1, a synthesis unit 2, a calculation unit 3, an iteration unit 4, and a processing unit 5.
The acquisition unit 1 is used for acquiring a plurality of pairs of real face images and target age characteristic vectors;
the synthesis unit 2 is used for a face generator based on an airspace attention mechanism to obtain a synthesized image according to each pair of real face images and target age feature vectors;
The calculating unit 3 is used for calculating a loss value of an image loss function according to each real face image and the corresponding synthesized image based on the face discriminator;
the iteration unit 4 is configured to iteratively adjust weights of the face generator and the face discriminator by using a loss gradient back propagation algorithm according to the loss value until convergence, so as to obtain a current face generator;
The processing unit 5 is configured to obtain a face image with a target age feature according to the face image to be processed and a corresponding target age feature vector based on the current face generator.
Preferably, the synthesizing unit 2 includes a splicing module and a synthesizing module.
The splicing module is used for splicing the real face image and the target age characteristic vector along the channel dimension aiming at each pair of the real face image and the target age characteristic vector to obtain a splicing result; and the synthesis module is used for a face generator based on an airspace attention mechanism, and obtaining a corresponding synthesized image according to the splicing result.
The invention also provides a face age-varying image countermeasure generation system, which comprises:
A processor; and
A memory arranged to store computer executable instructions that, when executed, cause the processor to:
Acquiring a plurality of pairs of real face images and target age feature vectors;
A face generator based on an airspace attention mechanism obtains a composite image according to each pair of real face images and target age feature vectors;
Calculating a loss value of an image loss function according to each real face image and the corresponding composite image based on the face discriminator;
Iteratively adjusting weights of the face generator and the face discriminator by using a loss gradient back propagation algorithm according to the loss value until convergence to obtain a current face generator;
and obtaining the face image with the target age characteristic according to the face image to be processed and the corresponding target age characteristic vector based on the current face generator.
In order to solve the technical problems, the invention also provides the following scheme:
The present invention also provides a computer-readable storage medium storing one or more programs that, when executed by an electronic device comprising a plurality of application programs, cause the electronic device to:
Acquiring a plurality of pairs of real face images and target age feature vectors;
A face generator based on an airspace attention mechanism obtains a composite image according to each pair of real face images and target age feature vectors;
Calculating a loss value of an image loss function according to each real face image and the corresponding composite image based on the face discriminator;
Iteratively adjusting weights of the face generator and the face discriminator by using a loss gradient back propagation algorithm according to the loss value until convergence to obtain a current face generator;
and obtaining the face image with the target age characteristic according to the face image to be processed and the corresponding target age characteristic vector based on the current face generator.
Compared with the prior art, the computer readable storage medium and the face age change image countermeasure generation system have the same beneficial effects as the face age change image countermeasure generation method, and are not repeated here.
Thus far, the technical solution of the present invention has been described in connection with the preferred embodiments shown in the drawings, but it is easily understood by those skilled in the art that the scope of protection of the present invention is not limited to these specific embodiments. Equivalent modifications and substitutions for related technical features may be made by those skilled in the art without departing from the principles of the present invention, and such modifications and substitutions will fall within the scope of the present invention.
Claims (7)
1. A face age-changing image countermeasure generation method, characterized by comprising:
Acquiring a plurality of pairs of real face images and target age feature vectors;
A face generator based on an airspace attention mechanism obtains a composite image according to each pair of real face images and target age feature vectors;
Calculating a loss value of an image loss function according to each real face image and the corresponding composite image based on the face discriminator;
Iteratively adjusting weights of the face generator and the face discriminator by using a loss gradient back propagation algorithm according to the loss value until convergence to obtain a current face generator;
based on the current face generator, obtaining a face image with target age characteristics according to the face image to be processed and the corresponding target age characteristic vector;
The real face image and the target age characteristic vector are divided into a real young face image and a target aging age characteristic vector, and a real old face image and a target aging age characteristic vector;
The loss value of the image loss function includes:
The antagonism loss value, the age estimation loss value and the face reconstruction loss value of the real young face image, and the antagonism loss value, the age estimation loss value and the face reconstruction loss value of the real old face image;
The loss value of the image loss function is calculated according to the following formula:
counter loss value of real-year old face image
Age estimation loss value of real-year old face image
Face reconstruction loss value of face image of real aged person
Counter loss value of true young face image
Age estimation loss value of real young face image
Face reconstruction loss value of true young face image
Wherein I y represents a true young face image, I o represents a true old face image, α y represents a target aging age feature vector, α o represents a target aging age feature vector, G r represents a face aging generator based on an airspace attention mechanism, G p represents a face aging generator based on an airspace attention mechanism, D r represents a young face discriminator, D p represents an old face discriminator, D r and D p superscript I represents a discrimination output for image authenticity, superscript α represents a regression output for a face age in an image, D represents a face discriminator, and G represents a face generator.
2. The face age-varying image countermeasure generation method according to claim 1, wherein the face generator based on the spatial domain attention mechanism obtains a composite image from each pair of the real face image and the target age feature vector, specifically comprising:
for each pair of real face images and target age feature vectors,
Splicing the real face image and the target age characteristic vector along the channel dimension to obtain a splicing result;
and a face generator based on an airspace attention mechanism obtains a corresponding synthetic image according to the splicing result.
3. The face age-varying image countermeasure generation method according to claim 2, wherein iteratively adjusting weights of the face generator and the face discriminator until convergence by using a loss gradient back propagation algorithm according to the loss value, specifically comprising:
the total loss value L is calculated according to the following formula:
Taking the total loss value L as an objective function according to Performing iterative optimization;
The weights of the generators G r and G p and the discriminators D r and D p are updated using a gradient back propagation algorithm until convergence.
4. A face age-changing image countermeasure generation system, the generation system comprising:
the acquisition unit is used for acquiring a plurality of pairs of real face images and target age characteristic vectors;
The synthesis unit is used for a face generator based on an airspace attention mechanism to obtain a synthesized image according to each pair of real face images and target age feature vectors; the real face image and the target age characteristic vector are divided into a real young face image and a target aging age characteristic vector, and a real old face image and a target aging age characteristic vector;
the computing unit is used for computing a loss value of an image loss function according to each real face image and the corresponding composite image based on the face discriminator; the loss value of the image loss function includes:
The antagonism loss value, the age estimation loss value and the face reconstruction loss value of the real young face image, and the antagonism loss value, the age estimation loss value and the face reconstruction loss value of the real old face image; the loss value of the image loss function is calculated according to the following formula:
counter loss value of real-year old face image
Age estimation loss value of real-year old face image
Face reconstruction loss value of face image of real aged person
Counter loss value of true young face image
Age estimation loss value of real young face image
Face reconstruction loss value of true young face image
Wherein I y represents a true young face image, I o represents a true old face image, α y represents a target aging age feature vector, α o represents a target aging age feature vector, G r represents a face aging generator based on an airspace attention mechanism, G p represents a face aging generator based on an airspace attention mechanism, D r represents a young face discriminator, D p represents an old face discriminator, D r and D p superscript I represents a discrimination output for image authenticity, superscript α represents a regression output for a face age in an image, D represents a face discriminator, and G represents a face generator;
The iteration unit is used for iteratively adjusting the weights of the face generator and the face discriminator by using a loss gradient back propagation algorithm according to the loss value until convergence to obtain a current face generator;
The processing unit is used for obtaining the face image with the target age characteristic according to the face image to be processed and the corresponding target age characteristic vector based on the current face generator.
5. The face age-varying image countermeasure generation system according to claim 4, wherein the synthesizing unit includes:
The splicing module is used for splicing the real face image and the target age characteristic vector along the channel dimension aiming at each pair of real face image and the target age characteristic vector to obtain a splicing result;
And the synthesis module is used for obtaining a corresponding synthesized image according to the splicing result by a face generator based on an airspace attention mechanism.
6. A face age-changing image countermeasure generation system, comprising:
A processor; and
A memory arranged to store computer executable instructions that, when executed, cause the processor to:
Acquiring a plurality of pairs of real face images and target age feature vectors;
A face generator based on an airspace attention mechanism obtains a composite image according to each pair of real face images and target age feature vectors; the real face image and the target age characteristic vector are divided into a real young face image and a target aging age characteristic vector, and a real old face image and a target aging age characteristic vector;
Calculating a loss value of an image loss function according to each real face image and the corresponding composite image based on the face discriminator; the loss value of the image loss function includes:
The antagonism loss value, the age estimation loss value and the face reconstruction loss value of the real young face image, and the antagonism loss value, the age estimation loss value and the face reconstruction loss value of the real old face image; the loss value of the image loss function is calculated according to the following formula:
counter loss value of real-year old face image
Age estimation loss value of real-year old face image
Face reconstruction loss value of face image of real aged person
Counter loss value of true young face image
Age estimation loss value of real young face image
Face reconstruction loss value of true young face image
Wherein I y represents a true young face image, I o represents a true old face image, α y represents a target aging age feature vector, α o represents a target aging age feature vector, G r represents a face aging generator based on an airspace attention mechanism, G p represents a face aging generator based on an airspace attention mechanism, D r represents a young face discriminator, D p represents an old face discriminator, D r and D p superscript I represents a discrimination output for image authenticity, superscript α represents a regression output for a face age in an image, D represents a face discriminator, and G represents a face generator;
Iteratively adjusting weights of the face generator and the face discriminator by using a loss gradient back propagation algorithm according to the loss value until convergence to obtain a current face generator;
and obtaining the face image with the target age characteristic according to the face image to be processed and the corresponding target age characteristic vector based on the current face generator.
7. A computer-readable storage medium storing one or more programs that, when executed by an electronic device comprising a plurality of application programs, cause the electronic device to:
Acquiring a plurality of pairs of real face images and target age feature vectors;
A face generator based on an airspace attention mechanism obtains a composite image according to each pair of real face images and target age feature vectors; the real face image and the target age characteristic vector are divided into a real young face image and a target aging age characteristic vector, and a real old face image and a target aging age characteristic vector;
Calculating a loss value of an image loss function according to each real face image and the corresponding composite image based on the face discriminator; the loss value of the image loss function includes:
The antagonism loss value, the age estimation loss value and the face reconstruction loss value of the real young face image, and the antagonism loss value, the age estimation loss value and the face reconstruction loss value of the real old face image; the loss value of the image loss function is calculated according to the following formula:
counter loss value of real-year old face image
Age estimation loss value of real-year old face image
Face reconstruction loss value of face image of real aged person
Counter loss value of true young face image
Age estimation loss value of real young face image
Face reconstruction loss value of true young face image
Wherein I y represents a true young face image, I o represents a true old face image, α y represents a target aging age feature vector, α o represents a target aging age feature vector, G r represents a face aging generator based on an airspace attention mechanism, G p represents a face aging generator based on an airspace attention mechanism, D r represents a young face discriminator, D p represents an old face discriminator, D r and D p superscript I represents a discrimination output for image authenticity, superscript α represents a regression output for a face age in an image, D represents a face discriminator, and G represents a face generator;
Iteratively adjusting weights of the face generator and the face discriminator by using a loss gradient back propagation algorithm according to the loss value until convergence to obtain a current face generator;
and obtaining the face image with the target age characteristic according to the face image to be processed and the corresponding target age characteristic vector based on the current face generator.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010441089.0A CN111612872B (en) | 2020-05-22 | 2020-05-22 | Face age change image countermeasure generation method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010441089.0A CN111612872B (en) | 2020-05-22 | 2020-05-22 | Face age change image countermeasure generation method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111612872A CN111612872A (en) | 2020-09-01 |
| CN111612872B true CN111612872B (en) | 2024-04-23 |
Family
ID=72199534
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010441089.0A Active CN111612872B (en) | 2020-05-22 | 2020-05-22 | Face age change image countermeasure generation method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111612872B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112287852B (en) * | 2020-11-02 | 2023-11-21 | 腾讯科技（深圳）有限公司 | Face image processing method, face image display method, face image processing device and face image display equipment |
| CN112766157B (en) * | 2021-01-20 | 2022-08-30 | 乐山师范学院 | Cross-age face image recognition method based on disentanglement representation learning |
| CN113436061B (en) * | 2021-07-01 | 2022-08-09 | 中科人工智能创新技术研究院(青岛)有限公司 | Face image reconstruction method and system |
| CN113569780B (en) * | 2021-08-03 | 2024-10-18 | 东南大学 | Face picture age conversion method based on gradient challenge and challenge model generation |
| CN113837930B (en) * | 2021-09-24 | 2024-02-02 | 重庆中科云从科技有限公司 | Face image synthesis method, device and computer readable storage medium |
| CN114022931B (en) * | 2021-10-29 | 2025-09-09 | 北京字节跳动网络技术有限公司 | Image processing method and device, electronic equipment and storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109308450A (en) * | 2018-08-08 | 2019-02-05 | 杰创智能科技股份有限公司 | A kind of face's variation prediction method based on generation confrontation network |
| CN109635745A (en) * | 2018-12-13 | 2019-04-16 | 广东工业大学 | A method of Multi-angle human face image is generated based on confrontation network model is generated |
| CN110322394A (en) * | 2019-06-18 | 2019-10-11 | 中国科学院自动化研究所 | Face age ageing image confrontation generation method and device based on attribute guidance |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10825219B2 (en) * | 2018-03-22 | 2020-11-03 | Northeastern University | Segmentation guided image generation with adversarial networks |
-
2020
- 2020-05-22 CN CN202010441089.0A patent/CN111612872B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109308450A (en) * | 2018-08-08 | 2019-02-05 | 杰创智能科技股份有限公司 | A kind of face's variation prediction method based on generation confrontation network |
| WO2020029356A1 (en) * | 2018-08-08 | 2020-02-13 | 杰创智能科技股份有限公司 | Method employing generative adversarial network for predicting face change |
| CN109635745A (en) * | 2018-12-13 | 2019-04-16 | 广东工业大学 | A method of Multi-angle human face image is generated based on confrontation network model is generated |
| CN110322394A (en) * | 2019-06-18 | 2019-10-11 | 中国科学院自动化研究所 | Face age ageing image confrontation generation method and device based on attribute guidance |
Non-Patent Citations (1)
| Title |
|---|
| Attribute Enhanced Face Aging with Wavelet-based Generative Adversarial Networks;Yunfan Liu et al;《arXiv》;第1-5页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111612872A (en) | 2020-09-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111612872B (en) | Face age change image countermeasure generation method and system | |
| CN111932444B (en) | Face attribute editing method based on generation countermeasure network and information processing terminal | |
| US12148129B2 (en) | Image dehazing method and system based on cyclegan | |
| CN113658051A (en) | A method and system for image dehazing based on recurrent generative adversarial network | |
| CN113723295A (en) | Face counterfeiting detection method based on image domain frequency domain double-flow network | |
| CN115409937A (en) | Method for building face video expression transfer model and expression transfer method and system based on integrated neural radiation field | |
| CN114936976B (en) | Memory perception module-based method for restoring image generated against network haze | |
| CN108171762A (en) | System and method for is reconfigured quickly in a kind of similar image of the compressed sensing of deep learning | |
| Cheng et al. | Towards pose-invariant lip-reading | |
| CN115914505A (en) | Video generation method and system based on voice-driven digital human model | |
| Liu et al. | Learning Hadamard-product-propagation for image dehazing and beyond | |
| WO2022166797A1 (en) | Image generation model training method, generation method, apparatus, and device | |
| Shankar et al. | Non-parallel emotion conversion using a deep-generative hybrid network and an adversarial pair discriminator | |
| CN112419455A (en) | Human action video generation method, system and storage medium based on human skeleton sequence information | |
| CN118568658B (en) | Multi-mode feature learning efficiency optimization method based on low-rank factor decomposition | |
| Chen et al. | Dense voxel 3d reconstruction using a monocular event camera | |
| CN118247180A (en) | Face image restoration method and system based on manifold constraint diffusion model | |
| CN111340173A (en) | Method and system for training generation countermeasure network for high-dimensional data and electronic equipment | |
| DE102021124428A1 (en) | TRAIN ENERGY-BASED VARIATIONAL AUTOENCODERS | |
| Li et al. | Facial attribute editing by latent space adversarial variational autoencoders | |
| CN112529772A (en) | Unsupervised image conversion method under zero sample setting | |
| CN116312922B (en) | A medical image diagnostic report steganography method based on convergent generative adversarial networks | |
| CN115908662B (en) | Speaker video generation model training and using method, device and equipment | |
| CN118334471A (en) | A medical image conversion method, system and device based on deep learning | |
| Orozco et al. | Refining amortized posterior approximations using gradient-based summary statistics |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |