Face anonymization method, system and device with controllability and medium

Abstract

The invention discloses a controllable face anonymization method, system and device and a medium, and particularly relates to the technical field of face recognition, and the technical key points are that the method comprises the steps: recognizing the face attribute features of an original face image through an attribute recognition model, performing modification splicing processing on the face attribute features according to a preset privacy protection intensity sequence to obtain a processed face attribute prediction vector; inputting the face attribute prediction vector into an identity mapping model to obtain a face identity feature vector, and inputting the face identity feature vector and the original face image into a face changing model for processing to obtain a face anonymization template image; constructing an availability vector representing the availability of the original face image, and converting the availability vector into an availability condition through a condition encoder network; and embedding the original face image into the face anonymization template image by using a reversible neural network, and controlling the reversible neural network through availability conditions to generate a final target anonymization image.

Description

Technical Field

[0001] The present invention relates to the field of face recognition technology, and in particular to a controllable face anonymization method, system, device and medium. Background Art

[0002] As an important biometric identification technology for identity authentication, face recognition has been widely used in many fields such as military, finance, public security and daily life. However, while people enjoy the convenience brought by face recognition, they also face the risk of personal privacy information being leaked. Because the rich personal information in face images may be leaked or abused during the face recognition process, and face images are a visual media that can directly expose personal identity information. Once leaked, they cannot be recovered. Therefore, it is very necessary to protect the privacy of faces.

[0003] In the technical field of face privacy protection, face anonymization is one of the main protection methods. Its core idea is to use technical means such as de-identification to eliminate or hide identity-sensitive information in images. Face anonymization can be divided into anonymization methods for human vision and anonymization methods for machine vision according to the objects it is facing. For example, researchers such as Hukkelas from the Norwegian University of Science and Technology proposed a model based on conditional generative adversarial networks, which generates anonymous face images that can retain the posture and background information of the original image; and researchers such as Li from the Department of Computer Science at Purdue University proposed a face de-identification method that measures the degree of anonymization. The method includes four stages: face attribute estimation, face confusion guided by privacy metrics, directional natural images, and adversarial perturbations. By using the GoogLeNet network to extract facial features and using the privacy-preserving attribute selection algorithm PPAS (Privacy-Preserving Attribute Selection), this method can improve the usability of anonymization and enhance privacy. When generating real and natural face images, the team used the StarGAN model and added attribute classification loss and image reconstruction loss to constrain the quality of anonymous images.

[0004] Anonymization methods for machine vision, such as: A research team from the Aristotle University of Thessaloniki proposed a new method based on adversarial sample generation to counter face recognition systems. This method is called the penalized fast gradient value method, which aims to deceive face recognition systems with minimal distortion, protect identity privacy, and maintain consistency of visual content before and after anonymization; You et al. proposed a reversible privacy-preserving recognition framework for the first time, which converts the original image into a pixelated form, which can then be restored through different network structures.

[0005] The team from Shanghai Jiao Tong University proposed a recoverable privacy protection method based on a deep generative model. Drawing on the idea of ​​feature deconstruction of the IdentityDP model, this method retains the detailed attribute features of anonymous images such as expression, posture, and lighting by decoupling identity features and attribute features, while maintaining the consistency of the identity of the de-anonymized image with the original image. The network structure includes an encoder, a generator, and an identity modification module. The encoder extracts the identity features and attribute features of the face, the identity modification module calculates the protected identity features, and finally the generator generates a de-identification result based on the identity features and attribute features.

[0006] Currently, most face anonymization methods do not fully consider the actual application scenarios of users. Basically, one method corresponds to a single protection (AI, human vision), and the anonymization result cannot be controlled, which greatly limits the user's use.

[0007] Therefore, the present invention aims to provide a controllable face anonymization method, system, device and medium to solve the above-mentioned related problems. Summary of the invention

[0008] The technical problem to be solved by the present invention is that most of the current face anonymization methods do not fully take into account the actual application scenarios of users. Basically, one method corresponds to a single protection, and the anonymization result cannot be controlled, which greatly limits the use of users. The purpose is to provide a controllable face anonymization method, system, device and medium. By setting different levels of privacy protection strength, identity mapping model and face replacement model, a face anonymization template image corresponding to the privacy protection strength can be obtained. The degree of protection of the image for human visual recognition increases with the increase of protection strength, thereby realizing dynamically adjustable anonymization for human vision; at the same time, in order to improve the availability of face anonymous template images By using the availability conditional coding network, the target anonymized image finally generated can choose whether to retain the identity features of the original image, thereby realizing dynamically adjustable anonymization for machine vision; at the same time, the semantic information of the original face image is embedded into the face anonymization template image through a reversible neural network, so that the target anonymized image finally generated can also be restored to the original face image with the help of a reversible neural network, realizing dynamically restorable anonymization, thereby solving the problem that most of the current face anonymization methods do not fully consider the actual application scenarios of users, and basically one method corresponds to a single protection, and the anonymization result cannot be controlled, which brings great restrictions to user use.

[0009] The present invention is achieved through the following technical solutions:

[0010] A controllable face anonymization method, the method comprising:

[0011] The facial attribute features of the original facial image are identified by using the attribute recognition model, and the facial attribute features are modified and spliced ​​according to a preset privacy protection strength sequence to obtain a processed facial attribute prediction vector;

[0012] The facial attribute prediction vector is input into the identity mapping model to obtain the facial identity feature vector, and then the facial identity feature vector and the original facial image are input into the face swap model to obtain the facial anonymization template image;

[0013] Constructing an availability vector that represents the availability of the original face image, and converting the availability vector into availability conditions through a conditional encoder network;

[0014] The original face image is embedded into the face anonymization template image using a reversible neural network, and the reversible neural network is controlled by availability conditions to generate the final target anonymized image.

[0015] Furthermore, the availability vector adopts a one-hot vector, and the availability vector includes an available vector and an unavailable vector. The reversible neural network adopts a flow-based generation model, and a set of bidirectional mapping functions are used in the reversible neural network to realize reversible transformation between input and output, and the reversible change process is controlled by availability conditions.

[0016] Furthermore, the attribute recognition model includes an age recognition model and a gender recognition model. Both the age recognition model and the gender recognition model are obtained by training a binary classification model, and the facial attribute features include age features and gender features; the identity mapping model is obtained by training a fully connected neural network.

[0017] Furthermore, the attribute recognition model is used to extract the facial attribute features of the original facial image, and the facial attribute features are modified and spliced ​​according to the preset privacy protection strength sequence to obtain the processed attribute prediction vector, which is specifically:

[0018] The original face image is input into the age recognition model and the gender recognition model respectively to obtain the age feature and the gender feature;

[0019] Setting a privacy protection strength sequence according to the facial attribute features, wherein the privacy protection strength sequence includes at least one level of privacy protection strength;

[0020] The age feature and gender feature are modified and concatenated according to the privacy protection strength sequence to obtain a processed attribute prediction vector.

[0021] Furthermore, the preset privacy protection strength sequence includes four levels of privacy protection strength.

[0022] The present invention also includes a controllable face anonymization system, which is used in any of the controllable face anonymization methods described above, and the system includes:

[0023] A prediction vector generation module is used to identify the facial attribute features of the original facial image using the attribute recognition model, and to modify and splice the facial attribute features according to a preset privacy protection strength sequence to obtain a processed facial attribute prediction vector;

[0024] The template image generation module is used to input the face attribute prediction vector into the identity mapping model to obtain the face identity feature vector, and then input the face identity feature vector and the original face image into the face replacement model to obtain the face anonymization template image;

[0025] Availability condition generation module, used to construct an availability vector representing the availability of the original face image, and convert the availability vector into availability conditions through a conditional encoder network;

[0026] The target image generation module is used to embed the original face image into the face anonymization template image by using a reversible neural network, and control the reversible neural network through availability conditions to generate a final target anonymization image.

[0027] Furthermore, the availability vector adopts a one-hot vector, and the availability vector includes an available vector and an unavailable vector. The reversible neural network adopts a flow-based generation model, and a set of bidirectional mapping functions are used in the reversible neural network to realize reversible transformation between input and output, and the reversible change process is controlled by availability conditions.

[0028] The present invention also includes a computer device, including a system memory and a processor, wherein the system memory stores a computer program, and the processor implements the steps of any one of the above-mentioned methods when executing the computer program.

[0029] The present invention also includes a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of any one of the methods described above are implemented.

[0030] The present invention also includes a computer program product comprising instructions. When the instructions are executed by a computer device cluster, the computer device cluster executes any of the above methods.

[0031] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0032] In the present invention, by setting different levels of privacy protection strength, identity mapping model and face-changing model, a face anonymization template image corresponding to the privacy protection strength can be obtained, and the degree of protection of the image for human visual recognition increases with the increase of protection strength, thereby realizing dynamically adjustable anonymization for human vision; at the same time, in order to improve the availability of the face anonymous template image, by using an availability conditional coding network, the target anonymization image finally generated can choose whether to retain the identity features of the original image, thereby realizing dynamically adjustable anonymization for machine vision; at the same time, the semantic information of the original face image is embedded into the face anonymization template image through a reversible neural network, so that the target anonymization image finally generated can also be restored to the original face image with the help of a reversible neural network, thereby realizing dynamically restorable anonymization, thereby solving the problem that most of the current face anonymization methods do not fully consider the actual application scenarios of users, basically one method corresponds to a single protection, and the anonymization result cannot be controlled, which brings great restrictions to the use of users. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, other relevant drawings can be obtained based on these drawings without creative work. In the drawings:

[0034] Figure 1 This is a method flow chart of a controllable face anonymization method in this embodiment;

[0035] Figure 2 Schematic diagram of a controllable face anonymization platform in this embodiment;

[0036] Figure 3 Schematic diagram of the structure of a controllable face anonymization system in this embodiment;

[0037] Figure 4 It is a structural schematic diagram of a computer device in this embodiment. DETAILED DESCRIPTION

[0038] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope of the present disclosure. Similarly, for the sake of clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0039] In the present disclosure, unless otherwise specified, the use of the terms "first", "second", etc. to describe various elements is not intended to limit the positional relationship, timing relationship, or importance relationship of these elements, and such terms are only used to distinguish one element from another element. In some examples, the first element and the second element may refer to the same instance of the element, and in some cases, based on the description of the context, they may also refer to different instances.

[0040] The terms used in the description of various examples in this disclosure are only for the purpose of describing specific examples and are not intended to be limiting. Unless the context clearly indicates otherwise, if the number of elements is not specifically limited, the element can be one or more. In addition, the term "and / or" used in this disclosure covers any one of the listed items and all possible combinations.

[0041] Example 1

[0042] See also Figure 1 As shown, this embodiment provides a controllable face anonymization method, the method comprising:

[0043] S1: Use the attribute recognition model to identify the facial attribute features of the original facial image, and modify and splice the facial attribute features according to a preset privacy protection strength sequence to obtain a processed facial attribute prediction vector;

[0044] Among them, the attribute recognition model includes an age recognition model and a gender recognition model. Both the age recognition model and the gender recognition model are obtained by training a deep neural network with good generalization to ensure the accuracy and availability of the face attribute prediction vector. The face attribute features include age features and gender features, which are as follows:

[0045] Acquire the original face image through an image acquisition device, wherein the image acquisition device may be a camera, a still camera or other device capable of acquiring face images, and no further restrictions are imposed here;

[0046] The original face image is input into the age recognition model and the gender recognition model respectively to obtain the age feature and the gender feature;

[0047] A privacy protection strength sequence is set according to the facial attribute characteristics, wherein the privacy protection strength sequence includes four levels of privacy protection strength;

[0048] It should be noted that, in this embodiment, the privacy protection strength is set according to the facial attribute characteristics (age and gender) based on human vision, and the four levels of privacy protection strength set specifically include: I: same gender and age, II: same gender but different ages, III: different genders and same age, IV: different genders but different ages.

[0049] The age feature and gender feature are modified and concatenated according to the privacy protection strength sequence to obtain a processed attribute prediction vector.

[0050] S2: Input the facial attribute prediction vector into the identity mapping model to obtain the facial identity feature vector, and then input the facial identity feature vector and the original facial image into the face swap model to obtain the facial anonymization template image;

[0051] It should be noted that, in this embodiment, the identity mapping model is obtained by training a fully connected neural network, and the identity mapping model is used to map the facial attribute prediction vector into a corresponding facial identity feature vector, which can be effectively used for the identity in face changing; the face changing model adopts the Simswap face changing program.

[0052] S3: construct an availability vector that represents the availability of the original face image, and convert the availability vector into availability conditions through a conditional encoder network;

[0053] It should be noted that, in this embodiment, the availability of the original face image refers to whether the image utility is retained. The availability vector adopts a two-dimensional one-hot vector, and the availability vector includes an available vector and an unavailable vector. A conditional encoder network is used to expand the dimension, which affects whether the target anonymized image finally generated retains the facial identity features of the original image, thereby achieving facial privacy protection for machine vision.

[0054] S4: The original face image is embedded into the face anonymization template image using a reversible neural network, and the reversible neural network is controlled by availability conditions to generate the final target anonymized image.

[0055] It should be noted that, in the present embodiment, the reversible neural network adopts a flow-based generative model, and uses a set of bidirectional mapping functions in the reversible neural network to realize reversible transformation between input and output, and the reversible change process is controlled by availability conditions. The construction process of the reversible neural network is specifically as follows: a reversible block is formed by alternately coupling multiple arbitrary networks, and the input and output of the reversible block are divided into two parts. Multiple reversible blocks are combined to construct a reversible neural network for identity transformation; the reversible neural network composed of multiple reversible blocks performs reversible transformation between the original features of the input and the new features of the output through a set of bidirectional mapping functions.

[0056] Specifically, in this embodiment, by setting different levels of privacy protection strength, identity mapping model and face-changing model, a face anonymization template image of corresponding privacy protection strength can be obtained, and the degree of protection of the image for human visual recognition increases with the increase of protection strength, thereby realizing dynamically adjustable anonymization for human vision; at the same time, in order to improve the availability of the face anonymous template image, by using the availability conditional coding network, the target anonymization image finally generated can choose whether to retain the identity features of the original image, thereby realizing dynamically adjustable anonymization for machine vision; at the same time, the semantic information of the original face image is embedded into the face anonymization template image through the reversible neural network, so that the target anonymization image finally generated can also be restored to the original face image with the help of the reversible neural network, thereby realizing dynamically restorable anonymization;

[0057] Therefore, the face anonymization method provided by the present invention has multiple privacy protection characteristics (reversibility, availability, diversity, etc.), and can be flexibly used in different real-life scenarios, thereby solving the problem that most current face anonymization methods do not fully take into account the actual application scenarios of users, and basically one method corresponds to a single protection, and the anonymization result cannot be controlled, which greatly restricts the user's use.

[0058] In another embodiment, the identity mapping model is trained using the idea of ​​adversarial training, the identity mapping model is used as a generator, a binary recognition model for distinguishing true and false identity features is constructed as a first discriminator, the attribute recognition model is used as a second discriminator, and a loss function is used to update the parameters of the identity mapping model. First, the existing image feature set is input into the binary recognition model to optimize the true and false identity feature identification ability of the binary recognition model. The existing image feature set includes real face images and face identity features of real face images. The first loss function used is specifically the loss between the face attribute prediction vector extracted by the attribute recognition model and the face identity feature vector. The identity mapping model is updated using the first loss function. The second loss function used is specifically the loss between the face identity features of the real face image and the face identity features generated by the generator. The binary recognition model is updated using the second loss function, specifically:

[0059] The expression of the first loss function is specifically:

[0060]

[0061] in, represents the total loss function of the generator; The adversarial loss is used to measure the ability of the identity feature vector generated by the identity mapping model to "fool" the discriminator; is the L1-norm loss function, which is used to measure the similarity between the attribute prediction vector and the identity feature vector; 1 , 2 They are The weight of and They are:

[0062]

[0063]

[0064] z represents a random face attribute prediction vector sampled from the latent distribution; G(z) represents the face identity feature vector generated by the generator based on the random face attribute prediction vector z; D(G(z)) represents the score of the discriminator on the generated face identity vector G(z); ‖·,·‖ 1 represents the L1 distance between two facial attribute features, F(∶)=E()+M() is the concatenated facial attribute prediction vector; E() represents the pre-trained facial age recognition model; M() represents the pre-trained facial gender recognition model.

[0065] The expression of the second loss function is specifically:

[0066]

[0067] in, Represents the total loss function of the binary recognition model; is the binary cross entropy loss, The loss of the binary recognition model to identify the true face identity features, The loss of generating face identity features for binary recognition model recognition, where and They are:

[0068]

[0069]

[0070] Where x represents a real face image; Q(x) represents the facial identity features extracted from the real face image x using a public face recognition model (FaceNet model); D(Q(x)) represents the score of the real face identity vector given by the binary recognition model; z represents a random face attribute prediction vector sampled from the latent distribution; G(z) represents the face identity vector generated by the generator based on the random face attribute prediction vector z; D(G(z)) represents the score of the generated face identity vector G(z) given by the discriminator.

[0071] In another embodiment, in order to ensure the similarity between the target anonymized image and the face anonymization template image finally generated, and the control effect of the availability vector, a third loss function is used to update the parameters of the reversible neural network and the encoder network. The third loss function at least includes: the perceptual loss between the generated target anonymized image and the face anonymization template image, the face identity information loss between the target anonymized image and the original face image under different availability vectors, and the reconstruction loss between the face image restored by the target anonymized image and the original face image, specifically:

[0072] The third loss function:

[0073]

[0074] in, represents the total loss function; The perceptual loss between the generated target anonymized image and the face anonymized template image; Represents the face identity information loss between the target anonymized image and the original face image under different availability vectors; represents the reconstruction loss between the face image restored from the target anonymized image and the original face image; 1 , 2 , 3 They are The weight of Specifically:

[0075]

[0076] Among them, X proc represents the input face anonymization template image, X out represents the target anonymized image; It is used to measure the similarity between the face anonymization template image and the target anonymization image in visual perception, where: and Specifically:

[0077]

[0078]

[0079] Where f() represents the features extracted from the pre-trained deep neural network; f(X out ) is the feature of the target anonymized image in the deep network; f(X proc ) is the feature of the face anonymization template image in the deep network; ‖·,·‖ 2 Used to measure the difference between two features; It is a traditional Euclidean loss function used to measure the pixel-level difference between the template image and the output image;

[0080] Among them, under the condition of preserving availability, the face identity information loss between the target anonymized image and the face anonymized template image is It is expressed as:

[0081]

[0082] Face identity information loss between target anonymized image and face anonymized template image under the condition of removing availability It is expressed as:

[0083]

[0084] Among them, X orig represents the original face image; F() represents the existing face feature extraction model; is the feature difference loss, Specifically:

[0085]

[0086] Among them, F 1 、F 2 Feature representation of the labeled anonymized image and the face anonymized template image;

[0087]

[0088] Among them, X rec Represents the face image restored by the target anonymized image through the reversible neural network; the loss function is used to measure the difference between the restored face image and the original face image, and L1-norm is used as the metric to ensure that the restored face image is as consistent as possible with the original face image to ensure the reversibility of the face image.

[0089] Example 2

[0090] See also Figure 2 As shown, this embodiment also provides a controllable face anonymization system application platform, which includes an image acquisition end, a server end and a user end. The image acquisition end is used to acquire original face images, the server end is used to embed a controllable face anonymization system to complete the face image anonymization processing, and the user end is used to display the target anonymized image; see Figure 3 As shown, a controllable face anonymization system provided by this embodiment is shown, and the system includes:

[0091] The prediction vector generation module 100 is used to identify the facial attribute features of the original facial image using the attribute recognition model, and modify and splice the facial attribute features according to a preset privacy protection strength sequence to obtain a processed facial attribute prediction vector;

[0092] The template image generation module 200 is used to input the face attribute prediction vector into the identity mapping model to obtain the face identity feature vector, and then input the face identity feature vector and the original face image into the face replacement model to obtain the face anonymization template image;

[0093] Availability condition generation module 300, used to construct an availability vector representing the availability of the original face image, and convert the availability vector into an availability condition through a conditional encoder network;

[0094] The target image generation module 400 is used to embed the original face image into the face anonymization template image using a reversible neural network, and control the reversible neural network through availability conditions to generate a final target anonymization image.

[0095] Furthermore, the availability vector adopts a one-hot vector, and the availability vector includes an available vector and an unavailable vector. The reversible neural network adopts a flow-based generation model, and a set of bidirectional mapping functions are used in the reversible neural network to realize reversible transformation between input and output, and the reversible change process is controlled by availability conditions.

[0096] It should be noted that the modules in the system of Example 2 correspond to the steps in the method of Example 1. The steps in the method of Example 1 have been described in detail in Example 1. In this Example 2, the contents of the modules in the system will not be described in detail.

[0097] Example 3

[0098] See also Figure 4 As shown, this embodiment further provides a computer device, including a system memory 1005 and a processor 1001, wherein the system memory 1005 stores a computer program, and the processor 1001 implements the steps of any of the above methods when executing the computer program.

[0099] It should be noted that the processor 1001 is used to execute the steps in the above method embodiments according to the instructions in the program code. Alternatively, the processor 1001 implements the functions of each module / unit in the above system / device embodiments when executing the computer program.

[0100] Specifically, in this embodiment, the computer program may be divided into one or more modules / units, one or more modules / units are stored in the system memory 1005, and are executed by the processor 1001 to complete the present application. One or more modules / units may be a series of computer program instruction segments that can complete specific functions, and the instruction segments are used to describe the execution process of the computer program in the terminal device.

[0101] The terminal device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The terminal device may include, but is not limited to, a processor 1001 and a system memory 1005. Those skilled in the art will appreciate that this does not constitute a limitation on the terminal device, and may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the terminal device may also include an input / output device 1003, a network access device 1002, a bus 1006, etc.

[0102] The processor 1001 may be a central processing unit (CPU), or other general-purpose processors 1001, digital signal processors 1001 (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor 1001 may be a microprocessor or any conventional processor, etc.

[0103] The system memory 1005 may be an internal storage unit of the terminal device, such as a hard disk or memory of the terminal device. The system memory 1005 may also be a storage device 1004 of the terminal device, such as a plug-in hard disk, a smart memory card (SmartMedia Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), etc. equipped on the terminal device. Further, the system memory 1005 may also include both the internal storage unit of the terminal device and the storage device 1004. The system memory 1005 is used to store computer programs and other programs and data required by the terminal device. The system memory 1005 may also be used to temporarily store data that has been output or is to be output.

[0104] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0105] Example 4

[0106] This embodiment provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of any one of the above methods are implemented.

[0107] Among them, the computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared or semiconductor system, system or device, or any combination of the above. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection with one or more wires, a portable computer disk, a hard disk. Random Access Memory (RAM), Read-Only Memory (ROM), Erasable Programmable Read Only Memory (EPROM), registers, hard disks, optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above, or any other form of computer-readable storage medium known in the art.

[0108] An exemplary storage medium is coupled to a processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can be located in an application-specific integrated circuit (ASIC). In an embodiment of the present invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system, or device.

[0109] Example 5

[0110] This embodiment also provides a computer program product including instructions. When the instructions are executed by a computer device cluster, the computer device cluster executes the method described in Embodiment 1.

[0111] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A controllable face anonymization method, characterized in that: Methods include: The facial attribute features of the original facial image are identified by using the attribute recognition model, and the facial attribute features are modified and spliced ​​according to a preset privacy protection strength sequence to obtain a processed facial attribute prediction vector; The facial attribute prediction vector is input into the identity mapping model to obtain the facial identity feature vector, and then the facial identity feature vector and the original facial image are input into the face swap model to obtain the facial anonymization template image; Constructing an availability vector that represents the availability of the original face image, and converting the availability vector into availability conditions through a conditional encoder network; The original face image is embedded into the face anonymization template image using a reversible neural network, and the reversible neural network is controlled by availability conditions to generate the final target anonymized image.

2. A controllable face anonymization method according to claim 1, characterized in that: The availability vector adopts a one-hot vector, which includes an available vector and an unavailable vector. The reversible neural network adopts a flow-based generation model, in which a set of bidirectional mapping functions are used to realize reversible transformation between input and output, and the reversible change process is controlled by availability conditions.

3. A controllable face anonymization method according to claim 1, characterized in that: The attribute recognition model includes an age recognition model and a gender recognition model. Both the age recognition model and the gender recognition model are obtained by training a binary classification model. The facial attribute features include age features and gender features. The identity mapping model is obtained by training a fully connected neural network.

4. A controllable face anonymization method according to claim 3, characterized in that: The attribute recognition model is used to extract the facial attribute features of the original face image, and the facial attribute features are modified and spliced ​​according to the preset privacy protection strength sequence to obtain the processed attribute prediction vector, which is specifically: Input the original face image into the age recognition model and the gender recognition model respectively to obtain age features and gender features; Setting a privacy protection strength sequence according to the facial attribute features, wherein the privacy protection strength sequence includes at least one level of privacy protection strength; The age feature and gender feature are modified and concatenated according to the privacy protection strength sequence to obtain a processed attribute prediction vector.

5. A controllable face anonymization method according to claim 4, characterized in that: The preset privacy protection strength sequence includes four levels of privacy protection strength.

6. A controllable face anonymization system, characterized in that: The system is used in a controllable face anonymization method as described in any one of claims 1 to 5, and the system comprises: A prediction vector generation module is used to identify the facial attribute features of the original facial image using the attribute recognition model, and to modify and splice the facial attribute features according to a preset privacy protection strength sequence to obtain a processed facial attribute prediction vector; The template image generation module is used to input the face attribute prediction vector into the identity mapping model to obtain the face identity feature vector, and then input the face identity feature vector and the original face image into the face replacement model to obtain the face anonymization template image; Availability condition generation module, used to construct an availability vector representing the availability of the original face image, and convert the availability vector into availability conditions through a conditional encoder network; The target image generation module is used to embed the original face image into the face anonymization template image by using a reversible neural network, and control the reversible neural network through availability conditions to generate a final target anonymization image.

7. A controllable face anonymization system according to claim 6, characterized in that: The availability vector adopts a one-hot vector, which includes an available vector and an unavailable vector. The reversible neural network adopts a flow-based generation model, in which a set of bidirectional mapping functions are used to realize reversible transformation between input and output, and the reversible change process is controlled by availability conditions.

8. A computer device comprising a system memory and a processor, wherein the system memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 5 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

10. A computer program product comprising instructions, characterized in that When the instructions are executed by a computer device cluster, the computer device cluster executes the method according to any one of claims 1 to 5.

Images