Face recognition decision method, face recognition decision model training method, related device and medium

Abstract

The embodiment of the specification discloses a face swiping decision method, a face swiping decision model training method, related devices and media. Wherein, the method comprises: obtaining target face swiping features corresponding to a target face swiping user based on a face swiping device, and obtaining target scene parameters corresponding to the face swiping device, then inputting the target face swiping features and the target scene parameters into a face swiping decision model, outputting a target face swiping result corresponding to the target face swiping user, the face swiping decision model is obtained by training based on face swiping data of known face swiping results in multiple scenes, the face swiping data includes face swiping features corresponding to multiple face swiping users in the scene respectively and scene parameters corresponding to the scene.

Description

Technical Field

[0001] This specification relates to the field of information processing technology, and in particular to a facial recognition decision-making method, a facial recognition decision-making model training method, related devices and media. Background Technology

[0002] The use of facial recognition devices is becoming increasingly widespread in offline commercial activities. Currently, these devices are placed in various scenarios such as retail stores, university cafeterias, company cafeterias, public transportation, and subway stations for facial recognition payment or access control. Facial recognition decision-making is the core step in the identity verification process. When a user initiates a facial recognition (identity verification) request, the facial recognition decision-making model deployed on the device analyzes the user's behavior to determine whether the verification is successful. Summary of the Invention

[0003] This specification provides a facial recognition decision-making method, a facial recognition decision-making model training method, related devices, and media. By utilizing the target scene parameters corresponding to the facial recognition device, the efficiency of facial recognition decision-making is adaptively improved in various scenarios, ensuring the accuracy of facial recognition decisions. The above technical solution is as follows:

[0004] Firstly, the embodiments of this specification provide a facial recognition decision-making method, including:

[0005] The system acquires the target facial recognition features corresponding to the target facial recognition user based on the facial recognition device; acquires the target scene parameters corresponding to the aforementioned facial recognition device; inputs the aforementioned target facial recognition features and the aforementioned target scene parameters into the facial recognition decision model, and outputs the target facial recognition result corresponding to the aforementioned target facial recognition user; the aforementioned facial recognition decision model is trained based on facial recognition data with known facial recognition results in multiple scenarios; the aforementioned facial recognition data includes the facial recognition features corresponding to each of the multiple facial recognition users in the aforementioned scenarios and the scene parameters corresponding to the aforementioned scenarios.

[0006] In one possible implementation, the target facial recognition features include target real-time facial recognition features and target offline general features; the acquisition of the target facial recognition features corresponding to the target facial recognition user based on the facial recognition device includes: acquiring the target real-time facial recognition features corresponding to the target facial recognition user based on the facial recognition device; and determining the target offline general features based on the target real-time facial recognition features.

[0007] In one possible implementation, the aforementioned real-time facial recognition features include N target facial recognition comparison results; the aforementioned N target facial recognition comparison results are the facial recognition comparison results corresponding to the N target users with the highest facial recognition comparison scores among the facial recognition comparison results between the target facial recognition image and the facial images of multiple users; the aforementioned target facial recognition image is obtained by capturing the aforementioned target facial recognition user based on the aforementioned facial recognition device; the aforementioned N is a positive integer.

[0008] In one possible implementation, the target facial recognition comparison result includes a target facial recognition comparison score; the target offline general features include offline general information corresponding to the M target users with the highest target facial recognition comparison scores among the N target users; the offline general information includes at least one of the following: the target user's first historical facial recognition information on the facial recognition device, the target user's historical activity information within the first preset range on the facial recognition device, the target user's second historical facial recognition information, and the target user's facial recognition information; the first historical facial recognition information includes the target user's historical facial recognition frequency and / or historical facial recognition comparison score range on the facial recognition device; the second historical facial recognition information includes the target user's historical facial recognition behavior information; and M is a positive integer less than or equal to N.

[0009] In one possible implementation, the above method is applied to the face recognition device; obtaining the target scene parameters corresponding to the face recognition device includes: receiving the server to determine the target scene parameters corresponding to the face recognition device based on the first target scene information corresponding to the face recognition device; the first target scene information includes at least one of the following: the openness information of the first target scene corresponding to the face recognition device, the age information and mask rate of the face recognition user in the first target scene, and the geographical location information corresponding to the face recognition device.

[0010] In one possible implementation, obtaining the target scene parameters corresponding to the facial recognition device includes: obtaining first target scene information corresponding to the facial recognition device; the first target scene information includes at least one of the following: openness information of the first target scene corresponding to the facial recognition device, age information and mask rate of the facial recognition user in the first target scene, and geographical location information corresponding to the facial recognition device; and determining the target scene parameters corresponding to the facial recognition device based on the first target scene information.

[0011] In one possible implementation, the first target scenario corresponding to the aforementioned facial recognition device is a scenario without historical facial recognition data; the aforementioned acquisition of the target scenario parameters corresponding to the aforementioned facial recognition device includes: determining the target scenario parameters corresponding to the aforementioned facial recognition device based on the same type of scenario corresponding to the aforementioned first target scenario; or determining the target scenario parameters corresponding to the aforementioned facial recognition device based on the scenario parameters of other scenarios within the second preset range of the aforementioned facial recognition device.

[0012] In one possible implementation, after inputting the target facial recognition features and the target scene parameters into the facial recognition decision model and outputting the target facial recognition result corresponding to the target facial recognition user, the method further includes: obtaining second target scene information corresponding to the facial recognition device; the second target scene information includes at least one of the following: openness information of the second target scene corresponding to the facial recognition device, age information and mask rate of the facial recognition user in the second target scene, and geographical location information of the facial recognition device in the second target scene; and updating the target scene parameters based on the second target scene information.

[0013] In one possible implementation, the aforementioned face recognition decision model includes a scene sharing layer and a scene parameter meta-learning unit; the step of inputting the aforementioned target face recognition features and the aforementioned target scene parameters into the face recognition decision model and outputting the target face recognition result corresponding to the aforementioned target face recognition user includes: inputting the aforementioned target face recognition features and the aforementioned target scene parameters into the face recognition decision model; the aforementioned scene sharing layer obtains target common features based on the aforementioned target face recognition features; the aforementioned scene parameter meta-learning unit obtains first target scene features based on the aforementioned target scene parameters; and outputting the target face recognition result corresponding to the aforementioned target face recognition user based on the aforementioned target common features and the aforementioned first target scene features.

[0014] Secondly, embodiments of this specification provide a method for training a facial recognition decision model. The method includes: acquiring facial recognition data with known facial recognition results in multiple scenarios; the facial recognition data includes facial recognition features corresponding to multiple facial recognition users in each of the multiple scenarios and scenario parameters corresponding to each scenario; and training a facial recognition decision model based on the facial recognition data.

[0015] In one possible implementation, the above-mentioned acquisition of facial recognition data with known facial recognition results in multiple scenarios includes: acquiring scene information corresponding to each of the multiple scenarios; the scene information includes at least one of the following: the openness information of the scene, the age information and mask rate of the facial recognition user in the scene, and the geographical location information corresponding to the scene; determining the scene parameters corresponding to each of the multiple scenarios based on the scene information corresponding to each of the multiple scenarios; and acquiring the facial recognition features corresponding to each of the multiple facial recognition users with known facial recognition results in each of the multiple scenarios.

[0016] In one possible implementation, the number of facial recognition users in each of the aforementioned scenarios exceeds the target number.

[0017] In one possible implementation, the aforementioned face recognition decision model includes a scene sharing layer and a scene parameter meta-learning unit; the scene sharing layer is used to learn the common features of face recognition under the aforementioned multiple scenes; the scene parameter meta-learning unit is used to display the scene information corresponding to each of the aforementioned multiple scenes, and to learn the correlation information between the aforementioned multiple scenes.

[0018] Thirdly, embodiments of this specification provide a facial recognition decision-making device, the device comprising:

[0019] The first acquisition module is used to acquire the target facial recognition features corresponding to the target facial recognition user based on the facial recognition device.

[0020] The second acquisition module is used to acquire the target scene parameters corresponding to the aforementioned face recognition device;

[0021] The face recognition decision module is used to input the target face recognition features and the target scene parameters into the face recognition decision model and output the target face recognition result corresponding to the target face recognition user. The face recognition decision model is trained based on face recognition data with known face recognition results in multiple scenes. The face recognition data includes the face recognition features corresponding to each of the multiple face recognition users in the above scenes and the scene parameters corresponding to the above scenes.

[0022] In one possible implementation, the target facial recognition features include real-time target facial recognition features and target offline general features; the first acquisition module includes:

[0023] The first acquisition unit is used to acquire the real-time facial recognition features of the target facial recognition user based on the facial recognition device.

[0024] The first determining unit is used to determine the offline general features of the target based on the real-time facial recognition features of the target.

[0025] In one possible implementation, the aforementioned real-time facial recognition features include N target facial recognition comparison results; the aforementioned N target facial recognition comparison results are the facial recognition comparison results corresponding to the N target users with the highest facial recognition comparison scores among the facial recognition comparison results between the target facial recognition image and the facial images of multiple users; the aforementioned target facial recognition image is obtained by capturing the aforementioned target facial recognition user based on the aforementioned facial recognition device; the aforementioned N is a positive integer.

[0026] In one possible implementation, the target facial recognition comparison result includes a target facial recognition comparison score; the target offline general features include offline general information corresponding to the M target users with the highest target facial recognition comparison scores among the N target users; the offline general information includes at least one of the following: the target user's first historical facial recognition information on the facial recognition device, the target user's historical activity information within the first preset range on the facial recognition device, the target user's second historical facial recognition information, and the target user's facial recognition information; the first historical facial recognition information includes the target user's historical facial recognition frequency and / or historical facial recognition comparison score range on the facial recognition device; the second historical facial recognition information includes the target user's historical facial recognition behavior information; and M is a positive integer less than or equal to N.

[0027] In one possible implementation, the above method is applied to the aforementioned facial recognition device; the second acquisition module is specifically used for:

[0028] The receiving server determines the target scene parameters corresponding to the face recognition device based on the first target scene information corresponding to the face recognition device; the first target scene information includes at least one of the following: the openness information of the first target scene corresponding to the face recognition device, the age information and mask rate of the face recognition user in the first target scene, and the geographical location information corresponding to the face recognition device.

[0029] In one possible implementation, the second acquisition module mentioned above includes:

[0030] The second acquisition unit is used to acquire the first target scene information corresponding to the facial recognition device; the first target scene information includes at least one of the following: the openness information of the first target scene corresponding to the facial recognition device, the age information and mask rate of the facial recognition user in the first target scene, and the geographical location information corresponding to the facial recognition device.

[0031] The second determining unit is used to determine the target scene parameters corresponding to the face recognition device based on the first target scene information.

[0032] In one possible implementation, the first target scenario corresponding to the aforementioned facial recognition device is a scenario without historical facial recognition data; the aforementioned second acquisition module includes:

[0033] The third determining unit is used to determine the target scene parameters corresponding to the facial recognition device based on the same type of scene corresponding to the first target scene; or

[0034] The fourth determining unit is used to determine the target scene parameters corresponding to the face recognition device based on scene parameters of other scenes within the second preset range of the face recognition device.

[0035] In one possible implementation, the facial recognition decision-making device further includes:

[0036] The third acquisition module is used to acquire the second target scene information corresponding to the above-mentioned face recognition device; the second target scene information includes at least one of the following: the openness information of the second target scene corresponding to the above-mentioned face recognition device, the age information and mask rate of the face recognition user in the above-mentioned second target scene, and the geographical location information of the above-mentioned face recognition device in the above-mentioned second target scene.

[0037] The update module is used to update the target scene parameters based on the second target scene information.

[0038] In one possible implementation, the aforementioned face recognition decision-making model includes a scene sharing layer and a scene parameter meta-learning unit;

[0039] The aforementioned facial recognition decision-making module includes:

[0040] The input unit is used to input the above-mentioned target face recognition features and the above-mentioned target scene parameters into the face recognition decision model. The above-mentioned scene sharing layer obtains the target common features based on the above-mentioned target face recognition features, and the above-mentioned scene parameter meta-learning unit obtains the first target scene features based on the above-mentioned target scene parameters.

[0041] The output unit is used to output the target face recognition result corresponding to the target face recognition user based on the above-mentioned common target features and the above-mentioned first target scene features.

[0042] Fourthly, embodiments of this specification provide a face recognition decision-making model training device, the device comprising:

[0043] The acquisition module is used to acquire facial recognition data of known facial recognition results in multiple scenarios. The facial recognition data includes the facial recognition features of multiple facial recognition users in each of the multiple scenarios and the scenario parameters corresponding to each scenario.

[0044] The training module is used to train the face recognition decision model based on the face recognition data mentioned above.

[0045] In one possible implementation, the above-mentioned acquisition module includes:

[0046] The first acquisition unit is used to acquire scene information corresponding to each of the multiple scenes; the scene information includes at least one of the following: the openness information of the scene, the age information and mask rate of the facial recognition user in the scene, and the geographical location information corresponding to the scene.

[0047] The determining unit is used to determine the scene parameters corresponding to each of the above multiple scenarios based on the scene information corresponding to each of the above multiple scenarios.

[0048] The second acquisition unit is used to acquire the facial features corresponding to each of the multiple facial recognition users with known facial recognition results in each of the above multiple scenarios.

[0049] In one possible implementation, the number of facial recognition users in each of the aforementioned scenarios exceeds the target number.

[0050] In one possible implementation, the aforementioned face recognition decision model includes a scene sharing layer and a scene parameter meta-learning unit; the scene sharing layer is used to learn the common features of face recognition under the aforementioned multiple scenes; the scene parameter meta-learning unit is used to display the scene information corresponding to each of the aforementioned multiple scenes, and to learn the correlation information between the aforementioned multiple scenes.

[0051] Fifthly, embodiments of this specification provide an electronic device, including: a processor and a memory;

[0052] The processor is connected to the memory; the memory is used to store executable program code; the processor runs a program corresponding to the executable program code by reading the executable program code stored in the memory, so as to execute the method provided by the first aspect or any possible implementation of the first aspect or the second aspect or any possible implementation of the embodiments of this specification.

[0053] Sixthly, embodiments of this specification provide a computer storage medium storing a plurality of instructions adapted for loading by a processor and executing the method steps provided by the first aspect or any possible implementation of the first aspect or the second aspect or any possible implementation of the second aspect of the embodiments of this specification.

[0054] Seventhly, embodiments of this specification provide a computer program product containing instructions that, when run on a computer or processor, cause the computer or processor to perform the method provided by the first aspect or any possible implementation of the first aspect or the second aspect or any possible implementation of the second aspect of the embodiments of this specification.

[0055] This embodiment first obtains the target facial recognition features corresponding to the target facial recognition user based on the facial recognition device, and also obtains the target scene parameters corresponding to the facial recognition device. Then, it inputs the target facial recognition features and the target scene parameters into a facial recognition decision model, outputting the target facial recognition result for the target facial recognition user. The facial recognition decision model is trained based on facial recognition data with known results from multiple scenarios. The facial recognition data includes the facial recognition features corresponding to each of the multiple facial recognition users in the scenario and the scene parameters corresponding to that scenario. Compared to modeling and deploying a corresponding facial recognition decision model for each scenario to make facial recognition decisions in each scenario, this embodiment does not require complex equipment. The configuration and real-time scene awareness of the facial recognition device are not required. Only a unified facial recognition decision model trained on facial recognition data with known facial recognition results in multiple scenarios needs to be maintained. This makes the deployment of the facial recognition decision model on the facial recognition device side much easier and greatly saves the operation and maintenance costs required for the facial recognition decision model. On the other hand, the embodiments in this specification use the target scene parameters corresponding to the usage scenario (first target scenario) of the facial recognition device during facial recognition as input to participate in facial recognition decision-making. This not only efficiently improves the adaptability of the facial recognition device in various scenarios and efficiently realizes differentiated decision-making of the facial recognition device in different usage scenarios, but also ensures the accuracy of facial recognition decision-making in various scenarios. Attached Figure Description

[0056] To more clearly illustrate the technical solutions in the embodiments of this specification, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0057] Figure 1 This is a schematic diagram illustrating the implementation process of a facial recognition decision-making method provided in related technologies.

[0058] Figure 2 A schematic diagram of the architecture of a facial recognition decision-making system provided as an exemplary embodiment of this specification;

[0059] Figure 3A This specification provides an exemplary embodiment of an application scenario diagram of a facial recognition decision-making method.

[0060] Figure 3B A schematic diagram illustrating an application scenario of another facial recognition decision-making method provided as an exemplary embodiment of this specification;

[0061] Figure 4 A schematic diagram illustrating the implementation process of a facial recognition decision-making method provided as an exemplary embodiment of this specification;

[0062] Figure 5 A flowchart illustrating a facial recognition decision-making method provided as an exemplary embodiment of this specification;

[0063] Figure 6 This is a schematic diagram illustrating an implementation process for obtaining target facial recognition features, provided as an exemplary embodiment of this specification.

[0064] Figure 7 This is a schematic diagram illustrating an implementation process for obtaining target scene parameters, provided as an exemplary embodiment of this specification.

[0065] Figure 8 A schematic diagram illustrating the implementation process of another facial recognition decision-making method provided as an exemplary embodiment of this specification;

[0066] Figure 9 A flowchart illustrating another facial recognition decision-making method provided as an exemplary embodiment of this specification;

[0067] Figure 10 A flowchart illustrating a face recognition decision model training method provided as an exemplary embodiment of this specification;

[0068] Figure 11 This specification provides a schematic diagram illustrating an implementation process for acquiring facial recognition data, as an exemplary embodiment.

[0069] Figure 12 A schematic diagram of the structure of a facial recognition decision-making device provided for an exemplary embodiment of this specification;

[0070] Figure 13 A schematic diagram of the structure of a face recognition decision model training device provided as an exemplary embodiment of this specification;

[0071] Figure 14 This is a schematic diagram of the structure of an electronic device provided as an exemplary embodiment of this specification. Detailed Implementation

[0072] The technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings.

[0073] The terms "first," "second," "third," etc., used in this specification, claims, and the foregoing drawings are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to such processes, methods, products, or apparatus.

[0074] It should be noted that the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, stored data, displayed data, etc.) and signals involved in the embodiments of this specification are all authorized by the user or fully authorized by all parties, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions. For example, historical activity information and historical facial recognition information involved in this specification were obtained with full authorization.

[0075] In related technologies, facial recognition devices may be placed in various commercial scenarios, and these scenarios are numerous and complex. The degree of openness or closedness of different scenarios, the age differences of facial recognition users, and the percentage of users wearing masks or covering their faces also vary. Therefore, to ensure the effectiveness of facial recognition decision-making in various scenarios such as retail stores, university cafeterias, and corporate cafeterias, etc., Figure 1 As shown, currently, it is mainly based on different usage scenarios of facial recognition devices (e.g.) Figure 1 Deploy the corresponding facial recognition decision-making model for each scenario (e.g., scenario A, scenario B, scenario C, etc.) Figure 1 The face recognition decision model (such as A, B, and C) is used to make face recognition decisions, thereby obtaining the target face recognition results corresponding to the target face recognition images of the target face recognition users in different scenarios.

[0076] However, on the one hand, the usage scenarios of facial recognition devices are not only complex but also change over time and with the movement of the devices. This necessitates that the facial recognition decision-making models in related technologies be retrained based on the facial recognition data in the changed scenarios—that is, deploying a new facial recognition decision-making model for each changed scenario—to achieve good facial recognition decision-making results. Therefore, the current approach of deploying different facial recognition decision-making models for different usage scenarios is not efficient in adapting to the various usage scenarios that may change at any time. On the other hand, due to the prevalence of emerging niche industries with limited and poorly covered facial recognition data, the corresponding facial recognition decision-making models may struggle to achieve good training results, thus making it difficult to ensure the effectiveness of facial recognition decisions in those scenarios.

[0077] Meanwhile, the deployment of the facial recognition decision model in the above-mentioned related technologies on the facial recognition device side will also be subject to the following constraints: (1) The facial recognition device's perception of the usage scenario is limited: In the related technologies, the facial recognition device can be given scene attributes by forcibly adding startup configuration parameters, thereby deploying the facial recognition decision model corresponding to the scene. However, this not only increases the difficulty of manually starting and configuring the facial recognition device, but also makes it difficult for the deployed facial recognition decision model to perceive the real usage scenario from the facial recognition device side due to the change of scene caused by the movement of the facial recognition device; (2) The number of deployed facial recognition decision models is limited: Although customizing facial recognition decision models for many scenarios can improve the facial recognition decision effect, the facial recognition decision model corresponding to each scenario needs to be maintained and managed separately. This not only makes the configuration and upgrade of the facial recognition device very difficult, but also greatly consumes manpower and maintenance costs.

[0078] Based on this, the embodiments of this specification provide a facial recognition decision-making method, a facial recognition decision-making model training method, related devices and media. Compared with the method of modeling and deploying corresponding facial recognition decision-making models for each scenario, this solution only needs to maintain a unified facial recognition decision-making model. By using the target scenario parameters corresponding to the usage scenario of the facial recognition device when the target user performs facial recognition (the first target scenario) as input to participate in the facial recognition decision-making, it can achieve adaptive improvement of the facial recognition device's decision-making efficiency in various scenarios and differentiated decision-making of the facial recognition device in different usage scenarios. It can also ensure the accuracy of facial recognition decision-making in various scenarios. Furthermore, it does not require complex device configuration or real-time scene perception of the facial recognition device, making it easier to deploy the facial recognition decision-making model on the facial recognition device side and greatly saving the operation and maintenance costs required for the facial recognition decision-making model.

[0079] Please refer to Figure 2 , Figure 2 This is a schematic diagram of the architecture of a facial recognition decision-making system provided as an exemplary embodiment of this specification.

[0080] like Figure 2 As shown, the facial recognition decision-making system may include: a facial recognition device 210 and a server 220. Wherein:

[0081] The facial recognition device 210 can be an Internet of Things (IoT) facial recognition device with facial recognition function (such as, but not limited to, facial recognition payment, facial recognition access) installed with a camera, or it can be a mobile phone, tablet computer, laptop computer or other device with user software and a camera installed. This specification does not limit this.

[0082] Optionally, when the facial recognition device 210 acquires a target facial recognition image of a target user, to ensure the accuracy of the facial recognition decision, it can first obtain the target facial recognition features corresponding to the target user and the target scene parameters corresponding to the facial recognition device 210 based on the target facial recognition image acquired by the facial recognition device 210. Then, the target facial recognition features and target scene parameters are input into the facial recognition decision model, and the target facial recognition result corresponding to the target user is output. This allows the facial recognition device 210 to efficiently output the corresponding facial recognition decision result according to its usage scenario, realizing the adaptive improvement of the facial recognition decision efficiency of the facial recognition device in various scenarios, as well as the differentiated decision-making of the facial recognition device in different usage scenarios. The above facial recognition decision model is trained based on facial recognition data with known facial recognition results in multiple scenarios. The facial recognition data includes the facial recognition features corresponding to multiple users in the scenario and the scene parameters corresponding to that scenario.

[0083] Optionally, the facial recognition device 210 can receive target scene parameters corresponding to the facial recognition device 210 from the server 220 via the network, or it can directly obtain the first target scene information corresponding to the facial recognition device 210 first, and then determine the target scene parameters corresponding to the facial recognition device 210 based on the first target scene information. This embodiment of the specification does not limit this. The first target scene information includes at least one of the following: the openness information of the first target scene corresponding to the facial recognition device, the age information and mask coverage of the facial recognition user in the first target scene, and the geographical location information corresponding to the facial recognition device.

[0084] Optionally, when the facial recognition device 210 acquires the target facial recognition image of the target user, to ensure the accuracy and efficiency of the facial recognition decision, a data connection can be established with the server 220 via the network. For example, the target facial recognition image and the first target scene information corresponding to the facial recognition device 210 can be sent to the server 220. Then, the device receives the facial recognition decision result obtained by the server 220 using the aforementioned facial recognition decision model, based on the target facial recognition features of the target user determined from the target facial recognition image, and the target scene parameters based on the first target scene information corresponding to the facial recognition device 210. Finally, the device determines whether the target user passes the facial recognition based on the aforementioned facial recognition decision result. This ensures both high success rate in identifying the user and low false recognition rate, and also allows for differentiated facial recognition decision results based on different scenarios, improving the adaptability of the facial recognition device 210 in various scenarios.

[0085] Server 220 can be a server capable of providing multiple facial recognition decision-making methods. It can receive target facial recognition images and corresponding first target scene information or target scene parameters from facial recognition device 210 via a network. Based on the target facial recognition images collected by facial recognition device 210, it obtains the target facial recognition features corresponding to the target user and determines the corresponding target scene parameters based on the first target scene information. Then, it inputs the target facial recognition features and target scene parameters into the facial recognition decision-making model and outputs the target facial recognition result for the target user. Server 220 can also send the target facial recognition result for the target user to facial recognition device 210 via a network, so that the target user can promptly know whether facial recognition was successful on the facial recognition device 210 with the user-version software installed. Server 220 can be, but is not limited to, a hardware server, a virtual server, or a cloud server.

[0086] The network can be a medium that provides a communication link between server 220 and any facial recognition device 210, or it can be the Internet, which includes network devices and transmission media, and is not limited to these. The transmission media can be a wired link, such as, but not limited to, coaxial cable, fiber optic cable, and digital subscriber line (DSL), or a wireless link, such as, but not limited to, wireless fidelity (WIFI), Bluetooth, and mobile device networks.

[0087] For example, Figure 2 The facial recognition device 210 in the middle can be Figure 3A The facial recognition device 310, such as Figure 3A As shown, in scenarios where facial recognition device 310 is used for offline facial recognition payment, the scenario information corresponding to facial recognition device 310 may differ at different times. For example, the user (development level) or whether the user is wearing a mask when using facial recognition device 310 (mask rate) may change. Therefore, in order for facial recognition device 310 to quickly adapt to changes in its usage scenario and thus complete facial recognition decisions more efficiently and accurately, a pre-trained facial recognition decision model provided in this embodiment can be deployed in the aforementioned facial recognition device 310. By inputting the target facial recognition features corresponding to the target facial recognition user and the target scenario parameters corresponding to the first target scenario of facial recognition device 310 into the facial recognition decision model, differentiated facial recognition decision results can be output according to different scenarios of facial recognition device 310, thereby improving the adaptability of facial recognition device 310 in various scenarios.

[0088] For example, Figure 2 The facial recognition device 210 in the middle can also be Figure 3BIn the face recognition access scenario shown, the face recognition device 340 can also be deployed with a pre-trained face recognition decision model provided in this embodiment. In this case, even if the target face recognition user 330 and... Figure 3A Even if the target user 320 for facial recognition payment is the same user, different facial recognition decision results will be output due to the different scenarios corresponding to facial recognition devices 340 and 310 (such as, but not limited to, the development level of the facial recognition devices, the age information of each facial recognition user on the facial recognition device, and the proportion of facial occlusion). Therefore, regardless of the facial recognition scenario and facial recognition device, only a unified facial recognition decision model needs to be maintained. By using the target scenario parameters corresponding to the usage scenario of the facial recognition device when the target facial recognition user makes facial recognition (the first target scenario) as input to participate in the facial recognition decision, differentiated decisions of the facial recognition device under different usage scenarios can be achieved, which greatly saves the operation and maintenance costs required for the facial recognition decision model.

[0089] Understandably, Figure 2 The facial recognition decision-making system shown can also be applied to various scenarios such as canteens, public transportation, subway stations, assisted driving scenarios, assisted teaching scenarios, and offline retail scenarios. The embodiments in this specification do not limit this application.

[0090] Understandably, Figure 2 The number of facial recognition devices 210 and servers 220 in the facial recognition decision-making system shown is only an example. In a specific implementation, the facial recognition decision-making system can contain any number of facial recognition devices and servers, and this specification does not specifically limit this. For example, but not limited to, facial recognition device 210 can be a facial recognition device cluster composed of multiple facial recognition devices, and server 220 can be a server cluster composed of multiple servers.

[0091] Next, combine Figures 2-3B This specification introduces a facial recognition decision-making method provided by an embodiment. Please refer to the following for details. Figure 4 This is a schematic diagram illustrating the implementation process of a facial recognition decision-making method provided in an exemplary embodiment of this specification. Figure 4As shown, when a user performs facial recognition on a facial recognition device, the system first acquires the user's real-time facial recognition features and offline general features based on the device. These features are then input into a facial recognition decision model 410 trained on facial recognition datasets corresponding to scenarios A, B, and C. This outputs the user's facial recognition decision results for each of these three scenarios. If only the facial recognition decision result for the current scenario is desired, a specific scenario branch can be routed by issuing a scenario ID. Furthermore, only when the issued scenario ID is one of the scenarios A, B, or C involved in the model's training can a targeted and accurate facial recognition decision result for that scenario be output. The scenario IDs are obtained by comprehensively clustering the scenario information (including but not limited to development level information, geographical location information, mask coverage, etc.) corresponding to each scenario, dividing the scenarios, and creating a table of scenario IDs. The aforementioned face recognition dataset includes real-time face recognition features and offline general features corresponding to multiple face recognition users in the same scene. The aforementioned face recognition decision model 410 can be an MMOE structure, including a connection layer, a shared bottom layer, expert network A, expert network B, parameter gates, and a tower layer. The connection layer is used to fuse the aforementioned real-time face recognition features and offline general features. The shared bottom layer is used to learn the common features between different scenes in the fused features. The expert networks A and B are used to complete the interaction of features between different scenes based on the common features between different scenes. The parameter gates and the tower layer are used to distinguish each scene branch and perform independent optimization, so that the face recognition decision result is more adapted to the specific scene.

[0092] Although Figure 4 The method shown can also solve the adaptive problem of face recognition decision in multiple scenarios. However, when deploying the unified face recognition decision model 410 with multiple scenario face recognition decision branches on the face recognition device side, as the number of scenarios increases, the number of branches of the face recognition decision model 410 will increase, the memory occupation of the model file on the face recognition device side will become uncontrollable, and the operation and maintenance cost of the model will also increase.

[0093] In order to solve the above-mentioned related technologies Figure 4 The problems existing in facial recognition decision-making in China will be discussed in the following sections. Figures 2-4 Taking the facial recognition decision made by the facial recognition device 210 as an example, this specification describes the facial recognition decision-making method provided in the embodiments. Please refer to the following for details. Figure 5 This is a flowchart illustrating a facial recognition decision-making method provided in an exemplary embodiment of this specification. Figure 5 As shown, this facial recognition decision-making method includes the following steps:

[0094] S502, obtain the target facial recognition features corresponding to the target facial recognition user based on the facial recognition device.

[0095] Specifically, in order for the facial recognition decision-making model to fully understand the differences between the target facial recognition image of the target user currently undergoing facial recognition on the device and the pre-stored facial images of multiple users, such as... Figure 6 As shown, the aforementioned target facial recognition features include real-time target facial recognition features, which in turn include N target facial recognition comparison results, where N is a positive integer. These N target facial recognition comparison results are the facial recognition comparison results corresponding to the N target users with the highest facial recognition comparison scores among the facial recognition comparison results between the target facial recognition image and the facial images of multiple users. These facial recognition comparison scores can be used to characterize the similarity between the target facial recognition image and the aforementioned facial images. The aforementioned target facial recognition image is obtained by capturing images of the target facial recognition user using a facial recognition device.

[0096] Understandably, the higher the facial recognition comparison score in the facial recognition comparison results, the more similar the target facial recognition image is to the facial image corresponding to that facial recognition comparison result. To a certain extent, the probability that the target facial recognition user is the user corresponding to that facial image is higher.

[0097] Furthermore, when a target user performs facial recognition at a facial recognition device, the device can first acquire the target user's real-time facial recognition features. Based on these features, the facial recognition decision model can fully understand the differences between the target facial recognition image and the pre-stored facial images of multiple users. This allows for a more accurate determination of the probability that the target user corresponds to at least one of the multiple users, ensuring the accuracy of the facial recognition decision.

[0098] Furthermore, the above-mentioned process of obtaining the target facial recognition real-time features corresponding to the target facial recognition user based on the facial recognition device can be as follows: first, the target facial recognition image of the target facial recognition user is obtained based on the camera installed on the facial recognition device; then, the target facial recognition image is compared with the facial images of multiple users pre-stored on the facial recognition device or in the server, for example, but not limited to, calculating the similarity between the target facial recognition image and the facial images of multiple users, thereby obtaining the facial recognition comparison results corresponding to the target facial recognition user and the multiple users. In the embodiments of this specification, the facial recognition comparison results corresponding to the target facial recognition user and the multiple users can be directly used as the target facial recognition real-time features corresponding to the target facial recognition user; in order to improve the efficiency of facial recognition decision-making and avoid interference from the facial recognition comparison results corresponding to facial images with poor similarity to the target facial recognition image, the facial recognition comparison results corresponding to the N target users with the highest facial recognition comparison scores or the facial recognition comparison results corresponding to the target users with facial recognition comparison scores greater than a preset score (i.e., the target facial recognition comparison results) can also be selected as the target facial recognition real-time features.

[0099] Understandably, in order for the facial recognition decision model to fully understand the differences between the target facial recognition user's image and the facial images of different users, and to avoid the situation where the target facial recognition user may have a generic face that affects the facial recognition decision, the aforementioned real-time features of the target facial recognition can include multiple target facial recognition comparison results, that is, the aforementioned N can be an integer greater than 1.

[0100] Specifically, to avoid situations where the target user has a generic face, leading to minimal differences in the real-time facial recognition features and thus affecting the facial recognition decision, it is crucial to ensure the correlation between the facial recognition decision and the facial recognition device. Figure 6As shown, the target facial recognition features mentioned above include not only real-time target facial recognition features but also target offline general features. The target facial recognition comparison results in the real-time features include target facial recognition comparison scores. The target offline general features include offline general information corresponding to the M target users with the highest target facial recognition comparison scores among the N target users. The offline general information includes at least one of the following: the target user's first historical facial recognition information on the facial recognition device, the target user's historical activity information within a first preset range of the facial recognition device, the target user's second historical facial recognition information, and the target user's facial recognition accuracy information. The first historical facial recognition information includes the target user's historical facial recognition frequency and / or historical facial recognition comparison score range on the facial recognition device, and the second historical facial recognition information includes the target user's historical facial recognition behavior information. M is a positive integer less than or equal to N. The facial recognition accuracy information includes facial comparison scores between the target user and other users within the preset area where the facial recognition device is located, used to determine whether the target user has a generic face and avoid misidentification.

[0101] Understandably, the higher the number of times a target user has used facial recognition technology in the first historical facial recognition information on the device, or the higher the target user's historical activity information within a first preset range on the device, or the higher the target user's historical facial recognition frequency, the greater the likelihood that the target user will use facial recognition technology on that device. Therefore, the facial recognition decision model is more likely to determine that the target user is the intended target, i.e., the probability of the facial recognition result indicating that the target user has successfully used facial recognition technology. If there is no facial recognition data in the first or second historical facial recognition information corresponding to the target user, meaning the target user has never used facial recognition technology or has never performed any facial recognition behavior, or if the target user's facial recognition comparison score is not within the target user's historical facial recognition comparison score range on the device, or if the target user's facial recognition information shows low facial recognition accuracy, then the likelihood of the target user being the intended target is relatively low. In other words, the facial recognition decision model can combine the aforementioned target offline general features to determine a target facial recognition result that is more correlated with the actual facial recognition situation of the device and the target user, greatly reducing the possibility of erroneous decisions (false recognition).

[0102] Understandably, the aforementioned target offline general features may include, in addition to including the offline general information corresponding to the M target users with the highest target face recognition comparison scores among the N target users, also include only the offline general information of the target users whose target face recognition comparison scores are greater than the target score. This specification embodiment does not limit this. When the target offline general features include the offline general information corresponding to each of the M target users, the face recognition decision model in this specification embodiment can output the probability that the target face recognition user is the corresponding target user among the M target users, i.e., the M face recognition pass rates (target face recognition results), based on the aforementioned target offline general features.

[0103] Furthermore, when a target user undergoes facial recognition at a facial recognition device, after obtaining the target user's real-time facial recognition features based on the device, the target user's offline general features can be further determined based on these real-time features. This allows the facial recognition decision model to fully understand the differences between the target facial recognition image and pre-stored facial images of multiple users, as well as the likelihood that a target user with a high facial recognition score compared to the target image will undergo facial recognition at the device. By combining the correlation between the target user and the facial recognition device, the influence of the target user or a generic face on the facial recognition decision can be eliminated, leading to more accurate decisions regarding the facial recognition result correlated with the device, further ensuring the accuracy of the facial recognition decision.

[0104] like Figure 5 As shown, in addition to S502, the facial recognition decision-making method also includes:

[0105] S504, obtain the target scene parameters corresponding to the facial recognition device.

[0106] Specifically, when a target user performs facial recognition at a facial recognition device, in addition to obtaining the target facial recognition features corresponding to the target user, in order for the facial recognition decision model to efficiently output more targeted facial recognition results based on the scene in which the facial recognition device is located, it is also necessary to further obtain the target scene parameters corresponding to the facial recognition device.

[0107] Optionally, if the facial recognition decision model in the embodiments of this specification is deployed on the facial recognition device side, when a target facial recognition user performs facial recognition on the device, the device can directly receive the first target scene information corresponding to the device from the server via the network to determine the target scene parameters corresponding to the device. This eliminates the need for complex device configuration and real-time scene perception on the device. Only a unified facial recognition decision model trained on facial recognition data with known results from multiple scenarios needs to be maintained. This simplifies the deployment of the facial recognition decision model on the device side, significantly reduces the cost of maintaining the model on the device, and improves the efficiency and adaptability of the device in various scenarios. The first target scene information includes at least one of the following: the openness information of the first target scene corresponding to the facial recognition device, the age information and mask coverage of the facial recognition user in the first target scene, and the geographical location information of the device. The openness information is used to characterize the changes in the facial recognition user in the first target scene. If the number of users using facial recognition devices in the primary target scenario varies significantly—for example, in offline retail scenarios, there are always different customers using facial recognition for payment—then it indicates that the primary target scenario corresponding to the facial recognition device is more open, meaning its development level is higher.

[0108] Optionally, such as Figure 7 As shown, when a target user uses facial recognition at a facial recognition device, the system first obtains the first target scene information corresponding to the device, and then determines the target scene parameters based on this information. The first target scene information includes at least one of the following: the openness information of the first target scene corresponding to the facial recognition device, the age information and mask-wearing rate of the user in the first target scene, and the geographical location information of the device. In other words, the first target scene information can be encoded into a corresponding vector to obtain the target scene parameters. This allows the system to reflect the differences in the target scene corresponding to the facial recognition device through the target scene parameters, and also facilitates efficient differentiated decision-making for the facial recognition device in different usage scenarios.

[0109] Understandably, the age information of the facial recognition user in the first target scenario may include, but is not limited to, the average age, age range, or age distribution of all facial recognition users on the facial recognition device. The mask rate may include, but is not limited to, the probability of wearing a mask or having one's face covered among all facial recognition users on the facial recognition device. The geographical location information corresponding to the facial recognition device may include the geographical location of the facial recognition device when the target facial recognition user performs facial recognition.

[0110] Optionally, when the first target scenario corresponding to the facial recognition device is a scenario without historical facial recognition data, i.e., no user has used the facial recognition device in the first target scenario (e.g., the facial recognition device is new or has been moved to another geographical location for facial recognition decision-making), to avoid the problem of poor consistency between the target scenario parameters determined based on the first target scenario information and the actual scenario parameters corresponding to the first target scenario due to insufficient information, the target scenario parameters corresponding to the facial recognition device can be directly determined based on scenarios of the same type as the first target scenario, or based on the scenario parameters of other scenarios within the second preset range of the facial recognition device. This allows setting similar scenario parameters for new scenarios without historical facial recognition data, solving the cold start problem of facial recognition decision-making in new scenarios, and further ensuring the effectiveness of facial recognition decision-making in various scenarios. The aforementioned target scenario parameters may be, but are not limited to, one or an average parameter of the scenario parameters of the aforementioned scenarios of the same type, or a scenario parameter of a certain scenario within the second preset range of the facial recognition device, or an average parameter of the scenario parameters of the aforementioned other scenarios.

[0111] Optionally, when the first target scenario corresponding to the facial recognition device is a new scenario without historical facial recognition data, without considering the effectiveness of facial recognition decision-making, a default target scenario parameter can be directly set to solve the cold start problem of the facial recognition device. Then, after there is some historical facial recognition data in the first target scenario corresponding to the facial recognition device, the openness information, the age information of the facial recognition user, and the mask rate, etc., of the first target scenario are determined based on the aforementioned historical facial recognition data. Finally, the originally set default target scenario parameter is updated according to the aforementioned first target scenario information. Thus, after solving the cold start problem, the adaptability of the facial recognition device in making facial recognition decisions can be further improved, and the accuracy of facial recognition decisions can be increased.

[0112] Understandably, S502 and S504 can be executed simultaneously or sequentially, and the embodiments in this specification do not limit this.

[0113] Please continue to refer to the following. Figure 5 ,like Figure 5 As shown, after obtaining the target facial recognition features corresponding to the target facial recognition user based on the facial recognition device in S502, and obtaining the target scene parameters corresponding to the facial recognition device in S504, the facial recognition decision method further includes:

[0114] S506: Input the target facial recognition features and target scene parameters into the facial recognition decision model, and output the target facial recognition result corresponding to the target facial recognition user.

[0115] Specifically, after obtaining the target facial recognition features corresponding to the target facial recognition user and the target scene parameters corresponding to the facial recognition device, these features and parameters can be input into the trained facial recognition decision model to output the target facial recognition result for the target user. This facial recognition decision model is trained based on facial recognition data with known results from multiple scenarios. The facial recognition data includes the facial recognition features of each user in the scenario and the scene parameters corresponding to that scenario.

[0116] Specifically, the aforementioned facial recognition decision-making model can include a scene sharing layer and a scene parameter meta-learning unit. After obtaining the target facial recognition features corresponding to the target facial recognition user and the target scene parameters corresponding to the facial recognition device, the target facial recognition features and target scene parameters can be input into the facial recognition decision-making model. The scene sharing layer obtains target common features based on the target facial recognition features, and the scene parameter meta-learning unit obtains first target scene features based on the target scene parameters. Then, based on the target common features and the first target scene features, the target facial recognition result corresponding to the target facial recognition user is output. Thus, the scene sharing layer can retain the common features between the first target scene and the scenes involved in training the facial recognition decision-making model, and the scene parameter meta-learning unit can explicitly model the first target scene information, learn the complex scene relationships between the first target scene and the scenes involved in training the facial recognition decision-making model, retain the characteristics of the first target scene, and efficiently realize differentiated decision-making of facial recognition devices in different usage scenarios.

[0117] For example, such as Figure 8 As shown, the aforementioned face recognition decision-making model can include a scene sharing layer, a scene parameter meta-learning unit, a connection layer, a scene embedding layer, and a parameter gate, in addition to these. After obtaining the target face recognition features (i.e., real-time and offline general features) corresponding to the target user's face recognition on the face recognition device, as well as the target scene parameters corresponding to the face recognition device, the connection layer of the face recognition decision-making model can fuse the real-time and offline general features. Then, the scene sharing layer extracts the common features from the fused features. Simultaneously, the target scene parameters are input into the scene embedding layer of the face recognition decision-making model. The scene embedding layer and parameter gate determine the dynamic network weights and biases in the face recognition decision-making model for the first target scene corresponding to the target scene parameters. Finally, based on the weights and biases combined with the features extracted by the scene sharing layer and the scene parameter meta-learning unit, the target face recognition result corresponding to the target user is output.

[0118] Optionally, the target facial recognition result may include the pass rate of the target facial recognition user when facial recognition is performed in the first target scenario of the facial recognition device.

[0119] Optionally, the above-mentioned target facial recognition result may also include the probability that the target facial recognition user is each of the above M target users, i.e., the M pass rates.

[0120] For example, in a facial recognition payment scenario, after obtaining the above M pass rates, the target facial recognition user can be further identified as the target user corresponding to the highest pass rate that is greater than the target threshold for facial recognition payment.

[0121] This embodiment first obtains the target facial recognition features corresponding to the target facial recognition user based on the facial recognition device, and also obtains the target scene parameters corresponding to the facial recognition device. Then, it inputs the target facial recognition features and the target scene parameters into a facial recognition decision model, outputting the target facial recognition result for the target facial recognition user. The facial recognition decision model is trained based on facial recognition data with known results from multiple scenarios. The facial recognition data includes the facial recognition features corresponding to each of the multiple facial recognition users in the scenario and the scene parameters corresponding to that scenario. Compared to modeling and deploying a corresponding facial recognition decision model for each scenario to make facial recognition decisions in each scenario, this embodiment does not require complex equipment. The configuration and real-time scene awareness of the facial recognition device are not required. Only a unified facial recognition decision model trained on facial recognition data with known facial recognition results in multiple scenarios needs to be maintained. This makes the deployment of the facial recognition decision model on the facial recognition device side much easier and greatly saves the operation and maintenance costs required for the facial recognition decision model. On the other hand, the embodiments in this specification use the target scene parameters corresponding to the usage scenario (first target scenario) of the facial recognition device during facial recognition as input to participate in facial recognition decision-making. This not only efficiently improves the adaptability of the facial recognition device in various scenarios and efficiently realizes differentiated decision-making of the facial recognition device in different usage scenarios, but also ensures the accuracy of facial recognition decision-making in various scenarios.

[0122] Please refer to the following. Figure 9 This is a flowchart illustrating another facial recognition decision-making method provided in an exemplary embodiment of this specification. Figure 9 As shown, the facial recognition decision-making method may include the following steps:

[0123] S902, based on the facial recognition device, obtain the target facial recognition features corresponding to the target facial recognition user.

[0124] Specifically, S902 is the same as S502, and will not be repeated here.

[0125] S904, obtain the target scene parameters corresponding to the facial recognition device.

[0126] Specifically, S904 is the same as S504, which will not be repeated here.

[0127] S906 inputs the target facial recognition features and target scene parameters into the facial recognition decision model and outputs the target facial recognition result corresponding to the target facial recognition user.

[0128] Specifically, S906 is identical to S506, which will not be repeated here.

[0129] S908, obtain the second target scene information corresponding to the facial recognition device.

[0130] Specifically, since the primary target scenario corresponding to the facial recognition device may change—for example, mask-wearing rates may change due to the pandemic, or the number of facial recognition users in open scenarios may also change—to ensure the effectiveness and accuracy of facial recognition decisions in various scenarios, the secondary target scenario information of the facial recognition device can be obtained every preset time period, or the secondary target scenario information corresponding to the facial recognition device can be obtained based on the facial recognition data of that target number of times the device performs facial recognition scans. This avoids the problem of inaccurate target scenario parameters caused by the actual scenario of the facial recognition device changing and the primary target scenario information not matching reality, thus affecting the accuracy and effectiveness of facial recognition decisions. The aforementioned secondary target scenario information includes at least one of the following: information on the openness of the secondary target scenario corresponding to the facial recognition device, age information and mask-wearing rates of facial recognition users in the secondary target scenario, and geographical location information of the facial recognition device in the secondary target scenario.

[0131] S910 updates the target scene parameters based on the second target scene information.

[0132] Specifically, after obtaining the second target scene information corresponding to the facial recognition device, this second target scene information can be encoded into corresponding scene parameters (vectors). Then, the target scene parameters corresponding to the facial recognition device are updated to the scene parameters encoded from the second target scene information. After the target scene parameters of the facial recognition device are updated, if a target user performs facial recognition on the device, the target facial recognition result will be determined based on the updated target scene parameters and the aforementioned target facial recognition features after obtaining the target facial recognition features corresponding to the target user.

[0133] In this embodiment, the second target scene information corresponding to the face recognition device is obtained periodically or after the amount of face recognition data from the face recognition device reaches a certain requirement. The target scene parameters are then updated based on the second target scene information after the change of the first target scene corresponding to the face recognition device. Thus, when the scene of the face recognition device changes, the face recognition data under the changed scene does not need to participate in the training of the face recognition decision model. The scene parameters corresponding to the face recognition device can be adjusted efficiently and quickly, which greatly facilitates the operation, maintenance and upgrading of the face recognition decision model.

[0134] Please refer to the following. Figure 10 This is a flowchart illustrating a face recognition decision-making model training method provided in an exemplary embodiment of this specification. Figure 10 As shown, the training method for this facial recognition decision-making model can include the following steps:

[0135] S1002, Obtain facial recognition data of known facial recognition results in multiple scenarios. The facial recognition data includes facial recognition features of multiple facial recognition users in each scenario and scenario parameters corresponding to each scenario.

[0136] Specifically, before making a face recognition decision using the face recognition decision model provided in the embodiments of this specification, face recognition data with known face recognition results in multiple scenarios can be obtained first, and the face recognition data corresponding to each of the above multiple scenarios can be used as training samples to train the face recognition decision model. The face recognition data includes the face recognition features corresponding to multiple face recognition users in each scenario and the scenario parameters corresponding to each scenario.

[0137] Specifically, such as Figure 11 As shown, the above implementation process for obtaining facial recognition data with known facial recognition results in multiple scenarios may include the following steps:

[0138] S1102, Obtain scene information corresponding to each of the multiple scenes.

[0139] Specifically, the aforementioned scenario information includes at least one of the following: the openness of the scenario, the age information of the users who use facial recognition in the scenario, the mask rate, and the geographical location information corresponding to the scenario, i.e., the geographical location information of the facial recognition device in that scenario.

[0140] Understandably, in order to ensure the training effect of the facial recognition decision model and avoid insufficient training of the facial recognition decision model due to insufficient data for parameter training in some small scenarios, the number of facial recognition users in each of the above scenarios is greater than the target number. This means that the above scenarios can all be scenarios where the amount of facial recognition data is greater than the target amount of data, that is, scenarios that can provide sufficient sample training data.

[0141] S1104, determine the scene parameters corresponding to each of the multiple scenes based on the scene information corresponding to each of the multiple scenes.

[0142] Specifically, the scene information corresponding to each of the above multiple scenarios can be encoded into corresponding vectors, thereby obtaining the scene parameters corresponding to each of the above multiple scenarios.

[0143] S1106, Obtain the facial recognition features corresponding to each of the multiple facial recognition users with known facial recognition results in multiple scenarios.

[0144] Specifically, the process of obtaining the facial features of multiple facial recognition users with known facial recognition results in each scenario in S1106 is similar to that in S502, and will not be repeated here.

[0145] S1004, a face recognition decision model trained based on face recognition data.

[0146] Specifically, after obtaining facial recognition data with known facial recognition results in multiple scenarios, an initial facial recognition decision model can be trained based on the aforementioned facial recognition data until the training objective is achieved (e.g., but not limited to, the accuracy of the facial recognition decision model in each scenario or other scenarios not involved in training is greater than the target accuracy threshold). Then, the training ends, and the trained facial recognition decision model is obtained.

[0147] Specifically, the aforementioned face recognition decision-making model includes a scene sharing layer and a scene parameter meta-learning unit. The scene sharing layer is used to learn the common features of face recognition in multiple scenarios, while the scene parameter meta-learning unit is used to display the scene information corresponding to each of the multiple scenarios and to learn the correlation information between multiple scenarios. Thus, the scene sharing layer can retain the common features between scenarios, and the scene parameter meta-learning unit can explicitly model the scene information of each scenario, learn the complex scene correlations between scenarios, and retain the characteristics of each scenario. This avoids the problem that small scenarios may not learn sufficiently due to the limited training data available for training, which may affect the accuracy of the face recognition decision-making model.

[0148] This specification's embodiments train a facial recognition decision model using facial recognition data with known results from multiple scenarios. This facial recognition data includes the facial features of multiple users in each scenario and the corresponding scenario parameters. Compared to modeling each scenario individually, this solution incorporates the scenario parameters into the training of the facial recognition decision model. This allows the model to learn the impact of facial features on facial recognition decisions in each scenario, while also learning the commonalities and differences between scenarios. This enables efficient and accurate facial recognition decision-making across different scenarios by requiring only one unified facial recognition decision model to operate and maintain, significantly reducing the operational costs of the facial recognition decision model.

[0149] Please refer to the following. Figure 12 , Figure 12 This specification provides an exemplary embodiment of a facial recognition decision-making device. For example... Figure 12 As shown, the facial recognition decision-making device 1200 includes:

[0150] The first acquisition module 1210 is used to acquire the target facial recognition features corresponding to the target facial recognition user based on the facial recognition device.

[0151] The second acquisition module 1220 is used to acquire the target scene parameters corresponding to the above-mentioned face recognition device;

[0152] The face recognition decision module 1230 is used to input the target face recognition features and the target scene parameters into the face recognition decision model and output the target face recognition result corresponding to the target face recognition user. The face recognition decision model is trained based on face recognition data with known face recognition results in multiple scenes. The face recognition data includes the face recognition features corresponding to each of the multiple face recognition users in the above scene and the scene parameters corresponding to the above scene.

[0153] In one possible implementation, the target facial recognition features include real-time target facial recognition features and target offline general features; the first acquisition module 1210 includes:

[0154] The first acquisition unit is used to acquire the real-time facial recognition features of the target facial recognition user based on the facial recognition device.

[0155] The first determining unit is used to determine the offline general features of the target based on the real-time facial recognition features of the target.

[0156] In one possible implementation, the aforementioned real-time facial recognition features include N target facial recognition comparison results; the aforementioned N target facial recognition comparison results are the facial recognition comparison results corresponding to the N target users with the highest facial recognition comparison scores among the facial recognition comparison results between the target facial recognition image and the facial images of multiple users; the aforementioned target facial recognition image is obtained by capturing the aforementioned target facial recognition user based on the aforementioned facial recognition device; the aforementioned N is a positive integer.

[0157] In one possible implementation, the target facial recognition comparison result includes a target facial recognition comparison score; the target offline general features include offline general information corresponding to the M target users with the highest target facial recognition comparison scores among the N target users; the offline general information includes at least one of the following: the target user's first historical facial recognition information on the facial recognition device, the target user's historical activity information within the first preset range on the facial recognition device, the target user's second historical facial recognition information, and the target user's facial recognition information; the first historical facial recognition information includes the target user's historical facial recognition frequency and / or historical facial recognition comparison score range on the facial recognition device; the second historical facial recognition information includes the target user's historical facial recognition behavior information; and M is a positive integer less than or equal to N.

[0158] In one possible implementation, the above method is applied to the aforementioned facial recognition device;

[0159] The second acquisition module 1220 is specifically used to: receive the server to determine the target scene parameters corresponding to the face recognition device based on the first target scene information corresponding to the face recognition device; the first target scene information includes at least one of the following: the openness information of the first target scene corresponding to the face recognition device, the age information and mask rate of the face recognition user in the first target scene, and the geographical location information corresponding to the face recognition device.

[0160] In one possible implementation, the second acquisition module 1220 includes:

[0161] The second acquisition unit is used to acquire the first target scene information corresponding to the facial recognition device; the first target scene information includes at least one of the following: the openness information of the first target scene corresponding to the facial recognition device, the age information and mask rate of the facial recognition user in the first target scene, and the geographical location information corresponding to the facial recognition device.

[0162] The second determining unit is used to determine the target scene parameters corresponding to the face recognition device based on the first target scene information.

[0163] In one possible implementation, the first target scenario corresponding to the aforementioned facial recognition device is a scenario without historical facial recognition data; the aforementioned second acquisition module 1220 includes:

[0164] The third determining unit is used to determine the target scene parameters corresponding to the facial recognition device based on the same type of scene corresponding to the first target scene; or

[0165] The fourth determining unit is used to determine the target scene parameters corresponding to the face recognition device based on scene parameters of other scenes within the second preset range of the face recognition device.

[0166] In one possible implementation, the facial recognition decision-making device 1200 further includes:

[0167] The third acquisition module is used to acquire the second target scene information corresponding to the above-mentioned face recognition device; the second target scene information includes at least one of the following: the openness information of the second target scene corresponding to the above-mentioned face recognition device, the age information and mask rate of the face recognition user in the above-mentioned second target scene, and the geographical location information of the above-mentioned face recognition device in the above-mentioned second target scene.

[0168] The update module is used to update the target scene parameters based on the second target scene information.

[0169] In one possible implementation, the aforementioned face recognition decision-making model includes a scene sharing layer and a scene parameter meta-learning unit;

[0170] The aforementioned facial recognition decision-making module 1230 includes:

[0171] The input unit is used to input the above-mentioned target face recognition features and the above-mentioned target scene parameters into the face recognition decision model. The above-mentioned scene sharing layer obtains the target common features based on the above-mentioned target face recognition features, and the above-mentioned scene parameter meta-learning unit obtains the first target scene features based on the above-mentioned target scene parameters.

[0172] The output unit is used to output the target face recognition result corresponding to the target face recognition user based on the above-mentioned common target features and the above-mentioned first target scene features.

[0173] The division of modules in the above-described facial recognition decision-making device is for illustrative purposes only. In other embodiments, the facial recognition decision-making device can be divided into different modules as needed to complete all or part of the functions of the facial recognition decision-making device. The implementation of each module in the facial recognition decision-making device provided in the embodiments of this specification can be in the form of a computer program. This computer program can run on a terminal or server. The program modules constituted by this computer program can be stored in the memory of the terminal or server. When the computer program is executed by a processor, it implements all or part of the steps of the facial recognition decision-making method described in the embodiments of this specification.

[0174] Please refer to the following. Figure 13 , Figure 13 This specification provides an exemplary embodiment of a facial recognition decision-making model training device. For example... Figure 13 As shown, the facial recognition decision-making model training device 1300 includes:

[0175] The acquisition module 1310 is used to acquire facial recognition data of known facial recognition results in multiple scenarios; the facial recognition data includes facial recognition features of multiple facial recognition users in each of the multiple scenarios and scenario parameters corresponding to each of the multiple scenarios.

[0176] Training module 1320 is used to train the face recognition decision model based on the face recognition data mentioned above.

[0177] In one possible implementation, the acquisition module 1310 includes:

[0178] The first acquisition unit is used to acquire scene information corresponding to each of the multiple scenes; the scene information includes at least one of the following: the openness information of the scene, the age information and mask rate of the facial recognition user in the scene, and the geographical location information corresponding to the scene.

[0179] The determining unit is used to determine the scene parameters corresponding to each of the above multiple scenarios based on the scene information corresponding to each of the above multiple scenarios.

[0180] The second acquisition unit is used to acquire the facial features corresponding to each of the multiple facial recognition users with known facial recognition results in each of the above multiple scenarios.

[0181] In one possible implementation, the number of facial recognition users in each of the aforementioned scenarios exceeds the target number.

[0182] In one possible implementation, the aforementioned face recognition decision model includes a scene sharing layer and a scene parameter meta-learning unit; the scene sharing layer is used to learn the common features of face recognition under the aforementioned multiple scenes; the scene parameter meta-learning unit is used to display the scene information corresponding to each of the aforementioned multiple scenes, and to learn the correlation information between the aforementioned multiple scenes.

[0183] The division of modules in the above-described facial recognition decision-making model training device is for illustrative purposes only. In other embodiments, the facial recognition decision-making model training device can be divided into different modules as needed to complete all or part of the functions of the above-described facial recognition decision-making model training device. The implementation of each module in the facial recognition decision-making model training device provided in the embodiments of this specification can be in the form of a computer program. This computer program can run on a terminal or server. The program modules constituted by this computer program can be stored in the memory of the terminal or server. When the computer program is executed by a processor, it implements all or part of the steps of the facial recognition decision-making model training method described in the embodiments of this specification.

[0184] Please see Figure 14 , Figure 14 This is a schematic diagram of the structure of an electronic device provided as an exemplary embodiment of this specification. For example... Figure 14 As shown, the electronic device 1400 may include: at least one processor 1410, at least one communication bus 1420, a user interface 1430, at least one network interface 1440, and a memory 1450. The communication bus 1420 can be used to enable communication between the aforementioned components.

[0185] The user interface 1430 may include a display screen and a camera. Optionally, the user interface 1430 may also include a standard wired interface and a wireless interface.

[0186] The network interface 1440 may optionally include a Bluetooth module, a Near Field Communication (NFC) module, a Wireless Fidelity (Wi-Fi) module, etc.

[0187] The processor 1410 may include one or more processing cores. The processor 1410 connects to various parts within the electronic device 1400 using various interfaces and lines. It executes various functions and processes data of the routing electronic device 1400 by running or executing instructions, programs, code sets, or instruction sets stored in the memory 1450, and by calling data stored in the memory 1450. Optionally, the processor 1410 may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 1410 may integrate one or a combination of several of the following: a Central Processing Unit (CPU), a Graphics Processing Unit (GPU), and a modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content required for display; and the modem handles wireless communication. It is understood that the modem may also be implemented as a separate chip without being integrated into the processor 1410.

[0188] The memory 1450 may include random access memory (RAM) or read-only memory (ROM). Optionally, the memory 1450 may include a non-transitory computer-readable medium. The memory 1450 may be used to store instructions, programs, code, code sets, or instruction sets. The memory 1450 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as an acquisition function, a facial recognition decision function, an update function, etc.), instructions for implementing the various method embodiments described above, etc.; the data storage area may store data involved in the various method embodiments described above, etc. Optionally, the memory 1450 may also be at least one storage device located remotely from the aforementioned processor 1410. Figure 14 As shown, the memory 1450, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and application programs.

[0189] Specifically, the electronic device 1400 can be the aforementioned facial recognition decision-making device, in Figure 14In the illustrated electronic device 1400, the user interface 1430 is mainly used to provide an input interface for the user, such as buttons or a camera on the facial recognition decision-making device, to obtain user-triggered commands; while the processor 1410 can be used to call the application program stored in the memory 1450 and specifically perform the following operations:

[0190] The system acquires the target facial recognition features corresponding to the target facial recognition user based on the facial recognition device; acquires the target scene parameters corresponding to the aforementioned facial recognition device; inputs the aforementioned target facial recognition features and the aforementioned target scene parameters into the facial recognition decision model, and outputs the target facial recognition result corresponding to the aforementioned target facial recognition user; the aforementioned facial recognition decision model is trained based on facial recognition data with known facial recognition results in multiple scenarios; the aforementioned facial recognition data includes the facial recognition features corresponding to each of the multiple facial recognition users in the aforementioned scenarios and the scene parameters corresponding to the aforementioned scenarios.

[0191] In some possible embodiments, the above-mentioned target facial recognition features include target real-time facial recognition features and target offline general features;

[0192] When the processor 1410 executes the process of obtaining the target facial recognition features corresponding to the target facial recognition user based on the facial recognition device, it is specifically used to perform: obtaining the target real-time facial recognition features corresponding to the target facial recognition user based on the facial recognition device; and determining the target offline general features based on the target real-time facial recognition features.

[0193] In some possible embodiments, the aforementioned target facial recognition real-time features include N target facial recognition comparison results; the aforementioned N target facial recognition comparison results are the facial recognition comparison results corresponding to the N target users with the highest facial recognition comparison scores among the facial recognition comparison results between the target facial recognition image and the facial images of multiple users; the aforementioned target facial recognition image is obtained by capturing the aforementioned target facial recognition user based on the aforementioned facial recognition device; the aforementioned N is a positive integer.

[0194] In some possible embodiments, the target facial recognition comparison result includes a target facial recognition comparison score; the target offline general features include offline general information corresponding to the M target users with the highest target facial recognition comparison scores among the N target users; the offline general information includes at least one of the following: the target user's first historical facial recognition information on the facial recognition device, the target user's historical activity information within the first preset range on the facial recognition device, the target user's second historical facial recognition information, and the target user's facial recognition information; the first historical facial recognition information includes the target user's historical facial recognition frequency and / or historical facial recognition comparison score range on the facial recognition device; the second historical facial recognition information includes the target user's historical facial recognition behavior information; and M is a positive integer less than or equal to N.

[0195] In some possible embodiments, the above-mentioned electronic device 1400 may be a facial recognition device;

[0196] When the processor 1410 executes the process of obtaining the target scene parameters corresponding to the facial recognition device, it is specifically used to perform: receiving the server to determine the target scene parameters corresponding to the facial recognition device based on the first target scene information corresponding to the facial recognition device; the first target scene information includes at least one of the following: the openness information of the first target scene corresponding to the facial recognition device, the age information and mask rate of the facial recognition user in the first target scene, and the geographical location information corresponding to the facial recognition device.

[0197] In some possible embodiments, when the processor 1410 executes the process of obtaining the target scene parameters corresponding to the facial recognition device, it is specifically used to perform: obtaining the first target scene information corresponding to the facial recognition device; the first target scene information includes at least one of the following: the openness information of the first target scene corresponding to the facial recognition device, the age information and mask rate of the facial recognition user in the first target scene, and the geographical location information corresponding to the facial recognition device; and determining the target scene parameters corresponding to the facial recognition device based on the first target scene information.

[0198] In some possible embodiments, the first target scenario corresponding to the above-mentioned facial recognition device is a scenario without historical facial recognition data;

[0199] When the processor 1410 executes the process of obtaining the target scene parameters corresponding to the facial recognition device, it is specifically used to perform: determining the target scene parameters corresponding to the facial recognition device based on the same type of scene corresponding to the first target scene; or determining the target scene parameters corresponding to the facial recognition device based on the scene parameters of other scenes within the second preset range of the facial recognition device.

[0200] In some possible embodiments, after the processor 1410 inputs the target facial recognition features and the target scene parameters into the facial recognition decision model and outputs the target facial recognition result corresponding to the target facial recognition user, it is further configured to: obtain second target scene information corresponding to the facial recognition device; the second target scene information includes at least one of the following: openness information of the second target scene corresponding to the facial recognition device, age information and mask rate of the facial recognition user in the second target scene, and geographical location information of the facial recognition device in the second target scene; and update the target scene parameters based on the second target scene information.

[0201] In some possible embodiments, the above-described face recognition decision model includes a scene sharing layer and a scene parameter meta-learning unit;

[0202] When the processor 1410 executes the process of inputting the target facial recognition features and the target scene parameters into the facial recognition decision model and outputting the target facial recognition result corresponding to the target facial recognition user, it specifically performs the following: inputting the target facial recognition features and the target scene parameters into the facial recognition decision model; the scene sharing layer obtains target common features based on the target facial recognition features; the scene parameter meta-learning unit obtains first target scene features based on the target scene parameters; and outputs the target facial recognition result corresponding to the target facial recognition user based on the target common features and the first target scene features.

[0203] In some possible embodiments, the electronic device 1400 may be the aforementioned face recognition decision model training device, and the processor 1410 may further perform the following: acquiring face recognition data of known face recognition results in multiple scenarios; the face recognition data includes face recognition features corresponding to multiple face recognition users in each of the multiple scenarios and scenario parameters corresponding to each scenario; and training a face recognition decision model based on the face recognition data.

[0204] In some possible embodiments, when the processor 1410 executes the process of acquiring face recognition data with known face recognition results in multiple scenarios, it is specifically used to perform the following: acquiring scene information corresponding to each of the multiple scenarios; the scene information includes at least one of the following: the openness information of the scene, the age information and mask rate of the face recognition user in the scene, and the geographical location information corresponding to the scene; determining the scene parameters corresponding to each of the multiple scenarios based on the scene information corresponding to each of the multiple scenarios; and acquiring the face recognition features corresponding to each of the multiple face recognition users with known face recognition results in each of the multiple scenarios.

[0205] In some possible embodiments, the number of facial recognition users in each of the above scenarios is greater than the target number.

[0206] In some possible embodiments, the face recognition decision model includes a scene sharing layer and a scene parameter meta-learning unit; the scene sharing layer is used to learn the common features of face recognition under the multiple scenarios; the scene parameter meta-learning unit is used to display the scene information corresponding to each of the multiple scenarios, and to learn the correlation information between the multiple scenarios.

[0207] This specification also provides a computer-readable storage medium storing instructions that, when executed on a computer or processor, cause the computer or processor to perform one or more steps in the above embodiments. If the constituent modules of the above-described facial recognition decision-making device and facial recognition decision-making training device are implemented as software functional units and sold or used as independent products, they can be stored in the above-described computer-readable storage medium.

[0208] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When these computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this specification are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in or transmitted through a computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. The aforementioned available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., Digital Versatile Discs (DVDs)), or semiconductor media (e.g., Solid State Disks (SSDs)).

[0209] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. The aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks. Unless otherwise specified, the technical features of this embodiment and its implementation can be combined arbitrarily.

[0210] The embodiments described above are merely preferred embodiments of this specification and are not intended to limit the scope of this specification. Any modifications and improvements made by those skilled in the art to the technical solutions of this specification without departing from the spirit of this specification should fall within the protection scope defined by the claims.

[0211] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims and specification may be performed in a different order than in the embodiments described in the specification and still achieve the desired results. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A facial recognition decision-making method, the method comprising: Based on the facial recognition device, obtain the target facial recognition features corresponding to the target facial recognition user; Obtain the target scene parameters corresponding to the facial recognition device; The target scene parameters are determined based on the first target scene information corresponding to the facial recognition device; the first target scene information includes at least one of the following: the openness information of the first target scene corresponding to the facial recognition device, the age information and mask rate of the facial recognition user in the first target scene, and the geographical location information corresponding to the facial recognition device; The openness information is used to characterize the changes in facial recognition users in the first target scenario; The target facial recognition features and the target scene parameters are input into the facial recognition decision model, and the target facial recognition result corresponding to the target facial recognition user is output. The facial recognition decision model is trained based on facial recognition data with known facial recognition results in multiple scenarios; the facial recognition data includes the facial recognition features of each facial recognition user in the scenario and the scenario parameters corresponding to the scenario. The facial recognition decision-making model includes a scene sharing layer and a scene parameter meta-learning unit; The step of inputting the target facial recognition features and the target scene parameters into the facial recognition decision model and outputting the target facial recognition result corresponding to the target facial recognition user includes: The target facial recognition features and the target scene parameters are input into the facial recognition decision model. The scene sharing layer obtains the target common features based on the target facial recognition features, and the scene parameter meta-learning unit obtains the first target scene features based on the target scene parameters. Based on the common features of the target and the first target scene features, the target face recognition result corresponding to the target face recognition user is output.

2. The method as described in claim 1, wherein the target facial recognition features include target real-time facial recognition features and target offline general features; The step of obtaining the target facial recognition features corresponding to the target facial recognition user based on the facial recognition device includes: Based on the facial recognition device, obtain the real-time facial recognition features corresponding to the target facial recognition user; The target's offline general features are determined based on the target's real-time facial recognition features.

3. The method as described in claim 2, wherein the real-time features of the target facial recognition include N target facial recognition comparison results; the N target facial recognition comparison results are the facial recognition comparison results of the N target users with the highest facial recognition comparison scores among the facial recognition comparison results between the target facial recognition image and the facial images of multiple users; the target facial recognition image is obtained by capturing the target facial recognition user based on the facial recognition device; and N is a positive integer.

4. The method as described in claim 3, wherein the target facial recognition comparison result includes a target facial recognition comparison score; the target offline general features include offline general information corresponding to the M target users with the highest target facial recognition comparison scores among the N target users; the offline general information includes at least one of the following: the target user's first historical facial recognition information on the facial recognition device, the target user's historical activity information within a first preset range on the facial recognition device, the target user's second historical facial recognition information, and the target user's facial recognition information; the first historical facial recognition information includes the target user's historical facial recognition frequency and / or historical facial recognition comparison score range on the facial recognition device; the second historical facial recognition information includes the target user's historical facial recognition behavior information; and M is a positive integer less than or equal to N.

5. The method according to any one of claims 1-4, wherein the method is applied to the facial recognition device; The step of obtaining the target scene parameters corresponding to the facial recognition device includes: The receiving server determines the target scene parameters corresponding to the face recognition device based on the first target scene information corresponding to the face recognition device.

6. The method according to any one of claims 1-4, wherein obtaining the target scene parameters corresponding to the facial recognition device includes: Obtain the first target scene information corresponding to the facial recognition device; The target scene parameters corresponding to the facial recognition device are determined based on the first target scene information.

7. The method according to any one of claims 1-4, wherein the first target scenario corresponding to the facial recognition device is a scenario without historical facial recognition data; The step of obtaining the target scene parameters corresponding to the facial recognition device includes: The target scene parameters corresponding to the facial recognition device are determined based on the same type of scene corresponding to the first target scene. or The target scene parameters corresponding to the face recognition device are determined based on the scene parameters of other scenes within the second preset range of the face recognition device.

8. The method according to any one of claims 1-4, wherein after inputting the target facial recognition features and the target scene parameters into the facial recognition decision model and outputting the target facial recognition result corresponding to the target facial recognition user, the method further includes: Obtain the second target scene information corresponding to the facial recognition device; The second target scene information includes at least one of the following: the openness information of the second target scene corresponding to the facial recognition device, the age information and mask rate of the facial recognition user in the second target scene, and the geographical location information of the facial recognition device in the second target scene; The target scene parameters are updated based on the second target scene information.

9. A method for training a facial recognition decision-making model, the method comprising: Acquire facial recognition data from known facial recognition results in multiple scenarios; the facial recognition data includes facial recognition features corresponding to multiple facial recognition users in each scenario and scenario parameters corresponding to each scenario; the scenario parameters are determined based on scenario information corresponding to the scenario; the scenario information includes at least one of the following: the openness information of the scenario, the age information and mask rate of facial recognition users in the scenario, and the geographical location information corresponding to the scenario; The openness information is used to characterize the changes in facial recognition users in the scenario. A face recognition decision model is trained based on the face recognition data; the face recognition decision model includes a scene sharing layer and a scene parameter meta-learning unit. The scene sharing layer is used to learn the common features of face recognition in the multiple scenes; the scene parameter meta-learning unit is used to display the scene information corresponding to each of the multiple scenes, and to learn the correlation information between the multiple scenes.

10. The method as described in claim 9, wherein acquiring facial recognition data with known facial recognition results in multiple scenarios includes: Obtain scene information corresponding to each of the multiple scenes; Determine the scene parameters corresponding to each of the multiple scenarios based on the scene information corresponding to each of the multiple scenarios. Obtain the facial recognition features corresponding to each of the multiple facial recognition users with known facial recognition results in each of the multiple scenarios.

11. The method as described in claim 9, wherein the number of facial recognition users in each of the plurality of scenarios is greater than the target number.

12. A facial recognition decision-making device, the device comprising: The first acquisition module is used to acquire the target facial recognition features corresponding to the target facial recognition user based on the facial recognition device. The second acquisition module is used to acquire the target scene parameters corresponding to the face recognition device; The target scene parameters are determined based on the first target scene information corresponding to the facial recognition device; the first target scene information includes at least one of the following: the openness information of the first target scene corresponding to the facial recognition device, the age information and mask rate of the facial recognition user in the first target scene, and the geographical location information corresponding to the facial recognition device; The openness information is used to characterize the changes in facial recognition users in the first target scenario; The face recognition decision module is used to input the target face recognition features and the target scene parameters into the face recognition decision model and output the target face recognition result corresponding to the target face recognition user. The facial recognition decision model is trained based on facial recognition data with known facial recognition results in multiple scenarios; the facial recognition data includes the facial recognition features of each facial recognition user in the scenario and the scenario parameters corresponding to the scenario. The facial recognition decision-making model includes a scene sharing layer and a scene parameter meta-learning unit; the facial recognition decision-making module is specifically used for: The target facial recognition features and the target scene parameters are input into the facial recognition decision model. The scene sharing layer obtains the target common features based on the target facial recognition features, and the scene parameter meta-learning unit obtains the first target scene features based on the target scene parameters. Based on the common features of the target and the first target scene features, the target face recognition result corresponding to the target face recognition user is output.

13. A facial recognition decision-making model training device, the device comprising: The acquisition module is used to acquire facial recognition data of known facial recognition results in multiple scenarios; the facial recognition data includes facial recognition features corresponding to multiple facial recognition users in each scenario and scenario parameters corresponding to each scenario; the scenario parameters are determined based on scenario information corresponding to the scenario; the scenario information includes at least one of the following: the openness information of the scenario, the age information and mask rate of facial recognition users in the scenario, and the geographical location information corresponding to the scenario; The openness information is used to characterize the changes in facial recognition users in the scenario. The training module is used to train a face recognition decision model based on the face recognition data; the face recognition decision model includes a scene sharing layer and a scene parameter meta-learning unit; The scene sharing layer is used to learn the common features of face recognition in the multiple scenes; the scene parameter meta-learning unit is used to display the scene information corresponding to each of the multiple scenes, and to learn the correlation information between the multiple scenes.

14. An electronic device, comprising: Processor and memory; The processor is connected to the memory; The memory is used to store executable program code; The processor runs a program corresponding to the executable program code stored in the memory to perform the method as described in any one of claims 1-8 or 9-11.

15. A computer storage medium storing a plurality of instructions adapted for loading by a processor and performing the steps of the method as claimed in any one of claims 1-8 or 9-11.

16. A computer program product comprising instructions that, when run on a computer or processor, cause the computer or processor to perform the method as described in any one of claims 1-8 or 9-11.

Images