Model optimization method, device, computer equipment, storage medium and program product

Abstract

The present application relates to a model optimization method, device, computer equipment, storage medium and program product. First plane images at multiple angles are obtained based on the acquired three-dimensional user portrait of the user, and second plane images corresponding to each first scene factor combination are obtained based on the background feature information of the user's target image and the facial feature information of each first plane image. The second plane images corresponding to each first scene factor combination are input into the face recognition model to obtain the first face recognition accuracy corresponding to each first scene factor combination, thereby optimizing the face recognition model according to the first face recognition accuracy corresponding to each first scene factor combination. The method fuses the background feature information of the user's target image and the facial feature information of each first plane image to obtain the second plane image of each first scene factor combination, enriching the real test scene, improving the face recognition accuracy, and enhancing the security of face recognition.

Description

Technical Field

[0001] The present application relates to the fields of information security technology and artificial intelligence, and in particular to a model optimization method, device, computer equipment, storage medium and program product. Background Art

[0002] The transformation of facial recognition services involves complex test scenarios. Typical test scenarios include recognition of facial makeup, expression recognition of joy, anger, sorrow, and happiness, and recognition of facial features with open mouth and closed eyes; as well as external environments with different lighting conditions such as backlight and darkness and different temperature changes; different skin colors such as black, yellow, brown, and white, twins, and pregnant women based on crowd attributes; and recognition of mask wearing under the global normalized epidemic situation, which makes facial recognition more difficult.

[0003] Existing face recognition technologies mainly improve the recognition ability of target objects through methods such as liveness detection, continuous detection and 3D detection, or perform security checks on biometric data through encryption signature authentication, etc. However, existing face recognition methods lack the design and detection of real test scenarios, resulting in low face recognition accuracy in real usage environments. Summary of the invention

[0004] Based on this, it is necessary to provide a model optimization method, device, computer equipment, storage medium and program product that can improve the accuracy of face recognition in response to the above technical problems.

[0005] In a first aspect, the present application provides a model optimization method, the method comprising:

[0006] Obtaining first plane images at multiple angles according to the acquired three-dimensional user portrait of the user;

[0007] Obtaining second plane images corresponding to each first scene factor combination according to background feature information of the target image of the user and face feature information of each first plane image;

[0008] Inputting the second plane images corresponding to each of the first scene factor combinations into a face recognition model to obtain the first face recognition accuracy rates corresponding to each of the first scene factor combinations;

[0009] The face recognition model is optimized according to the first face recognition accuracy rates corresponding to the first scene factor combinations.

[0010] In one embodiment, obtaining the second plane images corresponding to each first scene factor combination according to the background feature information of the target image of the user and the facial feature information of each first plane image includes:

[0011] Performing a change process on the facial feature information of each of the first plane images to obtain a plurality of third plane images;

[0012] Performing a change process on the background feature information of the target image of the user to obtain a plurality of fourth plane images;

[0013] According to the plurality of third planar images and the plurality of fourth planar images, second planar images corresponding to each combination of the first scene factors are obtained.

[0014] In one embodiment, optimizing the face recognition model according to the first face recognition accuracy corresponding to each first scene factor combination includes:

[0015] Determining a second face recognition accuracy rate from the first face recognition accuracy rates corresponding to each of the first scene factor combinations, wherein the second face recognition accuracy rate is a first face recognition accuracy rate that is less than a preset accuracy rate threshold;

[0016] The face recognition model is optimized according to a second scene factor combination, wherein the second scene factor combination is a first scene factor combination corresponding to the second face recognition rate.

[0017] In one embodiment, the optimizing the face recognition model according to the second scene factor combination includes:

[0018] Adding a new scene factor to the second scene factor combination to obtain a third scene factor combination;

[0019] Inputting the second plane image corresponding to the third scene factor combination into the face recognition model to obtain a third face recognition accuracy rate corresponding to the third scene factor combination;

[0020] The face recognition model is optimized according to the second face recognition accuracy and the third face recognition accuracy.

[0021] In one embodiment, optimizing the face recognition model according to the second face recognition accuracy and the third face recognition accuracy includes:

[0022] Executing a first iterative process, the first iterative process comprising: determining a deviation rate corresponding to the new scene factor according to the second face recognition accuracy rate and the third face recognition accuracy rate, and correcting the third face recognition accuracy rate according to the deviation rate to obtain a comparative face recognition accuracy rate, taking the maximum face recognition accuracy rate between the second face recognition accuracy rate and the comparative face recognition accuracy rate as a target face recognition accuracy rate for the third scene factor combination, and adding a new scene factor to the third scene factor combination to obtain a fourth scene factor combination;

[0023] Inputting the second plane image corresponding to the fourth scene factor combination into the face recognition model, obtaining a fourth face recognition accuracy rate corresponding to the fourth scene factor combination, replacing the second face recognition accuracy rate with the target face recognition accuracy rate, replacing the third face recognition accuracy rate with the fourth face recognition accuracy rate, and returning to execute the first iteration process until the number of added new scene factors is equal to the preset number of new scene factors;

[0024] For the same new scene factor, determining the ratio of the number of deviation rates that have a negative impact on the scene factor combination before the new scene factor is added to the total number of the second scene factor combinations;

[0025] If the ratio is greater than a preset ratio threshold, the face recognition model is optimized according to a new scene factor corresponding to the ratio greater than the preset ratio threshold.

[0026] In one of the embodiments, a plurality of image frames are extracted from the user's video at intervals of a preset number of frames;

[0027] Performing pixel transformation processing on each pixel point of each image frame to obtain an image frame after pixel transformation processing;

[0028] The target image is obtained by filtering out noise in the image frame after the pixel transformation process.

[0029] In one embodiment, the pixel transformation processing of each pixel point of each image frame includes at least one of the following:

[0030] Performing geometric transformation processing on each pixel point of each image frame;

[0031] Threshold processing is performed on each pixel point of each image frame to divide each pixel point of the image frame into different pixel value intervals.

[0032] In a second aspect, the present application also provides a model optimization device, the device comprising:

[0033] An acquisition module, used to obtain first plane images at multiple angles according to the acquired three-dimensional user portrait of the user;

[0034] A first determination module, configured to obtain second plane images corresponding to each first scene factor combination according to background feature information of the user's target image and face feature information of each of the first plane images;

[0035] A second determination module is used to input the second plane images corresponding to each of the first scene factor combinations into a face recognition model to obtain a first face recognition accuracy rate corresponding to each of the first scene factor combinations;

[0036] The optimization module is used to optimize the face recognition model according to the first face recognition accuracy corresponding to each combination of the first scene factors.

[0037] In a third aspect, the present application further provides a computer device, the computer device comprising a memory and a processor, the memory storing a computer program, and the processor implementing the following steps when executing the computer program:

[0038] Obtaining first plane images at multiple angles according to the acquired three-dimensional user portrait of the user;

[0039] Obtaining second plane images corresponding to each first scene factor combination according to background feature information of the target image of the user and face feature information of each first plane image;

[0040] Inputting the second plane images corresponding to each of the first scene factor combinations into a face recognition model to obtain the first face recognition accuracy rates corresponding to each of the first scene factor combinations;

[0041] The face recognition model is optimized according to the first face recognition accuracy rates corresponding to the first scene factor combinations.

[0042] In a fourth aspect, the present application further provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0043] Obtaining first plane images at multiple angles according to the acquired three-dimensional user portrait of the user;

[0044] Obtaining second plane images corresponding to each first scene factor combination according to background feature information of the target image of the user and face feature information of each first plane image;

[0045] Inputting the second plane images corresponding to each of the first scene factor combinations into a face recognition model to obtain the first face recognition accuracy rates corresponding to each of the first scene factor combinations;

[0046] The face recognition model is optimized according to the first face recognition accuracy rates corresponding to the first scene factor combinations.

[0047] In a fifth aspect, the present application further provides a computer program product, the computer program product comprising a computer program, which implements the following steps when executed by a processor:

[0048] Obtaining first plane images at multiple angles according to the acquired three-dimensional user portrait of the user;

[0049] Obtaining second plane images corresponding to each first scene factor combination according to background feature information of the target image of the user and face feature information of each first plane image;

[0050] Inputting the second plane images corresponding to each of the first scene factor combinations into a face recognition model to obtain the first face recognition accuracy rates corresponding to each of the first scene factor combinations;

[0051] The face recognition model is optimized according to the first face recognition accuracy rates corresponding to the first scene factor combinations.

[0052] The above-mentioned model optimization method, device, computer equipment, storage medium and program product obtain first plane images of multiple angles according to the acquired three-dimensional user portrait of the user, and obtain second plane images corresponding to each first scene factor combination according to the background feature information of the user's target image and the facial feature information of each first plane image, input the second plane images corresponding to each first scene factor combination into the face recognition model, obtain the first face recognition accuracy corresponding to each first scene factor combination, and thus optimize the face recognition model according to the first face recognition accuracy corresponding to each first scene factor combination. This method fuses the background feature information of the user's target image and the facial feature information of each first plane image to obtain second plane images of multiple first scene factor combinations, enriching the real test scene, and optimizing the face recognition model improves the face recognition accuracy of the real test scene and enhances the security of face recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 An application environment diagram of a model optimization method in an embodiment;

[0054] Figure 2 A schematic diagram of a flow chart of a model optimization method in one embodiment;

[0055] Figure 3 is a schematic diagram of a process of determining a second plane image in one embodiment;

[0056] Figure 4 A schematic diagram of a process for optimizing a face recognition model in one embodiment;

[0057] Figure 5 A schematic diagram of a process for optimizing a face model in another embodiment;

[0058] Figure 6 A schematic diagram of a process for optimizing a face model in yet another embodiment;

[0059] Figure 7 A schematic diagram of a process for determining a target image in one embodiment;

[0060] Figure 8 It is a structural block diagram of a model optimization device in one embodiment. DETAILED DESCRIPTION

[0061] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0062] Among the related technologies, the mainstream face recognition technology improves the recognition ability of living objects through methods such as liveness detection, continuous detection and 3D detection, or performs security checks on biometric data through encryption signature authentication. However, the lack of design and detection methods for real test scenarios makes it impossible for testers to fully verify the modification points, which is mainly reflected in the following two aspects:

[0063] The first aspect: there is a lack of test data and scenario manufacturing for the entire scenario, and only some pass verifications under normal test scenarios are verified. If multiple verifications fail, it is recommended to try verification through other channels.

[0064] The second aspect: only a trade-off can be made between system security and accuracy.

[0065] At present, due to the lack of a complete face recognition transformation test mechanism method, the user complaint rate for failing to pass the face scan is high in the real production environment. Due to the limitations of the current test method, the pass rate can only be improved by lowering the similarity threshold (such as from 98% to 95%), which is obviously a major risk to security. In response to the above technical problems, the present application provides a model optimization method, device, computer equipment, storage medium and program product.

[0066] The model optimization method provided in the embodiment of the present application can be applied to Figure 1 The application environment includes a computer device, which may be a server, and its internal structure diagram may be as shown in FIG. Figure 1 As shown. The computer device includes a processor, a memory and a network interface connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store user face recognition related data. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a model optimization method is implemented. The server can be implemented as an independent server or a server cluster consisting of multiple servers.

[0067] In one embodiment, Figure 2 As shown, a model optimization method is provided, which is applied to Figure 1 The computer device in the example is used to illustrate, including the following steps:

[0068] S201, obtaining first plane images at multiple angles according to the acquired three-dimensional user portrait of the user.

[0069] In this embodiment, the AI ​​intelligent completion algorithm can be used to match and fuse the collected photos with the head information library data to achieve head reconstruction and obtain the user's three-dimensional user portrait. Based on the user's three-dimensional user portrait, plane photos at different angles are collected as the first plane images at multiple angles.

[0070] S202, obtaining second plane images corresponding to each first scene factor combination according to background feature information of the user's target image and facial feature information of each first plane image.

[0071] In this embodiment, the facial feature information of each first plane image can be transformed to obtain each third plane image, and the background feature information of the user's target image and each third plane image can be fused to obtain the second plane image corresponding to each first scene factor combination. The background feature information of the user's target image can also be transformed to obtain each fourth plane image, and each fourth plane image can be fused with each third plane image to obtain the second plane image corresponding to each first scene factor combination.

[0072] S203: Input the second plane images corresponding to the first scene factor combinations into a face recognition model to obtain first face recognition accuracy rates corresponding to the first scene factor combinations.

[0073] Optionally, the face recognition model may be a liveness detection, a continuous detection, a 3D detection method, etc., which is not limited in the embodiments of the present application.

[0074] In this embodiment, the second plane image corresponding to each combination of scene factors is input into the face recognition model, and the 1:1 and 1:N scene tests of the face popular in the industry are mainly performed. The results of the recognition process are recorded, and the face recognition accuracy corresponding to each combination of scene factors is statistically calculated. For example, the second plane image corresponding to the scene factor combination of [closed eyes + mask] is input into the face recognition model, and the face recognition accuracy corresponding to the scene factor combination is 75%.

[0075] S204: Optimize the face recognition model according to the first face recognition accuracy rates corresponding to the first scene factor combinations.

[0076] In this embodiment, from the first face recognition accuracy rates corresponding to the first scene factor combinations, the first scene factor combinations corresponding to the first face recognition accuracy rates less than the preset threshold are selected, and the face recognition model is optimized according to the first scene factor combinations corresponding to the first face recognition accuracy rates less than the preset threshold. It is also possible to add new scene factors to the first scene factor combinations corresponding to the first face recognition accuracy rates less than the preset threshold, determine the scene factors that have a negative impact on the first face recognition accuracy rates less than the preset threshold, and optimize the scene factors with negative impacts, so as to achieve the purpose of optimizing the face recognition model.

[0077] In the above model optimization method, first plane images of multiple angles are obtained according to the three-dimensional user portrait of the user, and second plane images corresponding to each first scene factor combination are obtained according to the background feature information of the user's target image and the facial feature information of each first plane image. The second plane images corresponding to each first scene factor combination are input into the face recognition model to obtain the first face recognition accuracy corresponding to each first scene factor combination, thereby optimizing the face recognition model according to the first face recognition accuracy corresponding to each first scene factor combination. This method fuses the background feature information of the user's target image and the facial feature information of each first plane image to obtain second plane images of multiple first scene factor combinations, enriching the real test scene. Moreover, the optimization of the face recognition model improves the face recognition accuracy of the real test scene and enhances the security of face recognition.

[0078] Figure 3 FIG. 1 is a schematic diagram of a process for determining a second plane image in an embodiment. Figure 3 As shown, the embodiment of the present application relates to a possible implementation method of how to obtain the second plane image corresponding to each first scene factor combination according to the background feature information of the user's target image and the facial feature information of each first plane image, including the following steps:

[0079] S301, performing a change process on the facial feature information of each first plane image to obtain a plurality of third plane images.

[0080] Optionally, the facial feature information includes makeup information (heavy makeup, light makeup, beauty contact lens and foundation), expression information of joy, anger, sorrow, and happiness, facial features information with open mouth and closed eyes, black, yellow, brown, and white skin color information, or facial information of twins with a high degree of similarity and facial feature information of pregnant women.

[0081] In this embodiment, the first plane image can be input into a deep learning network, and the first plane image can be recognized by the deep learning network, so that the various regions of the first plane image can be divided to achieve the distinction between various parts of the face. Furthermore, by transforming the pixel values ​​of different regions of the first plane image, the change processing of facial feature information is achieved, thereby obtaining multiple third plane images.

[0082] Optionally, after using a machine learning algorithm to identify the first plane image, the pixel values ​​of different regions of the first plane image are transformed to achieve a change in facial feature information, thereby obtaining multiple third plane images.

[0083] Optionally, when the facial feature information of each first plane image is changed, only one of the feature information may be changed, or multiple feature information may be changed.

[0084] S302, performing change processing on the background feature information of the user's target image to obtain a plurality of fourth plane images.

[0085] Optionally, the background feature information includes lighting information such as backlight, darkness, different light sources and different colors of light, temperature information of different temperature changes, whether a mask is worn, etc.

[0086] In this embodiment, the change process of the background feature information of the user's target image may utilize the specific implementation method of S301 described above.

[0087] S303: Obtain second plane images corresponding to each first scene factor combination according to the plurality of third plane images and the plurality of fourth plane images.

[0088] In this embodiment, the third plane image and the fourth plane image may be randomly matched to fuse the image information, thereby obtaining the second plane image corresponding to each combination of scene factors.

[0089] Optionally, the third plane image with the scene factor of heavy makeup and the background feature information with the scene factor of darkness may be fused to obtain a second plane image with the scene factor combination of [heavy makeup + darkness]; or the third plane image with the scene factor of heavy makeup and the fourth plane image with the scene factors of darkness and backlight may be fused to obtain a second plane image with the scene factor combination of [heavy makeup + darkness + backlight].

[0090] In the embodiment of the present application, multiple third plane images are obtained by changing the facial feature information of each first plane image and multiple fourth plane images are obtained by changing the background feature information of the user's target image, and the multiple third plane images and the multiple fourth plane images are fused to obtain the second plane image corresponding to each scene factor combination. This method fuses the facial feature information with the background information to obtain image information of multiple scene factor combinations, enriching the test scenarios of face recognition.

[0091] Figure 4 FIG. 1 is a flow chart of optimizing a face recognition model in one embodiment. Figure 4 As shown, the embodiment of the present application relates to a possible implementation method of optimizing a face recognition model according to the first face recognition accuracy corresponding to each first scene factor combination, including the following steps:

[0092] S401, determining a second face recognition accuracy rate from the first face recognition accuracy rates corresponding to each first scene factor combination, wherein the second face recognition accuracy rate is a first face recognition accuracy rate that is less than a preset accuracy threshold.

[0093] Optionally, the preset accuracy threshold may be 50%, 60%, 90%, etc.

[0094] In this embodiment, according to the first face recognition accuracy rates corresponding to the first scene factor combinations obtained above, the first face recognition accuracy rate that is less than the preset accuracy threshold is determined, and the first face recognition accuracy rate that is less than the preset accuracy threshold is used as the second face recognition accuracy rate.

[0095] S402, optimizing the face recognition model according to the second scene factor combination, wherein the second scene factor combination is the first scene factor combination corresponding to the second face recognition rate.

[0096] In this embodiment, the first scene factor combination corresponding to the second face recognition accuracy rate can be directly optimized to optimize the face recognition model. The second scene factor combination can also be further analyzed, and by adding new scene factors, the scene factors in the second scene factor combination that have a negative impact on the second face recognition accuracy rate are determined, and the scene factors with negative impact are optimized to optimize the face recognition model.

[0097] In the embodiment of the present application, the second face recognition accuracy is determined from the first face recognition accuracy corresponding to each first scene factor combination, and the face recognition model is optimized according to the first scene factor combination corresponding to the second face recognition accuracy. In this method, the first scene factor combination corresponding to the first face recognition accuracy less than the preset accuracy threshold is optimized, and the scene combination with low accuracy is actively identified to avoid potential production risks and optimize the user experience.

[0098] Figure 5 FIG. 1 is a flow chart of optimizing a face model in another embodiment. Figure 5 As shown, the embodiment of the present application relates to a possible implementation method of how to optimize the face recognition model according to the second scene factor combination, including the following steps:

[0099] S501: Add a new scene factor to the second scene factor combination to obtain a third scene factor combination.

[0100] Optionally, the new scene factor may be a facial feature information factor, namely the facial features information factor, expression information factor, etc., or a background feature information factor, namely a lighting information factor, temperature information factor, etc.

[0101] In this embodiment, a new scene factor is added to the second scene factor combination, the second scene factor combination is [heavy makeup+darkness], and a new scene factor opening mouth is added, that is, the third scene factor combination is [heavy makeup+darkness+open mouth].

[0102] S502: Input the second plane image corresponding to the third scene factor combination into a face recognition model to obtain a third face recognition accuracy rate corresponding to the third scene factor combination.

[0103] In this embodiment, the specific implementation method can refer to S203.

[0104] S503, optimizing the face recognition model according to the second face recognition accuracy and the third face recognition accuracy.

[0105] In this embodiment, it is possible to directly determine whether the added new scene factor affects the second face recognition accuracy based on the second face recognition accuracy and the third face recognition accuracy, select the largest face recognition accuracy between the second face recognition accuracy and the third face recognition accuracy as the new face recognition accuracy, continue to add the second new scene factor, compare the new face recognition accuracy with the face recognition accuracy after adding the second new scene factor, and determine whether the added second new scene factor affects the new face recognition accuracy, until all influencing scene factors are determined.

[0106] In this embodiment, the face recognition model can also be optimized according to S601-S604.

[0107] In the embodiment of the present application, a third scene factor combination is obtained by adding a new scene factor to the second scene factor combination, and a second plane image corresponding to the third scene factor combination is input into a face recognition model to obtain a third face recognition accuracy rate corresponding to the third scene factor combination, and then the face recognition model is optimized according to the second face recognition accuracy rate and the third face recognition accuracy rate. This method obtains face recognition accuracy rates of different scene factor combinations by positively enumerating different scene factor combinations, and optimizes the face recognition model by face recognition accuracy rates of different scene factor combinations. The method is simple and efficient.

[0108] Figure 6 FIG. 1 is a flow chart of optimizing a face model in another embodiment. Figure 6 As shown, the embodiment of the present application relates to a possible implementation method of how to optimize the face recognition model according to the second face recognition accuracy and the third face recognition accuracy, including the following steps:

[0109] S601, execute the first iterative process, the first iterative process includes: determining the deviation rate corresponding to the new scene factor according to the second face recognition accuracy rate and the third face recognition accuracy rate, and correcting the third face recognition accuracy rate according to the deviation rate to obtain a comparative face recognition accuracy rate, taking the maximum face recognition accuracy rate between the second face recognition accuracy rate and the comparative face recognition accuracy rate as the target face recognition accuracy rate of the third scene factor combination, and adding the new scene factor to the third scene factor combination to obtain a fourth scene factor combination.

[0110] The deviation rate is used to describe the impact of the new scene factor on the correctness of the current scene factor combination, which can be positive, negative, or no impact. Optionally, if it is a positive impact, the deviation rate can be 0.8, if it is no impact, the value is 0.5, and if it is a negative impact, the value is 0.3.

[0111] In this embodiment, assuming that the third face recognition accuracy is 80%, which is less than the second face recognition accuracy of 82%, the deviation rate of the new scene factor for the second scene factor combination is 0.3, the third face recognition accuracy is added to the deviation rate, and the third face recognition accuracy is corrected to obtain a comparative face recognition accuracy of 83%. The largest face recognition accuracy is selected from the second face recognition accuracy and the comparative face recognition accuracy as the target face recognition accuracy of the third scene factor combination, that is, the target face recognition accuracy is 83%.

[0112] Furthermore, new scene factors are continuously added to the third scene factor combination to obtain a fourth scene factor combination.

[0113] S602, input the second plane image corresponding to the fourth scene factor combination into the face recognition model, obtain the fourth face recognition accuracy corresponding to the fourth scene factor combination, replace the second face recognition accuracy with the target face recognition accuracy, replace the third face recognition accuracy with the fourth face recognition accuracy, and return to execute the first iteration process until the number of new scene factors added is equal to the preset number of new scene factors.

[0114] In this embodiment, the second plane image corresponding to the fourth scene factor combination is input into the face recognition model to obtain the fourth face recognition accuracy corresponding to the fourth scene factor combination, the fourth face recognition accuracy is used as the third face recognition accuracy, and the target face recognition accuracy is used as the second face recognition accuracy, and the above S601 is re-executed to determine the deviation rate of the new scene factor according to the third face recognition accuracy and the second face recognition accuracy, that is, the deviation rate of the new scene factor for the third scene factor combination is determined according to the fourth face recognition accuracy and the target face recognition accuracy.

[0115] In this embodiment, according to the above-mentioned first iterative process, each new scene factor is gradually added to determine the impact of the new scene factor on the scene factor combination before the new scene factor is added, until all the preset new scene factors have been added to the second scene factor combination, and the iteration is completed.

[0116] S603: for the same new scene factor, determine the ratio of the number of deviation rates that have a negative impact on the scene factor combination before the new scene factor is added to the total number of the second scene factor combination.

[0117] In this embodiment, assuming that the same new scene factor is a backlighting scene factor, the total number of second scene factor combinations is 100, in second scene factor combination 1, the deviation rate at which the backlighting scene factor has a negative impact on the scene factor combination before the new scene factor is added, in second scene factor combination 2, the deviation rate at which the backlighting scene factor has a negative impact on the scene factor combination before the new scene factor is added, ..., among the 100 second scene factor combinations, the number of deviation rates at which the backlighting scene factor has a negative impact on the scene factor combination before the new scene factor is added is 80, and the proportion of the backlighting scene factor is 80%.

[0118] S604: If the ratio is greater than a preset ratio threshold, the face recognition model is optimized according to a new scene factor corresponding to the ratio greater than the preset ratio threshold.

[0119] Optionally, the preset ratio threshold may be 30%, 50%, etc., which is not limited in the embodiment of the present application.

[0120] In this embodiment, assuming that the preset ratio threshold is 30%, taking the above S603 as an example, if the ratio of the backlight scene factor is greater than the preset ratio threshold, an optimization suggestion is output, that is, the backlight scene factor needs to be optimized, so as to achieve the optimization of the face recognition model.

[0121] In the embodiment of the present application, by executing the first iterative process, new scene factors are continuously added to each second scene factor combination to optimize the face recognition accuracy, thereby finding all scene factors that have a negative impact on the face recognition model, so that subsequent personnel can optimize the face recognition model for the scene factors with negative impact. This method is more operational than a simple single-factor exhaustive simulation solution, and eliminates the situation where a single scene cannot be fitted or overfitted.

[0122] Figure 7 FIG. 1 is a flow chart of determining a target image in one embodiment. Figure 7 As shown, the following steps are included:

[0123] S701, extracting a plurality of image frames from a user's video at intervals of a preset number of frames.

[0124] Preferably, the user's video can be an original video from YouTube, Bilibili, Vlog, etc., which is more realistic and close to real life, replacing the single and cumbersome collection method using a camera or the like.

[0125] In this embodiment, a frame extraction tool can be used to collect frames of a user's video at regular intervals to obtain multiple image frames of the user. For videos from different sources, different categories can be distinguished according to the labels and information of the videos.

[0126] S702, performing pixel transformation processing on each pixel point of each image frame to obtain an image frame after pixel transformation processing.

[0127] In this embodiment, each image frame may be subjected to threshold processing, and each pixel point may be divided into different threshold intervals to obtain an image frame after pixel transformation processing. Alternatively, the size of each pixel point of each image frame may be changed to change the contrast of each image frame, thereby obtaining an image frame after pixel transformation processing.

[0128] S703, filtering out noise in the image frame after the pixel transformation process to obtain a target image.

[0129] In this embodiment, a Gaussian filter method may be used to remove noise in the image frame after pixel transformation processing to obtain a target image, or a median filter method may be used to filter out noise in the image frame after pixel transformation processing to obtain a target image.

[0130] In the embodiment of the present application, multiple image frames are extracted from the user's video at intervals of a preset number of frames, and pixel transformation processing is further performed on each pixel point of each image frame to obtain an image frame after pixel transformation processing, and then the noise in the image frame after pixel transformation processing is filtered out to obtain a target image. This method can make the content of the image itself more prominent and speed up the processing by performing pixel transformation and noise processing on each pixel point of each image frame.

[0131] Specifically, “S702, performing pixel transformation processing on each pixel point of each image frame to obtain an image frame after pixel transformation processing.” may include the following two implementation methods:

[0132] The first implementation method is to perform geometric transformation on each pixel of each image frame.

[0133] In this embodiment, when geometric transformation is performed on each pixel point of each image frame, the pixel value of the portrait is not changed, but the coordinate transformation is performed on the pixel value to change the arrangement relationship of the pixels, so that the characteristics of the image content itself are more prominent.

[0134] The second implementation manner: performing threshold processing on each pixel point of each image frame to divide each pixel point of the image frame into different pixel value intervals.

[0135] In this embodiment, since a single pixel of an image frame corresponds to 256 grayscales, the grayscales need to be limited, the 256 grayscales need to be divided into different intervals, and the pixels within the intervals need to be designated as a certain pixel.

[0136] Optionally, any one of the above methods may be used to perform pixel transformation processing on each pixel point of each image frame, or they may be used simultaneously.

[0137] It should be understood that, although the steps in the flowcharts involved in the above embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0138] Based on the same inventive concept, the embodiment of the present application also provides a model optimization device for implementing the model optimization method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more model optimization device embodiments provided below can refer to the limitations on the model optimization method above, and will not be repeated here.

[0139] In one embodiment, Figure 8 As shown, a model optimization device is provided, comprising: an acquisition module 11, a first determination module 12, a second determination module 13 and an optimization module 14, wherein:

[0140] An acquisition module 11 is used to obtain first plane images at multiple angles according to the acquired three-dimensional user portrait of the user;

[0141] A first determination module 12 is used to obtain second plane images corresponding to each first scene factor combination according to background feature information of the user's target image and face feature information of each first plane image;

[0142] A second determination module 13 is used to input the second plane image corresponding to each first scene factor combination into a face recognition model to obtain a first face recognition accuracy rate corresponding to each first scene factor combination;

[0143] The optimization module 14 is used to optimize the face recognition model according to the first face recognition accuracy corresponding to each first scene factor combination.

[0144] In one embodiment, the first determining module includes:

[0145] A first determining unit, configured to perform a change process on the facial feature information of each first plane image to obtain a plurality of third plane images;

[0146] A second determining unit, configured to perform a change process on the background feature information of the user's target image to obtain a plurality of fourth plane images;

[0147] The third determining unit is used to obtain the second plane image corresponding to each first scene factor combination according to the plurality of third plane images and the plurality of fourth plane images.

[0148] In one embodiment, the optimization module includes:

[0149] a fourth determining unit, configured to determine a second face recognition accuracy rate from the first face recognition accuracy rates corresponding to each first scene factor combination, wherein the second face recognition accuracy rate is a first face recognition accuracy rate that is less than a preset accuracy rate threshold;

[0150] The optimization unit is used to optimize the face recognition model according to the second scene factor combination, wherein the second scene factor combination is the first scene factor combination corresponding to the second face recognition rate.

[0151] In one embodiment, the optimization unit is further used to add a new scene factor to the second scene factor combination to obtain a third scene factor combination; input the second plane image corresponding to the third scene factor combination into the face recognition model to obtain a third face recognition accuracy rate corresponding to the third scene factor combination;

[0152] Optimize the face recognition model based on the second face recognition accuracy and the third face recognition accuracy.

[0153] In one embodiment, the optimization unit is further used to perform a first iterative process, which includes: determining a deviation rate corresponding to a new scene factor according to the second face recognition accuracy rate and the third face recognition accuracy rate, and correcting the third face recognition accuracy rate according to the deviation rate to obtain a comparative face recognition accuracy rate, taking the maximum face recognition accuracy rate between the second face recognition accuracy rate and the comparative face recognition accuracy rate as a target face recognition accuracy rate of a third scene factor combination, and adding a new scene factor to the third scene factor combination to obtain a fourth scene factor combination;

[0154] The second plane image corresponding to the fourth scene factor combination is input into the face recognition model to obtain the fourth face recognition accuracy corresponding to the fourth scene factor combination, and the target face recognition accuracy replaces the second face recognition accuracy, and the fourth face recognition accuracy replaces the third face recognition accuracy, and returns to execute the first iteration process until the number of new scene factors added is equal to the preset number of new scene factors; for the same new scene factor, determine the ratio of the number of deviation rates that have a negative impact on the scene factor combination before the new scene factor is added to the total number of second scene factor combinations; if the ratio is greater than a preset ratio threshold, optimize the face recognition model according to the new scene factor corresponding to the ratio greater than the preset ratio threshold.

[0155] In one embodiment, the apparatus further comprises:

[0156] An extraction module, used for extracting a plurality of image frames from a user's video at intervals of a preset number of frames;

[0157] A processing module, used for performing pixel transformation processing on each pixel point of each image frame to obtain an image frame after pixel transformation processing;

[0158] The filtering module is used to filter out the noise in the image frame after the pixel transformation process to obtain the target image.

[0159] In one embodiment, the processing module includes:

[0160] A first processing unit, used for performing geometric transformation processing on each pixel point of each image frame;

[0161] The second processing unit is used to perform threshold processing on each pixel point of each image frame to divide each pixel point of the image frame into different pixel value intervals.

[0162] Each module in the above-mentioned model optimization device can be implemented in whole or in part by software, hardware and a combination thereof. Each of the above-mentioned modules can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0163] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0164] Obtaining first plane images at multiple angles according to the acquired three-dimensional user portrait of the user;

[0165] Obtaining second plane images corresponding to each first scene factor combination according to background feature information of the user's target image and face feature information of each first plane image;

[0166] Inputting the second plane images corresponding to each first scene factor combination into a face recognition model to obtain the first face recognition accuracy corresponding to each first scene factor combination;

[0167] The face recognition model is optimized according to the first face recognition accuracy rates corresponding to the first scene factor combinations.

[0168] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0169] Performing a change process on the facial feature information of each first plane image to obtain a plurality of third plane images;

[0170] Performing a change process on the background feature information of the user's target image to obtain a plurality of fourth plane images;

[0171] According to the plurality of third planar images and the plurality of fourth planar images, second planar images corresponding to each combination of first scene factors are obtained.

[0172] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0173] Determining a second face recognition accuracy rate from the first face recognition accuracy rates corresponding to each first scene factor combination, wherein the second face recognition accuracy rate is a first face recognition accuracy rate that is less than a preset accuracy rate threshold;

[0174] The face recognition model is optimized according to the second scene factor combination, wherein the second scene factor combination is the first scene factor combination corresponding to the second face recognition rate. In one embodiment, when the processor executes the computer program, the following steps are also implemented:

[0175] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0176] Adding a new scenario factor to the second scenario factor combination to obtain a third scenario factor combination;

[0177] Inputting the second plane image corresponding to the third scene factor combination into the face recognition model to obtain a third face recognition accuracy rate corresponding to the third scene factor combination;

[0178] Optimize the face recognition model based on the second face recognition accuracy and the third face recognition accuracy.

[0179] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0180] Executing a first iteration process, the first iteration process comprising: determining a deviation rate corresponding to a new scene factor according to the second face recognition accuracy rate and the third face recognition accuracy rate, and correcting the third face recognition accuracy rate according to the deviation rate to obtain a comparative face recognition accuracy rate, taking the maximum face recognition accuracy rate between the second face recognition accuracy rate and the comparative face recognition accuracy rate as a target face recognition accuracy rate of a third scene factor combination, and adding a new scene factor to the third scene factor combination to obtain a fourth scene factor combination;

[0181] Input the second plane image corresponding to the fourth scene factor combination into the face recognition model, obtain the fourth face recognition accuracy corresponding to the fourth scene factor combination, replace the second face recognition accuracy with the target face recognition accuracy, replace the third face recognition accuracy with the fourth face recognition accuracy, and return to execute the first iteration process until the number of added new scene factors is equal to the preset number of new scene factors;

[0182] For the same new scenario factor, determine the ratio of the number of deviation rates that have a negative impact on the scenario factor combination before the new scenario factor is added to the total number of the second scenario factor combination;

[0183] If the ratio is greater than a preset ratio threshold, the face recognition model is optimized according to a new scene factor corresponding to the ratio greater than the preset ratio threshold.

[0184] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0185] Extract multiple image frames from the user's video at intervals of a preset number of frames;

[0186] Performing pixel transformation processing on each pixel point of each image frame to obtain an image frame after pixel transformation processing;

[0187] The target image is obtained by filtering out the noise in the image frame after pixel transformation processing.

[0188] In one embodiment, when the processor executes the computer program, the following steps are also implemented:

[0189] Performing geometric transformation processing on each pixel of each image frame;

[0190] Threshold processing is performed on each pixel point of each image frame to divide each pixel point of the image frame into different pixel value intervals.

[0191] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0192] Obtaining first plane images at multiple angles according to the acquired three-dimensional user portrait of the user;

[0193] Obtaining second plane images corresponding to each first scene factor combination according to background feature information of the user's target image and face feature information of each first plane image;

[0194] Inputting the second plane images corresponding to each first scene factor combination into a face recognition model to obtain the first face recognition accuracy corresponding to each first scene factor combination;

[0195] The face recognition model is optimized according to the first face recognition accuracy rates corresponding to the first scene factor combinations.

[0196] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0197] Performing a change process on the facial feature information of each first plane image to obtain a plurality of third plane images;

[0198] Performing a change process on the background feature information of the user's target image to obtain a plurality of fourth plane images;

[0199] According to the plurality of third planar images and the plurality of fourth planar images, second planar images corresponding to each combination of first scene factors are obtained.

[0200] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0201] Performing a change process on the facial feature information of each first plane image to obtain a plurality of third plane images;

[0202] Performing a change process on the background feature information of the user's target image to obtain a plurality of fourth plane images;

[0203] According to the plurality of third planar images and the plurality of fourth planar images, second planar images corresponding to each combination of first scene factors are obtained.

[0204] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0205] Determining a second face recognition accuracy rate from the first face recognition accuracy rates corresponding to each first scene factor combination, wherein the second face recognition accuracy rate is a first face recognition accuracy rate that is less than a preset accuracy rate threshold;

[0206] The face recognition model is optimized according to the second scene factor combination, wherein the second scene factor combination is the first scene factor combination corresponding to the second face recognition rate.

[0207] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0208] Adding a new scenario factor to the second scenario factor combination to obtain a third scenario factor combination;

[0209] Inputting the second plane image corresponding to the third scene factor combination into the face recognition model to obtain a third face recognition accuracy rate corresponding to the third scene factor combination;

[0210] Optimize the face recognition model based on the second face recognition accuracy and the third face recognition accuracy.

[0211] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0212] Executing a first iteration process, the first iteration process comprising: determining a deviation rate corresponding to a new scene factor according to the second face recognition accuracy rate and the third face recognition accuracy rate, and correcting the third face recognition accuracy rate according to the deviation rate to obtain a comparative face recognition accuracy rate, taking the maximum face recognition accuracy rate between the second face recognition accuracy rate and the comparative face recognition accuracy rate as a target face recognition accuracy rate of a third scene factor combination, and adding a new scene factor to the third scene factor combination to obtain a fourth scene factor combination;

[0213] Input the second plane image corresponding to the fourth scene factor combination into the face recognition model, obtain the fourth face recognition accuracy corresponding to the fourth scene factor combination, replace the second face recognition accuracy with the target face recognition accuracy, replace the third face recognition accuracy with the fourth face recognition accuracy, and return to execute the first iteration process until the number of added new scene factors is equal to the preset number of new scene factors;

[0214] For the same new scenario factor, determine the ratio of the number of deviation rates that have a negative impact on the scenario factor combination before the new scenario factor is added to the total number of the second scenario factor combination;

[0215] If the ratio is greater than a preset ratio threshold, the face recognition model is optimized according to a new scene factor corresponding to the ratio greater than the preset ratio threshold.

[0216] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0217] Extract multiple image frames from the user's video at intervals of a preset number of frames;

[0218] Performing pixel transformation processing on each pixel point of each image frame to obtain an image frame after pixel transformation processing;

[0219] The target image is obtained by filtering out the noise in the image frame after pixel transformation processing.

[0220] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0221] Performing geometric transformation processing on each pixel of each image frame;

[0222] Threshold processing is performed on each pixel point of each image frame to divide each pixel point of the image frame into different pixel value intervals.

[0223] In one embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the following steps:

[0224] Performing a change process on the facial feature information of each first plane image to obtain a plurality of third plane images;

[0225] Performing a change process on the background feature information of the user's target image to obtain a plurality of fourth plane images;

[0226] According to the plurality of third planar images and the plurality of fourth planar images, second planar images corresponding to each combination of first scene factors are obtained.

[0227] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0228] Performing a change process on the facial feature information of each first plane image to obtain a plurality of third plane images;

[0229] Performing a change process on the background feature information of the user's target image to obtain a plurality of fourth plane images;

[0230] According to the plurality of third planar images and the plurality of fourth planar images, second planar images corresponding to each combination of first scene factors are obtained.

[0231] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0232] Determining a second face recognition accuracy rate from the first face recognition accuracy rates corresponding to each first scene factor combination, wherein the second face recognition accuracy rate is a first face recognition accuracy rate that is less than a preset accuracy rate threshold;

[0233] The face recognition model is optimized according to the second scene factor combination, wherein the second scene factor combination is the first scene factor combination corresponding to the second face recognition rate.

[0234] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0235] Adding a new scenario factor to the second scenario factor combination to obtain a third scenario factor combination;

[0236] Inputting the second plane image corresponding to the third scene factor combination into the face recognition model to obtain a third face recognition accuracy rate corresponding to the third scene factor combination;

[0237] Optimize the face recognition model based on the second face recognition accuracy and the third face recognition accuracy.

[0238] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0239] Executing a first iteration process, the first iteration process comprising: determining a deviation rate corresponding to a new scene factor according to the second face recognition accuracy rate and the third face recognition accuracy rate, and correcting the third face recognition accuracy rate according to the deviation rate to obtain a comparative face recognition accuracy rate, taking the maximum face recognition accuracy rate between the second face recognition accuracy rate and the comparative face recognition accuracy rate as a target face recognition accuracy rate of a third scene factor combination, and adding a new scene factor to the third scene factor combination to obtain a fourth scene factor combination;

[0240] Input the second plane image corresponding to the fourth scene factor combination into the face recognition model, obtain the fourth face recognition accuracy corresponding to the fourth scene factor combination, replace the second face recognition accuracy with the target face recognition accuracy, replace the third face recognition accuracy with the fourth face recognition accuracy, and return to execute the first iteration process until the number of added new scene factors is equal to the preset number of new scene factors;

[0241] For the same new scenario factor, determine the ratio of the number of deviation rates that have a negative impact on the scenario factor combination before the new scenario factor is added to the total number of the second scenario factor combination;

[0242] If the ratio is greater than a preset ratio threshold, the face recognition model is optimized according to a new scene factor corresponding to the ratio greater than the preset ratio threshold.

[0243] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0244] Extract multiple image frames from the user's video at intervals of a preset number of frames;

[0245] Performing pixel transformation processing on each pixel point of each image frame to obtain an image frame after pixel transformation processing;

[0246] The target image is obtained by filtering out the noise in the image frame after pixel transformation processing.

[0247] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0248] Performing geometric transformation processing on each pixel of each image frame;

[0249] Threshold processing is performed on each pixel point of each image frame to divide each pixel point of the image frame into different pixel value intervals.

[0250] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0251] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0252] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0253] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A model optimization method, characterized in that: The method comprises: Obtaining first plane images at multiple angles according to the acquired three-dimensional user portrait of the user; Obtaining second plane images corresponding to each first scene factor combination according to background feature information of the target image of the user and face feature information of each first plane image; Inputting the second plane images corresponding to each of the first scene factor combinations into a face recognition model to obtain the first face recognition accuracy rates corresponding to each of the first scene factor combinations; Optimizing the face recognition model according to the first face recognition accuracy rates corresponding to each of the first scene factor combinations, including: determining a second face recognition accuracy rate from the first face recognition accuracy rates corresponding to each of the first scene factor combinations, wherein the second face recognition accuracy rate is a first face recognition accuracy rate that is less than a preset accuracy threshold; optimizing the face recognition model according to the second scene factor combination, wherein the second scene factor combination is the first scene factor combination corresponding to the second face recognition rate.

2. The method according to claim 1, characterized in that The obtaining, according to the background feature information of the target image of the user and the face feature information of each of the first plane images, the second plane images corresponding to each combination of the first scene factors comprises: Performing a change process on the facial feature information of each of the first plane images to obtain a plurality of third plane images; Performing a change process on the background feature information of the target image of the user to obtain a plurality of fourth plane images; According to the plurality of third planar images and the plurality of fourth planar images, second planar images corresponding to each combination of the first scene factors are obtained.

3. The method according to claim 1, characterized in that The optimizing the face recognition model according to the second scene factor combination includes: Adding a new scene factor to the second scene factor combination to obtain a third scene factor combination; Inputting the second plane image corresponding to the third scene factor combination into the face recognition model to obtain a third face recognition accuracy rate corresponding to the third scene factor combination; The face recognition model is optimized according to the second face recognition accuracy and the third face recognition accuracy.

4. The method according to claim 3, characterized in that The optimizing the face recognition model according to the second face recognition accuracy rate and the third face recognition accuracy rate includes: Executing a first iterative process, the first iterative process comprising: determining a deviation rate corresponding to the new scene factor according to the second face recognition accuracy rate and the third face recognition accuracy rate, and correcting the third face recognition accuracy rate according to the deviation rate to obtain a comparative face recognition accuracy rate, taking the maximum face recognition accuracy rate between the second face recognition accuracy rate and the comparative face recognition accuracy rate as a target face recognition accuracy rate for the third scene factor combination, and adding a new scene factor to the third scene factor combination to obtain a fourth scene factor combination; Inputting the second plane image corresponding to the fourth scene factor combination into the face recognition model, obtaining a fourth face recognition accuracy rate corresponding to the fourth scene factor combination, replacing the second face recognition accuracy rate with the target face recognition accuracy rate, replacing the third face recognition accuracy rate with the fourth face recognition accuracy rate, and returning to execute the first iteration process until the number of added new scene factors is equal to the preset number of new scene factors; For the same new scene factor, determining the ratio of the number of deviation rates that have a negative impact on the scene factor combination before the new scene factor is added to the total number of the second scene factor combinations; If the ratio is greater than a preset ratio threshold, the face recognition model is optimized according to a new scene factor corresponding to the ratio greater than the preset ratio threshold.

5. The method according to claim 1, characterized in that The method further comprises: Extracting a plurality of image frames from the user's video at intervals of a preset number of frames; Performing pixel transformation processing on each pixel point of each image frame to obtain an image frame after pixel transformation processing; The target image is obtained by filtering out noise in the image frame after the pixel transformation process.

6. The method according to claim 5, characterized in that The pixel transformation processing for each pixel point of each image frame includes at least one of the following: Performing geometric transformation processing on each pixel point of each image frame; Threshold processing is performed on each pixel point of each image frame to divide each pixel point of the image frame into different pixel value intervals.

7. A model optimization device, characterized in that: The device comprises: An acquisition module, used to obtain first plane images at multiple angles according to the acquired three-dimensional user portrait of the user; A first determination module, configured to obtain second plane images corresponding to each first scene factor combination according to background feature information of the user's target image and face feature information of each of the first plane images; A second determination module is used to input the second plane images corresponding to each of the first scene factor combinations into a face recognition model to obtain a first face recognition accuracy rate corresponding to each of the first scene factor combinations; An optimization module, used to optimize the face recognition model according to the first face recognition accuracy rates corresponding to each of the first scene factor combinations, includes: a fourth determination unit, used to determine a second face recognition accuracy rate from the first face recognition accuracy rates corresponding to each of the first scene factor combinations, wherein the second face recognition accuracy rate is a first face recognition accuracy rate that is less than a preset accuracy threshold; an optimization unit, used to optimize the face recognition model according to a second scene factor combination, wherein the second scene factor combination is a first scene factor combination corresponding to the second face recognition rate.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images