WO2020151315A1

WO2020151315A1 - Method and device for generating face recognition fusion model

Info

Publication number: WO2020151315A1
Application number: PCT/CN2019/117477
Authority: WO
Inventors: 戴磊
Original assignee: 平安科技（深圳）有限公司
Priority date: 2019-01-25
Filing date: 2019-11-12
Publication date: 2020-07-30
Also published as: CN110008815A

Abstract

A method and device for generating a face recognition fusion model, relating to the technical field of face recognition. The method comprises: using two face recognition models to respectively obtain feature vectors of face images in a positive sample and in a negative sample (S110); respectively obtaining, according to the feature vectors, included angles between the feature vectors of the face images in the positive sample and in the negative sample targeted by the two face recognition models to obtain confidences corresponding to the face recognition models (S120); comparing results of the confidences of the two face recognition models to obtain a combination of weight values of the two face recognition models according to a comparison result (S130); and calculating a weight of the two face recognition models according to the combination of the weight values to determine a face recognition fusion model (S140), so as to ensure that the face recognition accuracy of the face recognition fusion model generated by using the two face recognition models is higher than the accuracy of a single face recognition model.

Description

Method and device for generating face recognition fusion model

This application claims the priority of a Chinese patent application filed with the Chinese Patent Office on January 25, 2019, the application number is 201910075750.8, and the invention title is "Method and Apparatus for Generating Face Recognition Fusion Model", the entire content of which is incorporated by reference In this application.

Technical field

This application relates to the technical field of face recognition. Specifically, this application relates to a method and device for generating a face recognition fusion model.

Background technique

With the application of face recognition technology in security fields such as financial encryption unlocking, security inspection, and home security, users have continuously improved the accuracy of face recognition. At present, the method to improve the accuracy of face recognition is because of the different recognition characteristics of different models of face recognition. By adding two face recognition models in half, a fusion model of face recognition is obtained.

However, the inventor realizes that the face recognition fusion model obtained above, if the recognition accuracy of the two face recognition models are far apart, the generated fusion model may have a false recognition rate than the recognition model with the highest recognition accuracy alone. High defects.

Summary of the invention

In order to overcome the above technical problems, especially the fusion model generated in the prior art may have a higher false recognition rate than the recognition model with the highest recognition accuracy alone, the following technical solutions are proposed:

In the first aspect, this application provides a method for generating a face recognition fusion model, which includes the following steps: using two face recognition models to obtain the feature vector of each face image in a positive sample and a negative sample; Feature vectors, respectively obtaining the angles between the two face recognition models and the feature vectors of the face images in the positive and negative samples to obtain the confidence of the corresponding face recognition models; and recognizing the two faces The results of the model’s confidence are compared, and the combination of the weight values of the two face recognition models is obtained according to the comparison results; the weights of the two face recognition models are calculated according to the combination of the weight values to determine the face recognition Fusion model.

In the second aspect, the present application also provides a face recognition fusion model generation device, which includes: a feature vector obtaining module, which is used to use two face recognition models to obtain each face image in a positive sample and a negative sample. The feature vector; the confidence acquisition module is used to obtain the angle between the two face recognition models on the feature vector of the face image in the positive sample and the negative sample according to the feature vector to obtain the corresponding person The confidence of the face recognition model; a confidence comparison module, used to compare the results of the confidence of the two face recognition models, and obtain a combination of the weight values of the two face recognition models according to the comparison results; a face recognition fusion model The determining module is configured to calculate the weight of the two face recognition models according to the combination of the weight values to determine the face recognition fusion model.

In the third aspect, the present application also provides a computer device, which includes: one or more processors; a memory; one or more computer programs, wherein the one or more computer programs are stored in the memory and are Configured to be executed by the one or more processors, and the one or more computer programs are configured to execute a face recognition fusion model generation method, wherein the face recognition fusion model generation method includes: Two face recognition models are used to obtain the feature vector of each face image in the positive sample and the negative sample; according to the feature vector, the two face recognition models are used to obtain the positive sample and the negative sample. The included angle of the feature vector of the face image obtains the confidence of the corresponding face recognition model; the confidence results of the two face recognition models are compared, and the combination of the weight values of the two face recognition models is obtained according to the comparison result Calculating the weights of the two face recognition models according to the combination of the weight values to determine the face recognition fusion model.

In a fourth aspect, the present application also provides a computer-readable storage medium having a computer program stored on the computer-readable storage medium, and when the computer program is executed by a processor, a method for generating a face recognition fusion model is realized, wherein The method for generating the face recognition fusion model includes the following steps: using two face recognition models to obtain the feature vector of each face image in the positive sample and the negative sample respectively; according to the feature vector, obtain the The angle between the two face recognition models of the feature vectors of the face images in the positive sample and the negative sample to obtain the confidence of the corresponding face recognition model; the results of the confidence of the two face recognition models are compared Obtain a combination of weight values of the two face recognition models according to the comparison result; calculate the weights of the two face recognition models according to the combination of weight values, and determine the face recognition fusion model.

The method and device for generating a face recognition fusion model provided by this application overcomes the fact that the difference in recognition accuracy of two face recognition models is not considered in the prior art, and is based on the confidence of comparing the two face recognition models. As a result of the size of the degree, the two face recognition models are assigned weight values of different sizes to ensure that the face recognition fusion model generated by the two face recognition models has a higher accuracy of face recognition than a single face recognition model Degree should be high.

Description of the drawings

Fig. 1 is a flowchart of a method for generating a face recognition fusion model according to an embodiment of the present application;

2 is a flowchart of a method for generating a face recognition fusion model according to another embodiment of the present application;

FIG. 3 is a schematic diagram of an apparatus for generating a face recognition fusion model according to an embodiment of the application;

FIG. 4 is a schematic structural diagram of a computer device according to an embodiment of the application;

FIG. 5 is a schematic structural diagram of a terminal according to an embodiment of the application.

detailed description

In order to solve the above problems, the present application provides a method for generating a face recognition fusion model. Refer to FIG. 1. FIG. 1 is a flowchart of a method for generating a face recognition fusion model according to an embodiment. The method includes the following steps:

S110. Use two face recognition models to obtain the feature vector of each face image in the positive sample and the negative sample respectively.

In this step, different face images are collected, and positive samples and negative samples are respectively formed for the face images, and used as a training set for generating the face recognition fusion model.

The positive sample is that several collected face images belong to the same person; the negative sample is that several collected face images do not belong to the same person.

The confidence is used to reflect the recognition accuracy of the corresponding face recognition model; in this embodiment, it is the ability of the face recognition model to judge the similarity between the positive sample and the negative sample. Specifically, if the degree of separation between the positive and negative samples obtained by the face recognition model is higher, the confidence is higher. The higher the confidence of the corresponding face recognition model, the higher the accuracy of its face recognition.

In this step, the server uses the positive sample and the negative sample as a training set to calculate the confidence of the two face recognition models, and obtain the face recognition accuracy of the two face recognition models respectively. Wherein, the two face recognition models are used to generate a face recognition fusion model.

In the statement of the specific embodiment of the present application, the two face recognition models are defined as the face recognition model A and the face recognition model B respectively to distinguish them. Correspondingly, the confidence recognition model A labeled S _A, model B recognition confidence labeled S _B.

Take the face recognition model A as an example. The face recognition model A is used to recognize each image in all positive samples and each image in negative samples. When the face recognition model A recognizes each image, it will get the corresponding feature vector. Corresponding to each positive sample or negative sample, a corresponding number of feature vectors are obtained according to the number of positive and negative face images.

S120: According to the feature vector, respectively obtain the included angles of the two face recognition models to the feature vectors of the face images in the positive sample and the negative sample, to obtain the confidence of the corresponding face recognition model.

According to the feature vector of each face image in the positive sample and the negative sample obtained in step 110, the angle between the respective feature vectors is obtained. If the angle formed by the feature vector of the positive sample is smaller, the confidence of the face recognition model A is higher; otherwise, the confidence of the face recognition model A is lower. Conversely, the larger the angle formed by the feature vector of the negative sample, the higher the confidence of the face recognition model A; otherwise, the lower the confidence of the face recognition model A.

If both the positive sample and the negative sample include two face images, the included angles are respectively composed of two corresponding feature vectors. The calculation of the confidence of the corresponding face recognition model is to obtain the inner product after normalizing the two feature vectors. A face recognition model to be described concrete, which is calculated _{as: S A = (F1 / |} F1 |) · (F2 / | F2 |), wherein, F1, and F2 are the two feature vectors. Similarly, the same way to obtain the calculated face recognition model B S _B of confidence.

The included angle of the feature vector of the positive sample obtained by the face recognition model A for the positive sample should be smaller than the included angle of the feature vector obtained for the negative sample. If the degree of separation between the included angle of the feature vector of the positive sample and the included angle of the feature vector obtained from the negative sample is low, or even greater than the included angle of the feature vector obtained from the negative sample, then the face recognition model A The accuracy of image recognition is low. Otherwise, the face recognition model A has the highest image recognition accuracy.

In order to accurately test the recognition accuracy of face recognition model A and face recognition model B, multiple sets of positive and negative samples are used for testing. In this embodiment, the number of positive samples is 10,000, and the number of negative samples is 10,000. However, it should be noted that the above-mentioned number of positive and negative samples is only for taking as many test samples as possible, and the number of samples can be different.

In order to further verify the accuracy of the corresponding face recognition model, before the step of separately calculating the confidence of the two face recognition models in step S110, one of the face images of the positive sample and the negative sample When the face image belongs to the same person, the confidence of the positive sample and the negative sample of the two face recognition models are compared.

In this step, the verification basis of the two face recognition models, namely the positive sample and the negative sample, are unified, so that it is easier to measure the angle between the feature vectors corresponding to each face image, so as to verify the corresponding person Accuracy of face recognition model.

S130: Compare the confidence results of the two face recognition models, and obtain a combination of weight values of the two face recognition models according to the comparison result.

The server obtains the confidence results of the two face recognition models from step S110, and compares the results of the two results. According to the result of the comparison, different weight values are assigned to the two face recognition models.

Specifically, if S _A > S _B , then the weight value a assigned by the face recognition model A is higher than the weight value b of the face recognition model B. The weight value a and weight value b form a combination (a, b) of the two face recognition models. The following relationship exists between the weight value a and the weight value b: a+b=1, where aε[0,1] and bε[0,1].

S140: Calculate the weights of the two face recognition models according to the combination of the weight values, and determine the face recognition fusion model.

The server enters the value of the combination (a, b) of the two face recognition models into the weight calculation of the two face recognition models to obtain a fusion model about the two face recognition models.

The method for generating a face recognition fusion model provided by the present application overcomes the fact that the prior art does not take into account the difference in the recognition accuracy of the two face recognition models, and is based on the comparison of the confidence level of the two face recognition models. As a result, the two face recognition models are assigned weight values of different sizes to ensure that the face recognition fusion model generated by the two face recognition models has a higher face recognition accuracy than a single face recognition model .

For the step of obtaining the combination of the weight values of the two face recognition models according to the comparison result in step S130, the step includes: according to the result of the comparison of the confidence levels of the two face recognition models, reporting the two face recognition models to the Set the combination of each weight value.

Specifically, it is possible to first set different combinations of the weight value combinations (a, b) of the two face recognition models, and obtain the best weight value combination after calculation and comparison.

On the basis of the combination of the weight values set for the two face recognition models obtained above, for step S140, refer to FIG. 2, which is another embodiment of the method for generating a face recognition fusion model Flow chart, which includes the following steps:

S141. Obtain a combination of each of the weight values, and calculate the weight of the two face recognition models.

S142: According to the calculation result, correspondingly obtain and compare the confidence of each of the face recognition fusion models, and determine the face recognition fusion model with the highest confidence.

For the combination of the weight values, first take 0.6 for a and 0.4 for b, and compare the confidence of the face recognition fusion model with the face recognition fusion model where a is 0.7 and b is 0.3, and the interval is Built-in face recognition fusion model with the highest reliability. Then, the face recognition fusion model with the highest built-in confidence in this interval and the face recognition fusion model of the combination of the weight values of other (a, b) are used to compare the confidence, and so on. Finally, the face recognition fusion model with the highest confidence level is obtained, and the face recognition fusion model with the highest accuracy is obtained.

In order to improve the efficiency of determining the highest-confidence face recognition fusion model and its corresponding optimal weight combination, the above step S142 may further include: according to the range of the two weight values and the two weight values The value range of the sum is divided into equal scales, and the corresponding confidence is obtained, and the highest confidence is obtained after comparison.

Specifically, in the combination of weight values, since a∈[0,1], b∈[0,1], a+b=1, it can actually be regarded as searching in the interval of [0,1] The value of a that maximizes the auc of the roc curve, assuming that auc is a convex function in the interval [0,1] of a, the steps to improve efficiency are as follows:

(1) Divide the interval of [0,1] into equal divisions by 0.1 scale, and test the auc of each scale point (0,0.1,0.2,...,0.9,1) to find the first two scale points with the largest value;

(2) Divide the interval between the above two scale points into equal divisions of 0.01, and also test the auc of each scale point to find the first two scale points with the largest value;

(3) Repeat the above steps until the specified scale accuracy (such as 0.001).

For a certain set of test samples, the time complexity of the above steps is a constant 0 (1), so a certain amount of calculation is required, but this kind of test is generally performed offline, and the operation of the server does not cause calculation burden.

After step 140, according to the determined face recognition fusion model, a combination of weight values of (a, b) is correspondingly obtained.

The method for generating a face recognition fusion model provided in this application can be completed through a terminal. Specifically, the terminal divides the face image data in the memory and/or the face image data obtained on-site into positive samples and negative samples according to the user's mark. Then, according to the calculation of the confidence of the two face recognition models, a combination of corresponding weight values is formed, and after calculation and comparison, the face recognition fusion model with the highest confidence is obtained. The terminal correspondingly responds to the user's request for face recognition sent by the terminal according to the face recognition fusion model, and completes the corresponding face recognition task.

Based on the same inventive concept as the aforementioned method for generating a face recognition fusion model, an embodiment of the present application also provides an apparatus for generating a face recognition fusion model, as shown in FIG. 3, including:

The feature vector obtaining module 310 is configured to use the two face recognition models to obtain the feature vector of each face image in the positive sample and the negative sample respectively;

The confidence acquisition module 320 is configured to obtain the angles between the two face recognition models and the feature vectors of the face images in the positive sample and the negative sample according to the feature vector, to obtain the corresponding face recognition model Confidence;

The confidence comparison module 330 is configured to compare the confidence results of the two face recognition models, and obtain a combination of weight values of the two face recognition models according to the comparison results;

The face recognition fusion model determination module 340 is configured to calculate the weights of the two face recognition models according to the combination of the weight values to determine the face recognition fusion model.

Please refer to FIG. 4, which is a schematic diagram of the internal structure of a computer device in an embodiment. As shown in FIG. 5, the computer device includes a processor 410, a storage medium 420, a memory 430, and a network interface 440 connected through a system bus. Wherein, the storage medium 420 of the computer device stores an operating system, a database, and computer-readable instructions. The database may store control information sequences. When the computer-readable instructions are executed by the processor 410, the processor 410 can implement a In the method for generating a face recognition fusion model, the processor 410 can implement the feature vector obtaining module 310, the confidence obtaining module 320, and the confidence comparison in the device for generating a face recognition fusion model in the embodiment shown in FIG. 3 The module 330 and the face recognition fusion model determine the function of the module 340. The processor 410 of the computer device is used to provide calculation and control capabilities, and supports the operation of the entire computer device. The memory 430 of the computer device may store computer-readable instructions, and when the computer-readable instructions are executed by the processor 410, the processor 410 can make the processor 410 execute a method for generating a face recognition fusion model. The network interface 440 of the computer device is used to connect and communicate with the terminal. Those skilled in the art can understand that the structure shown in FIG. 4 is only a block diagram of part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may Including more or less parts than shown in the figure, or combining some parts, or having a different part arrangement.

If the computer device is a terminal, as shown in FIG. 5, for ease of description, only parts related to the embodiment of the present application are shown. For specific technical details that are not disclosed, please refer to the method part of the embodiment of the present application. The terminal can be any terminal device including a mobile phone, a tablet computer, a PDA (Personal Digital Assistant), a POS (Point of Sales, sales terminal), a car computer, etc. Take the terminal as a mobile phone as an example:

FIG. 5 shows a block diagram of a part of the structure of a mobile phone related to a terminal provided in an embodiment of the present application. 5, the mobile phone includes: a radio frequency (RF) circuit 510, a memory 520, an input unit 530, a display unit 540, a sensor 550, an audio circuit 560, a wireless fidelity (Wi-Fi) module 570, a processing 580, and power supply 590. Those skilled in the art can understand that the structure of the mobile phone shown in FIG. 5 does not constitute a limitation on the mobile phone, and may include more or less components than those shown in the figure, or combine some components, or arrange different components.

The following describes the components of the mobile phone in detail in conjunction with Figure 5:

The memory 520 may be used to store software programs and modules. The processor 580 executes various functional applications and data processing of the mobile phone by running the software programs and modules stored in the memory 520. The memory 520 may mainly include a storage program area and a storage data area. The storage program area may store an operating system, an application program required by at least one function (such as a voiceprint playback function, an image playback function, etc.), etc.; the storage data area may store Data (such as audio data, phone book, etc.) created based on the use of mobile phones. In addition, the memory 520 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one magnetic disk storage device, a flash memory device, or other volatile solid-state storage devices.

The input unit 530 can be used to receive inputted digital or character information, obtain and input a face image, and generate signal input related to user settings and function control of the mobile phone. Specifically, the input unit 530 may include a touch panel 531 and other input devices 532. The touch panel 531, also known as a touch screen, can collect the user's touch operations on or near it (for example, the user uses any suitable objects or accessories such as fingers, stylus, etc.) on the touch panel 531 or near the touch panel 531. Operation), and drive the corresponding connection device according to the preset program. Optionally, the touch panel 531 may include two parts: a touch detection device and a touch controller. Among them, the touch detection device detects the user's touch position, and detects the signal brought by the touch operation, and transmits the signal to the touch controller; the touch controller receives the touch information from the touch detection device, converts it into contact coordinates, and then sends it To the processor 580, and can receive and execute commands from the processor 580. In addition, the touch panel 531 can be implemented in multiple types such as resistive, capacitive, infrared, and surface acoustic wave. In addition to the touch panel 531, the input unit 530 may also include other input devices 532. Specifically, the other input device 532 may include, but is not limited to, one or more of a physical keyboard, function keys (such as volume control buttons, switch buttons, etc.), trackball, mouse, and joystick.

The display unit 540 may be used to display information input by the user or information provided to the user and various menus of the mobile phone. The display unit 540 may include a display panel 541. Optionally, the display panel 541 may be configured in the form of a liquid crystal display (Liquid Crystal Display, LCD), an organic light-emitting diode (Organic Light-Emitting Diode, OLED), etc. Further, the touch panel 531 can cover the display panel 541. When the touch panel 531 detects a touch operation on or near it, it transmits it to the processor 580 to determine the type of the touch event, and then the processor 580 responds to the touch event. The type provides corresponding visual output on the display panel 541. Although in FIG. 5, the touch panel 531 and the display panel 541 are used as two independent components to implement the input and input functions of the mobile phone, but in some embodiments, the touch panel 531 and the display panel 541 can be integrated. Realize the input and output functions of mobile phones.

The processor 580 is the control center of the mobile phone. It uses various interfaces and lines to connect various parts of the entire mobile phone. It executes by running or executing software programs and/or modules stored in the memory 520, and calling data stored in the memory 520. Various functions and processing data of the mobile phone can be used to monitor the mobile phone as a whole. Optionally, the processor 580 may include one or more processing units; preferably, the processor 580 may integrate an application processor and a modem processor, where the application processor mainly processes the operating system, user interface, and application programs, etc. , The modem processor mainly deals with wireless communication. It can be understood that the foregoing modem processor may not be integrated into the processor 580.

In the embodiment of the present application, the processor 580 included in the terminal also has the following functions: acquiring and using the positive and negative samples of the face image comparison combination, respectively calculating the confidence of the two face recognition models; The confidence results of the two face recognition models are compared, and the combination of the weight values of the two face recognition models is obtained according to the comparison results; the weights of the two face recognition models are calculated according to the combination of the weight values to determine the The fusion model of face recognition. That is, the processor 580 has the function of executing the method for generating a face recognition fusion model in any of the above embodiments, wherein the method for generating the face recognition fusion model includes: using two face recognition models to obtain positive samples respectively And the feature vector of each face image in the negative sample; according to the feature vector, the angle between the two face recognition models and the feature vector of the face image in the positive sample and the negative sample is obtained to obtain the corresponding The confidence of the face recognition model; compare the results of the confidence of the two face recognition models, and obtain a combination of the weight values of the two face recognition models according to the comparison result; according to the combination of the weight values, The weight calculation of the two face recognition models determines the face recognition fusion model. I won't repeat them here.

In one embodiment, the present application also proposes a storage medium storing computer-readable instructions. The storage medium is a volatile storage medium or a non-volatile storage medium. The computer-readable instructions are stored by one or more When the two processors are executed, one or more processors are made to perform the following steps: obtain and use the positive sample and the negative sample of the face image comparison combination, and calculate the confidence of the two face recognition models; The results of the confidence of the recognition models are compared, and the combination of the weight values of the two face recognition models is obtained according to the comparison results; the weights of the two face recognition models are calculated according to the combination of the weight values to determine the face Identify the fusion model.

Based on the foregoing embodiments, it can be seen that the greatest beneficial effect of this application lies in:

The method for generating a face recognition fusion model provided in this application overcomes the fact that the difference in recognition accuracy of two face recognition models is not considered in the prior art, and is based on the result of comparing the confidence levels of the two face recognition models , Assigning weight values of different sizes to the two face recognition models to ensure that the face recognition fusion model generated by the two face recognition models has a higher face recognition accuracy than a single face recognition model.

This application also provides the face recognition fusion model that determines the highest confidence level by calculating and comparing different combinations of the weight values. In this way, the face recognition fusion model has the highest accuracy for face recognition. Further, in order to improve the efficiency of determining the face recognition fusion model with the highest confidence level and its corresponding optimal weight combination, the value of the weight value is equal-scaled to obtain the corresponding two with the highest built-in confidence in the set interval. A combination of two weight values, and then further equal-scale division within the range of the highest combination of two weight values in the set interval, and by analogy, the face recognition with the highest accuracy for face recognition is obtained Fusion model.

In summary, through the use of the method and device for generating a face recognition fusion model, this application considers the face recognition accuracy of the face recognition model to be fused into the generation of the face recognition fusion model, which solves the problem of the generation in the prior art. The fusion model may have a higher false recognition rate than the recognition model with the highest recognition accuracy alone.

A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing relevant hardware through a computer program. The computer program can be stored in a computer readable storage medium. When executed, it may include the procedures of the above-mentioned method embodiments. Among them, the aforementioned storage medium may be a storage medium such as a magnetic disk, an optical disc, a read-only memory (Read-Only Memory, ROM), or a random access memory (Random Access Memory, RAM), etc.

The technical features of the above-mentioned embodiments can be combined arbitrarily. To make the description concise, all possible combinations of the various technical features in the above-mentioned embodiments are not described. However, as long as there is no contradiction in the combination of these technical features, All should be considered as the scope of this specification.

Claims

A method for generating a face recognition fusion model includes the following steps:

Use two face recognition models to obtain the feature vector of each face image in the positive sample and the negative sample;

According to the feature vector, respectively obtaining the included angles of the two face recognition models to the feature vectors of the face images in the positive sample and the negative sample to obtain the confidence of the corresponding face recognition model;

Comparing the confidence results of the two face recognition models, and obtaining a combination of weight values of the two face recognition models according to the comparison result;

According to the combination of the weight values, the weights of the two face recognition models are calculated to determine the face recognition fusion model.
The method according to claim 1,

The face image of the positive sample and one of the face images in the negative sample belong to the same person;

Before the step of obtaining, according to the feature vector, the angle between the two face recognition models and the feature vector of the face image in the positive sample and the negative sample, and obtaining the confidence of the corresponding face recognition model ,Also includes:

The two face recognition models are respectively compared with the confidence of the positive sample and the negative sample, and the accuracy of the corresponding face recognition model is verified.
The method according to claim 1,

The step of obtaining a combination of weight values of two face recognition models according to the comparison result includes:

According to the result of the comparison of the confidence levels of the two face recognition models, a combination of weight values is set to the two face recognition models.
According to the method of claim 3,

The step of calculating the weights of the two face recognition models according to the combination of the weight values, and determining the fusion model of the face recognition models includes:

Obtaining a combination of each of the weight values, and calculating the weight of the two face recognition models;

According to the calculation result, the confidence of each face recognition fusion model is correspondingly obtained and compared, and the face recognition fusion model with the highest confidence is determined.
According to the method of claim 4,

According to the result of the comparison of the confidence levels of the two face recognition models, after the step of setting the respective weight value combinations for the two face recognition models, the method further includes: according to different set weight value combinations , Respectively obtain the confidence of each corresponding face recognition fusion model;

Comparing the confidence level of each corresponding face recognition fusion model in the set interval to obtain the face recognition fusion model with the highest built-in confidence in the set interval;

The face recognition fusion model combining the highest face recognition fusion model and other weight values is used to compare the confidence.
According to the method of claim 5,

According to the calculation result, the step of correspondingly obtaining and comparing the confidence of each of the face recognition fusion models, and determining the face recognition fusion model with the highest confidence includes:

According to the value range of the two weight values and the value range of the sum of the two weight values, equal scale division is performed to obtain the values, and the corresponding confidence is obtained, and the highest confidence is obtained after comparison.
The method according to any one of claims 1-6,

The step of using two face recognition models to separately obtain the feature vector of each face image in the positive sample and the negative sample includes:

Using the two face recognition models to obtain feature vectors of each face image in several groups of positive samples and the negative samples;

Wherein, the positive sample and the negative sample each include two face images.
A device for generating a face recognition fusion model, including:

The feature vector obtaining module is configured to obtain the feature vector of each face image in the positive sample and the negative sample by using the two face recognition models;

The confidence acquisition module is used to obtain the angles between the two face recognition models and the feature vectors of the face images in the positive and negative samples according to the feature vectors to obtain the corresponding face recognition model Confidence;

A confidence comparison module, configured to compare the confidence results of the two face recognition models, and obtain a combination of weight values of the two face recognition models according to the comparison results;

The face recognition fusion model determination module is configured to calculate the weights of the two face recognition models according to the combination of the weight values to determine the face recognition fusion model.
A computer device including:

One or more processors;

Memory

One or more computer programs, wherein the one or more computer programs are stored in the memory and configured to be executed by the one or more processors, and the one or more computer programs are configured to execute A method of generating a face recognition fusion model:

Wherein, the method for generating the face recognition fusion model includes:

Use two face recognition models to obtain the feature vector of each face image in the positive sample and the negative sample;

According to the feature vector, respectively obtaining the included angles of the two face recognition models to the feature vectors of the face images in the positive sample and the negative sample to obtain the confidence of the corresponding face recognition model;

Comparing the confidence results of the two face recognition models, and obtaining a combination of weight values of the two face recognition models according to the comparison result;

According to the combination of the weight values, the weights of the two face recognition models are calculated to determine the face recognition fusion model.
9. The computer device according to claim 9, wherein the face image of the positive sample and one of the face images in the negative sample belong to the same person;

According to the feature vector, the angle between the two face recognition models and the feature vector of the face image in the positive sample and the negative sample is obtained respectively to obtain the confidence level of the corresponding face recognition model Before the steps, also include:

The two face recognition models are respectively compared with the confidence of the positive sample and the negative sample, and the accuracy of the corresponding face recognition model is verified.
The computer device according to claim 9, wherein the step of obtaining a combination of weight values of two face recognition models according to the comparison result comprises:

According to the result of the comparison of the confidence levels of the two face recognition models, a combination of weight values is set to the two face recognition models.
The computer device according to claim 11, wherein the step of calculating the weights of the two face recognition models according to the combination of the weight values, and determining the fusion model of the face recognition models comprises:

Obtaining a combination of each of the weight values, and calculating the weight of the two face recognition models;

According to the calculation result, the confidence of each face recognition fusion model is correspondingly obtained and compared, and the face recognition fusion model with the highest confidence is determined.
The computer device according to claim 12, after the step of setting a combination of weight values for the two face recognition models according to the result of the comparison of the confidence levels of the two face recognition models, further Including: obtaining the confidence of each corresponding face recognition fusion model according to the combination of different set weight values;

Comparing the confidence level of each corresponding face recognition fusion model in the set interval to obtain the face recognition fusion model with the highest built-in confidence in the set interval;

The face recognition fusion model combining the highest face recognition fusion model and other weight values is used to compare the confidence.
The computer device according to claim 13, wherein the step of correspondingly obtaining and comparing the confidence of each of the face recognition fusion models according to the calculation result, and determining the face recognition fusion model with the highest confidence, comprises :

According to the value range of the two weight values and the value range of the sum of the two weight values, equal scale division is performed to obtain the values, and the corresponding confidence is obtained, and the highest confidence is obtained after comparison.
The computer device according to any one of claims 9-14, wherein the step of using the two face recognition models to separately obtain the feature vector of each face image in the positive sample and the negative sample comprises:

Using the two face recognition models to obtain feature vectors of each face image in several groups of positive samples and the negative samples;

Wherein, the positive sample and the negative sample each include two face images.
A computer-readable storage medium in which a computer program is stored, and when the computer program is executed by a processor, a method for generating a face recognition fusion model is realized, wherein the generation of the face recognition fusion model The method includes the following steps:

Use two face recognition models to obtain the feature vector of each face image in the positive sample and the negative sample;

According to the feature vector, respectively obtaining the included angles of the two face recognition models to the feature vectors of the face images in the positive sample and the negative sample to obtain the confidence of the corresponding face recognition model;

Comparing the confidence results of the two face recognition models, and obtaining a combination of weight values of the two face recognition models according to the comparison result;

According to the combination of the weight values, the weights of the two face recognition models are calculated to determine the face recognition fusion model.
The computer-readable storage medium according to claim 16, wherein the face image of the positive sample and one of the face images in the negative sample belong to the same person;

Before the step of obtaining, according to the feature vector, the angle between the two face recognition models and the feature vector of the face image in the positive sample and the negative sample, and obtaining the confidence of the corresponding face recognition model ,Also includes:

The two face recognition models are respectively compared with the confidence of the positive sample and the negative sample, and the accuracy of the corresponding face recognition model is verified.
The computer-readable storage medium according to claim 16, wherein the step of obtaining a combination of weight values of two face recognition models according to the comparison result comprises:

According to the result of the comparison of the confidence levels of the two face recognition models, a combination of weight values is set to the two face recognition models.
The computer-readable storage medium according to claim 18, wherein the step of calculating the weights of the two face recognition models according to the combination of the weight values, and determining the fusion model of the face recognition models comprises:

Obtaining a combination of each of the weight values, and calculating the weight of the two face recognition models;

According to the calculation result, the confidence of each face recognition fusion model is correspondingly obtained and compared, and the face recognition fusion model with the highest confidence is determined.
The computer-readable storage medium according to claim 19, after the step of setting a combination of weight values for the two face recognition models according to the result of the comparison of the confidence levels of the two face recognition models, It also includes: obtaining the confidence of each corresponding face recognition fusion model according to the combination of different set weight values;

Comparing the confidence level of each corresponding face recognition fusion model in the set interval to obtain the face recognition fusion model with the highest built-in confidence in the set interval;

The face recognition fusion model combining the highest face recognition fusion model and other weight values is used to compare the confidence.