Face recognition method and device based on adaptive learning, equipment and medium

Abstract

The embodiment of the invention provides a face recognition method and device based on adaptive learning, equipment and a medium, and belongs to the technical field of financial science and technology. The method comprises the following steps: acquiring a current face image of a target user; comparing the current face feature vector of the target user with a historical face feature vector of the target user, and determining a feature vector variable quantity; adjusting a first target model parameter of the face recognition model through an incremental learning module under the condition that the feature vector variable quantity exceeds a first preset change threshold value; under the condition that the feature vector variable quantity exceeds a second preset change threshold value, adjusting a second target model parameter of the face recognition model through an online learning module; and optimizing the face recognition model according to the adjusted first target model parameter or the second target model parameter. According to the embodiment of the invention, self-adaptive learning can be carried out on the face change of the user, and the long-time accuracy of face recognition can be ensured in a financial scene.

Description

Technical Field

[0001] The present application relates to the field of financial technology (Fi ntech) technology, and in particular to a face recognition method and apparatus, device and medium based on adaptive learning. Background Art

[0002] Once deployed, current face recognition models are basically fixed, and their recognition accuracy will gradually decrease due to changes in user age, body shape, and habits. In order to maintain its sensitivity to user changes, existing face recognition models also need to be retrained regularly to maintain accuracy.

[0003] In financial scenarios, for example, for customer accounts or VIP customers that require long-term maintenance, the existing face recognition models have the above-mentioned technical defects, so there is a risk of recognition failure or the model needs to be retrained, which has a negative impact on customer service. Based on this, in financial scenarios, how to ensure the long-term accuracy of face recognition without retraining the model has become a technical problem that needs to be solved urgently. Summary of the invention

[0004] The main purpose of the embodiments of the present application is to propose a face recognition method and apparatus, equipment and medium based on adaptive learning, which can adaptively learn the changes in the user's face without retraining the model, thereby ensuring the long-term accuracy of face recognition. Especially in financial scenarios, when the user's face changes, the model parameters can be adjusted in real time through incremental learning and online learning techniques, so that the model can maintain its sensitivity to user changes without retraining, and can still maintain a high accuracy rate after long-term use, which can avoid the situation where the user's identity authentication fails, improve the real-time and accuracy of financial services, and improve customer satisfaction, thereby enhancing the competitiveness of financial enterprises.

[0005] To achieve the above-mentioned purpose, a first aspect of an embodiment of the present application proposes a face recognition method based on adaptive learning, the method comprising:

[0006] Get the current face image of the target user;

[0007] Inputting the current face image into a face recognition model for feature extraction to obtain a current face feature vector;

[0008] Compare the current facial feature vector with the historical facial feature vector of the target user to determine a change in the feature vector;

[0009] When the variation of the feature vector exceeds a first preset variation threshold, adjusting the first target model parameters of the face recognition model through an incremental learning module to obtain the adjusted first target model parameters;

[0010] When the variation of the feature vector exceeds a second preset variation threshold, adjusting the second target model parameters of the face recognition model through an online learning module, wherein the second preset variation threshold is greater than the first preset variation threshold;

[0011] The face recognition model is optimized according to the adjusted first target model parameters or the second target model parameters.

[0012] In some embodiments, when the change amount of the feature vector exceeds a first preset change threshold, adjusting the first target model parameters of the face recognition model through an incremental learning module to obtain the adjusted first target model parameters includes:

[0013] Identify a first target change region where the feature vector change exceeds the first preset change threshold;

[0014] Determining a first target convolutional layer in the face recognition model corresponding to the first target change region;

[0015] The first weight parameter and the first bias parameter of the first target convolutional layer are updated through the incremental learning module to obtain the updated first weight parameter and the first bias parameter.

[0016] In some embodiments, when the variation of the feature vector exceeds a second preset variation threshold, adjusting the second target model parameters of the face recognition model through an online learning module to obtain the adjusted second target model parameters includes:

[0017] Identify a second target change region where the feature vector change exceeds the second preset change threshold;

[0018] Determine a second target convolution layer, a batch normalization layer, and an activation function in the face recognition model corresponding to the second target change region;

[0019] The second weight parameter of the second target convolution layer, the second bias parameter of the second target convolution layer, the scaling parameter of the batch normalization layer, the offset parameter of the batch normalization layer and the adaptive parameter of the activation function are fine-tuned through the online learning module to obtain the fine-tuned second weight parameter, the second bias parameter, the scaling parameter, the offset parameter and the adaptive parameter.

[0020] In some embodiments, the method further comprises:

[0021] When the change in the feature vector exceeds a third preset change threshold, face recognition is performed again on the complete face of the target user, wherein the third preset change threshold is greater than the second preset change threshold.

[0022] In some embodiments, the method further comprises:

[0023] Inputting the current face image into the face recognition model for face recognition to obtain a face recognition result corresponding to the target user;

[0024] When it is determined that the face recognition result is abnormal or the confidence level of the face recognition result is lower than a preset confidence level, recording feedback data corresponding to the target user input;

[0025] The feedback data is analyzed by an error analysis module to obtain an error cause analysis result.

[0026] In some embodiments, the method further comprises:

[0027] Storing facial feature data corresponding to abnormal face recognition results or face recognition results with a confidence level lower than a preset confidence level as difficult example samples;

[0028] The face recognition model is trained based on the difficult samples.

[0029] In some embodiments, the face recognition model uses a lightweight convolutional neural network.

[0030] To achieve the above-mentioned purpose, a second aspect of an embodiment of the present application proposes a face recognition device based on adaptive learning, the device comprising:

[0031] An acquisition module is used to acquire the current face image of the target user;

[0032] A feature extraction module is used to input the current face image into a face recognition model for feature extraction to obtain a current face feature vector;

[0033] A comparison module, used to compare the current facial feature vector with the historical facial feature vector of the target user to determine the feature vector change amount;

[0034] A first adjustment module, configured to adjust the first target model parameters of the face recognition model through an incremental learning module to obtain the adjusted first target model parameters when the change amount of the feature vector exceeds a first preset change threshold;

[0035] A second adjustment module, configured to adjust the second target model parameters of the face recognition model through an online learning module when the change amount of the feature vector exceeds a second preset change threshold, wherein the adjusted second target model parameters, wherein the second preset change threshold is greater than the first preset change threshold;

[0036] An optimization module is used to optimize the face recognition model according to the adjusted first target model parameters or the second target model parameters.

[0037] To achieve the above objectives, a third aspect of an embodiment of the present application proposes an electronic device, which includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the method described in the first aspect is implemented.

[0038] To achieve the above-mentioned purpose, the fourth aspect of an embodiment of the present application proposes a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method described in the first aspect is implemented.

[0039] The face recognition method and device based on adaptive learning proposed in the present application, equipment and medium, obtain the current face image of the target user; input the current face image into the face recognition model for feature extraction to obtain the current face feature vector; compare the current face feature vector with the historical face feature vector of the target user to determine the feature vector change; when the feature vector change exceeds the first preset change threshold, adjust the first target model parameter of the face recognition model through the incremental learning module to obtain the adjusted first target model parameter; when the feature vector change exceeds the second preset change threshold, adjust the second target model parameter of the face recognition model through the online learning module, and the adjusted second target model parameter, wherein the second preset change threshold is greater than the first preset change threshold; optimize the face recognition model according to the adjusted first target model parameter or the second target model parameter. The current face image of the user is obtained and the feature extraction is performed on the current face image to obtain the current face feature vector. The current face feature vector is compared with the historical face feature vector to determine the feature vector change of the target user's face. For slightly changed facial features, incremental learning technology is used to adjust a small part of the model parameters in real time, and for significantly changed facial features, online learning technology is triggered to adjust the model parameters of the corresponding part in real time, ensuring that the model always adapts to the changes in the user's face without large-scale retraining. The embodiment of the present application realizes real-time adaptive adjustment of the model through incremental learning and online learning. The face recognition model is optimized according to the adjusted first target model parameters or second target model parameters. Based on this, the face recognition model of the embodiment of the present application can be automatically adjusted and optimized in long-term use, ensuring that the model always maintains a high recognition rate when dealing with changes in user facial features. The embodiment of the present application performs adaptive learning on the changes in the user's face without the need for retraining the model, ensuring the long-term accuracy of face recognition, especially in financial scenarios, when the user's face changes, the model parameters can be adjusted in real time through incremental learning and online learning technology, so that the model can maintain its sensitivity to user changes without retraining, and can still maintain a high accuracy after long-term use, which can avoid the situation where the user's identity authentication fails, improve the real-time and accuracy of financial services, improve customer satisfaction, and thus enhance the competitiveness of financial enterprises. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 is a flow chart of a face recognition method based on adaptive learning provided in an embodiment of the present application;

[0041] Figure 2 yes Figure 1 Flowchart of step S104 in FIG.

[0042] Figure 3yes Figure 1 Flow chart of step S105 in FIG.

[0043] Figure 4 is a flowchart of steps S401 to S403;

[0044] Figure 5 is a flowchart from step S501 to step S502;

[0045] Figure 6 is a structural schematic diagram of a face recognition device based on adaptive learning provided in an embodiment of the present application;

[0046] Figure 7 It is a schematic diagram of the hardware structure of the electronic device provided in the embodiment of the present application. DETAILED DESCRIPTION

[0047] 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.

[0048] It should be noted that, although the functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first", "second", etc. in the specification, claims and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0049] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0050] First, some nouns involved in this application are analyzed:

[0051] Artificial intelligence (AI) is a new technical science that studies and develops theories, methods, technologies and application systems for simulating, extending and expanding human intelligence. AI is a branch of computer science that attempts to understand the essence of intelligence and produce a new type of intelligent machine that can respond in a similar way to human intelligence. Research in this field includes robots, language recognition, image recognition, natural language processing and expert systems. AI can simulate the information process of human consciousness and thinking. AI is also a theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0052] Face recognition: It is a biometric technology that uses facial feature information to identify people. It uses a camera or a camera to collect images or video streams containing faces, and uses face detection technology to analyze whether there are faces. If there are faces, it further provides information about the location, size, and location of key facial organs of each face. Then, based on this information, it extracts each person's identity feature data and compares it with known facial features to identify the identity of each face.

[0053] Lightweight Convolutional Neural Network: It is a convolutional neural network (CNN) designed for resource-constrained devices (such as mobile devices or embedded systems), aiming to reduce the size and computational complexity of the model while maintaining or improving the performance of the model. The main idea of ​​this network design is to significantly reduce the number of parameters, computational workload, and memory usage of the model by optimizing the network structure and algorithm, thereby improving the inference speed and efficiency of the model. The design of lightweight convolutional neural networks is mainly aimed at the specific needs of mobile devices and embedded systems, such as low power consumption and real-time processing, so that these devices can run deep learning models more efficiently, provide faster response time and smoother user experience. The implementation methods of lightweight convolutional neural networks include but are not limited to the use of depthwise separable convolution, model compression technology, and reducing the amount of computation and the number of parameters by designing a more efficient network structure. These methods enable the model to significantly reduce storage and computing costs while maintaining a certain accuracy, thereby adapting to resource-constrained environments.

[0054] Once deployed, current face recognition models are basically fixed, and their recognition accuracy will gradually decrease due to changes in user age, body shape, and habits. In order to maintain its sensitivity to user changes, existing face recognition models also need to be retrained regularly to maintain accuracy.

[0055] In financial scenarios, for example, for customer accounts or VIP customers that require long-term maintenance, the existing face recognition models have the above-mentioned technical defects, so there is a risk of recognition failure or the model needs to be retrained, which has a negative impact on customer service. Based on this, in financial scenarios, how to ensure the long-term accuracy of face recognition without retraining the model has become a technical problem that needs to be solved urgently.

[0056] Based on this, the embodiment of the present application provides a face recognition method and apparatus, device and medium based on adaptive learning, which obtains the current face image of the target user; inputs the current face image into the face recognition model for feature extraction to obtain the current face feature vector; compares the current face feature vector with the historical face feature vector of the target user to determine the feature vector change; when the feature vector change exceeds the first preset change threshold, the first target model parameter of the face recognition model is adjusted by the incremental learning module to obtain the adjusted first target model parameter; when the feature vector change exceeds the second preset change threshold, the second target model parameter of the face recognition model is adjusted by the online learning module, and the adjusted second target model parameter, wherein the second preset change threshold is greater than the first preset change threshold; optimize the face recognition model according to the adjusted first target model parameter or the second target model parameter. The current face image of the user is obtained and the feature extraction is performed on the current face image to obtain the current face feature vector. The current face feature vector is compared with the historical face feature vector to determine the feature vector change of the target user's face. For slightly changed facial features, incremental learning technology is used to adjust a small part of the model parameters in real time, and for significantly changed facial features, online learning technology is triggered to adjust the model parameters of the corresponding part in real time, ensuring that the model always adapts to the changes in the user's face without large-scale retraining. The embodiment of the present application realizes real-time adaptive adjustment of the model through incremental learning and online learning. The face recognition model is optimized according to the adjusted first target model parameters or second target model parameters. Based on this, the face recognition model of the embodiment of the present application can be automatically adjusted and optimized in long-term use, ensuring that the model always maintains a high recognition rate when dealing with changes in user facial features. The embodiment of the present application performs adaptive learning on the changes in the user's face without the need for retraining the model, ensuring the long-term accuracy of face recognition, especially in financial scenarios, when the user's face changes, the model parameters can be adjusted in real time through incremental learning and online learning technology, so that the model can maintain its sensitivity to user changes without retraining, and can still maintain a high accuracy after long-term use, which can avoid the situation where the user's identity authentication fails, improve the real-time and accuracy of financial services, improve customer satisfaction, and thus enhance the competitiveness of financial enterprises.

[0057] The adaptive learning-based face recognition method, apparatus, device and medium provided in the embodiments of the present application are specifically illustrated through the following embodiments. First, the adaptive learning-based face recognition method in the embodiments of the present application is described.

[0058] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Artificial Intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0059] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0060] The face recognition method based on adaptive learning provided in the embodiment of the present application relates to the field of artificial intelligence technology. The face recognition method based on adaptive learning provided in the embodiment of the present application can be applied to a terminal, can also be applied to a server side, and can also be software running in a terminal or a server side. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc.; the server side can be configured as an independent physical server, or a server cluster or a distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application that implements a face recognition method based on adaptive learning, etc., but is not limited to the above forms.

[0061] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0062] It should be noted that in each specific implementation of the present application, when it comes to the need to perform relevant processing based on data related to user identity or characteristics such as user information, user behavior data, user historical data, and user location information, the user's permission or consent will be obtained first, and the collection, use, and processing of these data will comply with relevant laws, regulations, and standards. In addition, when the embodiment of the present application needs to obtain the user's sensitive personal information, the user's separate permission or consent will be obtained through a pop-up window or by jumping to a confirmation page. After clearly obtaining the user's separate permission or consent, the necessary user-related data for the normal operation of the embodiment of the present application will be obtained.

[0063] Figure 1 is an optional flowchart of the face recognition method based on adaptive learning provided in an embodiment of the present application. Figure 1 The method may include but is not limited to steps S101 to S106.

[0064] Step S101, obtaining the current face image of the target user;

[0065] Step S102, inputting the current face image into a face recognition model for feature extraction to obtain a current face feature vector;

[0066] Step S103, comparing the current facial feature vector with the historical facial feature vector of the target user to determine the feature vector change amount;

[0067] Step S104, when the variation of the feature vector exceeds a first preset variation threshold, adjusting the first target model parameters of the face recognition model through an incremental learning module to obtain adjusted first target model parameters;

[0068] Step S105, when the variation of the feature vector exceeds the second preset variation threshold, adjusting the second target model parameters of the face recognition model through the online learning module, wherein the second preset variation threshold is greater than the first preset variation threshold after the adjustment of the second target model parameters;

[0069] Step S106, optimizing the face recognition model according to the adjusted first target model parameters or the second target model parameters.

[0070] In step S101 of some embodiments, the current face image of the target user is obtained. Specifically, the current face image of the target user can be obtained through a camera. In a financial scenario, the customer can be photographed in real time through the face recognition system of the financial institution to obtain the current face image of the customer.

[0071] In step S102 of some embodiments, the current face image is input into the face recognition model for feature extraction to obtain the current face feature vector. The face recognition model is used to extract features from the current face image to obtain the current face feature vector. Among them, the face recognition model can use a lightweight convolutional neural network. In the scenarios of financial institutions such as banks, it is often necessary to deploy face recognition systems on terminal devices (such as ATM machines or self-service terminals). These devices have limited computing resources and cannot afford large and complex deep neural network models. Therefore, the use of lightweight convolutional neural networks, such as MobileNet or ShuffleNet, can significantly reduce computing costs and model volume while ensuring recognition accuracy. Lightweight convolutional neural networks can still provide efficient face recognition on embedded or resource-limited devices by designing a compact network structure and using fewer parameters. In addition, lightweight convolutional neural networks can also perform online learning and incremental updates more quickly, which helps the face recognition model to adapt in real time.

[0072] In step S103 of some embodiments, the current facial feature vector is compared with the historical facial feature vector of the target user to determine the amount of change in the feature vector. As the user's age, body shape and habits change, their facial features also change accordingly. The face recognition system identifies the part of the feature vector that has changed by comparing the feature vector differences between the user's current facial image and the historical facial features in the database. By analyzing these differences, the face recognition system can determine which areas of the face (such as eyebrows, eyes, nose, jaw, etc.) have changed. The face recognition system uses a local feature analysis algorithm, which determines which areas' feature vectors have shifted significantly based on changes in local areas (such as feature points). For example, if the eyebrows or eye areas have changed significantly, the face recognition system will focus on updating the feature parameters related to these areas.

[0073] In step S104 of some embodiments, when the change in the feature vector exceeds the first preset change threshold, the first target model parameters of the face recognition model are adjusted through the incremental learning module to obtain the adjusted first target model parameters. For facial features that undergo slight local changes, a specific small portion of model parameters can be adjusted in real time through the incremental learning module. Specifically, a small sample update mechanism can be used to update only a small portion of the parameters each time to ensure that the model always adapts to user changes without the need for large-scale retraining. For example, when the face recognition system identifies changes in certain areas, the incremental learning module will only fine-tune the weights and biases in the corresponding convolutional layers for these changed areas without affecting the parameters of other areas. This local update mechanism enables the model to quickly adapt to local facial changes, such as aging, changes in expression, etc.

[0074] In step S105 of some embodiments, when the feature vector change exceeds the second preset change threshold, the second target model parameters of the face recognition model are adjusted through the online learning module, and the adjusted second target model parameters, wherein the second preset change threshold is greater than the first preset change threshold. For facial features that have changed significantly, the online learning module is triggered to adjust the model parameters of the corresponding part in real time to ensure that the model always adapts to the user's facial changes without large-scale retraining. Every time the user authenticates the identity, the online learning module captures the current facial feature vector of the user's current facial image and calculates the difference between it and the historical facial feature vector already in the database. If the difference value exceeds a certain preset threshold, it means that the user's face has changed significantly, and the online learning process is triggered. This online learning process enables the model to quickly adapt to significant facial changes, such as significant facial changes caused by age changes, body shape changes, and habit changes.

[0075] In step S106 of some embodiments, the face recognition model is optimized according to the adjusted first target model parameters or second target model parameters. By adjusting the corresponding model parameters through incremental learning and online learning, the real-time adaptive adjustment of the model to face recognition is achieved. Therefore, the face recognition model can be automatically adjusted and optimized in long-term use, ensuring that the model always maintains a high recognition rate when dealing with changes in user facial features.

[0076] In financial scenarios, for example, the face recognition method of the embodiment of the present application is applicable to long-term account management and customer accounts that need to be maintained for a long time, such as exclusive services for high-net-worth customers, to ensure that the system can effectively identify and verify customer identities in the long term. For another example, the face recognition method of the embodiment of the present application is applicable to VIP customer identification, identifying and automatically providing customized services for VIP customers in financial institutions.

[0077] Steps S101 to S106 shown in the embodiment of the present application are to obtain the current face image of the target user; input the current face image into the face recognition model for feature extraction to obtain the current face feature vector; compare the current face feature vector with the historical face feature vector of the target user to determine the feature vector change; when the feature vector change exceeds the first preset change threshold, adjust the first target model parameter of the face recognition model through the incremental learning module to obtain the adjusted first target model parameter; when the feature vector change exceeds the second preset change threshold, adjust the second target model parameter of the face recognition model through the online learning module, and the adjusted second target model parameter, wherein the second preset change threshold is greater than the first preset change threshold; optimize the face recognition model according to the adjusted first target model parameter or the second target model parameter. The current face image of the user is obtained and the feature extraction of the current face image is performed to obtain the current face feature vector. The current face feature vector is compared with the historical face feature vector to determine the feature vector change of the target user's face. For slightly changed facial features, incremental learning technology is used to adjust a small part of the model parameters in real time, and for significantly changed facial features, online learning technology is triggered to adjust the model parameters of the corresponding part in real time, ensuring that the model always adapts to the changes in the user's face without large-scale retraining. The embodiment of the present application realizes real-time adaptive adjustment of the model through incremental learning and online learning. The face recognition model is optimized according to the adjusted first target model parameters or second target model parameters. Based on this, the face recognition model of the embodiment of the present application can be automatically adjusted and optimized in long-term use, ensuring that the model always maintains a high recognition rate when dealing with changes in user facial features. The embodiment of the present application performs adaptive learning on the changes in the user's face without the need for retraining the model, ensuring the long-term accuracy of face recognition, especially in financial scenarios, when the user's face changes, the model parameters can be adjusted in real time through incremental learning and online learning technology, so that the model can maintain its sensitivity to user changes without retraining, and can still maintain a high accuracy after long-term use, which can avoid the situation where the user's identity authentication fails, improve the real-time and accuracy of financial services, improve customer satisfaction, and thus enhance the competitiveness of financial enterprises.

[0078] See also Figure 2 In some embodiments, step S104 may include but is not limited to steps S201 to S203:

[0079] Step S201, identifying a first target change region where a feature vector change exceeds a first preset change threshold;

[0080] Step S202, determining a first target convolution layer in the face recognition model corresponding to the first target change region;

[0081] Step S203: Update the first weight parameter and the first bias parameter of the first target convolution layer through the incremental learning module to obtain the updated first weight parameter and the first bias parameter.

[0082] In some embodiments, a local feature analysis algorithm may be used to identify a first target change region where the feature vector change exceeds a first preset change threshold. The local feature analysis algorithm determines which regions have a large deviation in feature vectors by the change in local face regions (such as feature points). For example, if there is a local change in the eyebrow or eye region, the face recognition system will focus on updating the feature parameters related to these regions.

[0083] In some embodiments, the incremental learning module can update the model parameters according to the changed parts of the user's face each time the user authenticates. The incremental learning module adopts a small sample update mechanism, which only updates a small part of the parameters each time, ensuring that the model always adapts to the changes of the user without large-scale retraining.

[0084] In some embodiments, the face recognition system compares the feature vector differences between the current face image of the user and the historical face features in the database to identify the parts where the features have changed locally. By analyzing these differences, the face recognition system can determine which areas (such as eyebrows, eyes, nose, jaw, etc.) have changed.

[0085] In some embodiments, when the face recognition system identifies changes in certain areas, the incremental learning module will fine-tune the weights and biases in the corresponding convolutional layers only for these changed areas without affecting the parameters of other areas. This local update mechanism enables the model to quickly adapt to local changes in the face, such as aging, changes in expression, etc.

[0086] See also Figure 3 In some embodiments, step S105 may include but is not limited to steps S301 to S303:

[0087] Step S301, identifying a second target change region where the feature vector change exceeds a second preset change threshold;

[0088] Step S302, determining a second target convolution layer, a batch normalization layer, and an activation function in the face recognition model corresponding to the second target change region;

[0089] Step S303, fine-tuning the second weight parameter of the second target convolution layer, the second bias parameter of the second target convolution layer, the scaling parameter of the batch normalization layer, the offset parameter of the batch normalization layer, and the adaptive parameter of the activation function through the online learning module to obtain the fine-tuned second weight parameter, second bias parameter, scaling parameter, offset parameter, and adaptive parameter.

[0090] In some embodiments, when a user is authenticated, a face recognition model is called for identification. After identification, if significant changes are detected in the user's facial features, such as significant changes in the face caused by age, body shape or habits, the online learning module will be triggered in real time for fine-tuning.

[0091] In some embodiments, the online learning module captures the current facial feature vector of the user's current facial image each time the user authenticates the user, and calculates the difference between the current facial feature vector and the historical facial feature vector in the database. If the difference exceeds a certain preset threshold, it means that the user's face has changed significantly, and the online learning process is triggered.

[0092] In some embodiments, the online learning module fine-tunes the local convolution kernel weights of the face recognition model through small sample learning. The specific fine-tuning objects include:

[0093] 1. Convolutional layer weights and biases: In particular, the weights in high-level convolutional layers are locally updated based on new feature data.

[0094] 2. Parameters of the Batch Normalization (BN) layer: If the brightness and contrast of the face change, the parameters of the BN layer will be adjusted to adapt to the new lighting conditions.

[0095] 3. Parameter adjustment of activation function: For some activation functions with adaptive parameters (such as PReLU), the system can dynamically adjust these parameters according to new features to ensure consistent feature extraction effects under different conditions.

[0096] In some embodiments, when the change in the feature vector exceeds a third preset change threshold, face recognition is performed again on the complete face of the target user, wherein the third preset change threshold is greater than the second preset change threshold.

[0097] When it is recognized that the change of the user's facial features exceeds the third preset change threshold, the user is required to re-perform a complete facial capture to ensure the accuracy of facial recognition. It should be noted that the setting value of the third preset change threshold is greater than the second preset change threshold, that is, the case where the change of the user's facial features exceeds the third preset change threshold means that the degree of change of the user's facial features is higher than the above-mentioned significant change.

[0098] See also Figure 4 In some embodiments, the face recognition method based on adaptive learning may include but is not limited to steps S401 to S403:

[0099] Step S401, inputting the current face image into a face recognition model for face recognition, and obtaining a face recognition result corresponding to the target user;

[0100] Step S402, when it is determined that the face recognition result is abnormal or the confidence level of the face recognition result is lower than a preset confidence level, recording the feedback data input by the corresponding target user;

[0101] Step S403: Analyze the feedback data through the error analysis module to obtain error cause analysis results.

[0102] The face recognition method based on adaptive learning of the embodiment of the present application also has an abnormal situation feedback and analysis mechanism. During the recognition process, the special situations encountered (failure to recognize or inaccurate results) will be recorded in real time. For example, when the face recognition result is abnormal or a low confidence result occurs each time, the feedback data manually input by the user at that time will be recorded. And the feedback data can be analyzed by the error analysis module to obtain the error cause analysis result. The error cause of each user feedback is automatically analyzed by the error analysis module, for example, it is judged that the recognition problem is caused by factors such as changes in illumination, changes in facial posture or outdated system model parameters.

[0103] See also Figure 5 In some embodiments, the face recognition method based on adaptive learning of the present application may also include but is not limited to steps S501 to S502:

[0104] Step S501, storing facial feature data corresponding to abnormal face recognition results or face recognition results with a confidence level lower than a preset confidence level as difficult example samples;

[0105] Step S502: training a face recognition model based on hard examples.

[0106] The face recognition method based on adaptive learning in the embodiment of the present application also has an abnormal situation feedback and optimization mechanism. In each recognition operation of the user, the face recognition system will not only determine the confidence of the recognition through the algorithm, but also allow the user to manually feedback (for example, when the recognition fails or the result is inaccurate). This type of feedback data will be marked as an abnormal sample, and the system will record and analyze it in the background.

[0107] The face recognition system also automatically analyzes the cause of each user feedback error through the error analysis module to determine whether the recognition problem is caused by factors such as changes in lighting, facial posture, or outdated system model parameters. Subsequently, the face recognition system will store these feedback samples as "hard examples" and conduct special training when the model is updated to improve the model's performance in similar scenarios.

[0108] The feedback data input by users will be used in conjunction with the online learning mechanism. Each time the model is updated, the face recognition system will pay special attention to the difficult samples in the feedback and conduct intensive training based on this data. At the same time, the face recognition system will record the common features of multiple feedbacks and gradually optimize the parameters of the convolutional layer and decision layer.

[0109] By combining online learning with user feedback, the system can automatically adjust and optimize during long-term use, ensuring that the model always maintains a high recognition rate when responding to changes in user facial features.

[0110] In financial scenarios, for example, the face recognition method of the embodiment of the present application is applicable to long-term account management and customer accounts that need to be maintained for a long time, such as exclusive services for high-net-worth customers, to ensure that the system can effectively identify and verify customer identities in the long term. For another example, the face recognition method of the embodiment of the present application is applicable to VIP customer identification, identifying and automatically providing customized services for VIP customers in financial institutions.

[0111] See also Figure 6 The embodiment of the present application further provides a face recognition device based on adaptive learning, which can implement the above-mentioned face recognition method based on adaptive learning, and the device includes:

[0112] An acquisition module 610 is used to acquire a current face image of a target user;

[0113] The feature extraction module 620 is used to input the current face image into the face recognition model for feature extraction to obtain the current face feature vector;

[0114] A comparison module 630 is used to compare the current facial feature vector with the historical facial feature vector of the target user to determine the change amount of the feature vector;

[0115] A first adjustment module 640 is used to adjust the first target model parameters of the face recognition model through the incremental learning module to obtain the adjusted first target model parameters when the feature vector change exceeds a first preset change threshold;

[0116] A second adjustment module 650 is used to adjust the second target model parameters of the face recognition model through the online learning module when the feature vector change exceeds the second preset change threshold, wherein the second preset change threshold is greater than the first preset change threshold;

[0117] The optimization module 660 is used to optimize the face recognition model according to the adjusted first target model parameters or the second target model parameters.

[0118] In some embodiments of the present application, the acquisition module 610 acquires the current facial image of the target user; the feature extraction module 620 inputs the current facial image into the facial recognition model for feature extraction to obtain the current facial feature vector; the comparison module 630 compares the current facial feature vector with the historical facial feature vector of the target user to determine the change in the feature vector; the first adjustment module 640 adjusts the first target model parameters of the facial recognition model through the incremental learning module when the change in the feature vector exceeds the first preset change threshold to obtain the adjusted first target model parameters; the second adjustment module 650 adjusts the second target model parameters of the facial recognition model through the online learning module when the change in the feature vector exceeds the second preset change threshold, and the adjusted second target model parameters, wherein the second preset change threshold is greater than the first preset change threshold; the optimization module 660 optimizes the facial recognition model according to the adjusted first target model parameters or the second target model parameters.

[0119] In some embodiments of the present application, the current face image of the target user is obtained. Specifically, the current face image of the target user can be obtained through a camera. In a financial scenario, the customer can be photographed in real time through the face recognition system of the financial institution to obtain the current face image of the customer.

[0120] In some embodiments of the present application, the current face image is input into the face recognition model for feature extraction to obtain the current face feature vector. The face recognition model is used to extract features from the current face image to obtain the current face feature vector. Among them, the face recognition model can use a lightweight convolutional neural network. In the scenarios of financial institutions such as banks, it is often necessary to deploy face recognition systems on terminal devices (such as ATM machines or self-service terminals). These devices have limited computing resources and cannot afford large and complex deep neural network models. Therefore, the use of lightweight convolutional neural networks, such as MobileNet or ShuffleNet, can significantly reduce computing costs and model volume while ensuring recognition accuracy. Lightweight convolutional neural networks can still provide efficient face recognition on embedded or resource-limited devices by designing a compact network structure and using fewer parameters. In addition, lightweight convolutional neural networks can also perform online learning and incremental updates more quickly, which helps the face recognition model to adapt in real time.

[0121] In some embodiments of the present application, the current facial feature vector is compared with the historical facial feature vector of the target user to determine the amount of change in the feature vector. Due to changes in the user's age, body shape and habits, their facial features also change accordingly. The face recognition system identifies the part of the feature vector that has changed by comparing the feature vector differences between the user's current face image and the historical face features in the database. By analyzing these differences, the face recognition system can determine which areas of the face (such as eyebrows, eyes, nose, jaw, etc.) have changed. The face recognition system uses a local feature analysis algorithm, which determines which areas have undergone a large deviation in the feature vector by changes in local areas (such as feature points). For example, if the eyebrows or eye areas have changed significantly, the face recognition system will focus on updating the feature parameters related to these areas.

[0122] In some embodiments of the present application, when the change in the feature vector exceeds a first preset change threshold, the first target model parameters of the face recognition model are adjusted through the incremental learning module to obtain the adjusted first target model parameters. For facial features that undergo slight local changes, a specific small portion of model parameters can be adjusted in real time through the incremental learning module. Specifically, a small sample update mechanism can be used to update only a small portion of the parameters each time to ensure that the model always adapts to user changes without large-scale retraining. For example, when the face recognition system identifies changes in certain areas, the incremental learning module will only fine-tune the weights and biases in the corresponding convolutional layers for these changed areas without affecting the parameters of other areas. This local update mechanism enables the model to quickly adapt to local facial changes, such as aging, changes in expression, etc.

[0123] In some embodiments of the present application, when the feature vector change exceeds the second preset change threshold, the second target model parameters of the face recognition model are adjusted through the online learning module, and the adjusted second target model parameters, wherein the second preset change threshold is greater than the first preset change threshold. For facial features that have changed significantly, the online learning module is triggered to adjust the model parameters of the corresponding part in real time to ensure that the model always adapts to the user's facial changes without large-scale retraining. Every time the user authenticates the identity, the online learning module captures the current facial feature vector of the user's current facial image and calculates the difference between it and the historical facial feature vector already in the database. If the difference value exceeds a certain preset threshold, it means that the user's face has changed significantly, and the online learning process is triggered. This online learning process enables the model to quickly adapt to significant facial changes, such as significant changes in the user's face caused by age changes, body shape changes, and habit changes.

[0124] In some embodiments of the present application, the face recognition model is optimized according to the adjusted first target model parameters or second target model parameters. By adjusting the corresponding model parameters through incremental learning and online learning, the real-time adaptive adjustment of the model to face recognition is achieved. Therefore, the face recognition model can be automatically adjusted and optimized in long-term use, ensuring that the model always maintains a high recognition rate when dealing with changes in user facial features.

[0125] Based on this, in the face recognition device based on adaptive learning in the embodiment of the present application, the acquisition module 610 acquires the current face image of the target user; the feature extraction module 620 inputs the current face image into the face recognition model for feature extraction to obtain the current face feature vector; the comparison module 630 compares the current face feature vector with the historical face feature vector of the target user to determine the change in the feature vector; the first adjustment module 640 adjusts the first target model parameters of the face recognition model through the incremental learning module when the change in the feature vector exceeds the first preset change threshold to obtain the adjusted first target model parameters; the second adjustment module 650 adjusts the second target model parameters of the face recognition model through the online learning module when the change in the feature vector exceeds the second preset change threshold, and the adjusted second target model parameters, wherein the second preset change threshold is greater than the first preset change threshold; the optimization module 660 optimizes the face recognition model according to the adjusted first target model parameters or the second target model parameters. The present application obtains the current face image of the target user; inputs the current face image into the face recognition model for feature extraction to obtain the current face feature vector; compares the current face feature vector with the historical face feature vector of the target user to determine the feature vector change; when the feature vector change exceeds the first preset change threshold, the first target model parameter of the face recognition model is adjusted by the incremental learning module to obtain the adjusted first target model parameter; when the feature vector change exceeds the second preset change threshold, the second target model parameter of the face recognition model is adjusted by the online learning module, and the adjusted second target model parameter, wherein the second preset change threshold is greater than the first preset change threshold; optimizes the face recognition model according to the adjusted first target model parameter or the second target model parameter. The current face image of the user is obtained and the feature extraction of the current face image is performed to obtain the current face feature vector. The current face feature vector is compared with the historical face feature vector to determine the feature vector change of the target user's face. For facial features that have slightly changed, incremental learning technology is used to adjust a small part of the model parameters in real time. For facial features that have changed significantly, online learning technology is triggered to adjust the model parameters of the corresponding part in real time, ensuring that the model always adapts to the user's facial changes without large-scale retraining. The embodiment of the present application realizes real-time adaptive adjustment of the model through incremental learning and online learning. The face recognition model is optimized according to the adjusted first target model parameters or second target model parameters. Based on this, the face recognition model of the embodiment of the present application can be automatically adjusted and optimized during long-term use, ensuring that the model always maintains a high recognition rate when responding to changes in user facial features.The embodiments of the present application perform adaptive learning on changes in the user's face without the need for retraining of the model, thereby ensuring the long-term accuracy of face recognition. Especially in financial scenarios, when the user's face changes, the model parameters can be adjusted in real time through incremental learning and online learning techniques, so that the model can maintain its sensitivity to user changes without retraining, and can still maintain high accuracy after long-term use, which can avoid situations where user identity authentication fails, improve the real-time and accuracy of financial services, and improve customer satisfaction, thereby enhancing the competitiveness of financial enterprises.

[0126] The specific implementation of the face recognition device based on adaptive learning is basically the same as the specific implementation of the face recognition method based on adaptive learning described above, and will not be repeated here.

[0127] The embodiment of the present application also provides an electronic device, the electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the above-mentioned face recognition method based on adaptive learning when executing the computer program. The electronic device can be any intelligent terminal including a tablet computer, a car computer, etc.

[0128] See also Figure 7 , Figure 7 The hardware structure of an electronic device of another embodiment is illustrated, and the electronic device includes:

[0129] The processor 701 can be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application.

[0130] The memory 702 may be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 702 can store an operating system and other application programs. When the technical solution provided by the embodiment of this specification is implemented by software or firmware, the relevant program code is stored in the memory 702, and the processor 701 is called to execute the face recognition method based on adaptive learning of the embodiment of this application, that is, by obtaining the current face image of the target user; inputting the current face image into the face recognition model for feature extraction to obtain the current face feature vector; comparing the current face feature vector with the historical face feature vector of the target user to determine the feature vector change; when the feature vector change exceeds the first preset change threshold, adjusting the first target model parameter of the face recognition model through the incremental learning module to obtain the adjusted first target model parameter; when the feature vector change exceeds the second preset change threshold, adjusting the second target model parameter of the face recognition model through the online learning module, the adjusted second target model parameter, wherein the second preset change threshold is greater than the first preset change threshold; optimizing the face recognition model according to the adjusted first target model parameter or the second target model parameter. The current face image of the user is obtained, and the feature extraction of the current face image is performed to obtain the current face feature vector. The current facial feature vector is compared with the historical facial feature vector to determine the change in the feature vector of the target user's face. For facial features that have changed slightly, incremental learning technology is used to adjust a specific small part of the model parameters in real time. For facial features that have changed significantly, online learning technology is triggered to adjust the model parameters of the corresponding part in real time, ensuring that the model always adapts to the user's facial changes without large-scale retraining. The embodiment of the present application realizes real-time adaptive adjustment of the model through incremental learning and online learning. The face recognition model is optimized according to the adjusted first target model parameters or the second target model parameters. Based on this, the face recognition model of the embodiment of the present application can be automatically adjusted and optimized during long-term use, ensuring that the model always maintains a high recognition rate when responding to changes in user facial features. The embodiments of the present application perform adaptive learning on changes in the user's face without the need for retraining of the model, thereby ensuring the long-term accuracy of face recognition. Especially in financial scenarios, when the user's face changes, the model parameters can be adjusted in real time through incremental learning and online learning techniques, so that the model can maintain its sensitivity to user changes without retraining, and can still maintain high accuracy after long-term use, which can avoid situations where user identity authentication fails, improve the real-time and accuracy of financial services, and improve customer satisfaction, thereby enhancing the competitiveness of financial enterprises.

[0131] The input / output interface 703 is used to implement information input and output.

[0132] The communication interface 704 is used to realize the communication interaction between this device and other devices. Communication can be realized through wired methods (such as USB, network cable, etc.) or wireless methods (such as mobile network, WI FI, Bluetooth, etc.).

[0133] The bus transmits information between various components of the device (eg, the processor 701 , the memory 702 , the input / output interface 703 , and the communication interface 704 ).

[0134] The processor 701 , the memory 702 , the input / output interface 703 and the communication interface 704 are connected to each other in communication within the device via a bus.

[0135] An embodiment of the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the above-mentioned face recognition method based on adaptive learning.

[0136] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0137] The face recognition method based on adaptive learning, the face recognition device based on adaptive learning, the electronic device and the storage medium provided in the embodiment of the present application are as follows: obtaining the current face image of the target user; inputting the current face image into the face recognition model for feature extraction to obtain the current face feature vector; comparing the current face feature vector with the historical face feature vector of the target user to determine the feature vector change; when the feature vector change exceeds the first preset change threshold, adjusting the first target model parameter of the face recognition model through the incremental learning module to obtain the adjusted first target model parameter; when the feature vector change exceeds the second preset change threshold, adjusting the second target model parameter of the face recognition model through the online learning module, the adjusted second target model parameter, wherein the second preset change threshold is greater than the first preset change threshold; optimizing the face recognition model according to the adjusted first target model parameter or the second target model parameter. Obtaining the current face image of the user and extracting the features of the current face image to obtain the current face feature vector. Comparing the current face feature vector with the historical face feature vector, thereby determining the feature vector change of the target user's face. For slightly changed facial features, incremental learning technology is used to adjust a small part of the model parameters in real time, and for significantly changed facial features, online learning technology is triggered to adjust the model parameters of the corresponding part in real time, ensuring that the model always adapts to the changes in the user's face without large-scale retraining. The embodiment of the present application realizes real-time adaptive adjustment of the model through incremental learning and online learning. The face recognition model is optimized according to the adjusted first target model parameters or second target model parameters. Based on this, the face recognition model of the embodiment of the present application can be automatically adjusted and optimized in long-term use, ensuring that the model always maintains a high recognition rate when dealing with changes in user facial features. The embodiment of the present application performs adaptive learning on the changes in the user's face without the need for retraining the model, ensuring the long-term accuracy of face recognition, especially in financial scenarios, when the user's face changes, the model parameters can be adjusted in real time through incremental learning and online learning technology, so that the model can maintain its sensitivity to user changes without retraining, and can still maintain a high accuracy after long-term use, which can avoid the situation where the user's identity authentication fails, improve the real-time and accuracy of financial services, improve customer satisfaction, and thus enhance the competitiveness of financial enterprises.

[0138] It will be appreciated by those skilled in the art that all or some of the steps and systems in the disclosed method above may be implemented as software, firmware, hardware and appropriate combinations thereof. Some physical components or all physical components may be implemented as software executed by a processor, such as a central processing unit, a digital signal processor or a microprocessor, or may be implemented as hardware, or may be implemented as an integrated circuit, such as an application specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include a computer storage medium (or a non-transitory medium) and a communication medium (or a temporary medium). As known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as a computer-readable program, a data structure, a program module or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disk (DVD) or other optical disk storage, magnetic cassette, magnetic tape, disk storage or other magnetic storage device, or any other medium that may be used to store desired information and may be accessed by a computer. Furthermore, it is well known to those skilled in the art that communication media generally contain computer-readable programs, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

[0139] The embodiments described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0140] Those skilled in the art will appreciate that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0141] The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0142] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices may be implemented as software, firmware, hardware, or a suitable combination thereof.

[0143] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0144] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0145] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the above units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0146] The units described above as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0147] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0148] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including multiple instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM), disk or optical disk and other media that can store programs.

[0149] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but the scope of the rights of the present invention is not limited thereto. Any modification, equivalent substitution and improvement made by a person skilled in the art without departing from the scope and essence of the present invention should be within the scope of the rights of the present invention.

Claims

1. A face recognition method based on adaptive learning, characterized in that: The method comprises: Get the current face image of the target user; Inputting the current face image into a face recognition model for feature extraction to obtain a current face feature vector; Compare the current facial feature vector with the historical facial feature vector of the target user to determine a change in the feature vector; When the variation of the feature vector exceeds a first preset variation threshold, adjusting the first target model parameters of the face recognition model through an incremental learning module to obtain the adjusted first target model parameters; When the variation of the feature vector exceeds a second preset variation threshold, adjusting the second target model parameters of the face recognition model through an online learning module, wherein the second preset variation threshold is greater than the first preset variation threshold; The face recognition model is optimized according to the adjusted first target model parameters or the second target model parameters.

2. The method according to claim 1, characterized in that When the variation of the feature vector exceeds a first preset variation threshold, adjusting the first target model parameter of the face recognition model through an incremental learning module to obtain the adjusted first target model parameter includes: Identify a first target change region where the feature vector change exceeds the first preset change threshold; Determining a first target convolutional layer in the face recognition model corresponding to the first target change region; The first weight parameter and the first bias parameter of the first target convolutional layer are updated through the incremental learning module to obtain the updated first weight parameter and the first bias parameter.

3. The method according to claim 1, characterized in that When the variation of the feature vector exceeds a second preset variation threshold, adjusting the second target model parameters of the face recognition model through an online learning module to obtain the adjusted second target model parameters includes: Identify a second target change region where the feature vector change exceeds the second preset change threshold; Determining a second target convolution layer, a batch normalization layer, and an activation function in the face recognition model corresponding to the second target change region; The second weight parameter of the second target convolution layer, the second bias parameter of the second target convolution layer, the scaling parameter of the batch normalization layer, the offset parameter of the batch normalization layer and the adaptive parameter of the activation function are fine-tuned through the online learning module to obtain the fine-tuned second weight parameter, the second bias parameter, the scaling parameter, the offset parameter and the adaptive parameter.

4. The method according to claim 1, characterized in that: The method further comprises: When the change in the feature vector exceeds a third preset change threshold, face recognition is performed again on the complete face of the target user, wherein the third preset change threshold is greater than the second preset change threshold.

5. The method according to claim 1, characterized in that The method further comprises: Inputting the current face image into the face recognition model for face recognition to obtain a face recognition result corresponding to the target user; When it is determined that the face recognition result is abnormal or the confidence level of the face recognition result is lower than a preset confidence level, recording feedback data corresponding to the target user input; The feedback data is analyzed by an error analysis module to obtain an error cause analysis result.

6. The method according to claim 5, characterized in that The method further comprises: Storing facial feature data corresponding to abnormal face recognition results or face recognition results with a confidence level lower than a preset confidence level as difficult example samples; The face recognition model is trained based on the difficult samples.

7. The method according to any one of claims 1 to 6, characterized in that: The face recognition model adopts a lightweight convolutional neural network.

8. A face recognition device based on adaptive learning, characterized in that: The device comprises: An acquisition module is used to acquire the current face image of the target user; A feature extraction module is used to input the current face image into a face recognition model for feature extraction to obtain a current face feature vector; A comparison module, used to compare the current facial feature vector with the historical facial feature vector of the target user to determine the feature vector change amount; A first adjustment module, configured to adjust the first target model parameters of the face recognition model through an incremental learning module to obtain the adjusted first target model parameters when the change amount of the feature vector exceeds a first preset change threshold; A second adjustment module, configured to adjust the second target model parameters of the face recognition model through an online learning module when the change amount of the feature vector exceeds a second preset change threshold, wherein the adjusted second target model parameters, wherein the second preset change threshold is greater than the first preset change threshold; An optimization module is used to optimize the face recognition model according to the adjusted first target model parameters or the second target model parameters.

9. An electronic device, characterized in that: The electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the face recognition method based on adaptive learning as described in any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the face recognition method based on adaptive learning according to any one of claims 1 to 7 is implemented.

Images