Desensitization model optimization method, electronic equipment and terminal equipment

Abstract

The embodiment of the invention relates to the technical field of data processing, and discloses a desensitization model optimization method, electronic equipment and terminal equipment, and the method comprises the steps that the electronic equipment can determine the number of non-desensitized images needing to be sent to the terminal equipment; the electronic equipment can send the non-desensitized images of the image number to the terminal equipment in advance according to the image number when the terminal equipment is started and identification of the non-desensitized images is not executed. The electronic equipment can also receive the non-desensitized images sent by the electronic equipment in real time according to the number of the images in the process of processing the non-desensitized images by the user. The terminal equipment can determine the label of each non-desensitized image in the process of identifying the non-desensitized images, and sends the labels to the electronic equipment. The electronic equipment can utilize the tags to optimize the desensitization model. By applying the technical scheme of the invention, the omission ratio of the desensitization model can be reduced, and the privacy protection effect can be improved.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of data processing technology, and specifically to a method for optimizing a desensitization model, an electronic device, and a terminal device. Background Art

[0002] In order to protect customer privacy, facial images collected by the camera on the vehicle need to be desensitized using a desensitization model in the cloud. However, when the desensitization model processes real data, there may be problems with missed detection, resulting in the leakage of user privacy. Therefore, reducing the missed detection rate of the desensitization algorithm has become an urgent problem to be solved. Summary of the invention

[0003] In view of the above problems, an embodiment of the present application provides a method for optimizing a desensitization model, an electronic device, and a terminal device, which are used to solve the problem of reducing the missed detection rate of the desensitization algorithm.

[0004] According to a first aspect of an embodiment of the present application, a method for optimizing a desensitization model is provided, the method comprising:

[0005] Determine the number of images that are not desensitized and send them to the terminal device, and send the number of images that are not desensitized to the terminal device; the images that are not desensitized are images that are identified by the desensitization model as not requiring desensitization processing;

[0006] Receive the label of the non-desensitized image sent by the terminal device, and optimize the desensitization model according to the label of the non-desensitized image.

[0007] According to a second aspect of an embodiment of the present application, a method for optimizing a desensitization model is provided, the method comprising:

[0008] Receiving a non-desensitized image under the number of images sent by the electronic device; and determining a label of the non-desensitized image; the non-desensitized image is an image that is identified by the desensitization model as not requiring desensitization processing;

[0009] The label of the non-desensitized image is sent to the electronic device; wherein the label of the non-desensitized image is used to optimize the desensitization model.

[0010] According to a third aspect of an embodiment of the present application, there is provided a device for optimizing a desensitization model, comprising:

[0011] A sending module, used to determine the number of images that are not desensitized and send them to the terminal device, and send the number of images that are not desensitized to the terminal device;

[0012] A receiving module is used to receive the label of the non-desensitized image sent by the terminal device, and optimize the desensitization model according to the label of the non-desensitized image.

[0013] According to a fourth aspect of an embodiment of the present application, there is provided a device for optimizing a desensitization model, comprising:

[0014] A receiving module, used for receiving a non-desensitized image under the number of images sent by the electronic device; and determining a label of the non-desensitized image;

[0015] A sending module is used to send the label of the non-desensitized image to the electronic device; wherein the label of the non-desensitized image is used to optimize the desensitization model.

[0016] According to the fifth aspect of the embodiments of the present application, there is provided an electronic device, comprising: a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other through the communication bus; the memory is used to store at least one executable instruction, and the executable instruction enables the processor to execute the method in the first aspect and any possible design of the first aspect.

[0017] According to the sixth aspect of an embodiment of the present application, a terminal device is provided, comprising: a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other through the communication bus; the memory is used to store at least one executable instruction, and the executable instruction enables the processor to execute the method in the second aspect and any possible design of the second aspect.

[0018] According to the seventh aspect of the embodiments of the present application, a computer-readable storage medium is provided, wherein at least one executable instruction is stored in the storage medium, and when the executable instruction is run on an electronic device, the electronic device executes the method in the first aspect and any possible design of the first aspect, or, when the executable instruction is run on a terminal device, the terminal device executes the method in the second aspect and any possible design of the second aspect.

[0019] The embodiment of the present application determines the number of non-desensitized images that need to be sent to the terminal device; based on the number of images, when the terminal device is turned on and the recognition of the non-desensitized images is not performed, the non-desensitized images of the number of images are pre-sent to the terminal device; when the user performs the processing of the non-desensitized images, based on the number of images, the non-desensitized images sent by the electronic device in real time are received; when the user views and processes the images, the real-time and smooth effect of the images is ensured, thereby improving the user experience. In addition, the embodiment of the present application can also optimize the desensitization model by obtaining the labels of the desensitized images calibrated by the user in the terminal device, and using these labels to optimize the desensitization model, thereby reducing the missed detection rate of the desensitization model and improving the privacy protection effect.

[0020] The above description is only an overview of the technical solution of the embodiment of the present application. In order to more clearly understand the technical means of the embodiment of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the embodiment of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The accompanying drawings are only used to illustrate the embodiments and are not to be considered as limiting the present application. In addition, the same reference symbols are used to represent the same components throughout the accompanying drawings. In the accompanying drawings:

[0022] Figure 1 A signaling interaction diagram of a first embodiment of the optimization method of a desensitization model provided by the present application is shown;

[0023] Figure 2 A schematic diagram showing a flow chart of a second embodiment of the optimization method of the desensitization model provided by the present application;

[0024] Figure 3 A schematic diagram of the structure of the optimization device for the desensitization model provided by the present application is shown;

[0025] Figure 4 A schematic diagram of the structure of the optimization device for the desensitization model provided by the present application is shown;

[0026] Figure 5 A schematic diagram showing the structure of an embodiment of an electronic device provided by the present application is shown;

[0027] Figure 6 A schematic structural diagram of an embodiment of a terminal device provided in the present application is shown. DETAILED DESCRIPTION

[0028] The exemplary embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present application are shown in the drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments set forth herein.

[0029] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0030] With the increasing privacy requirements, the privacy and security issues of smart car cockpit technology have become increasingly important, and anonymization detection has become an indispensable privacy protection method in smart car cockpits. At present, in vehicles, the in-car cameras and the external radar sensors are often from different brands, and the anonymization algorithms they apply are also different, resulting in a lack of unified system management for the collection and desensitization testing of facial feature values ​​inside and outside the car. In addition, due to the limitations of machine learning algorithms, sometimes faces are not recognized and desensitization is missed. For example, when 100 face images are collected, the machine learning algorithm determines that 99 of them include faces and performs desensitization, while the remaining image may not be desensitized because the machine learning algorithm cannot recognize the face in the image. This image may cause face leakage. Therefore, how to reduce the missed detection rate of the desensitization algorithm has become an urgent problem to be solved. If the recognition accuracy of the machine learning algorithm needs to be further improved, it is necessary to further judge the recognition results manually, and further train the machine learning algorithm based on the judgment results. However, according to regulatory requirements, these un-desensitized photos can only be viewed by the car owner, and the cloud is only forwarded but not retained. Therefore, adding car owners to the optimization plan of machine learning algorithms is an important part of achieving machine learning algorithm optimization.

[0031] The desensitization model used in this application is a machine learning model. The optimization method of the desensitization model of this application can forward the processed result as an un-desensitized image to the user's terminal device after processing the collected initial image using the desensitization model. The user can view the un-desensitized image through the terminal device after agreeing to participate in the algorithm optimization plan. The user can judge the displayed photo to confirm whether this un-desensitized image is an image that does not need to be desensitized, or an image that needs to be desensitized but has not been successfully desensitized. Subsequently, the terminal device can transmit the image features of the image that needs to be desensitized but has not been successfully desensitized back to the cloud for further optimization of the desensitization model. This application introduces a car owner participation mechanism and optimizes the anonymization algorithm using the car owner's manual judgment, which effectively makes up for the limitations of machine recognition. At the same time, the collected pictures are accurately labeled and classified, which significantly improves the processing efficiency and accuracy. In addition, the end-to-end transmission method is adopted to ensure the security and privacy protection of un-desensitized pictures during the transmission process. Furthermore, the linkage enterprise points system successfully stimulates the enthusiasm of users to participate and strongly promotes the continuous optimization of the algorithm.

[0032] Figure 1 The signaling interaction diagram of the first embodiment of the optimization method of the desensitization model of the present application is shown, and the method is executed by an electronic device and a terminal device. Figure 1 As shown, the method comprises the following steps:

[0033] Step 110: The electronic device determines the number of images that are not desensitized and sends them to the terminal device, and sends the number of images of the non-desensitized images to the terminal device. The non-desensitized images are images that the desensitization model identifies as not requiring desensitization processing.

[0034] Among them, the electronic device can first determine the number of non-desensitized images that need to be sent to the terminal device. Optionally, each terminal device can correspond to a user. For each terminal device, the electronic device can determine a different number of images. And for different terminal devices, the non-desensitized images sent by the electronic device are images of the user corresponding to the terminal device. Optionally, the electronic device can be connected to the terminal device for communication. Optionally, the electronic device can be a background management platform such as a cloud or a service platform.

[0035] Once the number of images to be sent is determined, the electronic device can select the number of non-desensitized images from the stored non-desensitized images based on the number of images. The electronic device can send the selected number of non-desensitized images to the terminal device. Optionally, the non-desensitized images can be images that are identified as not requiring desensitization after being processed by a desensitization model. Optionally, the non-desensitized images transmitted back to the terminal device by the electronic device can be images that are encrypted by the terminal device and sent to the electronic device. After receiving the encrypted image uploaded by the terminal device, the electronic device can store the image in a dedicated database.

[0036] Optionally, the non-desensitized image is an image that is determined not to require desensitization processing after being processed using the desensitization model.

[0037] Optionally, in the storage area of ​​the electronic device, multiple processing tags may be set for each image. The processing tag may be used to indicate the processing result of the image in the electronic device. Optionally, after the electronic device receives the image uploaded by the terminal device, the processing tag may be "pending". Optionally, when the image is processed by the desensitization model in the electronic device, the processing tag after the desensitization is completed may be "desensitized", and the tag of the image identified as not requiring desensitization may be "non-desensitized". Optionally, when the image is sent to the terminal device, the processing tag of the image may be "manually processed". Optionally, the electronic device may select an image with a label of "non-desensitized" from the images in the storage area, and send it to the terminal device as a non-desensitized image. Optionally, there may be images of different users in the electronic device. In the process of sending non-desensitized images to the terminal device, the electronic device only sends the image of the user corresponding to the terminal device to the terminal device.

[0038] In one example, the number of images sent to the terminal device can be determined based on multiple factors. Optionally, the electronic device can determine the number of images based on the terminal information of the terminal device. Optionally, the electronic device can send a terminal information acquisition request to the terminal device when it is determined that the number of images needs to be calculated. Optionally, the electronic device can periodically calculate the number of images. Optionally, the period can be a preset value. Or, the period can be determined by user settings. For example, the period can be once a day, once a week, once a month, etc. Optionally, the electronic device can obtain the terminal information of the terminal device sent by the terminal device.

[0039] The terminal information includes at least one of the following: a historical operation log of the terminal device, performance information of the terminal device, and a real-time operation log of the terminal device. The operation log may be a record of the operation of the user's terminal device during the manual recognition of non-desensitized images. Specifically, the operation log is used to record detailed information such as user operation behavior, image processing status, and data transmission in real time. Optionally, the operation log may include the user's operation behavior, image processing status, and data transmission status, so as to facilitate tracing and troubleshooting when problems arise.

[0040] Among them, the historical operation log can be a record of completed operations. For example, if a manual identification operation is currently being performed, the historical operation log can be an operation log before the current operation is started. Alternatively, the historical operation log can also be an operation log of a preset time length before the current moment. For example, the operation log can be an operation log up to 24:00 on the previous day. Among them, the real-time operation log is a log of the manual identification operation currently being performed. Among them, the performance information of the terminal device may include the performance parameters of the terminal device at the current moment and within a certain time length before the current moment. Alternatively, the performance information of the terminal device may also include the performance parameters of the terminal device during the previous manual identification operation. Optionally, the performance information can be obtained by regularly collecting system performance indicators through a built-in monitoring tool. For example, the performance information may include response time, resource utilization, etc.

[0041] Step 120: The terminal device receives the non-desensitized images in the number of images sent by the electronic device, and determines the labels of the non-desensitized images.

[0042] Among them, the terminal device can receive the non-desensitized images sent in advance by the electronic device according to the number of images when the terminal device is turned on and the recognition of the non-desensitized images is not performed. And, the terminal device can also receive the non-desensitized images sent in real time by the electronic device according to the number of images when the user performs the processing of the non-desensitized images. The terminal device can determine the label of each non-desensitized image during the recognition of the non-desensitized images. In one example, the terminal device can obtain the label of the non-desensitized image in response to the user's calibration operation. The terminal device can first obtain the user's judgment result on the non-desensitized image. The judgment result can be used to indicate whether the non-desensitized image contains sensitive information. If the non-desensitized image contains sensitive information, the terminal device can further obtain the label of the non-desensitized image. The label is usually used to indicate the reason why the sensitive information in the non-desensitized image is not recognized.

[0043] Optionally, the terminal device can display the non-massified image in the form of a task card pop-up window. The user can open or close the task card through the floating window icon. The user can view the image and complete the recognition by swiping the card left and right. For example, if it does not contain sensitive information, you can swipe the card to the right. If it contains sensitive information, you can swipe the card to the left. Optionally, the task card pop-up window can be triggered in the application of the terminal device.

[0044] Step 130: The terminal device sends the label of the non-desensitized image to the electronic device, wherein the label of the non-desensitized image is used to optimize the desensitization model.

[0045] Among them, the terminal device can send these labels to the electronic device after completing the determination of the labels of the non-desensitized images. During the sending process, the terminal device may take additional encryption measures to ensure the security of the label data during the transmission process. At the same time, the terminal device may record information such as the sending time and sending status for subsequent data tracking and auditing. Optionally, after obtaining the label of the non-desensitized image, the electronic device can use the label of the non-desensitized image for desensitization model optimization processing. Optionally, the desensitization model is a machine learning model for processing sensitive information. The use of the desensitization model can retain the value and availability of the data as much as possible while protecting the privacy of the user. Specifically, the label of the non-desensitized image can be used to train, verify and test the desensitization model.

[0046] In one example, the terminal device may also send the image features of the non-desensitized image to the electronic device while sending the label of the non-desensitized image to the electronic device. The image features can be obtained by processing the non-desensitized image using a preset feature extraction model in the terminal device. Optionally, the terminal device may perform feature extraction on the non-desensitized image after obtaining the label of the non-desensitized image. The image features of the non-desensitized image will also be used to optimize the desensitization model.

[0047] Step 140: The electronic device receives the label of the non-desensitized image sent by the terminal device, and optimizes the desensitization model according to the label of the non-desensitized image.

[0048] Among them, after receiving the labels of the non-desensitized images sent by the terminal device, the electronic device can use these labels to optimize the desensitization model. The optimization process may include adjusting the parameters of the desensitization model, updating the weights of the model, and introducing new features. The electronic device can improve the performance of the desensitization model by iterating and optimizing the desensitization model, making it more accurate and efficient when processing sensitive information. Specifically, the optimization of the desensitization model can be based on performance monitoring data, and targeted optimization strategies can be formulated and implemented, such as adjusting algorithm parameters, increasing server resources, etc., to ensure that the system can efficiently process large amounts of data and user requests.

[0049] In one example, the process of optimizing the desensitization model by the electronic device according to the label of the non-desensitized image may specifically include the following steps:

[0050] Step 141: Classify the non-desensitized images according to the labels of the non-desensitized images, wherein the labels of the non-desensitized images are marked by the user.

[0051] After receiving the labels of the non-desensitized images from the terminal device, the electronic device can classify the non-desensitized images according to the labels of the non-desensitized images. The labels of the non-desensitized images can be labels calibrated by the user.

[0052] Step 142: If the number of non-desensitized images in the classification is greater than the first threshold, the desensitization model is optimized using the labels of the non-desensitized images in the classification.

[0053] Among them, the electronic device can count the number of non-desensitized images in the category corresponding to each label. If the number of non-desensitized images in a certain category exceeds a preset first threshold, the electronic device will use the labels of the non-desensitized images under the category to optimize the desensitization model. Optionally, in the optimization process, in order to ensure the balance of model training, the electronic device usually selects a certain number of non-desensitized images for training. When the number of non-desensitized images including multiple categories is greater than the first threshold, the electronic device can select a certain number of non-desensitized images for each category for model optimization.

[0054] In one example, since a large number of numbers are usually required to achieve optimization training during the desensitization model training process. Therefore, after determining a classified un-desensitized image, the electronic device can also generate a second amount of simulation data based on the un-desensitized image in the classification. Optionally, the electronic device can use a model such as a generative adversarial network to generate the simulation data. The electronic device can use un-desensitized images and simulation data to achieve additional training and optimization of the desensitization algorithm model.

[0055] In the present application, the electronic device can first determine the number of non-desensitized images that need to be sent to the terminal device. Based on the number of images, the electronic device can pre-send the non-desensitized images of the number of images to the terminal device when the terminal device is turned on and the recognition of the non-desensitized images is not performed. The electronic device can also receive the non-desensitized images sent by the electronic device in real time according to the number of images during the user's processing of the non-desensitized images. The terminal device can determine the label of each non-desensitized image during the recognition of the non-desensitized images, and send these labels to the electronic device. The electronic device can use these labels to optimize the desensitization model. In the present application, by using the labels of these non-desensitized images, the desensitization model is optimized, the missed detection rate of the desensitization model is reduced, and the privacy protection effect is improved. In addition, the present application ensures the real-time and smooth effect of the image when the user views and processes the image by caching the non-desensitized images of the number of images to the terminal device, thereby improving the user experience. In addition, this application makes full use of the participation of car owners to improve the accuracy of the anonymization algorithm, protects the user's right to know and right to choose, and enhances the user's trust in the privacy protection measures of automobile companies. The entire process meets the end-to-end data collection requirements, and while no images are retained on electronic devices, it effectively optimizes the company's anonymization algorithm through feedback in the form of task card pop-up windows.

[0056] It should be noted that the above process needs to be executed after the user confirms to join the algorithm optimization plan. For example, after determining that the user has joined the algorithm optimization plan, the electronic device can periodically calculate the number of images and send the number of non-desensitized images to the terminal device for caching when the terminal device is turned on.

[0057] On the basis of the above embodiment, the electronic device can also receive the initial image uploaded by the terminal device in real time. In this case, the terminal device can be a vehicle computer. The initial image can be an image captured by the vehicle's sensor. After receiving the initial image, the electronic device can use a desensitization model to process the initial image so that the initial image can be guaranteed not to leak sensitive information during subsequent processing. In addition, the electronic device can also mark the initial image that the desensitization model identifies and determines does not need to be desensitized, and then send it to the terminal device for identification and processing by the user. This specific process may include:

[0058] Step 210: Receive the initial image sent by the terminal device.

[0059] The initial image may be an encrypted image.

[0060] Step 220: Use the desensitization model to process the initial image to obtain a desensitized image and a non-desensitized image. The non-desensitized image is an image that the desensitization model identifies as not requiring desensitization. The desensitized image is an image that the desensitization model identifies as requiring desensitization and after the desensitization operation is completed.

[0061] Among them, after the electronic device receives the initial image from the terminal device, it immediately starts the built-in desensitization model to process the image. The desensitization model can intelligently analyze the content of the initial image and identify whether the initial image contains sensitive information. If it contains sensitive information, the desensitization model can directly desensitize the initial image. The electronic device can use the desensitization model to desensitize the initial image to obtain a desensitized image. If it does not contain sensitive information, the desensitization model will not desensitize the initial image. The initial image that is not desensitized is a non-desensitized image.

[0062] Step 230: Input the non-demensified image into the recognition model to obtain a sensitive probability value. The sensitive probability value represents the probability of containing sensitive information.

[0063] Among them, after the user marks more non-desensitized images that are incorrectly identified by the sensitive model in the terminal device, the electronic device can more effectively use these marked data to optimize its sensitive identification model. To this end, the electronic device first carefully screens the non-desensitized images, removes those that obviously do not contain sensitive information, and sends those non-desensitized images that are judged to have a high possibility of sensitive information according to the current model to the user terminal. This strategy not only reduces the redundant workload of user review, but also improves the accuracy and efficiency of users in identifying sensitive information, thereby promoting the continuous optimization of the sensitive identification model.

[0064] Therefore, the electronic device can input the un-desensitized image into the recognition model. The recognition model is used to identify the sensitivity probability value of the presence of sensitive information in the un-desensitized image. If the sensitivity probability value is higher, it is more likely that sensitive information exists in the un-desensitized image. Otherwise, if the sensitivity probability value is lower, it is less likely that sensitive information exists in the un-desensitized image.

[0065] Step 240: If the sensitivity probability value is greater than the second threshold, the non-desensitized image is determined as the non-desensitized image to be sent to the terminal device.

[0066] Among them, a second threshold may be set in the electronic device. The setting of the second threshold may be determined based on experience. The electronic device may compare the sensitive probability value and the second threshold. If the sensitive probability value is greater than the second threshold, the electronic device may determine the non-desensitized image as the non-desensitized image to be sent to the terminal device. Otherwise, if the sensitive probability value is less than or equal to the second threshold, the electronic device may determine that the non-desensitized image does not need to be sent to the terminal device.

[0067] The present application uses the recognition model to identify the sensitivity probability value of the non-desensitized image, thereby achieving further recognition of the non-desensitized image, reducing the probability of the non-desensitized image that obviously does not contain sensitive information being transmitted to the terminal device, and improving the effectiveness of user recognition.

[0068] On the basis of the above embodiment, when the terminal device is a vehicle computer, the terminal device also needs to perform the following steps:

[0069] Step 310: Obtain the driving status of the vehicle.

[0070] The terminal device can obtain the driving status of the vehicle in real time. The driving status can be called through an interface to obtain the driving status of the vehicle at the current moment.

[0071] Step 320: If it is determined based on the driving status that the vehicle is stationary, a prompt message is sent to prompt the user to view and process the non-desensitized pictures.

[0072] Among them, a safety judgment logic may be set in the terminal device. The terminal device can use the safety judgment logic to judge the driving status and determine whether the vehicle is in a stationary state. If the vehicle is in a stationary state, the terminal device can send a prompt message. The prompt message prompts the user to view and process the non-desensitized pictures. Specifically, the terminal device can ask the user whether it is necessary to push the non-desensitized image when it is determined that the vehicle is stationary and not in a driving state. Optionally, the non-desensitized image can be acquired by pushing a task card in a task card pop-up window.

[0073] The determination that the vehicle is stationary can improve the safety of the user, ensure the user's driving safety, and avoid affecting driving safety.

[0074] On the basis of the above embodiment, the terminal device can also send a collection request to the user. If the user agrees to the collection, it is considered that the user agrees to join the machine learning algorithm optimization plan. Optionally, the content of the collection request may include: Dear car owner, we are optimizing the anonymization algorithm of the vehicle and sincerely invite you to participate. You will receive photos that have not been desensitized. You only need to confirm whether this photo needs to be desensitized. Your participation will help improve the privacy and security of the vehicle, and you can also get points rewards. This application encourages user participation through a points system, which helps to improve the implementation effect of the scheme. Optionally, two buttons can be set corresponding to the collection request. The two buttons can correspond to agreeing to participate and not participating temporarily. When the user chooses to agree to participate, the user joins the algorithm optimization plan. Otherwise, the user does not join the algorithm optimization plan. The terminal device will not receive the un-desensitized image pushed by the electronic device. Optionally, the task card pop-up window can use an intuitive layout and clear text, a concise and clear interface to convey the key information of the image desensitization judgment task to the user. And the "agree to participate" and "not participate" clear selection buttons can facilitate users to make decisions quickly. For example, when a user uses a car manufacturer app and opens it, the system will first retrieve the driving status in the car manufacturer app interface. The driving status may include information such as whether the car is in driving process and whether the vehicle is unlocked. The car manufacturer app will only display the non-desensitized image sent to the cache by the electronic device when it determines that the user is not currently in driving process. The system will only push the task card pop-up window when the user's safety is ensured.

[0075] In the terminal device, the un-desensitized image appears in the form of a card in the conference pop-up window. Among them, a concise and clear pop-up window will appear in a user-friendly form to introduce the user to the task of image desensitization judgment and explain the importance of this task for optimizing algorithms and improving privacy and security. Users face two options: agree to participate or not participate for the time being. If the user chooses to agree to participate, the system will use end-to-end encryption technology to ensure the secure transmission of data, and randomly push photos that have not been desensitized to the user in the form of cards for judgment.

[0076] On the basis of the above embodiments, when the user finds that the un-desensitized image contains sensitive information, the terminal device will immediately display labels, such as fisheye distortion, portrait inversion, etc., to help the user determine the specific reasons why the algorithm fails to recognize the face. Subsequently, the terminal device will conduct an in-depth analysis of the image and extract multiple features such as color, texture, and shape at the pixel level. This process covers a comprehensive analysis of the pixel information of the image, including the extraction of features such as color distribution, brightness, and contrast, such as obtaining color features by calculating the histogram of each color channel. At the same time, the terminal device will also use a variety of algorithms to capture the texture features in the image, such as grayscale co-occurrence matrix and local binary patterns (Local Binary Patterns, LBP), and detect the contours and edges of objects to calculate shape features, such as perimeter, area, and aspect ratio. And in terms of texture feature extraction, the terminal device flexibly uses a variety of algorithms such as grayscale co-occurrence matrix and local binary patterns to fully capture subtle texture changes in the image. Through advanced image edge detection algorithms, we can accurately identify the outline and edges of objects, and further accurately calculate shape feature parameters such as perimeter, area, aspect ratio, etc., providing strong support for subsequent image analysis and processing. Electronic devices can also use deep learning technology to identify key elements and patterns in non-desensitized images, and pay special attention to the geometric features of the face, such as the position and proportion of the eyes, nose, and mouth. In addition, the impact of factors such as the shooting environment and lighting conditions of the non-desensitized images on facial feature recognition will be deeply analyzed. The terminal device can also use mathematical models to quantitatively evaluate the brightness and contrast of the image to ensure an accurate grasp of the overall brightness and darkness of the image. In addition, the terminal device will also consider the impact of factors such as shooting angle and clarity on the features, and make necessary adjustments and corrections accordingly. Finally, the feature values ​​extracted by the terminal device will be sorted, encoded and converted into a data format suitable for transmission (such as JSON), and then uploaded to the electronic device through a secure network connection.

[0077] On the basis of the above embodiments, in order to ensure the stable operation of the system, the electronic device may also regularly perform performance monitoring and optimization to ensure that a large amount of image data and user requests can be processed in a timely manner.

[0078] Figure 2 A flowchart of another embodiment of the optimization method of the desensitization model of the present application is shown. When the terminal information includes the historical operation log of the terminal device and the performance information of the terminal device, the electronic device can determine the number of images according to the terminal information of the terminal device. The method is executed by the electronic device. Figure 2 As shown, the method comprises the following steps:

[0079] Step 410: Determine the processing speed of the non-desensitized image according to the historical operation log, and determine the cache time required to cache the non-desensitized image according to the performance information. The historical operation log represents the operation log of the terminal device processing the non-desensitized image before the current operation.

[0080] Among them, the electronic device can obtain the historical operation log. The historical operation log includes the operation log of the user processing the non-desensitized image before the current operation. The electronic device will intelligently analyze and determine the processing speed of the non-desensitized image based on the data processing recorded in the historical operation log. At the same time, the electronic device will also accurately calculate the reasonable cache time required to cache the non-desensitized image to the terminal device based on the performance information.

[0081] In one example, the calculation of the cache time may include the following steps:

[0082] Step 411: Determine the first time required to send the un-desensitized image to the terminal device according to the size of the un-desensitized image and the network transmission speed in the performance information. The un-desensitized image is an encrypted image.

[0083] The electronic device may determine the first time according to the ratio of the size of the un-desensitized image to the network transmission speed. The size of the un-desensitized image is a fixed size. The network transmission speed may be determined according to the average network transmission speed in the performance information. The un-desensitized image is an encrypted image. The electronic device is usually provided with a storage space for the un-desensitized image.

[0084] Step 412: Determine the second time required to process the non-desensitized image based on the data processing efficiency in the performance information.

[0085] Among them, the processing of the non-desensitized image may include decoding, decryption, etc. The electronic device may determine the time required for the terminal device to decode and decrypt the non-desensitized image based on the performance information of the terminal device. Optionally, the electronic device may use the average of the processing times of multiple desensitized images in the performance information as the second time.

[0086] Step 413: Determine a third time required to write the processed non-desensitized image into the cache according to the cache efficiency in the performance information.

[0087] The terminal device also needs to write the processed non-desensitized image into the cache. The electronic device can also obtain the time required for the terminal device to write the processed non-desensitized image into the cache through the performance information.

[0088] Step 414: Determine the cache time according to the first time, the second time and the third time.

[0089] The electronic device may determine the sum of the above three times as the cache time. The cache time may be calculated by respectively calculating the mean of each stage to obtain a more accurate calculation of the cache time.

[0090] Step 420: Determine the number of non-desensitized images that need to be cached in the terminal device based on the processing speed and cache time.

[0091] Among them, in order to ensure the smoothness when the user starts to process the non-desensitized images and to ensure that there is no situation of waiting for the non-desensitized images to be loaded during the processing, the electronic device needs to further calculate the optimal number of non-desensitized images that need to be cached on the terminal device based on the cache time and processing speed to ensure the efficiency of data processing.

[0092] In one example, in the specific calculation process of the number of images, the electronic device may also consider the repetition rate. The repetition rate is the ratio of repeated display of non-desensitized images. The specific process includes:

[0093] Step 421: Determine the ratio of the cache time to the processing speed to determine the number of processable images.

[0094] The electronic device may first determine the ratio of the buffering time to the processing speed as the number of processable images. The number of processable images may be used to indicate the number of images that the user can process within the buffering time.

[0095] Step 422: Determine the number of non-desensitized images that need to be cached in the terminal device based on a preset repetition rate and the number of processable images.

[0096] Among them, the preset repetition rate is used to indicate the proportion of non-desensitized images that need to appear repeatedly. The electronic device can determine the first parameter according to the product of 1 and the repetition rate. The electronic device can divide the processable images by the value of the first parameter as the number of non-desensitized images that need to be cached to the terminal device. For example, when the repetition rate is 30%, the number of images is the value of the processable images divided by 1.3.

[0097] In the present application, the electronic device can determine the user's processing speed for the non-desensitized image based on the data processing recorded in the historical operation log. At the same time, the electronic device can also accurately calculate the reasonable cache time required to cache the non-desensitized image to the terminal device based on the performance information. The electronic device can also further calculate the optimal number of non-desensitized images that need to be cached on the terminal device based on the cache time and processing speed. The present application can determine the image data corresponding to the terminal device through the cache time and processing speed, and can more accurately implement the cache of each terminal device, reduce unnecessary cache of non-desensitized images, and improve the utilization rate of the cache of the terminal device.

[0098] On the basis of the above embodiment, when the terminal information includes the real-time operation log of the terminal device, the electronic device can determine the number of images sent to the terminal device in real time according to the terminal information of the terminal device. The method is executed by the electronic device. Figure 2 As shown, the method comprises the following steps:

[0099] Step 510: Determine the number of unprocessed non-desensitized images remaining in the buffer of the terminal device according to the real-time operation log. The real-time operation log represents the operation log of the terminal device processing the non-desensitized images in real time during this operation.

[0100] Among them, the electronic device can obtain the real-time operation log of the terminal device, which records in detail the specific situation of the terminal device processing the non-desensitized images in real time during this operation. Subsequently, the electronic device will count the number of non-desensitized images processed during the current operation based on the log information. Since the terminal device will only send the label of the non-desensitized image to the electronic device when the non-desensitized image contains sensitive information. Therefore, the electronic device cannot directly determine the number of non-desensitized images that need to be supplemented based on the label sent by the terminal device. Therefore, the electronic device can periodically obtain the real-time operation log.

[0101] The electronic device can count the number of desensitized images that have been sent to the terminal device. The electronic device can determine the number of unprocessed non-desensitized images remaining in the cache of the terminal device based on the difference between the number of desensitized images that have been sent to the terminal device and the number of processed non-desensitized images.

[0102] Step 520: Determine the number of images based on the initial number of un-desensitized images in the cache of the terminal device and the number of remaining unprocessed un-desensitized images.

[0103] Among them, the electronic device can also obtain the initial number of non-desensitized images in the cache. The initial number can be the number of non-desensitized images initially sent to the terminal device obtained by calculation by the electronic device. Specifically, the electronic device can calculate the initial number according to step 420. The electronic device determines the number of images based on the difference between the initial number and the number of remaining unprocessed non-desensitized images. The number of images is the number of images that need to be added to the terminal device.

[0104] This application ensures the smoothness of image display when users are processing non-desensitized images, thereby improving user experience, through real-time management of the number of non-desensitized images remaining on the terminal device.

[0105] Based on the above embodiments, the electronic device can correct the feature deviation caused by the shooting angle by using the principle of image geometric transformation based on the shooting angle consideration. In addition, the electronic device can also appropriately enhance or eliminate the blurred image based on the clarity evaluation and adjustment through the clarity evaluation algorithm.

[0106] Moreover, during the data transmission process, the electronic device and the terminal device can also organize and encode the extracted feature values ​​into JSON format to facilitate network transmission and server parsing. In addition, the electronic device and the terminal device can also use a high-intensity encryption algorithm to encrypt the feature value data in transmission to ensure data confidentiality and integrity. The electronic device and the terminal device can ensure the security of data transmission by establishing a stable and encrypted network connection channel.

[0107] In the electronic device, a special database can also be built to store the received eigenvalue data, and a reasonable data structure and index can be designed to improve the query and analysis efficiency. In the database, different types of eigenvalue data can also be classified and marked to facilitate subsequent algorithm optimization and data analysis.

[0108] On the basis of the above embodiments, in order to further improve the enthusiasm of users to participate in the model optimization plan, the terminal device can promote user participation in algorithm optimization by linking the relevant points system of the enterprise, thereby attracting more users to participate. Optionally, the electronic device can customize the upper limit of the number of pictures pushed to a user in one day. Optionally, the terminal device can push related tasks to the user when the user uses the car company app. In addition, the terminal device can also push the anonymized image recognition task to the user by calling the driving status in the car company app interface, and when it is determined that the user is not driving and does not affect the user's safety.

[0109] Figure 3 The schematic diagram of the structure of an embodiment of the optimization device of the desensitization model of the present application is shown. Figure 3 As shown, in an electronic device, the optimization device 600 of the desensitization model includes: a sending module 610 and a receiving module 620.

[0110] The sending module 610 is used to determine the number of images that are not desensitized and send them to the terminal device, and send the number of images of the non-desensitized images to the terminal device. The non-desensitized images are images that the desensitization model identifies as not requiring desensitization processing.

[0111] The receiving module 620 is used to receive the label of the non-desensitized image sent by the terminal device, and optimize the desensitization model according to the label of the non-desensitized image.

[0112] Optionally, the sending module 610 is configured to:

[0113] The number of images is determined based on the terminal information of the terminal device.

[0114] The terminal information includes at least one of the following: a historical operation log of the terminal device, performance information of the terminal device, and a real-time operation log of the terminal device.

[0115] Optionally, the historical operation log represents the operation log of the terminal device processing the non-desensitized image before the current operation. Optionally, the sending module 610 is used to:

[0116] Based on the historical operation logs, determine the processing speed of the non-desensitized images. And based on the performance information, determine the cache time required to cache the non-desensitized images.

[0117] Based on the processing speed and cache time, determine the number of non-desensitized images that need to be cached to the terminal device.

[0118] Optionally, the sending module 610 is configured to:

[0119] The first time required to send the un-desensitized image to the terminal device is determined according to the size of the un-desensitized image and the network transmission speed in the performance information. The un-desensitized image is an encrypted image.

[0120] Based on the data processing efficiency in the performance information, determine the second time required to process the non-desensitized image.

[0121] A third time required to write the processed non-desensitized image into the cache is determined according to the cache efficiency in the performance information.

[0122] A cache time is determined according to the first time, the second time and the third time.

[0123] Optionally, the sending module 610 is configured to:

[0124] Determine the ratio of cache time to processing speed as the number of images that can be processed.

[0125] According to the preset repetition rate and the number of processable images, the number of non-desensitized images that need to be cached to the terminal device is determined.

[0126] Optionally, the real-time operation log represents the operation log of the terminal device processing the non-desensitized image in real time during this operation. The sending module 610 is used to:

[0127] Based on the real-time operation log, determine the number of unprocessed, non-desensitized images remaining in the cache of the terminal device.

[0128] The number of images is determined based on the initial number of un-desensitized images in the cache of the terminal device and the number of remaining unprocessed un-desensitized images.

[0129] Optionally, the receiving module 620 is configured to:

[0130] Classify the non-desensitized images according to their labels;

[0131] If the number of non-desensitized images in the classification is greater than the first threshold, the desensitization model is optimized using the labels of the non-desensitized images in the classification;

[0132] Among them, the labels of non-desensitized images are marked by users.

[0133] Optionally, the receiving module 620 is configured to:

[0134] generating a second amount of simulation data based on the non-desensitized images in the classification;

[0135] The desensitization algorithm model is optimized using non-desensitized images and simulation data.

[0136] Optionally, the receiving module 620 is further configured to:

[0137] Receive the initial image sent by the terminal device.

[0138] The initial image is processed using the desensitization model to obtain a desensitized image and a non-desensitized image.

[0139] Among them, the non-desensitized image is the image that the desensitization model identifies as not needing desensitization. The desensitized image is the image that the desensitization model identifies as needing desensitization and after the desensitization operation is completed.

[0140] Optionally, the receiving module 620 is configured to:

[0141] The unmasked image is input into the recognition model to obtain the sensitive probability value, which represents the probability of containing sensitive information.

[0142] If the sensitivity probability value is greater than the second threshold, the non-desensitized image is determined as the non-desensitized image to be sent to the terminal device.

[0143] The optimization device 600 of the desensitization model provided in the embodiment of the present application can execute the above method embodiment. Its specific implementation principle and technical effects can be found in the above method embodiment, and this embodiment will not be repeated here.

[0144] Figure 4 The schematic diagram of the structure of an embodiment of the optimization device of the desensitization model of the present application is shown. Figure 4 As shown, in the terminal device, the optimization device 700 of the desensitization model includes: a receiving module 710 and a sending module 720.

[0145] The receiving module 710 is used to receive the non-desensitized images under the number of images sent by the electronic device, and determine the labels of the non-desensitized images. The non-desensitized images are images that the desensitization model identifies as not requiring desensitization processing.

[0146] The sending module 720 is used to send the label of the non-desensitized image to the electronic device. The label of the non-desensitized image is used to optimize the desensitization model.

[0147] Optionally, the receiving module 710 is configured to:

[0148] In response to the user's calibration operation, the label of the non-desensitized image is obtained.

[0149] Optionally, the sending module 720 is configured to:

[0150] The terminal information of the terminal device is sent to the electronic device. The terminal information includes at least one of the following: a historical operation log of the terminal device, performance information of the terminal device, and a real-time operation log of the terminal device. The terminal information is used to determine the number of images.

[0151] Optionally, the sending module 720 is configured to:

[0152] Use the preset feature extraction model to process the non-desensitized images to obtain the image features of the non-desensitized images.

[0153] The image features of the un-desensitized image are sent to the electronic device, wherein the image features of the un-desensitized image are used to optimize the desensitization model.

[0154] Optionally, when the terminal device is a vehicle computer, the desensitization model optimization device 700 further includes:

[0155] The display module is used to obtain the driving status of the vehicle. If the vehicle is determined to be stationary according to the driving status, a prompt message is sent to prompt the user to view and process the non-desensitized image.

[0156] The optimization device 700 of the desensitization model provided in the embodiment of the present application can execute the above method embodiment. Its specific implementation principle and technical effects can be found in the above method embodiment, and this embodiment will not be repeated here.

[0157] Figure 5 The schematic diagram of the structure of the embodiment of the electronic device of the present application is shown, and the specific embodiment of the present application does not limit the specific implementation of the electronic device. Figure 5 As shown, the electronic device 800 may include: a processor (processor) 802 , a communication interface (Communications Interface) 804 , a memory (memory) 806 , and a communication bus 808 .

[0158] The processor 802, the communication interface 804, and the memory 806 communicate with each other via a communication bus 808. The communication interface 804 is used to communicate with other devices such as a client or other server network elements. The processor 802 is used to execute a program 810, which can specifically execute the above-mentioned steps in the above-mentioned method embodiment.

[0159] Specifically, the program 810 may include program code including computer executable instructions.

[0160] The processor 802 may be a central processing unit (CPU), or an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application. The one or more processors included in the electronic device may be processors of the same type, such as one or more CPUs; or may be processors of different types, such as one or more CPUs and one or more ASICs.

[0161] The memory 806 is used to store the program 810. The memory 806 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0162] Program 810 can be specifically called by processor 802 to enable the electronic device to execute the above method. Its implementation method and technical effect are similar, and this embodiment will not be repeated here.

[0163] Figure 6 The schematic diagram of the structure of the embodiment of the terminal device of the present application is shown, and the specific embodiment of the present application does not limit the specific implementation of the terminal device. Figure 6 As shown, the terminal device 900 may include: a processor (processor) 902 , a communication interface (Communications Interface) 904 , a memory (memory) 906 , and a communication bus 908 .

[0164] The processor 902, the communication interface 904, and the memory 906 communicate with each other via a communication bus 908. The communication interface 904 is used to communicate with other devices such as a client or other server network elements. The processor 902 is used to execute a program 910, which can specifically execute the above-mentioned steps in the above-mentioned method embodiment.

[0165] Specifically, the program 910 may include program code, which includes computer-executable instructions.

[0166] The processor 902 may be a central processing unit (CPU), or an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application. The one or more processors included in the terminal device may be processors of the same type, such as one or more CPUs; or may be processors of different types, such as one or more CPUs and one or more ASICs.

[0167] The memory 906 is used to store the program 910. The memory 906 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0168] Program 910 can be specifically called by processor 902 to enable the terminal device to execute the above method. Its implementation method and technical effect are similar, and this embodiment will not be repeated here.

[0169] The present application also provides a computer-readable storage medium, in which a computer program is stored. When the computer program is executed by a processor, it is used to implement the methods provided in the various embodiments described above.

[0170] Among them, the computer-readable storage medium can be a computer storage medium or a communication medium. The communication medium includes any medium that is convenient for transmitting a computer program from one place to another. The computer storage medium can be any available medium that can be accessed by a general-purpose or special-purpose computer. For example, a computer-readable storage medium is coupled to a processor so that the processor can read information from the computer-readable storage medium and write information to the computer-readable storage medium. Of course, the computer-readable storage medium can also be a component of the processor. The processor and the computer-readable storage medium can be located in an application-specific integrated circuit (ASIC). In addition, the ASIC can be located in a user device. Of course, the processor and the computer-readable storage medium can also exist in a communication device as discrete components.

[0171] Specifically, the computer-readable storage medium may be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. The storage medium may be any available medium that can be accessed by a general or special-purpose computer.

[0172] The present application also provides a computer program product, which includes a computer program stored in a computer-readable storage medium. At least one processor of a device can read the computer program from the computer-readable storage medium, and at least one processor executes the computer program so that the device implements the methods provided in the various embodiments described above.

[0173] The algorithms or displays provided herein are not inherently related to any particular computer, virtual system or other device. In addition, the embodiments of the present application are not directed to any particular programming language.

[0174] In the specification provided herein, a large number of specific details are described. However, it is understood that the embodiments of the present application can be practiced without these specific details. Similarly, in order to simplify the application and help understand one or more of the various inventive aspects, in the above description of the exemplary embodiments of the present application, the various features of the embodiments of the present application are sometimes grouped together into a single embodiment, figure, or description thereof. Wherein, the claims that follow the specific embodiment are thus clearly incorporated into the specific embodiment, wherein each claim itself is used as a separate embodiment of the application.

[0175] Those skilled in the art will appreciate that the modules in the devices in the embodiments may be adaptively changed and arranged in one or more devices different from the embodiments. The modules or units or components in the embodiments may be combined into one module or unit or component, and further may be divided into a plurality of submodules or subunits or subcomponents. Except that at least some of such features and / or processes or units are mutually exclusive.

[0176] It should be noted that the above embodiments illustrate the present application rather than limit the present application, and that those skilled in the art may design alternative embodiments without departing from the scope of the appended claims. In the claims, any reference symbol between brackets shall not be constructed as a limitation on the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "one" or "an" preceding an element does not exclude the presence of multiple such elements. The present application may be implemented by means of hardware including several different elements and by means of a suitably programmed computer. In a unit claim that lists several devices, several of these devices may be embodied by the same hardware item. The use of the words first, second, and third, etc. does not indicate any order. These words may be interpreted as names. The steps in the above embodiments, unless otherwise specified, should not be understood as limitations on the order of execution.

Claims

1. A method for optimizing a desensitization model, characterized in that: The method comprises: Determine the number of images that are not desensitized and send them to the terminal device, and send the number of images that are not desensitized to the terminal device; the images that are not desensitized are images that are identified by the desensitization model as not requiring desensitization processing; Receive the label of the non-desensitized image sent by the terminal device, and optimize the desensitization model according to the label of the non-desensitized image.

2. The method according to claim 1, characterized in that Determine the number of images that are not desensitized and sent to the terminal device, including: Determining the number of images according to terminal information of the terminal device; The terminal information includes at least one of the following: a historical operation log of the terminal device, performance information of the terminal device, and a real-time operation log of the terminal device.

3. The method according to claim 2, characterized in that The historical operation log represents the operation log of the terminal device processing the non-desensitized image before the current operation; Determining the number of images according to the terminal information of the terminal device includes: Determine, based on the historical operation log, a processing speed for processing the non-desensitized image; and determine, based on the performance information, a cache time required to cache the non-desensitized image; According to the processing speed and the cache time, the number of the non-desensitized images that need to be cached in the terminal device is determined.

4. The method according to claim 3, characterized in that Determining, according to the performance information, a cache time required for caching the non-desensitized image, including: Determine, according to the size of the un-desensitized image and the network transmission speed in the performance information, a first time required to send the un-desensitized image to the terminal device; the un-desensitized image is an encrypted image; Determining a second time required to process the non-desensitized image according to the data processing efficiency in the performance information; Determining, according to the cache efficiency in the performance information, a third time required to write the processed non-desensitized image into the cache; The cache time is determined according to the first time, the second time and the third time.

5. The method according to claim 3, characterized in that: Determining the number of the non-desensitized images that need to be cached in the terminal device according to the processing speed and the cache time includes: Determine the ratio of the cache time to the processing speed as the number of processable images; According to the preset repetition rate and the number of processable images, the number of the non-desensitized images that need to be cached in the terminal device is determined.

6. The method according to claim 2, characterized in that The real-time operation log represents the operation log of the terminal device processing the non-desensitized image in real time during this operation; Determining the number of images according to the terminal information of the terminal device includes: Determine, according to the real-time operation log, the number of the unprocessed non-desensitized images remaining in the cache of the terminal device; The number of images is determined based on the initial number of the non-desensitized images in the cache of the terminal device and the number of the remaining unprocessed non-desensitized images.

7. The method according to any one of claims 1 to 6, characterized in that The desensitization model is optimized according to the label of the non-desensitized image, including: Classifying the non-desensitized image according to the label of the non-desensitized image; If the number of the non-desensitized images in the classification is greater than a first threshold, the desensitization model is optimized using the labels of the non-desensitized images in the classification; The label of the non-desensitized image is marked by the user.

8. A method for optimizing a desensitization model, characterized in that: The method comprises: Receiving a non-desensitized image under the number of images sent by the electronic device; and determining a label of the non-desensitized image; the non-desensitized image is an image that is identified by the desensitization model as not requiring desensitization processing; The label of the non-desensitized image is sent to the electronic device; wherein the label of the non-desensitized image is used to optimize the desensitization model.

9. The method according to claim 8, characterized in that The method further comprises: Sending terminal information of a terminal device to an electronic device; wherein the terminal information includes at least one of the following: a historical operation log of the terminal device, performance information of the terminal device, and a real-time operation log of the terminal device; the terminal information is used to determine the number of images.

10. The method according to claim 8 or 9, characterized in that: The method further comprises: Using a preset feature extraction model to process the non-desensitized image to obtain image features of the non-desensitized image; The image features of the non-desensitized image are sent to an electronic device; wherein the image features of the non-desensitized image are used to optimize the desensitization model.

Images