An update method, device, server and storage medium for an AI service platform

Abstract

The embodiment of the present invention provides an update method, device, server and storage medium of an AI service platform, wherein the server in the AI ​​service platform includes a first type of AI model and a second type of AI model, wherein the first type of AI model includes multiple first AI models, each of which does not have a function to be updated and all users of the AI ​​service platform can use its function, and the second type of AI model includes multiple second AI models, each of which has a function to be updated and only some users can use its function, and the method includes: obtaining business data generated by a user when using the function to be updated of at least one second AI model; and updating at least one second AI model based on the business data. The business data generated by the user is real data of a real application scenario, and the second AI model trained by the server based on the business data can output accurate processing results for the application scenario, which can improve the accuracy of the processing results.

Description

Technical Field

[0001] The present invention relates to the field of AI service technology, and in particular to an update method, device, server and storage medium for an AI service platform. Background Art

[0002] The AI ​​(Artificial Intelligence) service platform is a network platform that provides AI model training and releases AI services for enterprise users. At present, the AI ​​service platform releases AI services in the following way: first obtain the original data, generate sample data for AI model training based on the original data, and then train the AI ​​model based on the sample data. After the training is completed, the corresponding service of the AI ​​model is released online for enterprise users to choose to use.

[0003] Generally speaking, the original data of an AI model is the business data related to the functions of the AI ​​model obtained by the AI ​​service platform. However, the business data is not necessarily the real data of the scenario in which the service corresponding to the AI ​​model is actually applied.

[0004] For example, for an AI model that analyzes personnel mobility through video surveillance, its actual application scenario is a subway station that is yet to be opened. The raw data obtained by the AI ​​service platform may be surveillance video data from other subway stations that have already been opened. Since the locations and angles of surveillance cameras at different subway stations are different, and the flow of personnel is also different, if surveillance video data from other subway stations that have already been opened is used as raw data, the trained AI model will not be able to accurately analyze personnel mobility when applied to the subway station that is yet to be opened, and the accuracy of the processing results will not be very high. Summary of the invention

[0005] The purpose of the embodiments of the present invention is to provide an update method, device, server and storage medium for an AI service platform, so as to improve the accuracy of the processing results of the AI ​​service. The specific technical solution is as follows:

[0006] In a first aspect, an embodiment of the present invention provides an update method for an AI service platform, which is applied to a server in the AI ​​service platform, wherein the server includes a first type of AI model and a second type of AI model, wherein the first type of AI model includes multiple first AI models, each of which is configured to have no function to be updated and all users of the AI ​​service platform can use its function, and the second type of AI model includes multiple second AI models, each of which is configured to have a function to be updated and only some users can use its function, and the method includes:

[0007] Acquire business data generated by a user when using a function to be updated of at least one second AI model among the multiple second AI models;

[0008] The at least one second AI model is updated based on the business data.

[0009] Optionally, the step of obtaining business data generated by the user when using the function to be updated of the at least one second AI model includes:

[0010] Pushing a link of the to-be-updated function of the at least one second AI model to a user who has logged in to the AI ​​service platform, so that the user can access the to-be-updated function through the link;

[0011] The business data uploaded by the user during accessing the function to be updated is obtained.

[0012] Optionally, before the step of pushing a link of the to-be-updated function of the at least one second AI model to a user who has logged in to the AI ​​service platform, the method further includes:

[0013] receiving a user login request, wherein the user login request includes first user information;

[0014] Determining whether the first user information matches pre-recorded user information of a target user;

[0015] If there is a match, it is determined that the user corresponding to the user login request is the target user, wherein the target user is the user to whom the link of the to-be-updated function of the at least one second AI model needs to be pushed.

[0016] Optionally, before the step of pushing a link of the to-be-updated function of the at least one second AI model to a user who has logged in to the AI ​​service platform, the method further includes:

[0017] receiving a user login request, wherein the user login request includes second user information;

[0018] Based on the second user information, determining whether the user corresponding to the user login request is an administrative user;

[0019] If yes, displaying an AI service list including the to-be-updated function of the at least one second AI model;

[0020] The step of pushing a link of the to-be-updated function of at least one second AI model to a user who has logged in to the AI ​​service platform includes:

[0021] When a push instruction is obtained from the management user based on the AI ​​service list, a link to the updated function of the at least one second AI model is pushed to the user, wherein the push instruction is issued when the management user determines that the user who has logged in to the AI ​​service platform is the target user, and the target user is the user to whom the link to the updated function of the at least one second AI model needs to be pushed.

[0022] Optionally, the step of pushing a link of the to-be-updated function of the at least one second AI model to a trial user who has logged in to the AI ​​service platform includes:

[0023] Send push information to a terminal device of a user who has logged in to the AI ​​service platform, wherein the push information includes a link address of the function to be updated of the at least one second AI model.

[0024] Optionally, after the at least one second AI model is updated, the method further includes:

[0025] Removing the at least one second AI model after the update from the second type of AI models, and adding it to the first type of AI models;

[0026] When the user logs in to the AI ​​service platform, the functional interface of the at least one second AI model that has been updated is displayed.

[0027] Optionally, the step of updating the at least one second AI model based on the business data includes:

[0028] Generate multiple training samples based on the business data;

[0029] Preprocessing each of the training samples according to the function to be updated to obtain a processed training sample;

[0030] The at least one second AI model is trained based on the processed training samples until the output result of the at least one second AI model meets a preset condition, and the training is stopped to obtain an updated second AI model.

[0031] In a second aspect, an embodiment of the present invention provides an updating device for an AI service platform, which is applied to a server in the AI ​​service platform, wherein the server includes a first type of AI model and a second type of AI model, wherein the first type of AI model includes multiple first AI models, each of which is configured to have no function to be updated and all users of the AI ​​service platform can use its function, and the second type of AI model includes multiple second AI models, each of which is configured to have a function to be updated and only some users can use its function, and the device includes:

[0032] A business data acquisition module, used to acquire business data generated when a user uses a function to be updated of at least one second AI model among the multiple second AI models;

[0033] The AI ​​model updating module is used to update the at least one second AI model based on the business data.

[0034] In a third aspect, an embodiment of the present invention provides a server, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other via the communication bus;

[0035] Memory, used to store computer programs;

[0036] The processor is used to implement any method step described in the first aspect when executing the program stored in the memory.

[0037] In a fourth aspect, an embodiment of the present invention provides 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 steps described in any one of the first aspects are implemented.

[0038] Beneficial effects of the embodiments of the present invention:

[0039] In the solution provided by the embodiment of the present invention, the server in the AI ​​service platform may include a first type of AI model and a second type of AI model, the first type of AI model includes multiple first AI models, each of which is configured to have no function to be updated and all users of the AI ​​service platform can use its function, the second type of AI model includes multiple second AI models, each of which is configured to have a function to be updated and only some users can use its function, the server can obtain the business data generated when the user uses the function to be updated of at least one second AI model among the multiple second AI models, and then update the at least one second AI model based on the business data. In this way, since the user can use the function to be updated of at least one second AI model, that is, the user in the scenario to which the at least one second AI model is to be applied, the business data generated when the user uses the function to be updated of the at least one second AI model is the real data of the scenario to be applied, and then the server can update the at least one second AI model based on the business data to obtain the updated second AI model, and the updated second AI model can output accurate processing results for the application scenario, and the user can also obtain accurate processing results when using the AI ​​service function provided by the updated second AI model, which can greatly improve the accuracy of the processing results of the AI ​​service. Of course, it is not necessary to achieve all of the advantages described above at the same time to implement any product or method of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other embodiments can be obtained based on these drawings without paying creative work.

[0041] Figure 1 A flowchart of an update method of an AI service platform provided by an embodiment of the present invention;

[0042] Figure 2 Based on Figure 1 A flow chart of a method for determining a target user in the illustrated embodiment;

[0043] Figure 3(a) is based on Figure 1 A flow chart of a display method of an AI service list in the illustrated embodiment;

[0044] FIG3( b ) is a schematic diagram of a user of the AI ​​service platform based on the embodiment shown in FIG3( a );

[0045] Figure 4 Based on Figure 1A flow chart of the functional interface display method of the illustrated embodiment;

[0046] Figure 5 Based on Figure 4 A schematic diagram of an AI service page of the illustrated embodiment;

[0047] Figure 6 for Figure 4 A specific flow chart of step S401 in the illustrated embodiment;

[0048] Figure 7 for Figure 1 A specific flow chart of step S102 in the illustrated embodiment;

[0049] Figure 8 A schematic diagram of the structure of an updating device for an AI service platform provided in an embodiment of the present invention;

[0050] Fig. 9 for Figure 8 A specific structural diagram of the business data acquisition module 810 in the illustrated embodiment;

[0051] Fig.10 A schematic diagram of the structure of a server provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0053] In order to improve the accuracy of the processing results of AI services, embodiments of the present invention provide an update method, device, server, computer-readable storage medium, and computer program product for an AI service platform. First, an update method for an AI service platform provided by an embodiment of the present invention is introduced.

[0054] An updating method for an AI service platform provided in an embodiment of the present invention can be applied to a server in an AI service platform, which is a network platform that can provide users with AI model training and publish AI services. It can provide users with a variety of AI services, for example, face recognition services, vehicle violation recognition services, passenger flow statistics services, etc., which are not specifically limited here.

[0055] like Figure 1As shown, an update method of an AI service platform is applied to a server in the AI ​​service platform, the server includes a first type of AI model and a second type of AI model, the first type of AI model includes a plurality of first AI models, each of which is configured to have no function to be updated and all users of the AI ​​service platform can use its function, the second type of AI model includes a plurality of second AI models, each of which is configured to have a function to be updated and only some users can use its function, the method includes:

[0056] S101, obtaining business data generated when a user uses a function to be updated of at least one second AI model among the multiple second AI models;

[0057] S102: Update the at least one second AI model based on the business data.

[0058] It can be seen that in the solution provided by the embodiment of the present invention, the server in the AI ​​service platform includes a first type of AI model and a second type of AI model, the first type of AI model includes multiple first AI models, each of which is configured to have no function to be updated and all users of the AI ​​service platform can use its function, and the second type of AI model includes multiple second AI models, each of which is configured to have a function to be updated and only some users can use its function, and the server can obtain the business data generated when the user uses the function to be updated of at least one second AI model among the multiple second AI models, and then update the at least one second AI model based on the business data. In this way, since the user can be a user who uses the function to be updated of at least one second AI model, that is, a user in the scenario to which the at least one second AI model is to be applied, the business data generated when the user uses the function to be updated of the at least one second AI model is the real data of the scenario to be applied, and then the server can update the at least one second AI model based on the business data to obtain an updated second AI model, and the updated second AI model can output accurate processing results for the application scenario, and the user can also obtain accurate processing results when using the AI ​​service function provided by the updated second AI model, which can greatly improve the accuracy of the processing results of the AI ​​service.

[0059] The AI ​​service platform can provide users with a variety of AI functions, each AI function corresponds to an AI model, so the above-mentioned server can include a first type of AI model and a second type of AI model. The first type of AI model can include multiple first AI models, each of which is configured to have no function to be updated and all users of the AI ​​service platform can use its function. The first AI model can be called a mature AI model, which can provide users with mature AI services. The first AI model is configured not to have a function to be updated, that is, the first AI model is a trained AI model that can provide users with accurate processing results, rather than an AI model in the training stage.

[0060] The above-mentioned second category of AI models may include multiple second AI models, each of which is configured to have a function to be updated and only some users can use its function. That is to say, the second AI model has not yet reached the most ideal state, and the processing results it provides may not be accurate enough, and further training and updating are required. It can be called an immature AI model. Since the second AI model needs further training and updating, it can be set so that only some users can use its function. In this way, the business data generated by the user when using the function to be updated of the second AI model can be obtained, while avoiding providing inaccurate processing results to all users and affecting the user experience. Among them, users who can use the function to be updated of the second AI model can be called trial users.

[0061] The second AI model may be an AI model that has been trained with sample data generated by business data related to the function to be updated of the second AI model but has not reached the most ideal state. The first AI model and the second AI model may be deep learning models such as convolutional neural network models, deep trust network models, and stacked autoencoder network models, or other machine learning models other than deep learning models, which are not specifically limited here.

[0062] In order to improve the accuracy of the processing results output by the second AI model and make it a mature AI model that does not have any functions to be updated and whose functions can be used by all users of the AI ​​service platform, it is necessary to obtain business data of the actual scenarios to which the second AI model is to be applied, and then train the second AI model based on the business data.

[0063] Therefore, in the above step S101, the server can obtain the business data generated by the user when using the function to be updated of at least one second AI model among the multiple second AI models. In the process of using the function to be updated, the user can upload business data, which is the data that needs to be processed by the second AI model, and it must be the business data of the actual scenario to be applied by the function to be updated uploaded by the user. For example, if the scenario to be applied to the function to be updated is the detection of people at the entrance of subway station A, then the business data uploaded by the user can be images or videos taken by the monitoring equipment installed at the entrance of subway station A.

[0064] Furthermore, in the above step S102, the server can update at least one second AI model based on the business data. After obtaining the above business data, the server can generate training samples based on the business data, and use the training samples to train the at least one second AI model until the second AI model converges to obtain an updated second AI model. The updated second AI model is a mature AI model that can provide accurate processing results for users.

[0065] As an implementation manner of an embodiment of the present invention, the step of obtaining the business data generated when the user uses the to-be-updated function of at least one second AI model among the multiple second AI models may include:

[0066] Push a link of the to-be-updated function of the at least one second AI model to a user who has logged in to the AI ​​service platform, so that the user can access the to-be-updated function through the link; and obtain the business data uploaded by the user during the process of accessing the to-be-updated function.

[0067] If the user who has logged in to the AI ​​service platform is a trial user, the server can push the link of the to-be-updated function of the at least one second AI model to the trial user, so that the trial user can access the to-be-updated function through the link. The link to the AI ​​service can be a URL (Uniform Resource Locator) corresponding to the AI ​​service, or other forms that can link to the to-be-updated function.

[0068] In order to facilitate the determination of whether the user logging into the AI ​​service platform is a trial user, the trial user can be determined in advance and the user identification of the trial user can be recorded, where the user identification can be the user name used to log in to the AI ​​service platform, user ID (Identity document, identity identification number), etc., as long as the user identity can be uniquely identified, and no specific limitation is made here.

[0069] The above trial users are users in the scenarios where the function to be updated is to be applied, that is, users in the actual scenarios where the function to be updated is to be applied after it matures. For example, if the scenario where the function to be updated is to be applied is the detection of people at the entrance of subway station A, then the trial users can be the staff of subway station A. For another example, if the scenario where the function to be updated is to be applied is the identification of vehicles at intersection B, then the trial users can be the management personnel of the transportation department to which intersection B belongs.

[0070] After the trial user obtains the link of the function to be updated, he can access the function to be updated through the link. Then, when accessing the function to be updated, the server can also obtain the business data uploaded by the user, and then update the second AI model.

[0071] It can be seen that in this embodiment, the server can push the link of the to-be-updated function of the second AI model to the user who has logged in to the AI ​​service platform, so that the user can access the to-be-updated function through the link, and then obtain the business data uploaded by the user during the process of accessing the to-be-updated function. In this way, it is convenient for the user to access the to-be-updated function, thereby obtaining the business data uploaded by the user during the process of accessing the to-be-updated function, and completing the update of the second AI model.

[0072] As an implementation of an embodiment of the present invention, before the step of pushing a link of the to-be-updated function of at least one second AI model to a user who has logged in to the AI ​​service platform, Figure 2 As shown, the above method may also include:

[0073] S201, receiving a user login request;

[0074] When a user logs in to the AI ​​service platform through a terminal device, a user login request is issued, and the server can receive the user login request. The user login request may include user information, and the server may use the user information as the first user information. The first user information may include information that can identify the user, such as a user name and a user ID.

[0075] S202, determining whether the first user information matches the pre-recorded user information of the target user, if so, executing step S203; if not, determining that the user corresponding to the user login request is a common user;

[0076] S203: Determine that the user corresponding to the user login request is a target user.

[0077] In order to facilitate the determination of whether the user who logs in to the AI ​​service platform is the target user, the server can pre-record the user information of each trial user. Then, when receiving a user login request, the first user information included in the user login request is matched with the pre-recorded user information of the trial user. The target user is the user to whom the link of the to-be-updated function of the at least one second AI model is to be pushed, that is, the trial user.

[0078] If the first user information matches the user information of the pre-recorded trial user, it means that the user who issued the user login request is the target user, then the server can execute step S203, that is, determine that the user corresponding to the user login request is the target user, and then continue to execute the above-mentioned step of pushing the link of the to-be-updated function of at least one second AI model to the user who has logged in to the AI ​​service platform.

[0079] If the first user information does not match the user information of the pre-recorded target user, it means that the user who issued the user login request is not the pre-determined target user, then the server can determine that the user who issued the user login request is an ordinary user, and then display an ordinary AI service page. The ordinary AI service page may include links to the first AI models, but not to the second AI model, so as to prevent ordinary users from using the second AI model to obtain processing results with insufficient accuracy, thereby affecting user experience and corporate image.

[0080] In one implementation, in order to facilitate the determination of the link of the second AI model that needs to be pushed to the user who has logged in to the AI ​​service platform, when the server pre-records the user information of the target user, it can also record the second AI model corresponding to the target user. The user information of the target user and the corresponding second AI model can be recorded in a table, for example, as shown in the following table:

[0081]

[0082]

[0083] In this way, if the server determines that the first user information matches the user information U3, the server can determine that the user corresponding to the user login request is the target user, and then push the link of the second AI model 3 to the target user.

[0084] It can be seen that in this embodiment, before pushing the link of at least one second AI model to the user who has logged in to the AI ​​service platform, the server can receive a user login request, determine whether the first user information included in the user login request matches the user information of the pre-recorded target user, and if they match, determine that the user corresponding to the user login request is the target user. In this way, when the user logs in to the AI ​​service platform, the server can automatically identify the target user without manual identification, saving manual identification costs and improving the identification efficiency of the target user.

[0085] As an implementation of an embodiment of the present invention, before the step of pushing a link of the to-be-updated function of at least one second AI model to a user who has logged in to the AI ​​service platform, as shown in FIG3(a), the method may further include:

[0086] S301, receiving a user login request;

[0087] When a user logs in to the AI ​​service platform through a terminal device, a user login request is issued, and the server can receive the user login request. The user login request may include user information, and the server may use the user information as the second user information. The second user information may include information that can identify the user, such as a user name and a user ID.

[0088] S302, based on the second user information, determine whether the user corresponding to the user login request is a management user, if yes, execute step S303; if not, determine that the user corresponding to the user login request is a common user.

[0089] In order to facilitate determining whether the user logging into the AI ​​service platform is an administrative user of the AI ​​service platform, the server may pre-record the user information of each administrative user, and then upon receiving a user login request, match the second user information included in the user login request with the pre-recorded user information of the administrative user.

[0090] If the second user information matches the user information of the pre-recorded management user, it means that the user who issued the user login request is the management user, and the server can execute step S303. If the second user information does not match the user information of the pre-recorded management user, it means that the user who issued the user login request is not the pre-determined management user, and the server can determine that the user who issued the user login request is an ordinary user, and then display the ordinary AI service page.

[0091] In one embodiment, as shown in FIG3(b), users of the AI ​​service platform can be divided into three types, namely trial users, ordinary users, and management users. The AI ​​service platform can be a server or a cluster of multiple servers. In order to distinguish the three types of users, the server can pre-record the correspondence between each user information and the user type, so that after receiving a user login request, it can determine whether the user who issued the user login request is a common user or a management user based on the user information included in the user login request.

[0092] S303: Display an AI service list including the to-be-updated functions of the at least one second AI model.

[0093] After the server determines that the user currently logged in to the AI ​​service platform is an administrative user, it can display the AI ​​service list including the above functions to be updated, so that the administrative user can view each function to be updated. Of course, the AI ​​service list can also include various mature AI services. In one embodiment, the function name corresponding to each AI model can be displayed in the AI ​​service list.

[0094] Accordingly, the step of pushing a link of the to-be-updated function of the at least one second AI model to a user who has logged in to the AI ​​service platform may include:

[0095] In case of obtaining a push instruction issued by the management user based on the AI ​​service list, a link to the to-be-updated function of the at least one second AI model is pushed to the user.

[0096] After the management user logs in to the AI ​​service platform, when the management user sees that other users have logged in to the AI ​​service platform, the management user can determine whether the other user is the target user based on the user information of the pre-recorded target user, and then issue a push instruction based on the above AI service list when the other user is determined to be the target user. In other words, the push instruction is issued when the management user determines that the user who has logged in to the AI ​​service platform is the target user.

[0097] The server can then obtain the push instruction, and then push the link of the function to be updated to the target user. In one embodiment, the function to be updated in the above AI service list may correspond to a trigger interface, such as a trigger button, etc. The management user can issue a push instruction through the trigger interface.

[0098] For example, based on the above table, if the management user determines that the user information of the user who logs in to the AI ​​service platform matches the user information U1, then the management user can determine that the user is the target user, and can determine that the corresponding second AI model is the second AI model 1 according to the above table, and then the management user can issue a push instruction through the trigger interface corresponding to the function to be updated of the second AI model 1. The server can then push the link of the function to be updated of the second AI model 1 to the target user.

[0099] It can be seen that in this embodiment, before the above step of pushing the link of the to-be-updated function of at least one second AI model to the user who has logged in to the AI ​​service platform, the server can receive a user login request, and based on the second user information included in the user login request, determine whether the user corresponding to the user login request is a management user, and if so, display an AI service list including the to-be-updated function of the at least one second AI model, and then, when the push instruction issued by the management user based on the AI ​​service list is obtained, push the link of the to-be-updated function of the at least one second AI model to the target user (for example, a trial user). In this way, compared with the current AI service platform, there is no need to configure a different AI service page for each user, and there is no need to make major improvements to the AI ​​service platform, and the link of the to-be-updated service of the second AI model can be pushed to the trial user, and the configuration cost of the AI ​​service platform is relatively low.

[0100] As an implementation of an embodiment of the present invention, the step of pushing a link of the to-be-updated function of at least one second AI model to a user who has logged in to the AI ​​service platform may include:

[0101] Send push information to the terminal device of the user who has logged in to the AI ​​service platform.

[0102] The server may push the link address of the to-be-updated function of the at least one second AI model to the user by sending a push message to the terminal device of the user who has logged in to the AI ​​service platform, wherein the push message includes the link address of the to-be-updated function. For example, the push message may be sent to the terminal device of the user who has logged in to the AI ​​service platform by a short message, email, instant messaging message, or the like.

[0103] In this way, the user's terminal device can display the received push information. In this case, the specific display form of the link address of the function to be updated can be the specific URL corresponding to the function to be updated. The user can enter the function interface of the function to be updated of the second AI model by clicking on the URL, thereby achieving access to the function to be updated.

[0104] In another embodiment, the specific display form of the link address of the function to be updated can be a button, text, icon, etc. in the AI ​​service page of the AI ​​service platform. In this way, the user can trigger the button, text, icon, etc. displayed in the AI ​​service page by clicking, etc., and then enter the function interface of the function to be updated of the second AI model, thereby realizing access to the function to be updated.

[0105] It can be seen that in this embodiment, the server can send push information to the terminal device of the user who has logged in to the AI ​​service platform, and can push the link address of the function to be updated to the user, thereby obtaining the business data of the actual application scenario generated by the user's access to the function to be updated.

[0106] As an implementation method of the embodiment of the present invention, Figure 4 As shown, after the second AI model is updated, the method may further include:

[0107] S401, removing the at least one second AI model after updating from the second type of AI models, and adding it to the first type of AI models;

[0108] After training the at least one second AI model to obtain the updated second AI model, in order to enable all ordinary users to access the AI ​​service functions of the updated second AI model, since the first type of AI model is a type of model whose functions can be used by all users, the server can remove the updated second AI model from the second type of AI model and add it to the first type of AI model. As an implementation method, the various models included in the first type of AI model and the second type of AI model can be recorded in the form of a service list, so the server can add the function link of the updated second AI model to the service list corresponding to the first type of AI model. Among them, the service list may include links to the functions of various mature AI models.

[0109] S402: When the user logs in to the AI ​​service platform, a functional interface of the at least one second AI model that has been updated is displayed.

[0110] When a user logs in to the AI ​​service platform, since the at least one updated second AI model has been trained and can provide accurate processing results to users, and it has been removed from the second category of AI models and added to the first category of AI models, the server can display the functional interface of the at least one updated second AI model, so that all users can use the functions provided by the at least one updated second AI model.

[0111] In one embodiment, the server can send the above-mentioned service list to the user's terminal device, and then, after the user's terminal device receives the service list, it can display the service list in the AI ​​service page, and the user can view the functions of each mature AI model included in the service list, and can also view the link to the function of the second AI model after the update is completed that is added to the service list, and then can use the function of the second AI model after the update is completed.

[0112] Since the first type of AI model can provide satisfactory services for users and can output processing results with high accuracy, ordinary users, trial users and management users can all view the functions provided by the first type of AI model. Therefore, in this embodiment, users who log in to the AI ​​service platform can include ordinary users, trial users and management users, without distinguishing between specific types of users.

[0113] It can be seen that in this embodiment, the server can remove the at least one second AI model after the update from the second category of AI models and add it to the first category of AI models, and then display the function interface of the at least one second AI model after the update when the user logs in to the AI ​​service platform. In this way, the user can view the function interface of the at least one second AI model after the update when the user logs in to the AI ​​service platform, so as to access the function of the at least one second AI model after the update.

[0114] In order to prevent ordinary users from guessing the specific URL of the function to be updated, which may lead to an unsatisfactory user experience when accessing the function to be updated, as an implementation method of an embodiment of the present invention, the generation rules of the specific URL corresponding to the first AI model and the specific URL corresponding to the function to be updated may be different.

[0115] In this way, the specific URL corresponding to the first AI model and the specific URL corresponding to the function to be updated will show different patterns, and ordinary users will not be able to infer the specific URL corresponding to the function to be updated based on the specific URL corresponding to the first AI model they see. In order to further ensure that ordinary users cannot infer the specific URL corresponding to the function to be updated based on the specific URL corresponding to the first AI model they see, the specific URL corresponding to the function to be updated can be composed of a string with higher complexity.

[0116] For example, the AI ​​service page can be Figure 5As shown, it includes a service list 510, a service search input box 520 and a search button 530 for the first AI model. The function names of the first AI model are displayed in the service list 510, which are face recognition, mask recognition, and vehicle recognition, respectively. The corresponding specific URLs are https: / / AI0001.com, https: / / AI0002.com, and https: / / AI0004.com, respectively. Ordinary users can access the corresponding function interface of the first AI model by clicking on the function name of the first AI model, entering the function name in the service search input box 520, and clicking the search button 530. When the function interface of the first AI model is displayed, the specific URL corresponding to the first AI model will be displayed in the URL address bar 540 above the interface, and the user can also see the URL.

[0117] If the generation rule of the specific URL corresponding to the function to be updated of the second AI model is the same as the generation rule of the specific URL corresponding to the first AI model, the specific URL corresponding to the function to be updated may be https: / / AI0003.com. Then, after ordinary users view https: / / AI0001.com, https: / / AI0002.com and https: / / AI0004.com, they may think that there should be https: / / AI0003.com, and then enter "https: / / AI0003.com" in the URL address bar 540 to access the function interface of the function to be updated. Therefore, the generation rule of the specific URL corresponding to the function to be updated of the second AI model needs to be different from the generation rule of the specific URL corresponding to the first AI model. For example, the specific URL corresponding to the first AI model is generated according to the rule of name + serial number, and the specific URL corresponding to the function to be updated of the second AI model is generated using the rule of randomly generated special character long garbled code, thereby presenting different rules to avoid erroneous access caused by the specific URL corresponding to the function to be updated of the second AI model being predicted by ordinary users.

[0118] Correspondingly, such as Figure 6 As shown, the step of removing the at least one second AI model after the update from the second type of AI model and adding it to the first type of AI model may include:

[0119] S601, generating a specific URL corresponding to the at least one second AI model after the update is completed based on the generation rule corresponding to the first type of AI model;

[0120] Since the generation rules of the specific URL corresponding to the first type of AI model are different from the specific URL corresponding to the above-mentioned function to be updated, after the second AI model is trained to become a mature AI model, the server can generate the specific URL corresponding to the second AI model based on the generation rules corresponding to the first type of AI model.

[0121] For example, the first AI model includes the first AI model corresponding to the specific URLs https: / / AI0001.com, https: / / AI0002.com and https: / / AI0004.com. The server can generate the specific URL https: / / AI0003.com corresponding to at least one second AI model obtained after the training is completed according to the generation rules corresponding to the first AI model.

[0122] S602: Add the link of the at least one second AI model after the update is completed to the service list of the first type of AI model, and record the specific URL corresponding to the second AI model after the update is completed.

[0123] After generating the specific URL corresponding to at least one second AI model after the update is completed, the server can add the link of at least one second AI model after the update (which can be a function name) to the service list of the first type of AI model. After the user logs in to the AI ​​service platform, he can view the service list of the first type of AI model, and also view the link corresponding to the second AI model after the update is completed, and then access the function interface of at least one second AI model after the update through the specific URL corresponding to the link of at least one second AI model after the update is completed.

[0124] It can be seen that in this embodiment, the specific URL corresponding to the first type of AI model may be different from the generation rule of the specific URL corresponding to the function to be updated. The server can generate a specific URL corresponding to at least one second AI model after the update is completed based on the generation rule corresponding to the first type of AI model, add the link of the at least one second AI model after the update is completed to the service list of the first type of AI model, and record the specific URL corresponding to the at least one second AI model after the update is completed, so as to remove the at least one second AI model after the update is completed from the second type of AI model and add it to the first type of AI model.

[0125] As an implementation method of the embodiment of the present invention, Figure 7 As shown, the step of updating the at least one second AI model based on the business data may include:

[0126] S701, generating a plurality of training samples based on the business data;

[0127] After the server obtains the business data generated by the user when using the to-be-updated function of the at least one second AI model, it can generate multiple training samples based on the business data. For example, if the business data generated by the user when using the to-be-updated function of the second AI model is a surveillance video, the server can use each frame of the surveillance video as a training sample.

[0128] S702, pre-processing each of the training samples according to the function to be updated to obtain a processed training sample;

[0129] If the samples corresponding to the function to be updated need to be labeled, the above preprocessing can be a labeling process, for example, the function to be updated is target recognition in an image. In one embodiment, the training samples can be automatically labeled to obtain labels, thereby obtaining processed training samples. In another embodiment, after the server obtains the training samples, the training samples can be saved, and then the training samples can be manually labeled when necessary to obtain labels, thereby obtaining processed training samples, which are all reasonable.

[0130] If the samples corresponding to the function to be updated do not need to be labeled, the above preprocessing is the processing method corresponding to the function to be updated. For example, if the function to be updated is to improve the resolution of the image, then the preprocessing can be to improve the resolution of the training samples.

[0131] S703: Train the at least one second AI model based on the processed training samples until an output result of the at least one second AI model meets a preset condition, stop the training, and obtain an updated second AI model.

[0132] After obtaining the processed training samples, the server can use the processed training samples to train the at least one second AI model until the at least one second AI model converges. At this time, if the output result of the second AI model meets the preset conditions, the training can be stopped to obtain the updated AI model. The specific training method can adopt any training method in the field of model training, for example, a gradient descent algorithm, a stochastic gradient descent algorithm, etc. can be adopted.

[0133] As an implementation method, the server can input the processed training sample into the second AI model, and the second AI model will process the processed training sample based on the current model parameters to obtain an output result. The server can adjust the model parameters according to the output result of the second AI model. In one implementation method, if the training sample has a corresponding label, the server can compare the output result of the second AI model with the label of the training sample to obtain the difference between the two, and then adjust the model parameters according to the difference.

[0134] Until the processing result output by the at least one second AI model meets the preset condition, wherein the preset condition can be determined according to the function to be implemented by the at least one second AI model, and can be the accuracy reaching the preset accuracy, or the number of iterations of the training sample reaching the preset number, etc., which are not specifically limited here. At this time, the second AI model can already accurately process the input data, thereby outputting a processing result with an accuracy that meets the requirements, so it can be determined that the second AI model at this time has converged, and the updated second AI model is obtained.

[0135] Among them, the above-mentioned preset accuracy and preset number of times can be determined according to the actual requirements of the accuracy of the processing results of the at least one second AI model. For example, the preset accuracy can be 90%, 95%, 98%, etc., and the preset number of times can be 8000, 10000, 15000, etc., which are not specifically limited here.

[0136] After obtaining the updated second AI model, the server can remove the updated second AI model from the second category of AI models and add it to the first category of AI models. In this way, all users of the AI ​​service platform can use the functions of the updated second AI model to achieve the purpose of providing users with satisfactory AI services.

[0137] It can be seen that in this embodiment, the server can generate multiple training samples based on business data, pre-process each training sample according to the function to be updated, obtain the processed training sample, and then train at least one second AI model based on the processed training sample until the output result of the at least one second AI model meets the preset conditions, stop training, and obtain the updated second AI model. In this way, the server can train and obtain the updated second AI model that can provide users with satisfactory AI services.

[0138] Corresponding to the above-mentioned AI service platform update method, an embodiment of the present invention further provides an AI service platform update device. The following introduces an AI service platform update device provided by an embodiment of the present invention.

[0139] like Figure 8As shown, an updating device for an AI service platform is applied to a server in the AI ​​service platform, the server includes a first type of AI model and a second type of AI model, the first type of AI model includes a plurality of first AI models, each of which is configured to have no function to be updated and all users of the AI ​​service platform can use its function, the second type of AI model includes a plurality of second AI models, each of which is configured to have a function to be updated and only some users can use its function, the device includes:

[0140] A business data acquisition module 810, used to acquire business data generated when a user uses a function to be updated of at least one second AI model among the plurality of second AI models;

[0141] The AI ​​model updating module 820 is used to update the at least one second AI model based on the business data.

[0142] It can be seen that in the solution provided by the embodiment of the present invention, the server in the AI ​​service platform may include a first type of AI model and a second type of AI model, the first type of AI model includes multiple first AI models, each of which is configured to have no function to be updated and all users of the AI ​​service platform can use its function, the second type of AI model includes multiple second AI models, each of which is configured to have a function to be updated and only some users can use its function, the server can obtain the business data generated when the user uses the function to be updated of at least one second AI model among the multiple second AI models, and then update the at least one second AI model based on the business data. In this way, since the user can use the function to be updated of at least one second AI model, that is, the user in the scenario to which the at least one second AI model is to be applied, the business data generated when the user uses the function to be updated of the at least one second AI model is the real data of the scenario to be applied, and then the server can update the at least one second AI model based on the business data to obtain the updated second AI model, and the updated second AI model can output accurate processing results for the application scenario, and the user can also obtain accurate processing results when using the AI ​​service function provided by the updated second AI model, which can greatly improve the accuracy of the processing results of the AI ​​service.

[0143] As an implementation method of the embodiment of the present invention, Fig. 9 As shown, the business data acquisition module 810 may include:

[0144] A link pushing unit 811 is used to push a link of the to-be-updated function of the at least one second AI model to a user who has logged in to the AI ​​service platform, so that the user can access the to-be-updated function through the link;

[0145] The data acquisition unit 812 is used to acquire the business data uploaded by the user when accessing the function to be updated.

[0146] As an implementation of the embodiment of the present invention, the above device may further include:

[0147] A first request receiving module, configured to receive a user login request before pushing a link of the to-be-updated function of the at least one second AI model to a user who has logged in to the AI ​​service platform;

[0148] The user login request includes first user information.

[0149] An information matching module, used to determine whether the first user information matches the pre-recorded user information of the target user;

[0150] The first user determination module is configured to determine that the user corresponding to the user login request is the target user if the first user information matches the pre-recorded user information of the target user.

[0151] Among them, the target user is the user to whom the link of the to-be-updated function of the at least one second AI model needs to be pushed.

[0152] As an implementation of the embodiment of the present invention, the above device may further include:

[0153] A second request receiving module, configured to receive a user login request before pushing a link of the to-be-updated function of the at least one second AI model to a user who has logged in to the AI ​​service platform;

[0154] The user login request includes second user information.

[0155] A second user determination module, configured to determine whether the user corresponding to the user login request is an administrative user based on the second user information;

[0156] a list display module, configured to display an AI service list including a function to be updated of the at least one second AI model if the user corresponding to the user login request is a management user;

[0157] The link push unit 811 may include:

[0158] The first push subunit is used to push the link of the to-be-updated function of the at least one second AI model to the user when a push instruction issued by the management user based on the AI ​​service list is obtained.

[0159] Among them, the push instruction is issued when the management user determines that the user who has logged in to the AI ​​service platform is the target user, and the target user is the user to whom the link of the to-be-updated function of the at least one second AI model needs to be pushed.

[0160] As an implementation of the embodiment of the present invention, the link push unit 811 may include:

[0161] The second push sub-unit is used to send push information to the terminal device of the user who has logged in to the AI ​​service platform, wherein the push information includes the link address of the function to be updated of the at least one second AI model.

[0162] As an implementation of the embodiment of the present invention, the above device may further include:

[0163] a classification change module, configured to remove the at least one second AI model after the update from the second category of AI models and add it to the first category of AI models after the update of the at least one second AI model is completed;

[0164] An interface display module is used to display the functional interface of the at least one second AI model that has been updated when a user logs in to the AI ​​service platform.

[0165] As an implementation of an embodiment of the present invention, the AI ​​model updating module 820 may include:

[0166] A training sample generating unit, configured to generate a plurality of training samples based on the service data;

[0167] A preprocessing unit, used for preprocessing each of the training samples according to the function to be updated to obtain a processed training sample;

[0168] A model training unit is used to train the at least one second AI model based on the processed training samples until the output result of the at least one second AI model meets a preset condition, stop training, and obtain the updated second AI model.

[0169] The embodiment of the present invention further provides a server, which can be a service in the above-mentioned AI service platform, such as Fig.10 As shown, the server may include a processor 1001, a communication interface 1002, a memory 1003 and a communication bus 1004, wherein the processor 1001, the communication interface 1002, and the memory 1003 communicate with each other via the communication bus 1004.

[0170] Memory 1003, used for storing computer programs;

[0171] The processor 1001 is used to implement the update method steps of the AI ​​service platform described in any of the above embodiments when executing the program stored in the memory 1003.

[0172] It can be seen that in the solution provided by the embodiment of the present invention, the server may include a first type of AI model and a second type of AI model, the first type of AI model includes multiple first AI models, each of which is configured to have no function to be updated and all users of the AI ​​service platform can use its function, the second type of AI model includes multiple second AI models, each of which is configured to have a function to be updated and only some users can use its function, the server can obtain the business data generated when the user uses the function to be updated of at least one second AI model among the multiple second AI models, and then update the at least one second AI model based on the business data. In this way, since the user can use the function to be updated of at least one second AI model, that is, the user in the scenario to which the at least one second AI model is to be applied, the business data generated when the user uses the function to be updated of the at least one second AI model is the real data of the scenario to be applied, and then the server can update the at least one second AI model based on the business data to obtain the updated second AI model, and the updated second AI model can output accurate processing results for the application scenario, and the user can also obtain accurate processing results when using the AI ​​service function provided by the updated second AI model, which can greatly improve the accuracy of the processing results of the AI ​​service.

[0173] The communication bus mentioned in the above server can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The communication bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0174] The communication interface is used for communication between the above server and other devices.

[0175] The memory may include a random access memory (RAM) or a non-volatile memory (NVM), such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.

[0176] The above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0177] In another embodiment provided by the present invention, a computer-readable storage medium is further provided, in which a computer program is stored. When the computer program is executed by a processor, the updating method steps of the AI ​​service platform described in any of the above embodiments are implemented.

[0178] In another embodiment provided by the present invention, a computer program product including instructions is also provided, which, when executed on a computer, enables the computer to execute the steps of the method for updating the AI ​​service platform described in any of the above embodiments.

[0179] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented by software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media integrated. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state hard disk Solid State Disk (SSD)), etc.

[0180] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0181] Each embodiment in this specification is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the update device, server, computer-readable storage medium, and computer program product embodiments of the AI ​​service platform, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments.

[0182] The above description is only a preferred embodiment of the present invention and is not intended to limit the protection scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention are included in the protection scope of the present invention.

Claims

1. A method for updating an AI service platform, characterized in that: A server applied to an AI service platform, the server comprising a first type of AI model and a second type of AI model, the first type of AI model comprising a plurality of first AI models, each of which is configured to have no function to be updated and all users of the AI ​​service platform can use its function, the second type of AI model comprising a plurality of second AI models, each of which is configured to have a function to be updated and only some users can use its function, the some users being users in a scenario to which at least one second AI model is to be applied, the method comprising: Acquire business data generated by a user when using a function to be updated of at least one second AI model among the multiple second AI models; Updating the at least one second AI model based on the business data; The step of obtaining the business data generated when the user uses the to-be-updated function of at least one second AI model among the multiple second AI models includes: Pushing a link of the to-be-updated function of the at least one second AI model to a user who has logged in to the AI ​​service platform, so that the user can access the to-be-updated function through the link; The business data uploaded by the user during accessing the function to be updated is obtained.

2. The method according to claim 1, characterized in that Before the step of pushing a link of the to-be-updated function of the at least one second AI model to a user who has logged in to the AI ​​service platform, the method further includes: receiving a user login request, wherein the user login request includes first user information; Determining whether the first user information matches pre-recorded user information of a target user; If there is a match, it is determined that the user corresponding to the user login request is the target user, wherein the target user is the user to whom the link to the updated function of the at least one second AI model needs to be pushed.

3. The method according to claim 1, characterized in that Before the step of pushing a link of the to-be-updated function of the at least one second AI model to a user who has logged in to the AI ​​service platform, the method further includes: receiving a user login request, wherein the user login request includes second user information; Based on the second user information, determining whether the user corresponding to the user login request is an administrative user; If yes, displaying an AI service list including the to-be-updated function of the at least one second AI model; The step of pushing a link of the to-be-updated function of the at least one second AI model to a user who has logged in to the AI ​​service platform includes: When a push instruction is obtained from the management user based on the AI ​​service list, a link to the updated function of the at least one second AI model is pushed to the user, wherein the push instruction is issued when the management user determines that the user who has logged in to the AI ​​service platform is the target user, and the target user is the user to whom the link to the updated function of the at least one second AI model needs to be pushed.

4. The method according to claim 1, characterized in that: The step of pushing a link of the to-be-updated function of the at least one second AI model to a trial user who has logged in to the AI ​​service platform includes: Send push information to a terminal device of a user who has logged in to the AI ​​service platform, wherein the push information includes a link address of the function to be updated of the at least one second AI model.

5. The method according to any one of claims 1 to 4, characterized in that: After the at least one second AI model is updated, the method further includes: Removing the at least one second AI model after the update from the second type of AI models, and adding it to the first type of AI models; When the user logs in to the AI ​​service platform, the functional interface of the at least one second AI model that has been updated is displayed.

6. The method according to any one of claims 1 to 4, characterized in that: The step of updating the at least one second AI model based on the business data includes: Generate multiple training samples based on the business data; Preprocessing each of the training samples according to the function to be updated to obtain a processed training sample; The at least one second AI model is trained based on the processed training samples until the output result of the at least one second AI model meets a preset condition, and the training is stopped to obtain an updated second AI model.

7. An updating device for an AI service platform, characterized in that: A server applied to an AI service platform, the server comprising a first type of AI model and a second type of AI model, the first type of AI model comprising a plurality of first AI models, each of which is configured to have no function to be updated and all users of the AI ​​service platform can use its function, the second type of AI model comprising a plurality of second AI models, each of which is configured to have a function to be updated and only some users can use its function, the some users being users in a scenario to which at least one second AI model is to be applied, the device comprising: A business data acquisition module, used to acquire business data generated when a user uses a function to be updated of at least one second AI model among the multiple second AI models; An AI model updating module, configured to update the at least one second AI model based on the business data; Wherein, the business data acquisition module includes: A link pushing unit, configured to push a link of the to-be-updated function of the at least one second AI model to a user who has logged in to the AI ​​service platform, so that the user can access the to-be-updated function through the link; The data acquisition unit is used to acquire the business data uploaded by the user during the process of accessing the function to be updated.

8. A server, characterized in that: It includes a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus; Memory, used to store computer programs; A processor, for implementing the method steps described in any one of claims 1 to 6 when executing a program stored in a memory.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method steps described in any one of claims 1 to 6 are implemented.

Images