Information processing device, information processing method, and information processing program

Abstract

To enable effortless selection of a machine learning program needed by a user by providing profile information to AI logic.SOLUTION: A server device 10 is configured to: store multiple analysis models and profile information indicative of respective execution conditions for the multiple analysis models; and, upon receipt of a request from an operator O, extract an analysis model executable in an execution environment requested by the operator O from the multiple analysis models according to the profile information.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to an information processing device, an information processing method, and an information processing program. [Background technology]

[0002] AI (Artificial Intelligence) systems that use programs created through machine learning to process images and physical information acquired by sensors and detect something are already in widespread use. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Special Publication No. 2017-529633 Summary of the Invention [Problem to be solved by the invention]

[0004] However, currently, when a user tries to run a program created by machine learning in an execution environment prepared by the user, information such as what kind of input signal each program requires, how much calculation is required to execute one detection process, how frequently detection results can be output in the execution environment prepared by the user, and whether the detection results output by the program match the information required by the user is not provided in a unified format.As a result, users are required to manually check the detailed description of each program to determine whether it can be run in the execution environment prepared by the user and whether it can output the detection results required at the frequency required by the user, thereby selecting selectable programs, which is an obstacle to the distribution of machine learning programs in the market.

[0005] The present invention has been made in consideration of the above, and aims to enable users to easily select the machine learning program they need by adding profile information to AI logic. [Means for solving the problem]

[0006] An information processing device according to one embodiment of the present invention includes a memory unit that stores multiple machine learning models along with profile information indicating the execution conditions of each of the multiple machine learning models, and an extraction unit that, when a user request is received, extracts a machine learning model from the multiple machine learning models that is executable in the execution environment requested by the user based on the profile information.

[0007] In an information processing method according to one embodiment of the present invention, when a computer receives a user request, it executes a process to extract a machine learning model from among the multiple machine learning models that is executable in the execution environment requested by the user, based on profile information indicating the execution conditions of each of the multiple machine learning models stored together with the multiple machine learning models.

[0008] An information processing program according to one embodiment of the present invention causes a computer to execute a process in which, when a user request is received, the computer extracts a machine learning model from among the multiple machine learning models that is executable in the execution environment requested by the user, based on profile information indicating the execution conditions of each of the multiple machine learning models stored together with the multiple machine learning models. [Effects of the Invention]

[0009] According to the present invention, by adding profile information to AI logic, it is possible to easily select the machine learning program that the user requires. [Brief explanation of the drawings]

[0010] [Figure 1]1 is a diagram illustrating a configuration example and a processing example of a monitoring system according to an embodiment; [Figure 2] FIG. 10 is a diagram illustrating a specific example 1 of input / output signal adjustment processing according to the embodiment. [Figure 3] FIG. 10 is a diagram illustrating a specific example 2 of the input / output signal adjustment process according to the embodiment. [Figure 4] FIG. 10 is a diagram illustrating a specific example 3 of the input / output signal adjustment process according to the embodiment. [Figure 5] 1 is a block diagram showing an example of the configuration of each device in a monitoring system according to an embodiment; [Figure 6] FIG. 2 is a diagram illustrating an example of an analysis model storage unit of a server device according to an embodiment. [Figure 7] FIG. 4 is a diagram illustrating an example of a candidate information storage unit of the server device according to the embodiment. [Figure 8] FIG. 10 is a diagram showing a specific example of profile information according to the embodiment. [Figure 9] FIG. 10 is a diagram illustrating a specific example of a display screen of an operator terminal according to the embodiment. [Figure 10] FIG. 2 is a diagram illustrating an example of an analysis model storage unit of the AI ​​execution device according to the embodiment. [Figure 11] FIG. 2 is a diagram illustrating an example of a selection result storage unit of the AI ​​execution device according to the embodiment. [Figure 12] FIG. 2 is a diagram illustrating an example of a monitoring data storage unit of the AI ​​execution device according to the embodiment. [Figure 13] 1 is a flowchart illustrating an example of the overall flow of a monitoring system according to an embodiment. [Figure 14] FIG. 2 is a diagram illustrating an example of a hardware configuration according to an embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0011] An information processing device, an information processing method, and an information processing program according to an embodiment of the present invention will be described in detail below with reference to the accompanying drawings. Note that the present invention is not limited to the embodiments described below.

[0012] The configuration and processing of the monitoring system 100 according to the embodiment, the configuration and processing of each device of the monitoring system 100, and the processing flow of the monitoring system 100 will be described below in order, and finally the effects of the embodiment will be described.

[0013] 1. Configuration and Processing of Monitoring System 100 The configuration and processing of a monitoring system 100 according to an embodiment will be described in detail using Fig. 1. Fig. 1 is a diagram showing an example of the configuration and processing of the monitoring system 100 according to an embodiment. Below, an example of the overall configuration of the monitoring system 100, an example of the processing of the monitoring system 100, and the effects of the monitoring system 100 will be described. Note that in the embodiment, the detection of agricultural pests in farmland will be described as an example, but the field of use is not limited thereto, and the system can also be applied to monitoring parks, roads, rivers, etc., and the analysis targets to be detected can be not only animals but also people, construction machinery, and robots.

[0014] (1-1. Example of the overall configuration of the monitoring system 100) The monitoring system 100 includes a server device 10, an operator terminal 20, an AI execution device 30, and monitoring equipment 40. The server device 10, the operator terminal 20, and the AI ​​execution device 30 are connected to each other via a predetermined communication network (not shown) so that they can communicate with each other via wired or wireless connections. The predetermined communication network can be any of a variety of communication networks, such as the Internet or a dedicated line.

[0015] (1-1-1. Server device 10) The server device 10 is an information processing device that generates candidate information to be provided to an operator O and receives a designation of an AI logic selected by the operator O. For example, the server device 10 obtains profile information of the AI ​​logic from many development companies that develop and own the AI ​​logic, and stores the profile information in a storage area of ​​the server device 10. Similarly, the server device 10 may obtain the program body of the AI ​​logic from the development company and store it in the server device 10. The server device 10 may also store access information that enables access to the AI ​​logic body stored in a database (not shown) on the development company side connected via the Internet or the like. The server device 10 may be realized in a cloud environment, an on-premise environment, an edge environment, or the like.

[0016] (1-1-2. Operator terminal 20) The operator terminal 20 is a manager terminal used by an operator O who is the manager of a facility, equipment, or area (appropriately referred to as a "monitored location") that is the monitoring site. Note that the monitoring system 100 shown in FIG. 1 may include multiple operator terminals 20.

[0017] (1-1-3. AI execution device 30) The AI ​​execution device 30 is a device that installs and runs the AI ​​logic program for the selected AI logic. For example, the AI ​​execution device 30 can be realized in a cloud environment, an on-premise environment, an edge environment, or the like.

[0018] (1-1-4. Monitoring equipment 40) Monitoring device 40 is equipment used for monitoring that is installed at a monitored location, and is realized by camera 40A, sensor 40B, etc. Here, camera 40A is, for example, a photographing device such as a security camera or a monitoring camera that is installed at the monitored location. Sensor 40B is, for example, a measuring device such as a thermometer, hygrometer, sound level meter, intrusion detection sensor, human presence sensor, gas detector, etc. that is installed at the monitored location.

[0019] (1-2. Example of overall processing of monitoring system 100) The following describes the overall processing of the above-described monitoring system 100. Note that the processing of steps S1 to S9 below may be executed in a different order. Also, some of the processing of steps S1 to S9 below may be omitted.

[0020] (1-2-1. Resource information transmission process) First, operator O operates operator terminal 20 to specify resource information of the execution environment that will be used as a condition for selecting candidates for server device 10 (step S1). Here, the resource information of the execution environment includes the maximum capacity value of the input device connected to the execution environment, the size of the memory area for storing the AI ​​logic program, and the computing capacity of the processing device that executes the AI ​​logic. Furthermore, the resource information of the execution environment matches the content of the resources possessed by AI execution device 30.

[0021] (1-2-2. Candidate information request processing) Second, the operator O operates the operator terminal 20 to request candidate information from the server device 10 (step S2). For example, the operator terminal 20, through an operation by the operator O, requests candidate information from the server device 10 indicating candidates for AI logic that can be executed in the execution environment in which the operator O wants to set the AI ​​logic. The operator O may also specify an analysis category (e.g., object detection, face recognition) indicating the classification of the analysis target of the AI ​​logic, the required output value of the detection result, and the frequency at which the detection result is to be output.

[0022] Here, AI logic refers to an analysis method implemented by an analysis model AM, which is a trained machine learning model that outputs the analysis target.

[0023] (1-2-3. Profile information loading process) Third, the server device 10 reads a plurality of pieces of profile information stored in its own storage area (step S3). For example, the server device 10 reads the profile information assigned to the stored model data of the AI ​​logic.

[0024] Here, profile information indicates the execution conditions of the AI ​​logic, and is text information that specifies, for example, the detection target of the AI ​​logic, the input signal, the amount of calculation required to perform one detection, the amount of memory required to store the program, the meaning of the detection results to be output, the format of the output signal, and the terms of use including usage fees.

[0025] (1-2-4. Candidate information generation process) Fourth, the server device 10 generates candidate information (step S4). For example, the server device 10 extracts AI logics that can be executed in the execution environment requested by the operator O, and generates candidate information that includes identification information of the extracted AI logics. At this time, the server device 10 extracts AI logics that can be executed in the execution environment and that can output the detection results required by the user at the required frequency, based on the loaded profile information and resource information received from the operator terminal 20, etc.

[0026] (1-2-5. Candidate information transmission process) Fifth, the server device 10 transmits the candidate information to the operator terminal 20 (step S5). For example, the server device 10 transmits the candidate information including identification information of the AI ​​logic executable in the requested execution environment to the operator terminal 20.

[0027] (1-2-6. Candidate information display process) Sixth, the operator terminal 20 displays the candidate information (step S6). For example, the operator terminal 20 displays on the monitor an AI logic selection screen that presents a list of identification information for the AI ​​logics as the candidate information. At this time, the operator terminal 20 may also display, based on the information provided by the server device 10, performance information that indicates the expected performance when each AI logic is executed under the conditions indicated by the resource information, and usage conditions including the usage fee for each AI logic.

[0028] (1-2-7. Selection result input process) Seventh, the operator O inputs the selection result to the operator terminal 20 (step S7). For example, the operator O makes a decision based on the performance information and usage conditions, including usage fees, of each AI logic displayed on the AI ​​logic selection screen, and selects the identification information of the AI ​​logic determined to be optimal, thereby inputting it as the selection result of the AI ​​logic to be used for analysis. Note that the operator terminal 20 may automatically perform the selection of the AI ​​logic by specifying selection conditions related to usage conditions, including performance information and usage fees, in advance.

[0029] Here, when selecting an AI logic, the operator O may select a combination of multiple AI logics on the premise that multiple AI logics having different properties will be used in combination.

[0030] For example, when two AI logics are used in combination in the monitoring system 100, an analysis is first performed by "AI logic A" on one piece of input information in the AI ​​execution device 30. At this time, in the monitoring system 100, "AI logic B" can use, in addition to the input information, the output information that is the detection result of "AI logic A" as an analysis target. Then, in the monitoring system 100, the output of the detection result of "AI logic B" is considered to be the output of one detection result from the system. In the monitoring system 100, the same applies when three or more AI logics are combined, where multiple AI logics are executed in sequence, and the AI ​​logic executed in a certain sequence can use the output results of the AI ​​logic executed up to that sequence as input in addition to the input information acquired by the system.

[0031] When the profile information reading process of step S3 above is executed for multiple AI logics, the server device 10 reads the detection target, input signal, amount of calculation required to perform one detection, amount of memory required to store the program, meaning of the detection result to be output, format of the output signal, usage conditions including usage fees, etc. for each of the multiple AI logics.

[0032] When the candidate information generation process of step S4 above is performed for multiple AI logics, all AI logics must be individually executable under the conditions specified by the resource information. In addition, even if all AI logics are installed, the resources specified by the conditions specified by the resource information must not be insufficient. Furthermore, even if the processing is executed with the computational performance specified as being available to the system under the conditions specified by the resource information, the total execution time for sequentially executing the processing of all AI logics must be within a time that allows output at the frequency required by the user. The server device 10 outputs combinations of AI logics that satisfy these conditions as candidate information.

[0033] When the candidate information transmission process of step S5 above is executed for multiple AI logics, the server device 10 outputs the combination of AI logics that is determined to satisfy the conditions in that combination as candidate information and transmits it to the operator terminal 20.

[0034] When the candidate information display process of step S6 above is performed for multiple AI logics, the operator terminal 20 may display usage conditions including the total performance information and total usage fees to be achieved for each combination.

[0035] When the selection result input process of step S7 above is performed for multiple AI logics, the operator O inputs the selection result selected from the candidate information presented as a combination of AI logics to the operator terminal 20. At this time, the operator O must also specify the order in which each AI logic is to be executed, and for AI logic that is executed in a certain order, it is possible to specify as input information the value that was specified as the output result of the AI ​​logic that was executed up to that point.

[0036] (1-2-8. Selection result sending process) Eighth, the operator terminal 20 transmits the selection result, which is the identification information of the selected AI logic, to the AI ​​execution device 30, and instructs the AI ​​execution device 30 to install the selected AI logic (step S8). For example, the operator terminal 20 transmits the identification information of the AI ​​logic input to the operator terminal 20 by the operator O to the AI ​​execution device 30 via the server device 10 as the selection result.

[0037] (1-2-9. AI logic setting process) Ninth, the AI ​​execution device 30 installs the AI ​​logic (step S9). For example, the AI ​​execution device 30 installs an application of the AI ​​logic on the resources that the AI ​​execution device 30 has.

[0038] At this time, the AI ​​execution device 30 can also refer to the profile information of the AI ​​logic to be installed and execute input / output signal adjustment processing to adjust input signals and output signals. Here, specific examples 1 to 3 of the input / output signal adjustment processing will be explained using Figures 2 to 4.

[0039] (Example 1) Specific example 1 of adjusting input and output signals so that one AI logic can be executed will be described using FIG. 2. FIG. 2 is a diagram showing specific example 1 of input / output signal adjustment processing according to an embodiment. As shown in FIG. 2(1), the AI ​​execution device 30 adjusts the monitoring data (e.g., image data, measurement data) collected from the monitoring device 40 based on the input signal specifications indicated by the profile information so as to convert it into a format that can be input to "AI logic A." Also, as shown in FIG. 2(2), the AI ​​execution device 30 adjusts the output signal output from "AI logic A" so as to convert it into a format specified by the operator O based on the output signal specifications indicated by the profile information.

[0040] (Example 2) Specific example 2 of adjusting input and output signals so that two or more AI logics can be serially linked and executed will be described using FIG. 3. FIG. 3 is a diagram showing specific example 2 of input / output signal adjustment processing according to an embodiment. In the example of FIG. 3, processing of serially linking and executing two AI logics in the order of "AI logic A" and "AI logic B" will be described. As shown in FIG. 3(1), the AI ​​execution device 30 adjusts the monitoring data collected from the monitoring device 40 so as to convert it into a format that can be input to "AI logic A." Also, as shown in FIG. 3(2), the AI ​​execution device 30 adjusts the output signal output from "AI logic A" so as to convert it into a format that can be input to "AI logic B." Also, as shown in FIG. 3(3), the AI ​​execution device 30 adjusts the output signal output from "AI logic B" so as to convert it into a format specified by the operator O.

[0041] (Example 3) Using FIG. 4, a specific example 3 will be described in which input and output signals are adjusted so that two or more AI logics can be linked and executed in parallel. FIG. 4 is a diagram showing a specific example 3 of input / output signal adjustment processing according to an embodiment. In the example of FIG. 4, a process for linking and executing two AI logics, "AI logic A" and "AI logic B," in parallel will be described. As shown in FIG. 4(1), the AI ​​execution device 30 adjusts the monitoring data collected from the monitoring device 40 so that it is converted into a format that can be input to both "AI logic A" and "AI logic B." Furthermore, as shown in FIG. 4(2), the AI ​​execution device 30 adjusts the output signals output from "AI logic A" and "AI logic B" so that they are both converted into a format specified by the operator O.

[0042] (1-2-10. Monitoring data collection process) Tenth, the AI ​​execution device 30 collects monitoring data from the monitoring equipment 40 (step S10). For example, the AI ​​execution device 30 collects image data of still images taken every second from a camera 40A installed at the monitored location. The AI ​​execution device 30 also collects temperature measurement data measured every second from a sensor 40B installed at the monitored location.

[0043] At this time, the AI ​​execution device 30 saves the collected monitoring data together with the shooting time and measurement time. Note that the image data may be image data of a moving image or data including audio data. Furthermore, the measurement data may be data on humidity, noise, intrusion detection signals, human presence signals, gas detection signals, etc.

[0044] (1-2-11. Monitoring data analysis processing) Eleventh, the AI ​​execution device 30 analyzes the monitoring data (step S11). For example, the AI ​​execution device 30 refers to the selection result and analyzes the monitoring data using the AI ​​logic selected by the operator O. At this time, the AI ​​execution device 30 inputs the monitoring data into the analysis model AM corresponding to each AI logic, and detects an event corresponding to the output analysis target.

[0045] (1-3. Effects of the monitoring system 100) Below, the problems with the monitoring system 100P according to the reference technology will be explained, and then the effects of the monitoring system 100 will be explained.

[0046] (1-3-1. Problems with the 100P monitoring system) A monitoring system 100P according to the reference technology is a technology for provisioning services or resources in a cloud service for successful application execution, which detects a request to execute an application in a cloud service, and in response to the detected request, reads a descriptor record for the application from a descriptor file, the descriptor record being specific to the cloud service and providing details of the environmental resources or services required to execute the application. The monitoring system 100P also translates the resource and service requirements into actions to be taken in the cloud service environment to provision the resources or services required for the application, mediates the translated actions to occur in a predetermined sequence based on the details provided in the application's descriptor record, provides a status of the translated actions, and uses the status to determine whether the resources or services required for successful application execution in the cloud service have been provisioned.

[0047] However, it is difficult for the monitoring system 100P to automatically select AI logic and configure the device during execution. For example, in the monitoring system 100P, information such as input values, such as image data and measurement data to be analyzed, output values ​​output by the AI ​​logic, computational resources required to execute the AI ​​logic, and the time required to obtain the output values, is not available from the AI ​​logic itself. Therefore, when installing AI logic, the operator O must obtain the above information separately and adjust the execution environment before executing the AI ​​logic. Furthermore, when selecting AI logic in the monitoring system 100P, it is necessary to confirm whether the input values ​​required by the AI ​​logic can be prepared so that they can be used in the execution environment, whether sufficient computational resources can be secured (if multiple AI logics are executed, how much computational resources can be provided for each AI logic), and whether the output values ​​contain the analysis results required by the operator O. Therefore, the operator O must obtain the above information separately before selecting AI logic.

[0048] Furthermore, when installing AI logic in a user-prepared environment, Surveillance System 100P must convert input signals from cameras and sensors available in the execution environment into signals required by the AI ​​logic, allocate computing resources to enable the AI ​​logic to output detection results with sufficient frequency, and convert the output values, which are the unique codes of each individual program, into the code system required by the company.These tasks must be handled one by one through custom modifications to the program that controls the activation of the AI ​​logic, which is a factor in the increased costs of using AI logic.

[0049] Furthermore, in monitoring system 100P, even when the output value of one AI logic is used as the input value of another AI logic, both the output value and the input value are designed with unique specifications, so customized development is required to link the two AI logics.

[0050] (1-3-2. Overview of the monitoring system 100) The monitoring system 100 executes the following processes. First, the operator O specifies resource information of the execution environment to the server device 10 via the operator terminal 20. Second, the operator terminal 20, through an operation by the operator O, requests candidate information indicating executable AI logic candidates from the server device 10 and specifies the execution environment in which the AI ​​logic is to be set. Third, the server device 10 reads profile information attached to the saved model data of the AI ​​logic. Fourth, the server device 10 extracts AI logic requested by the operator O that is executable in the specified execution environment and capable of outputting the detection results required by the operator O at the frequency required by the operator O, and generates candidate information including identification information of the extracted AI logic. Fifth, the server device 10 transmits the candidate information including identification information of the executable AI logic to the operator terminal 20. Sixth, the operator terminal 20 displays a setting screen on the monitor presenting a list of identification information of the AI ​​logic as candidate information. Seventh, the operator O inputs the selection of the AI ​​logic to be used for analysis to the operator terminal 20. Eighth, the operator terminal 20 transmits the selection of the AI ​​logic to the AI ​​execution device 30 as a selection result. Ninth, the AI ​​execution device 30 installs the AI ​​logic application in the specified execution environment and adjusts input and output signals using the profile information. Tenth, the AI ​​execution device 30 collects monitoring data from the monitoring equipment 40. Eleventh, the AI ​​execution device 30 analyzes the monitoring data using the installed AI logic and detects events.

[0051] As described above, monitoring system 100 stores multiple machine learning models along with profile information indicating the execution conditions for each of the multiple machine learning models, and upon receiving a user's request, extracts from the multiple machine learning models, based on the profile information, a machine learning model that is executable in an execution environment prepared by the user and that can output the detection results required by the user at the frequency required by the user. Furthermore, when monitoring system 100 installs a machine learning model selected by the user into the execution environment prepared by the user, based on the profile information indicating the execution conditions of the machine learning model, it automatically performs the following settings: converting input signals of the execution environment into a signal format required by the selected machine learning model; securing computational resources in the execution environment required for outputting detection results at the frequency required by the user; and converting output values ​​of the machine learning model into a format required by the user.

[0052] (1-3-3. Effects of the monitoring system 100) The monitoring system 100 has the following advantages. First, when selecting an AI logic to be executed in an AI execution environment requested by an operator O, the monitoring system 100 can extract an AI logic that is executable and outputs the output value requested by the operator O at the frequency required by the operator O, thereby enabling automatic extraction of selectable AI logic candidates. Second, when installing the AI ​​logic selected by the operator O, the monitoring system 100 can automatically set the format and period of the input signal to one that the AI ​​logic can process, set the execution timing of the AI ​​processing by understanding the period of the input signal and the time interval from input to output, and configure the output value to be converted into a value required for subsequent processing before output. Third, it becomes possible to realize combination applications by combining multiple AI logics, such as using the output value of one AI logic as the input value of the next AI logic, or processing collected input values ​​using the output value of an AI logic before using them as the input value of the next AI logic.

[0053] As described above, the monitoring system 100 enables the operator O to easily select the AI ​​logic he or she needs from among numerous AI logics, and automatically install and execute the AI ​​logic. That is, the monitoring system 100 allows the user to easily select the machine learning program he or she needs and further simplifies the installation of the selected program by adding profile information to the AI ​​logic. Furthermore, the monitoring system 100 makes it easy to find AI logic that can run in the execution environment for the machine learning model prepared by the user and that can output the detection results the user needs at the frequency the user needs, and can automatically configure the execution environment when installing the selected AI logic.

[0054] 2. Configuration and Processing of Each Device in Monitoring System 100 The configuration and processing of each device included in the monitoring system 100 shown in Fig. 1 will be described using Fig. 5. Fig. 5 is a block diagram showing an example configuration of each device of the monitoring system 100 according to an embodiment. Below, an example configuration of the entire monitoring system 100 according to an embodiment will be described, and then an example configuration and processing of the server device 10, the operator terminal 20, the AI ​​execution device 30, and the monitoring equipment 40 will be described in detail.

[0055] (2-1. Example of the overall configuration of the monitoring system 100) An example of the overall configuration of the monitoring system 100 shown in Fig. 1 will be described using Fig. 5. As shown in Fig. 5, the monitoring system 100 has a server device 10, an operator terminal 20, an AI execution device 30, and monitoring equipment 40. The server device 10, the operator terminal 20, the AI ​​execution device 30, and the monitoring equipment 40 are communicatively connected by a communication network N realized by the Internet, a dedicated line, or the like.

[0056] The server device 10 is installed in a cloud environment, an on-premise environment, an edge environment, etc. The operator terminal 20 is installed in a monitoring room or the like of a facility or equipment managed by an operator O. The AI ​​execution device 30 is installed in a cloud environment, an on-premise environment, an edge environment, etc. The monitoring device 40 is installed at a monitored site, which is the monitoring site.

[0057] (2-2. Configuration Example and Processing Example of Server Device 10) An example of the configuration and processing of the server device 10 will be described with reference to Fig. 5. The server device 10 is an information processing device, and includes an input unit 11, an output unit 12, a communication unit 13, a storage unit 14, and a control unit 15.

[0058] (2-2-1. Input section 11) The input unit 11 controls input of various information to the server device 10. For example, the input unit 11 is realized by a mouse, a keyboard, etc., and accepts input of various information to the server device 10.

[0059] (2-2-2. Output section 12) The output unit 12 controls the output of various information from the server device 10. For example, the output unit 12 is realized by a display or the like, and displays various information stored in the server device 10.

[0060] (2-2-3. Communications Department 13) The communication unit 13 controls data communication with other devices. For example, the communication unit 13 performs data communication with each communication device via a router, etc. The communication unit 13 can also perform data communication with a terminal (not shown).

[0061] (2-2-4. Storage section 14) The storage unit 14 stores various information referenced by the control unit 15 when it operates and various information acquired when the control unit 15 operates. The storage unit 14 stores multiple machine learning models as well as profile information indicating the execution conditions of each of the multiple machine learning models, and includes an analysis model storage unit 14a and a candidate information storage unit 14b. Here, the storage unit 14 may be realized by, for example, a semiconductor memory element such as a random access memory (RAM) or a flash memory, or a storage device such as a hard disk or an optical disk. Note that, in the example of FIG. 5, the storage unit 14 is installed inside the server device 10, but it may also be installed outside the server device 10, or multiple storage units may be installed.

[0062] (2-2-4-1. Analysis model storage unit 14a) The analytical model storage unit 14a stores an analytical model AM. For example, the analytical model storage unit 14a stores the analytical model AM, which is a machine learning model received by a receiving unit 15a of the control unit 15 (described later) and set by a setting unit 15c. Here, an example of data stored in the analytical model storage unit 14a will be described with reference to FIG. 6. FIG. 6 is a diagram illustrating an example of the analytical model storage unit 14a of the server device 10 according to the embodiment. In the example of FIG. 6, the analytical model storage unit 14a has items such as "AI logic," "profile information," and "analysis model."

[0063] "AI logic" refers to identification information for identifying AI logic that can be set by the server device 10, such as an identification number or identification symbol for the AI ​​logic. "Profile information" refers to the execution conditions for the AI ​​logic, such as text information that specifies the output target, input signal, computing power, memory, executable file, output signal, etc. of the AI ​​logic. "Analysis model" refers to model data for a machine learning model, such as data that includes execution data for executing the algorithm of the analysis model AM, such as image analysis that detects events from image data and sensor analysis that detects events from measurement data, as well as model parameters and hyperparameters that are setting values.

[0064] Figure 6 shows an example in which multiple trained machine learning models are stored in the analysis model storage unit 14a, such as {profile information: "profile information A", analysis model: "analysis model A"} for the AI ​​logic identified by "AI logic A" {profile information: "profile information B", analysis model: "analysis model B"} for the AI ​​logic identified by "AI logic B" {profile information: "profile information B", analysis model: "analysis model B"} for the AI ​​logic identified by "AI logic C" {profile information: "profile information C", analysis model: "analysis model C"} for the AI ​​logic identified by "AI logic D" {profile information: "profile information D", analysis model: "analysis model D"} for the AI ​​logic identified by "AI logic E" {profile information: "profile information E", analysis model: "analysis model E"}.

[0065] (2-2-4-2. Candidate information storage unit 14b) The candidate information storage unit 14b stores candidate information. For example, the candidate information storage unit 14b stores candidate information including identification information of AI logics extracted by the extraction unit 15b of the control unit 15, which will be described later. Here, an example of data stored in the candidate information storage unit 14b will be described with reference to FIG. 7. FIG. 7 is a diagram illustrating an example of the candidate information storage unit 14b of the server device 10 according to the embodiment. In the example of FIG. 7, the candidate information storage unit 14b has items such as "candidate information," "AI logic," "performance information," and "usage conditions."

[0066] "Candidate information" is identification information for identifying candidates for AI logic that can be executed in a specified execution environment, extracted by the extraction unit 15b of the control unit 15 described below, and is an identification number or identification symbol for the candidate AI logic or candidate group. "AI logic" indicates identification information for identifying the AI ​​logic specified by the user, operator O, and is, for example, the identification number or identification symbol of the AI ​​logic. "Performance information" indicates information regarding the performance that is expected to be achieved when the AI ​​logic identified by "AI logic" is executed in an execution environment, such as information about processing time and maximum output frequency. "Conditions of use" indicates the conditions for using the AI ​​logic identified by "AI logic", and is, for example, information about usage fees, usage restrictions, and usage period.

[0067] That is, in FIG. 7, the AI ​​logic associated with the candidate identified by "candidate information #1" is {AI logic: "AI logic A", "AI logic B", "AI logic C"}, the performance information is {Performance information: "Performance information 1-A", "Performance information 1-B", "Performance information 1-C"}, the usage conditions are {Usage conditions: "Usage conditions 1-A", "Usage conditions 1-B", "Usage conditions 1-C"}, and the AI ​​logic associated with the candidate identified by "candidate information #2" is {AI logic: "AI logic 2- An example is shown in which data such as the AI ​​logic associated with the candidate identified by "candidate information #3" is {AI logic: "AI logic B"}, performance information {performance information: "performance information 3-B"}, and usage conditions are {usage conditions: "usage conditions 3-B"}, etc. is stored in the candidate information storage unit 14b.

[0068] (2-2-5. Control unit 15) The control unit 15 controls the entire server device 10. The control unit 15 has a reception unit 15a, an extraction unit 15b, and a setting unit 15c. Here, the control unit 15 can be realized by, for example, an electronic circuit such as a CPU (Central Processing Unit) or an MPU (Micro Processing Unit), or an integrated circuit such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array).

[0069] (2-2-5-1. Reception section 15a) The reception unit 15a receives various types of information. The reception unit 15a may store the received various types of information in the storage unit 14. The AI ​​logic reception process, the candidate information request reception process, and the AI ​​logic selection reception process will be described below.

[0070] (AI logic reception processing) The reception unit 15a executes an AI logic reception process. For example, the reception unit 15a receives model data of a machine learning model, including execution data for executing an algorithm of the analysis model AM, such as image analysis for detecting an event from image data or sensor analysis for detecting an event from measurement data, model parameters that are setting values, hyperparameters, etc. At this time, the reception unit 15a receives the model data of the analysis model AM to which profile information has been assigned from a predetermined database, and stores the model data in the storage unit 14.

[0071] To explain a specific example of the AI ​​logic reception process, the reception unit 15a receives, from a predetermined database, the following as an analysis model AM for image analysis that enables object detection: {AI logic: "AI logic A", profile information: "profile information A", analysis model: "analysis model A"}, {AI logic: "AI logic B", profile information: "profile information B", analysis model: "analysis model B"}, {AI logic: "AI logic C", profile information: "profile information C", analysis model: "analysis model C"}, {AI logic: "AI logic D", profile information: "profile information D", analysis model: "analysis model D"}, {AI logic: "AI logic E", profile information: "profile information E", analysis model: "analysis model E"}, ...

[0072] (Resource information reception process) The reception unit 15a executes a resource information reception process. For example, the reception unit 15a receives resource information of the execution environment that is a condition for selecting an AI logic candidate, transmitted from the operator terminal 20. At this time, the reception unit 15a receives, as resource information, the maximum capability value of the input device connected to the execution environment, the size of the memory area for storing the AI ​​logic program, the computational capability of the processing device that executes the AI ​​logic, and the like.

[0073] A specific example of the resource information reception process will be described. The reception unit 15a receives "resource information #1" as resource information input by the operator O operating the operator terminal 20.

[0074] (Candidate information request acceptance process) The reception unit 15a executes a candidate information request process. For example, the reception unit 15a receives a request for candidate information indicating candidates for AI logics, transmitted from the operator terminal 20. At this time, the reception unit 15a can also receive an analysis category indicating a classification of an analysis target of the AI ​​logic, output information indicating an output value of a required detection result, an output frequency indicating how often the detection result is output, and the like.

[0075] To explain a specific example of the candidate information request reception process, the reception unit 15a receives {transmission request: "transmission request #1", analysis category: "analysis category #2", output information: "output information #1", output frequency: "output frequency #1"} as a request to transmit candidate information input by the operator O by operating the operator terminal 20.

[0076] (AI logic selection acceptance processing) The reception unit 15a executes an AI logic selection reception process. For example, the reception unit 15a receives a selection of an AI logic, thereby receiving, as a selection result, a selection of an analysis model AM from a plurality of analysis models AM used in the AI ​​logic. At this time, the reception unit 15a receives, as a selection result, a selection of an operator O from a plurality of analysis models AM extracted by the extraction unit 15b.

[0077] To explain a specific example of the AI ​​logic selection reception process, the reception unit 15a receives as a selection result a selection of using one AI logic {AI logic: "AI logic A"} input by the operator O by operating the AI ​​logic selection screen displayed on the monitor of the operator terminal 20, and transmits the selection result to the AI ​​execution device 30. The reception unit 15a also receives as a selection result a selection of using two AI logics in series {AI logic 1: "AI logic A", AI logic 2: "AI logic B"} input by the operator O by operating the AI ​​logic selection screen displayed on the monitor of the operator terminal 20, and transmits the selection result to the AI ​​execution device 30. The reception unit 15a also receives as a selection result a selection of using two AI logics in parallel {AI logic 1-1: "AI logic A", AI logic 1-2: "AI logic B"} input by the operator O by operating the AI ​​logic selection screen displayed on the monitor of the operator terminal 20, and transmits the selection result to the AI ​​execution device 30.

[0078] (2-2-5-2. Extraction part 15b) The extraction unit 15b extracts various types of information. The extraction unit 15b may store the extracted various types of information in the storage unit 14. The profile information reading process, the candidate information generating process, and the candidate information transmitting process will be described below.

[0079] (Profile information loading process) The extraction unit 15b executes a profile information reading process. For example, the extraction unit 15b refers to profile information that defines an output target, input signal, calculation amount, memory, executable file, output signal, etc. of the AI ​​logic, which is stored in the analysis model storage unit 14a of the storage unit 14.

[0080] Specific examples of profile information will be described using FIG. 8. FIG. 8 is a diagram showing specific examples of profile information according to an embodiment. Below, specific example 1, which is profile information that defines the output target of the AI ​​logic, specific example 2, which is profile information that defines the input signal of the AI ​​logic, specific example 3, which is profile information that defines the computational amount of the AI ​​logic, specific example 4, which is profile information that defines the memory of the AI ​​logic, specific example 5, which is profile information that defines the executable file of the AI ​​logic, and specific example 6, which is profile information that defines the output signal of the AI ​​logic, will be described. Note that the profile information is structured text information such as JSON (JavaScript (registered trademark) Object Notation), XML (Extensible Markup Language), YAML (YAML Ain't Markup Language), etc., but the format and type are not particularly limited.

[0081] (Example 1: Output target) As shown in Figure 8(1), the profile information specifies the output target of the AI ​​logic. Figure 8(1) shows an example in which, for "Category," which indicates the classification of the output target of the AI ​​logic, "Main," which indicates the primary classification, is "Object Detection," which enables object detection, and "Sub-1," which indicates the secondary classification, is "Agricultural Pest Detection," which enables agricultural pest detection. In other words, the output target of the AI ​​logic is specified as "Main" for general targets in each analytical field, and if the AI ​​logic has performed learning that strengthens the analytical capabilities of a specific target, that target is specified as "Sub," but the specification of the output target is not particularly limited.

[0082] Here, the profile information that specifies the output target of the AI ​​logic may be "object detection" or "agricultural pest detection" as image analysis that analyzes image data, as well as "person detection," "animal detection," "behavior detection," "face recognition," "person identification by gait," "gender and age estimation," "clothing and equipment detection," "detection of persons requiring assistance," "robot detection," "work vehicle detection," etc. Furthermore, the profile information that specifies the output target of the AI ​​logic may be "temperature abnormality," "humidity abnormality," "noise abnormality," "intrusion detection abnormality," "person detection abnormality," "gas detection abnormality," etc. as sensor analysis that analyzes measurement data.

[0083] (Example 2: Input signal) As shown in Figure 8(2), the profile information specifies the input signal of the AI ​​logic. Figure 8(2) shows an example in which, for the input "Input" required by the AI ​​logic, the "Type" indicating the input category of image data is "still image," the "Resolution" indicating the number of pixels (data granularity of the image data) is set to "512 x 512" indicating the initial setting, "Resolution2" indicating other compatible settings is set to "640 x 480," and the "FPS" indicating the frequency of capturing image data is set to "100 ms" in frames per second (FPS).

[0084] Here, the profile information defining the input signal to the AI ​​logic may be image data such as "still images," or moving image data such as "videos," and their capture frequency (e.g., FPS). Also, the profile information defining the input signal may be measurement data such as "temperature," "humidity," "noise," "intrusion detection signals," "human presence signals," "gas detection signals," and their capture frequency.

[0085] (Example 3: Computational complexity) As shown in Figure 8(3), the profile information specifies the specifications of the computing device required to execute the AI ​​logic and the amount of calculation required to perform one detection. Figure 8(3) shows an example where, for "Performance," which indicates the amount of calculation by the AI ​​logic, the "Calculation Type," which indicates the calculation type of the AI ​​logic, is "Neural Network," the "Arithmetic Accuracy," which indicates the calculation accuracy of the AI ​​logic, is "FP16," and the "Operation Volume," which indicates the amount of calculation by the AI ​​logic to perform one detection, is "75 Giga Operations."

[0086] Here, the computational complexity of AI logic refers to the computational complexity required for the AI ​​logic to execute one output, and assuming that computational resources with a certain level of performance are available, the time required for those computational resources to execute that computational complexity indicates the time it takes to use that AI logic to output an analysis result from an input. This value is used to calculate how many outputs can be executed per unit time, and this value becomes the output frequency.

[0087] Furthermore, the profile information that specifies the computational capabilities of the AI ​​logic may have calculation precision of "FP32," "FP26," "INT8," etc. in addition to "FP16."

[0088] (Example 4: Memory) As shown in Figure 8(4), the profile information specifies the memory of the AI ​​logic. Figure 8(4) shows an example in which the "Memory Footprint" indicating the memory footprint required by the AI ​​logic is "450MB" as the "Size" indicating the memory footprint capacity.

[0089] Here, the profile information that specifies the memory of the AI ​​logic is specified separately from the computational complexity of the AI ​​logic shown in specific example 4, and is used, for example, to take into account shared resources when multiple AI logics are executed simultaneously.

[0090] (Example 5: Executable file) As shown in Figure 8(5), the profile information specifies the execution file of the AI ​​logic. Figure 8(5) shows an example in which the "AI Operation" that indicates the AI ​​operation used by the AI ​​logic has a "Type" that indicates the classification of the description method for the AI ​​operation as "ONNX (Open Neural Network Exchange)," and the "Detail" that indicates the details of the AI ​​operation has a "Version" that indicates the version as "2.0," and a "Subset" that indicates the subset as "RenesasRZ."

[0091] Here, the profile information that specifies the executable file of the AI ​​logic may be "ONNX" as a classification of the description method of the AI ​​operation, or may be "TensorFlow" or a unique classification, etc. Furthermore, the profile information that specifies the executable file of the AI ​​logic may be specified as a subset of the AI ​​operation with an arbitrary name.

[0092] (Example 6: Output signal) As shown in Figure 8(6), the profile information defines the output signal of the AI ​​logic. Figure 8(6) shows an example in which, for "Output," which indicates an output signal that the AI ​​logic can output, "Output1," which indicates a first output signal, has a "Value" indicating the output value of "001" and a "Meaning" indicating the meaning corresponding to the output value of "Dog." "Output2," which indicates a second output signal, has a "Value" indicating the output value of "002" and a "Meaning" indicating the meaning corresponding to the output value of "Cat." "Output3," which indicates a third output signal, has a "Value" indicating the output value of "003" and a "Meaning" indicating the meaning corresponding to the output value of "Bear." That is, the output target of the AI ​​logic is defined in a format in which a "Meaning" indicating the meaning corresponding to the output value is assigned to a "Value" indicating the output value expressed by a unique label, but the format of the defined output target is not particularly limited.

[0093] Here, the profile information that defines the output signal of the AI ​​logic is expressed in English as "Dog," "Cat," "Bear," etc., to indicate the meaning corresponding to the output value, but it may also be expressed in other languages.

[0094] (Candidate information generation process) The extraction unit 15b executes a candidate information generation process. When the extraction unit 15b receives a request from the operator O, it extracts, from among a plurality of analysis models AM, an analysis model AM that is executable in the execution environment requested by the operator O, based on the profile information. At this time, the extraction unit 15b identifies the computational amount of the analysis model AM indicated by the loaded profile information, extracts a feasible output frequency by referring to resource information indicating the computational resources, etc., of the execution environment requested by the operator O, and if this exceeds the output frequency requested by the user, generates candidate information including identification information of the extracted analysis model AM.

[0095] To explain a specific example of the candidate information generation process, the extraction unit 15b identifies the computational amount and memory of each analysis model AM indicated by ``profile information A,'' ``profile information B,'' ``profile information C,'' ``profile information D,'' and ``profile information E'' as profile information of the analysis model AM stored in the analysis model storage unit 14a, refers to ``resource information #1'' as resource information of ``execution environment #1,'' which is the execution environment requested by operator O stored in the candidate information storage unit 14b, extracts ``AI logic A,'' ``AI logic B,'' and ``AI logic C'' as analysis models AM that are executable in ``execution environment #1'' and whose total achievable output frequency exceeds the output frequency requested by the user, generates {AI logic: ``AI logic A,'' ``AI logic B,'' ``AI logic C''} as candidate information to be provided to operator O, and stores it in the candidate information storage unit 14b.

[0096] (Candidate information sending process) The extraction unit 15b executes a candidate information transmission process. For example, the extraction unit 15b transmits, to the operator terminal 20 of the operator O, candidate information including the identification information of the generated analysis model AM.

[0097] To explain a specific example of the candidate information transmission process, the extraction unit 15b refers to {AI logic: "AI logic A", "AI logic B", "AI logic C"} from the candidate information storage unit 14b as candidate information to be provided to the operator O, transmits it to the operator terminal 20 used by the operator O, and displays it on the input / output unit 21 of the operator terminal 20 as an AI logic selection screen.

[0098] (2-2-5-3. Setting section 15c) The setting unit 15c executes various settings. Note that the setting unit 15c may acquire various information from the storage unit 14. The AI ​​logic setting instruction process will be described below.

[0099] (AI logic setting instruction processing) The setting unit 15c executes an AI logic setting instruction process. For example, the setting unit 15c sets the analysis model AM, the selection of which has been accepted by the accepting unit 15a from the operator O, in the execution environment requested by the operator O via the AI ​​execution device 30. At this time, the setting unit 15c refers to the selection result, acquires model data of the analysis model AM corresponding to the selection result, and installs the model data in the requested execution environment via the AI ​​execution device 30.

[0100] In this way, the operator X selects to use one candidate from among the multiple candidates presented by the server device 10, and instructs the server device 10 to automatically install all of the AI ​​logics included in the single AI logic or combination of AI logics indicated by that candidate into the AI ​​execution device 30.

[0101] To explain a specific example of the AI ​​logic setting instruction process, the setting unit 15c causes the AI ​​execution device 30 to refer to {AI logic: "AI logic A", "AI logic B"} as the selection results in the selection result storage unit 32b of the AI ​​execution device 30, acquire "analysis model A" and "analysis model B" as model data from the analysis model storage unit 14a, and set "analysis model A" and "analysis model B" to an executable state.

[0102] At this time, when operator O designates AI execution device 30 as the destination of automatic installation, setting unit 15c sets "analysis model A" and "analysis model B" to an executable state on the resources of AI execution device 30. Note that setting unit 15c assumes that the AI ​​execution device 30 to be the target of automatic installation has the same specifications as the execution environment used when the candidate information was generated, but the specifications, i.e., specs, of the AI ​​execution device 30 to be the target of automatic installation are not particularly limited.

[0103] (2-3. Configuration Example and Processing Example of Operator Terminal 20) 5 again, a description will be given of an example of the configuration and processing of the operator terminal 20. The operator terminal 20 is a posting device and a viewing device, and includes an input / output unit 21, a transmitting / receiving unit 22, and a communication unit .

[0104] (2-3-1. Input / output section 21) The input / output unit 21 controls the input of various information to the operator terminal 20. For example, the input / output unit 21 is realized by a mouse, a keyboard, a touch panel, or the like, and accepts input of various information to the operator terminal 20. The input / output unit 21 also controls the display of various information from the operator terminal 20. For example, the input / output unit 21 is realized by a display, or the like, and displays various information stored in the operator terminal 20. The setting screen display process will be described below.

[0105] (Settings screen display process) The input / output unit 21 executes a setting screen display process. For example, the input / output unit 21 presents a list of a plurality of analysis models AM and displays an AI logic selection screen as a setting screen for receiving input for selecting an analysis model AM.

[0106] To explain a specific example of the setting screen display process, the input / output unit 21 displays {AI logic: "AI logic A", "AI logic B", "AI logic C"} corresponding to "transmission request #1" as candidate information transmitted from the server device 10.

[0107] (2-3-2. Transmitter / receiver 22) The transmitting / receiving unit 22 transmits various types of information. For example, the transmitting / receiving unit 22 transmits a request to transmit candidate information and resource information input by the operator O to the server device 10. The transmitting / receiving unit 22 also transmits the selection result input by the operator O via the AI ​​logic selection screen to the server device 10.

[0108] The transmitting / receiving unit 22 receives various types of information. For example, the transmitting / receiving unit 22 receives candidate information transmitted from the server device 10.

[0109] (2-3-3. Communications Department 23) The communication unit 23 controls data communication with other devices. For example, the communication unit 23 performs data communication with each communication device via a router, etc. The communication unit 23 can also perform data communication with a terminal (not shown).

[0110] (2-3-4. Specific Examples of Display Screens on Operator Terminal 20) Here, a specific example of a display screen output by the input / output unit 21 of the operator terminal 20 will be described with reference to Fig. 9. Fig. 9 is a diagram showing a specific example of a display screen of the operator terminal 20 according to the embodiment. The "AI logic selection screen" will be described below.

[0111] (2-3-4-1. AI logic selection screen) 9, the operator terminal 20 displays an "AI logic selection screen," which is a setting screen that accepts input of a selection of an AI logic (i.e., an analysis model AM). Here, the operator O can input a selection of the AI ​​logic that he or she wishes to set in the execution environment by clicking one of the displayed buttons, "AI logic A," "AI logic B," and "AI logic C."

[0112] (2-3-4-2. Other) Operator O can input a selection of multiple AI logics that he / she wants to set in the execution environment by clicking two or more buttons from the displayed "AI Logic A," "AI Logic B," and "AI Logic C." At this time, Operator O can also input a selection to set multiple AI logics by connecting them in series, or input a selection to set multiple AI logics by connecting them in parallel.

[0113] (2-4. Configuration and Processing Examples of AI Execution Device 30) 5, an example of the configuration and processing of the AI ​​execution device 30 will be described. The AI ​​execution device 30 has a communication unit 31, a storage unit 32, and a control unit 33.

[0114] (2-4-1. Communications Department 31) The communication unit 31 controls data communication with other devices. For example, the communication unit 31 performs data communication with each communication device via a router or the like. The communication unit 31 can also perform data communication with a terminal (not shown).

[0115] (2-4-2. Storage section 32) The storage unit 32 stores various types of information referenced by the control unit 33 when it operates, as well as various types of information acquired when the control unit 33 operates. The storage unit 32 stores multiple machine learning models as well as profile information indicating the execution conditions for each of the multiple machine learning models, and has an analysis model storage unit 32a, a selection result storage unit 32b, and a monitoring data storage unit 32c. Here, the storage unit 32 can be realized, for example, by a semiconductor memory element such as RAM or flash memory, or a storage device such as a hard disk or optical disk. Note that, although the storage unit 32 is installed inside the AI ​​execution device 30 in the example of FIG. 5, it may be installed outside the AI ​​execution device 30, or multiple storage units may be installed.

[0116] (2-4-2-1. Analysis model storage unit 32a) The analytical model storage unit 32a stores an analytical model AM. For example, the analytical model storage unit 32a stores the analytical model AM, which is a machine learning model received by a transceiver unit 33a of the control unit 33 (described later) and set or executed by an execution unit 33b. Here, an example of data stored in the analytical model storage unit 32a will be described with reference to FIG. 10. FIG. 10 is a diagram showing an example of the analytical model storage unit 32a of the AI ​​execution device 30 according to the embodiment. In the example of FIG. 10, the analytical model storage unit 32a has items such as "AI logic," "profile information," and "analysis model."

[0117] "AI logic" refers to identification information for identifying AI logic that can be set or executed by the AI ​​execution device 30, such as an identification number or symbol for the AI ​​logic. "Profile information" refers to the execution conditions for the AI ​​logic, such as text information that specifies the detection target, input signal, amount of calculation, memory, executable file, output signal, etc. "Analysis model" refers to model data for a machine learning model, such as data that includes execution data for executing the algorithm of the analysis model AM, such as image analysis that detects events from image data or sensor analysis that detects events from measurement data, as well as model parameters and hyperparameters that are setting values.

[0118] Figure 10 shows an example in which multiple trained machine learning models are stored in the analysis model storage unit 32a, such as {profile information: "profile information A", analysis model: "analysis model A"} for the AI ​​logic identified by "AI logic A", {profile information: "profile information C", analysis model: "analysis model C"} for the AI ​​logic identified by "AI logic C".

[0119] (2-4-2-2. Selection result storage unit 32b) The selection result storage unit 32b stores the selection results for installation in the execution environment. For example, the selection result storage unit 32b stores the selection results, including the identification information of the AI ​​logic, input by the operator O via a setting screen and received by the transmitter / receiver 33a of the control unit 33, which will be described later. Here, an example of data stored in the selection result storage unit 32b will be described with reference to FIG. 11. FIG. 11 is a diagram showing an example of the selection result storage unit 32b of the AI ​​execution device 30 according to the embodiment. In the example of FIG. 11, the selection result storage unit 32b has items such as "AI logic," "input information," "detection result format," "detection result code," "execution frequency," and "format after conversion of the detection result."

[0120] "AI logic" refers to identification information for identifying the selected AI logic, such as the identification number or symbol of the AI ​​logic selected by operator O. "Input information" refers to information about the input signal to be input to the selected AI logic, such as the format of the input signal that needs to be input to the AI ​​logic selected by operator O. "Detection result format" refers to information about the format of the output target output from the selected AI logic, such as the format of the output signal of the detection target detected by the AI ​​logic selected by operator O. "Detection result code" refers to information about the code of the output target output from the selected AI logic, such as the code of the output signal of the detection target detected by the AI ​​logic selected by operator O. "Execution frequency" refers to the frequency at which the selected AI logic is executed, such as the number of outputs per unit time specified by operator O when executing the AI ​​logic selected by operator O. "Detection result converted format" refers to information about the format of the output target output from the selected AI logic after conversion, such as the converted format when the AI ​​logic selected by operator O outputs the detection result so that it can be used in subsequent processing, or the format of the output signal so that the user can use the detection result of this AI logic in subsequent processing.

[0121] (2-4-2-3. Monitoring data storage unit 32c) The monitoring data storage unit 32c stores monitoring data. For example, the monitoring data storage unit 32c stores monitoring data received by a transceiver unit 33a of the control unit 33, which will be described later. Here, an example of data stored in the monitoring data storage unit 32c will be described with reference to FIG. 12. FIG. 12 is a diagram showing an example of the monitoring data storage unit 32c of the AI ​​execution device 30 according to the embodiment. In the example of FIG. 12, the monitoring data storage unit 32c has items such as "monitoring device," "monitoring target location," "time," and "monitoring data."

[0122] "Monitoring equipment" refers to identification information for identifying the photographing equipment or measuring equipment, such as the identification number or symbol of camera 40A or sensor 40B. "Monitored location" refers to identification information for identifying the facility, equipment, or section in the monitored area where the photographing equipment or measuring equipment is installed, such as the identification number or symbol of the facility, equipment, or section. "Time" refers to the time of photographing or measurement, and is expressed, for example, in years, months, days, hours, minutes, and seconds. "Monitoring data" refers to the monitoring data acquired during the photographing or measurement time, such as image data of still images, image data of moving images, image data of moving images including audio data, temperature measurement data, humidity measurement data, noise measurement data, intrusion detection signal measurement data, motion detection signal measurement data, and gas detection signal measurement data, all acquired every second.

[0123] That is, Figure 12 shows an example in which data such as {time: "time #1", monitoring data: "monitoring data #1-1"}, {time: "time #2", monitoring data: "monitoring data #1-2"}, {time: "time #3", monitoring data: "monitoring data #1-3"} is stored in the monitoring data storage unit 32c for the monitoring device 40 identified by "monitoring device #1" and the monitored location identified by "monitoring location #1".

[0124] (2-4-3. Control unit 33) The control unit 33 is responsible for overall control of the server device 10. The control unit 33 has a transmission / reception unit 33a and an execution unit 33b. Here, the control unit 33 can be realized by, for example, an electronic circuit such as a CPU or an MPU, or an integrated circuit such as an ASIC or an FPGA.

[0125] (2-4-3-1. Transmitter / receiver 33a) The transmitting / receiving unit 33a receives various types of information. For example, the transmitting / receiving unit 33a receives a selection result transmitted from the server device 10. The transmitting / receiving unit 33a can also receive a selection result transmitted from the operator terminal 20.

[0126] The transmitting / receiving unit 33a receives model data of the analytical model AM transmitted from the server device 10. It can also receive model data of the analytical model AM from a predetermined database.

[0127] The transmitter / receiver 33a collects monitoring data. The monitoring data collection process (image data collection process, measurement data collection process) will be described below.

[0128] (Monitoring data collection process) The transmitter / receiver 33a executes a monitoring data collection process, for example, by collecting monitoring data acquired by monitoring devices 40 installed at monitored locations.

[0129] To explain a specific example, the transmitter / receiver 33a collects monitoring data such as {time: "time #1", monitoring data: "monitoring data #1-1"}, {time: "time #2", monitoring data: "monitoring data #1-2"}, {time: "time #3", monitoring data: "monitoring data #1-3"} as monitoring data acquired by the monitoring device 40-1 installed at the "monitored location #1", and stores the data in the monitoring data storage unit 32c.

[0130] (Image data collection and processing) The transmitter / receiver 33a executes an image data collection process as the monitoring data collection process. For example, the transmitter / receiver 33a collects image data acquired by a photographing device installed at a monitoring target base as the monitoring data. At this time, the transmitter / receiver 33a collects, for example, image data of a still image, image data of a moving image, image data of a moving image including audio data, etc., acquired every second by the camera 40A.

[0131] (Measurement data collection processing) The transmitter / receiver 33a executes a measurement data collection process as the monitoring data collection process. For example, the transmitter / receiver 33a collects, as monitoring data, measurement data acquired by measuring devices installed at the monitored base. At this time, the transmitter / receiver 33a collects, for example, temperature measurement data, humidity measurement data, noise measurement data, intrusion detection signal measurement data, human presence signal measurement data, gas detection signal measurement data, etc., acquired every second by the sensor 40B.

[0132] (2-4-3-2. Executive Unit 33b) The execution unit 33b executes various processes. Note that the execution unit 33b may acquire various information from the storage unit 32. The AI ​​logic setting process, the input / output signal adjustment process, and the event detection process will be described below.

[0133] (AI logic setting process) The execution unit 33b executes the AI ​​logic setting process. For example, the execution unit 33b sets the analysis model AM received by the transmission / reception unit 33a in the execution environment requested by the operator O. At this time, the execution unit 33b refers to the selection result, receives model data of the analysis model AM corresponding to the selection result, and installs the model data in the requested execution environment.

[0134] To explain a specific example of the AI ​​logic setting process, the execution unit 33b refers to {AI logic: "AI logic A", "AI logic B"} as the selection result of the selection result storage unit 32b, refers to "analysis model A" and "analysis model B" as model data from the analysis model storage unit 32a, and sets "analysis model A" and "analysis model B" to an executable state on the resources of the AI ​​execution device 30.

[0135] (Input / output signal conditioning processing) The execution unit 33b executes input / output signal adjustment processing. Below, specific example 1, which is input / output signal adjustment processing for a single AI logic, specific example 2, which is input / output signal adjustment processing for multiple AI logics in series, and specific example 3, which is input / output signal adjustment processing for multiple AI logics in parallel, will be described.

[0136] (Example 1: Single AI logic) As a specific example 1, the execution unit 33b is configured to convert, based on the profile information, an input signal acquired in the execution environment into a form that can be input to the selected analysis model AM, and to convert the output signal of the selected analysis model AM into a form requested by the operator O.

[0137] For example, when setting "AI Logic A" in "Execution Environment #1," the execution unit 33b refers to the input signal specifications indicated in the "Profile Information A" of "AI Logic A" and adjusts the data to convert the monitoring data collected from the monitoring device 40 into an input category, data granularity, and acquisition frequency that can be input to "AI Logic A," and refers to the output signal specifications indicated in the "Profile Information A" of "AI Logic A" and adjusts the data to convert the output signal output from "AI Logic A" into the output value specified by the operator O and the meaning corresponding to the output value.

[0138] (Example 2: Serial AI logic) As a second specific example, the execution unit 33b is configured to convert the output signal of a selected analysis model AM into an input signal that can be input to another selected analysis model AM, based on the profile information.

[0139] For example, when "AI Logic A" and "AI Logic B" are set in series in the AI ​​execution device 30, the execution unit 33b refers to the input signal specifications indicated in the "profile information A" of "AI Logic A" and adjusts the monitoring data collected from the monitoring device 40 to be converted into an input category, data granularity, and acquisition frequency that can be input to "AI Logic A," refers to the output signal specifications indicated in the "profile information A" of "AI Logic A" and the input signal specifications indicated in the "profile information B" of "AI Logic B," and adjusts the output signal output from "AI Logic A" to be converted into an input category, data granularity, and acquisition frequency that can be input to "AI Logic B," and refers to the output signal specifications indicated in the "profile information B" of "AI Logic B," and adjusts the output signal output from "AI Logic B" to be converted into an output value specified by the operator O and a meaning corresponding to the output value.

[0140] (Example 3: Parallel AI logic) As a third specific example, the execution unit 33b is configured to convert, based on the profile information, an input signal acquired in the execution environment into a form that can be input to both the selected analysis model AM and another selected analysis model AM.

[0141] For example, when "AI Logic A" and "AI Logic B" are set in parallel in the AI ​​execution device 30, the execution unit 33b refers to the input signal specifications indicated by the "profile information A" of "AI Logic A" and the input signal specifications indicated by the "profile information B" of "AI Logic B," and adjusts the monitoring data collected from the monitoring device 40 to convert it into an input category, granularity, and acquisition frequency that can be input to both "AI Logic A" and "AI Logic B," and refers to the output signal specifications indicated by the "profile information A" of "AI Logic A" and the output signal specifications indicated by the "profile information B" of "AI Logic B," and adjusts the output signals output from "AI Logic A" and "AI Logic B" to convert both into output values ​​specified by the operator O and meanings corresponding to the output values.

[0142] (Event detection processing) The execution unit 33b executes an event detection process, for example, to detect an event that has occurred at each of a plurality of monitored locations based on the collected monitoring data.

[0143] To explain a specific example, the execution unit 33b analyzes the monitoring data {monitoring device: "monitoring device #1", monitored location: "monitored location #1", time: "time #1", monitoring data: "monitoring data #1-1"} collected from monitoring device 40-1 installed at monitored location 1, detects an event that occurred at monitored location 1, and outputs the detection result {monitoring device: "monitoring device #1", monitored location: "monitored location #1", time: "time #1", event: "event #1-1"}. In addition, the execution unit 33b analyzes the monitoring data {monitoring device: "monitoring device #2", monitored location: "monitored location #2", time: "time #3", monitoring data: "monitoring data #2-3"} collected from the monitoring device 40-2 installed at the monitored location 2, detects an event that occurred at the monitored location 2, and outputs the detection result {monitoring device: "monitoring device #2", monitored location: "monitored location #2", time: "time #3", event: "event #2-3"}.

[0144] At this time, the execution unit 33b may acquire the detection result output by inputting {monitoring device: "monitoring device #1", monitored site: "monitored site #1", time: "time #1", monitoring data: "monitoring data #1-1"} as the monitoring data collected from the monitoring device 40-1 installed at the monitored site 1 into "analysis model A" as the analysis model AM stored in the analysis model storage unit 32a. Also, the execution unit 33b may acquire the detection result output by inputting {monitoring device: "monitoring device #2", monitored site: "monitored site #2", time: "time #3", monitoring data: "monitoring data #2-3"} as the monitoring data collected from the monitoring device 40-2 installed at the monitored site 2 into "analysis model B" as the analysis model AM stored in the analysis model storage unit 32a.

[0145] (Image data analysis processing) The execution unit 33b executes an image data analysis process as the event detection process. For example, the execution unit 33b detects an event based on the collected image data. In this case, the execution unit 33b analyzes image data of still images captured by the camera 40A every second, and detects events such as animal intrusions and animal behavior based on differences in the image data. The execution unit 33b can also detect an event that has occurred using an analysis model AM that outputs an event when image data of still images captured by the camera 40A every second is input.

[0146] (Measurement data analysis processing) The execution unit 33b executes a measurement data analysis process as the event detection process. For example, the execution unit 33b detects an event based on the collected measurement data. At this time, the execution unit 33b analyzes measurement data such as temperature, humidity, noise, intrusion detection signal, human presence signal, and gas detection signal acquired every second by the sensor 40B, and detects an abnormal state or the like at the monitored location as an event based on a specified value of each measurement data. The execution unit 33b can also detect an event that has occurred using an analysis model AM that outputs an event when measurement data such as temperature, humidity, noise, intrusion detection signal, human presence signal, and gas detection signal acquired every second by the sensor 40B is input.

[0147] 3. Flow of each process in the monitoring system 100 The processing flow of the monitoring system 100 according to the embodiment will be described with reference to Fig. 13. Fig. 13 is a flowchart showing an example of the overall flow of the monitoring system 100 according to the embodiment. Note that the processing of steps S101 to S111 below can also be executed in a different order. Furthermore, some of the processing of steps S101 to S111 below may be omitted.

[0148] (3-1. Resource information transmission process) First, the operator terminal 20 executes a resource information transmission process (step S101). For example, the operator terminal 20 transmits the resource information of the execution environment input by the operator O to the server device 10.

[0149] (3-2. Candidate information request processing) Second, the operator terminal 20 executes a candidate information request process (step S102). For example, the operator terminal 20 transmits a request to transmit candidate information input by the operator O to the server device 10 together with the specified execution environment, analysis category, etc.

[0150] (3-3. Profile information loading process) Third, the server device 10 executes a profile information reading process (step S103). For example, the server device 10 reads the profile information (e.g., output target, input signal, calculation amount, memory, executable file, output signal) attached to the saved model data of the AI ​​logic.

[0151] (3-4. Candidate information generation process) Fourth, the server device 10 executes a candidate information generation process (step S104). For example, the server device 10 extracts AI logics that can be executed in the execution environment requested by the operator O based on the profile information and resource information, and generates candidate information that includes identification information of the extracted AI logics.

[0152] (3-5. Candidate information transmission process) Fifth, the server device 10 executes a candidate information transmission process (step S105). For example, the server device 10 transmits, to the operator terminal 20, candidate information including identification information of AI logics that can be executed in the requested execution environment.

[0153] (3-6. Candidate information display process) Sixth, the operator terminal 20 executes a candidate information display process (step S106). For example, the operator terminal 20 displays on the monitor an AI logic selection screen that presents a list of identification information of AI logics as candidate information.

[0154] (3-7. Selection result input processing) Seventh, the operator O executes a selection result input process (step S107). For example, the operator O selects an AI logic from a list of identification information of the AI ​​logic, thereby inputting the selection of the AI ​​logic to be used for analysis as the selection result.

[0155] (3-8. Selection result sending process) Eighth, the operator terminal 20 executes a selection result transmission process (step S108). For example, the operator terminal 20 transmits the selection result, including the identification information of the AI ​​logic input by the operator O to the operator terminal 20, to the server device 10. In addition, the server device 10 transmits the received selection result, as well as the model data of the AI ​​logic, the profile information of the AI ​​logic, the output information, and the output frequency to the AI ​​execution device 30.

[0156] (3-9. AI logic setting process) Ninth, the AI ​​execution device 30 executes the AI ​​logic setting process (step S109). For example, if the operator O has designated the AI ​​execution device 30 as the execution environment, the AI ​​execution device 30 installs the AI ​​logic application on the resources of the AI ​​execution device 30. The AI ​​execution device 30 may also refer to the profile information of the AI ​​logic for which the application is to be installed, and adjust the input and output signals.

[0157] (3-10. Monitoring data collection processing) Tenth, the AI ​​execution device 30 executes a monitoring data collection process (step S110). For example, the AI ​​execution device 30 collects monitoring data acquired by monitoring devices 40 installed at monitored locations.

[0158] (3-11. Monitoring data analysis processing) Eleventh, the AI ​​execution device 30 executes a monitoring data analysis process (step S111). For example, the AI ​​execution device 30 uses the AI ​​logic set in the execution environment to analyze the monitoring data and detect events that have occurred at the monitored base.

[0159] 4. Effects of the embodiment Finally, the effects of the embodiment will be described below: Effects 1 to 9 corresponding to the processing according to the embodiment will be described below.

[0160] (4-1. Effect 1) First, in the processing according to the above-described embodiment, the server device 10 stores multiple analysis models AM along with profile information indicating the execution conditions of each of the multiple analysis models AM, and when a request from an operator O is received, extracts, based on the profile information, an analysis model AM that is executable in the execution environment requested by the operator O from among the multiple analysis models AM. Therefore, in this processing, by adding profile information to the AI ​​logic, the user can easily select the machine learning program they need.

[0161] (4-2. Effect 2) Second, in the processing according to the embodiment described above, the server device 10 receives a selection by the operator O from among the extracted analysis models AM, and sets the analysis model AM whose selection has been received in the execution environment requested by the operator O. Therefore, in this processing, by automatically installing the AI ​​logic application in the execution environment requested by the operator O, and by assigning profile information to the AI ​​logic, the user can easily select and set the machine learning program they require.

[0162] (4-3. Effect 3) Third, in the processing according to the above-described embodiment, the AI ​​execution device 30 converts, based on the profile information, input signals acquired in the execution environment into a format that can be input to the selected analysis model AM, and configures the AI ​​execution device 30 to convert the output signals of the selected analysis model AM into a format requested by the operator O. Therefore, in this processing, by automatically adjusting input and output signals when installing an AI logic application in the execution environment requested by the operator O, and by assigning profile information to the AI ​​logic, the user can easily select and set the machine learning program they need.

[0163] (4-4. Effect 4) Fourth, in the processing according to the above-described embodiment, the AI ​​execution device 30 is configured to convert the output signal of a selected analysis model AM into an input signal that can be input to another selected analysis model AM, based on the profile information. Therefore, in this processing, when multiple AI logic applications are serially connected and installed in the execution environment requested by the operator O, the input / output signals are automatically adjusted, and by assigning profile information to the AI ​​logic, the user can easily select and set the machine learning program they need.

[0164] (4-5. Effect 5) Fifth, in the processing according to the above-described embodiment, the AI ​​execution device 30 is configured to convert, based on the profile information, input signals acquired in the execution environment into a format that can be input to both the selected analysis model AM and other selected analysis models AM. Therefore, in this processing, when multiple AI logic applications are connected in parallel and installed in the execution environment requested by the operator O, input / output signals are automatically adjusted, and by assigning profile information to the AI ​​logic, the user can easily select and set the machine learning program they need.

[0165] (4-6. Effect 6) Sixth, in the process according to the above-described embodiment, the profile information includes at least one of information specifying the output target, input signal, computational complexity, memory, executable file, and output signal. Therefore, in this process, by adding information in a common format to the AI ​​logic, the profile information is added to the AI ​​logic, allowing the user to easily select and set the machine learning program they need.

[0166] (4-7. Effect 7) Seventh, in the process according to the embodiment described above, the AI ​​execution device 30 inputs the monitoring data collected from the monitoring devices 40 installed at the monitored locations into the machine learning model whose selection has been accepted, and detects the output analysis target as an event. Therefore, in this process, by adding profile information to the AI ​​logic, the user can easily select and set the machine learning program they need, and can effectively detect events that occur.

[0167] (4-8. Effect 8) Eighth, in the process according to the above-described embodiment, the monitoring data is image data acquired by camera 40A installed at the monitored location. Therefore, in this process, by adding profile information to the AI ​​logic, the user can easily select and set the machine learning program they need, and can effectively detect events that have occurred based on image analysis.

[0168] (4-9. Effect 9) Ninth, in the process according to the above-described embodiment, the monitoring data is measurement data acquired by a sensor 40B installed at a monitored site. Therefore, in this process, by adding profile information to the AI ​​logic, the user can easily select and set the machine learning program they need, and can effectively detect events that have occurred based on sensor analysis.

[0169] [5. System] The information including the processing procedures, control procedures, specific names, various data and parameters shown in the above documents and drawings can be changed arbitrarily unless otherwise specified.

[0170] Furthermore, the components of each device shown in the figure are functional concepts and do not necessarily have to be physically configured as shown. In other words, the specific form of distribution and integration of each device is not limited to that shown. In other words, all or part of them can be functionally or physically distributed and integrated in any unit depending on various loads, usage conditions, etc.

[0171] Furthermore, all or any part of the processing functions performed by each device may be realized by a CPU and a program analyzed and executed by the CPU, or may be realized as hardware using wired logic.

[0172] [6. Hardware] Next, an example of the hardware configuration of a server device 10, which is an information processing device, will be described. Note that other devices may also have a similar hardware configuration. FIG. 14 is a diagram showing an example of the hardware configuration according to an embodiment. As shown in FIG. 14, the server device 10 includes a communication device 10a, an HDD (Hard Disk Drive) 10b, a memory 10c, and a processor 10d. The components shown in FIG. 14 are connected to each other via a bus or the like.

[0173] The communication device 10a is a network interface card or the like, and communicates with other servers. The HDD 10b stores programs and databases that operate the functions shown in FIG.

[0174] The processor 10d reads out a program that executes the same processes as the respective processing units shown in FIG. 5 from the HDD 10b or the like and loads it into the memory 10c, thereby operating a process that executes the respective functions described in FIG. 5 or the like. For example, this process executes the same functions as the respective processing units of the server device 10. Specifically, the processor 10d reads out a program having the same functions as the reception unit 15a, extraction unit 15b, setting unit 15c, or the like from the HDD 10b or the like. Then, the processor 10d executes a process that executes the same processes as the reception unit 15a, extraction unit 15b, setting unit 15c, or the like.

[0175] In this way, the server device 10 operates as a device that executes various processing methods by reading and executing a program. The server device 10 can also realize functions similar to those of the above-described embodiment by reading the program from a recording medium using a media reader and executing the read program. Note that the program in these other embodiments is not limited to being executed by the server device 10. For example, the present invention can also be applied in the same way to cases where another computer or server executes the program, or where these execute the program in cooperation with each other.

[0176] This program can be distributed via a network such as the Internet. In addition, this program can be recorded on a computer-readable recording medium such as a hard disk, a flexible disk (FD), a CD-ROM, a magneto-optical disk (MO), or a digital versatile disk (DVD), and can be executed by being read from the recording medium by a computer.

[0177] [7. Other] Some examples of combinations of the disclosed technical features are set out below.

[0178] (1) An information processing device comprising: a memory unit that stores multiple machine learning models along with profile information indicating the execution conditions of each of the multiple machine learning models; and an extraction unit that, when a user request is received, extracts a machine learning model from the multiple machine learning models that is executable in the execution environment requested by the user based on the profile information.

[0179] (2) The information processing device described in (1) further includes a reception unit that receives the user's selection from among the extracted machine learning models, and a setting unit that sets the machine learning model whose selection is received in the execution environment requested by the user.

[0180] (3) The information processing device described in (2), wherein the setting unit converts the input signal acquired in the execution environment into a form that can be input to the selected machine learning model based on the profile information, and configures the setting unit to convert the output signal of the selected machine learning model into a form requested by the user.

[0181] (4) The information processing device described in (2) or (3), wherein the setting unit is configured to convert the output signal of the selected machine learning model into an input signal that can be input to another selected machine learning model based on the profile information.

[0182] (5) An information processing device described in any one of (2) to (4), wherein the setting unit is configured to convert an input signal acquired in the execution environment into a form that can be input to both the selected machine learning model and another selected machine learning model based on the profile information.

[0183] (6) The information processing device according to any one of (1) to (5), wherein the profile information includes at least one of information defining an output target, an input signal, a calculation amount, a memory, an executable file, and an output signal.

[0184] (7) An information processing device described in any one of (2) to (6), which inputs monitoring data collected from monitoring equipment installed at a monitored site into the machine learning model whose selection has been accepted, and sets the machine learning model in an execution device that detects the output analysis target as an event.

[0185] (8) The information processing device according to (7), wherein the monitoring data is image data acquired by photographing equipment installed at the monitored location.

[0186] (9) The information processing device according to (7) or (8), wherein the monitoring data is measurement data acquired by a measuring device installed at the monitoring target site.

[0187] (10) An information processing method in which, when a computer receives a user's request, it extracts a machine learning model from among the multiple machine learning models that is executable in the execution environment requested by the user, based on profile information indicating the execution conditions of each of the multiple machine learning models stored together with the multiple machine learning models.

[0188] (11) An information processing program that causes a computer to execute a process in which, when a user's request is received, a machine learning model that can be executed in the execution environment requested by the user is extracted from among the multiple machine learning models based on profile information that indicates the execution conditions of each of the multiple machine learning models and is stored together with the multiple machine learning models. [Explanation of symbols]

[0189] 10 Server device 11 Input section 12 Output section 13 Communications Department 14 Storage section 14a Analysis model storage section 14b Candidate information storage unit 15 Control Unit 15a Reception 15b Extraction part 15c Setting section 20 Operator terminal 21 Input / output section 22 Transmitter / Receiver 23 Communications Department 30 AI execution device 31 Communications Department 32 Storage section 32a Analysis model memory section 32b Selection result storage unit 32c Monitoring data storage unit 33 Control Unit 33a Transmitter / Receiver 33b Executive Department 40 Surveillance equipment 40A Camera 40B Sensor 100 Surveillance System

Claims

1. a storage unit that stores a plurality of machine learning models and profile information indicating execution conditions for each of the plurality of machine learning models; an extraction unit that, when a user request is received, extracts, from the plurality of machine learning models, a machine learning model that is executable in the execution environment requested by the user, based on the profile information; and An information processing device comprising:

2. a reception unit that receives a selection by the user from among the extracted machine learning models; a setting unit that sets the machine learning model whose selection has been accepted in the execution environment requested by the user; The information processing device according to claim 1 , further comprising:

3. The setting unit converting an input signal acquired in the execution environment into a form that can be input to the selected machine learning model based on the profile information, and configuring the system so that an output signal of the selected machine learning model is converted into a form requested by the user; The information processing device according to claim 2 .

4. The setting unit and configuring the machine learning system to convert an output signal of the selected machine learning model into an input signal in a format that can be input to another selected machine learning model based on the profile information. The information processing device according to claim 2 .

5. The setting unit and configuring the execution environment to convert, based on the profile information, an input signal acquired in the execution environment into a form that can be input to both the selected machine learning model and another selected machine learning model. The information processing device according to claim 2 .

6. The profile information includes at least one of information defining an output target, an input signal, a calculation amount, a memory, an executable file, and an output signal.

6. The information processing device according to claim 1.

7. The setting unit inputting the monitoring data collected from the monitoring devices installed at the monitored site into the machine learning model whose selection has been accepted, and causing an execution device that detects the output analysis target as an event to set the machine learning model; The information processing device according to claim 2 .

8. The monitoring data is image data acquired by a photographing device installed at the monitoring target site. The information processing device according to claim 7 .

9. The monitoring data is measurement data acquired by a measuring device installed at the monitoring target site. The information processing device according to claim 7 .

10. The computer When a request from a user is received, extracting a machine learning model that is executable in the execution environment requested by the user from among the plurality of machine learning models based on profile information that indicates execution conditions for each of the plurality of machine learning models and that is stored together with the plurality of machine learning models; An information processing method that performs processing.

11. On the computer, When a request from a user is received, extracting a machine learning model that is executable in the execution environment requested by the user from among the plurality of machine learning models based on profile information that indicates execution conditions for each of the plurality of machine learning models and that is stored together with the plurality of machine learning models; An information processing program that executes processing.

Images