Method and device for evaluating interaction strength of virtual objects

Through feature encoding of virtual object attribute data and multi-layer output layer superposition processing, the target probability distribution of virtual objects is generated, which solves the accuracy of virtual object interaction intensity evaluation in the prior art, and improves the authenticity and complex data of the evaluation results. Fitting ability.

CN113886768BActive Publication Date: 2024-11-05SHENZHEN TENCENT INFORMATION TECH CO LTD
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
CN202111172374.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-10-08
Publication Date
2024-11-05
Estimated Expiration
2041-10-08

AI Technical Summary

Technical Problem

The existing target numerical evaluation scheme cannot accurately evaluate the interaction intensity of virtual objects in virtual scenarios, and the prediction results do not match the actual application, so it cannot adapt to the complex interaction of virtual objects in different scenarios.

Method used

By obtaining the attribute data of the virtual object, using the interaction intensity evaluation model for feature encoding, using at least two sub-output layers to predict the interaction intensity separately, and superimposing the probability distribution of each sub-output layer through the second sub-output layer to generate the virtual object Target probability distribution to improve assessment accuracy.

Benefits of technology

It improves the accuracy of the evaluation of virtual objects' interaction intensity, makes the evaluation results more in line with the real data in actual applications, and provides the target numerical and probability distribution of the interaction intensity of virtual objects.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN113886768B_ABST
    Figure CN113886768B_ABST
Patent Text Reader

Abstract

The present application provides a method, device, equipment, computer-readable storage medium and computer program product for evaluating the interaction strength of virtual objects, including: obtaining the attribute data of virtual objects in a virtual scene; feature encoding the attribute data of virtual objects through the feature coding layer of the interaction strength evaluation model to obtain corresponding feature vectors; based on the feature vectors, respectively predicting the interaction strength of virtual objects through at least two first sub-output layers included in the interaction strength evaluation model to obtain the probability distribution of the interaction strength of virtual objects; superimposing the probability distribution of the interaction strength output by each first sub-output layer through the second sub-output layer of the interaction strength evaluation model to obtain the target probability distribution of the interaction strength of virtual objects. Through the present application, the probability distribution of the interaction strength of virtual objects can be predicted, which is more in line with the real data in practical applications and improves the accuracy of the evaluation results.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to Internet technology, and in particular to a method, device, electronic device, computer-readable storage medium and computer program product for evaluating the interaction strength of a virtual object. Background Art

[0002] In the related art, when evaluating the interactive strength of props or characters in the game, the attribute values ​​of new props or characters are used as the feature data of the model input, the model is trained, and a single predicted target value is output as the standard for evaluating the interactive strength of props or characters. However, in actual applications, taking the prediction of the time required for a specific car to complete a certain track in the game as an example, the time spent by players of a certain level in different game matches is not a single fixed value. The time required is sometimes more and sometimes less. That is, for the same props or character attribute values, the corresponding interactive strength value is not unique, and the method of predicting the target value in the related art has low accuracy. Summary of the invention

[0003] The embodiments of the present application provide a method, apparatus, device, computer-readable storage medium and computer program product for evaluating the interaction strength of virtual objects, which can predict the probability distribution of the interaction strength of virtual objects, are more consistent with real data in practical applications, and improve the accuracy of the evaluation results.

[0004] The technical solution of the embodiment of the present application is implemented as follows:

[0005] The present application embodiment provides a method for evaluating the interaction strength of a virtual object, including:

[0006] Obtaining attribute data of virtual objects in the virtual scene;

[0007] Through the feature coding layer of the interaction strength evaluation model, feature coding is performed on the attribute data of the virtual object to obtain a corresponding feature vector;

[0008] Based on the feature vector, the interaction intensity of the virtual object is predicted respectively through at least two first sub-output layers included in the interaction intensity evaluation model to obtain a probability distribution of the interaction intensity of the virtual object;

[0009] The probability distributions of the interaction strength output by each of the first sub-output layers are superimposed through the second sub-output layer of the interaction strength evaluation model to obtain a target probability distribution of the interaction strength of the virtual object.

[0010] An embodiment of the present application provides a device for evaluating the interaction strength of a virtual object, including:

[0011] An acquisition module, used to acquire attribute data of virtual objects in a virtual scene;

[0012] A feature encoding module, used to perform feature encoding on the attribute data of the virtual object through a feature encoding layer of an interaction strength evaluation model to obtain a corresponding feature vector;

[0013] The first output module is used to predict the interaction strength of the virtual object based on the feature vector through at least two first sub-output layers included in the interaction strength evaluation model, so as to obtain a probability distribution of the interaction strength of the virtual object.

[0014] The second output module is used to superimpose the probability distributions of the interaction strength output by each of the first sub-output layers through the second sub-output layer of the interaction strength evaluation model to obtain a target probability distribution of the interaction strength of the virtual object.

[0015] In the above scheme, the first output module is also used to perform the following processing through each first sub-output layer of the interaction strength evaluation model: obtaining the mapping relationship between the probability distribution of the interaction strength and the characteristic vector of the attribute data; determining the probability distribution of the interaction strength of the virtual object based on the characteristic vector of the attribute data of the virtual object and the mapping relationship.

[0016] In the above scheme, the first output module is also used to obtain a first mapping relationship between the characteristic vector of the attribute data and the interaction intensity parameter, wherein the interaction intensity parameter includes the mean of the interaction intensity and the standard deviation of the interaction intensity; determine the second mapping relationship between the probability distribution of the interaction intensity and the interaction intensity parameter; based on the first mapping relationship and the second mapping relationship, determine the mapping relationship between the probability distribution of the interaction intensity and the characteristic vector of the attribute data.

[0017] In the above scheme, the second output module is also used to obtain the weights corresponding to each of the first sub-output layers; based on the obtained weights, the probability distribution of the interaction intensity output by each of the first sub-output layers is weighted summed to obtain the target probability distribution of the interaction intensity of the virtual object.

[0018] In the above scheme, the second output module is also used to perform weighted summation processing on the mapping relationships output by each of the first sub-output layers based on the acquired weights when the probability distribution of the interaction intensity is represented by a mapping relationship between probability and interaction intensity, so as to obtain the target probability distribution of the interaction intensity of the virtual object.

[0019] In the above scheme, the device also includes a training module, which is used to obtain training sample data carrying labels; wherein the training sample data is attribute data of a virtual object in a virtual scene, and the label is the real interaction strength of the virtual object; the training sample data is feature encoded through the feature encoding layer of the interaction strength assessment model to obtain a corresponding sample feature vector; based on the sample feature vector, the interaction strength of the virtual object is predicted respectively through at least two first sub-output layers included in the interaction strength assessment model to obtain a first probability distribution of the interaction strength of the virtual object; through the second sub-output layer of the interaction strength assessment model, the first probability distribution of the interaction strength output by each first sub-output layer is superimposed to obtain a second probability distribution of the interaction strength of the virtual object; based on the second probability distribution, the interaction strength of the virtual object when the probability is maximum is determined; the difference between the determined interaction strength and the real interaction strength is obtained, and the model parameters of the interaction strength assessment model are updated based on the difference.

[0020] In the above scheme, the second output module is also used to output the target probability distribution of the interaction intensity of the virtual object in at least one of the following ways: a mapping relationship between probability and interaction intensity, a probability distribution diagram of interaction intensity, and a plurality of interaction intensity probability pairs; wherein the interaction intensity probability pair includes: the interaction intensity of the virtual object and the corresponding probability.

[0021] In the above scheme, the virtual object includes a vehicle in a virtual scene, the interaction intensity includes the time taken by the vehicle to travel the target distance, and the first output module is further used to predict the time taken by the vehicle to travel the target distance based on the feature vector through at least two first sub-output layers included in the interaction intensity evaluation model to obtain the probability distribution of the time; the second output module is further used to superimpose the probability distribution of the time output by each of the first sub-output layers through the second sub-output layer of the interaction intensity evaluation model to obtain the target probability distribution of the time taken by the vehicle to travel the target distance.

[0022] In the above scheme, the device also includes an integration module, which is used to obtain a target probability distribution of the interaction intensity of the virtual object; integrate the target probability distribution based on a target evaluation interval of the interaction intensity; and use the result of the integration as the probability that the interaction intensity belongs to the target evaluation interval.

[0023] An embodiment of the present application provides an electronic device, including:

[0024] A memory for storing executable instructions;

[0025] The processor is used to implement the method for evaluating the interaction strength of a virtual object provided in an embodiment of the present application when executing the executable instructions stored in the memory.

[0026] An embodiment of the present application provides a computer-readable storage medium storing executable instructions for causing a processor to execute instructions to implement the method for evaluating the interactive strength of a virtual object provided in the embodiment of the present application.

[0027] An embodiment of the present application provides a computer program product, including a computer program. When the computer program is executed by a processor, the method for evaluating the interaction strength of a virtual object provided in the embodiment of the present application is implemented.

[0028] The embodiments of the present application have the following beneficial effects:

[0029] When evaluating the interaction strength of virtual objects in a virtual scene, the acquired attribute data of the virtual objects is predicted, and after obtaining at least two probability distributions of the interaction strength of the virtual objects, the at least two probability distributions are superimposed to obtain a target probability distribution of the interaction strength of the virtual objects. In this way, by obtaining the target probability distribution of the interaction strength of the virtual objects, the evaluation results are more consistent with the real data in actual applications, thereby improving the accuracy of the evaluation results. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 A schematic diagram of the structure of a virtual object interaction strength evaluation system 100 provided in an embodiment of the present application;

[0031] Figure 2 is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application;

[0032] Figure 3 is a flow chart of a method for evaluating the interactive strength of a virtual object provided in an embodiment of the present application;

[0033] Figure 4A This is an optional structural diagram of a feature coding layer provided in an embodiment of the present application;

[0034] Figure 4B This is an optional structural diagram of a feature coding layer provided in an embodiment of the present application;

[0035] Figure 5A is an optional flowchart of a feature vector determination process provided in an embodiment of the present application;

[0036] Figure 5B is an optional flowchart of a feature vector determination process provided in an embodiment of the present application;

[0037] Figure 6is an optional structural diagram of the output layer provided in an embodiment of the present application;

[0038] Figure 7 It is an optional flow chart of obtaining the mapping relationship between the probability distribution of interaction intensity and the feature vector of attribute data provided in the embodiment of the present application;

[0039] Figure 8 is a schematic diagram of the probability distribution of the interaction intensity of virtual objects provided in an embodiment of the present application;

[0040] Fig. 9 is a schematic diagram of a target probability distribution of interaction intensity of a virtual object provided in an embodiment of the present application;

[0041] Fig.10 It is an optional flow chart of the interaction intensity evaluation method provided in the embodiment of the present application;

[0042] Fig.11 is a schematic diagram of target probability distribution of interaction intensity of virtual objects provided in an embodiment of the present application;

[0043] Fig.12 It is an optional flow chart of the interaction intensity assessment model training method provided in the embodiment of the present application;

[0044] Fig.13 is a schematic diagram of training sample data provided in an embodiment of the present application;

[0045] Fig.14 It is an optional structural diagram of the interaction intensity evaluation model provided in the embodiment of the present application;

[0046] Fig.15 It is a schematic diagram of the overall structure of the interaction intensity evaluation model provided in the embodiment of the present application;

[0047] Fig.16 It is an optional flow chart of a method for evaluating the interaction strength of a virtual object provided in an embodiment of the present application;

[0048] Fig.17 It is a comparative diagram of the target probability distribution and the actual distribution statistics provided in the embodiment of the present application. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of this application.

[0050] In the following description, reference is made to “some embodiments”, which describe a subset of all possible embodiments, but it can be understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0051] In the following description, the terms "first\second\third" are only used to distinguish similar objects, and do not represent a specific order for the objects. It is understandable that "first\second\third" can be interchanged with a specific order or sequence where permitted, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein. In the following description, the term "plurality" refers to at least two.

[0052] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0053] Before further describing the embodiments of the present application in detail, the nouns and terms involved in the embodiments of the present application are explained. The nouns and terms involved in the embodiments of the present application are subject to the following interpretations.

[0054] 1) Probability distribution: It is used to describe the probability law of the value of a random variable, indicating the possibility of an event occurring in an experiment.

[0055] 2) Gaussian distribution: Also known as normal distribution, its probability density function is a bell-shaped curve, that is, most of the data is concentrated at the mean value, and a small part is at both ends.

[0056] 3) Standardization: that is, through certain mathematical transformation methods, the original data is converted according to a certain ratio so that it falls into a small specific interval, such as the interval of [0, 1] or [-1, 1], eliminating the differences in the properties, dimensions, orders of magnitude and other characteristic attributes between different variables, and converting it into a dimensionless relative value, that is, a standardized value, so that the values ​​of each indicator are at the same order of magnitude, which facilitates the comprehensive analysis and comparison of indicators of different units or orders of magnitude.

[0057] During the implementation of the embodiments of the present application, the applicant discovered that the related technology has the following problems:

[0058] The existing target value evaluation scheme is mainly based on a large number of live network data sets. The data sets contain a large number of numerical attribute and target value pairs (that is, a set of data containing multiple attribute values ​​corresponds to a target value). Then, a machine learning algorithm is used to train the model, and the trained model is used to predict the target value. Its main features are: the attribute value data set is used as feature data X, the target value data set is used as target data Y, and regression algorithms such as logistic regression, linear regression, and Lasso regression are used for model training. In practical applications, the numerical attribute x of the virtual object in the virtual scene is used as the model input to predict the target value y. The predicted target value y is equivalent to the evaluation standard of the mutual strength of virtual objects in the virtual scene.

[0059] For existing target numerical evaluation schemes, firstly, the traditional machine learning algorithm model used is relatively simple, equivalent to a single hidden layer neural network, which is difficult to fit complex data; at the same time, for a specific set of numerical attributes x, only a single target value y can be predicted. In practical applications, the target value y usually presents a certain probability distribution, such as {p1: y1, p1: y1, ..., p n :y n}, where y n For a specific event, p n For the appearance of y n Therefore, the existing target numerical evaluation scheme is not consistent with practical applications and cannot make an accurate evaluation of the interaction strength of virtual objects in virtual scenes.

[0060] The embodiments of the present application provide a method, device, electronic device, computer-readable storage medium and computer program product for evaluating the interaction strength of virtual objects, which increase the depth of the model. Compared with traditional machine learning algorithms, the model has better fitting capabilities for complex data. At the same time, by outputting the probability distribution of the evaluation results of the interaction strength, it is also possible to predict the target value of the virtual object interaction strength and the probability distribution of the target value, which is more in line with real data in practical applications and improves the accuracy of the evaluation results.

[0061] The following describes an exemplary application of the device for evaluating the interaction strength of virtual objects provided in an embodiment of the present application. The device for evaluating the interaction strength of virtual objects provided in an embodiment of the present application can be implemented as various types of user terminals such as laptop computers, tablet computers, desktop computers, set-top boxes, mobile devices (for example, mobile phones, portable music players, personal digital assistants, dedicated messaging devices, portable gaming devices), and can also be implemented as a server.

[0062] Based on the above explanations of the nouns and terms involved in the embodiments of the present application, the following describes the evaluation system for the interactive strength of virtual objects provided by the embodiments of the present application. Figure 1 , Figure 11 is a schematic diagram of the structure of the evaluation system 100 for the interaction strength of virtual objects provided in an embodiment of the present application. To support an exemplary application, an interaction strength evaluation client 401 is provided on the terminal 400, and the terminal 400 is connected to the server 200 via the network 300, wherein the network 300 may be a wide area network or a local area network, or a combination of the two, and a wireless or wired link is used to realize data transmission.

[0063] The terminal 400 is used to, in response to an input operation on the attribute data of the virtual object in the virtual scene, obtain the attribute data of the virtual object in the virtual scene, and in response to an interaction strength evaluation instruction for the virtual object in the virtual scene, send the attribute data of the virtual object to the server 200;

[0064] The server 200 is used to perform feature encoding on the attribute data of the virtual object through the feature encoding layer of the interaction strength evaluation model to obtain a corresponding feature vector; based on the feature vector, the interaction strength of the virtual object is predicted through at least two first sub-output layers included in the interaction strength evaluation model to obtain a probability distribution of the interaction strength of the virtual object; through the second sub-output layer of the interaction strength evaluation model, the probability distribution of the interaction strength output by each first sub-output layer is superimposed to obtain a target probability distribution of the interaction strength of the virtual object;

[0065] The terminal 400 is further used to receive and output the target probability distribution of the interaction strength of the virtual object sent by the server 200.

[0066] In some embodiments, the server 200 may be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The terminal 400 may be a smart phone, a tablet computer, a laptop computer, a desktop computer, a set-top box, and a mobile device (e.g., a mobile phone, a portable music player, a personal digital assistant, a dedicated messaging device, a portable gaming device, an intelligent speaker, and a smart watch), etc., but is not limited thereto. The terminal device and the server may be directly or indirectly connected by wired or wireless communication, which is not limited in the embodiments of the present application.

[0067] Next, an electronic device that implements the method for evaluating the interactive strength of a virtual object provided in an embodiment of the present application is described. In practical applications, the electronic device can be Figure 1 The server 200 or terminal 400 shown, see Figure 2 , Figure 2 is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application, Figure 2 The electronic device shown includes: at least one processor 410, a memory 450, at least one network interface 420 and a user interface 430. The various components in the terminal 400 are coupled together via a bus system 440. It is understood that the bus system 440 is used to achieve connection and communication between these components. In addition to the data bus, the bus system 440 also includes a power bus, a control bus and a status signal bus. However, for the sake of clarity, the bus system 440 is not shown in FIG. Figure 2 Various buses are labeled as bus system 440 .

[0068] Processor 410 can be an integrated circuit chip with signal processing capabilities, such as a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc., where the general-purpose processor can be a microprocessor or any conventional processor, etc.

[0069] The user interface 430 includes one or more output devices 431 that enable presentation of media content, including one or more speakers and / or one or more visual display screens. The user interface 430 also includes one or more input devices 432, including user interface components that facilitate user input, such as a keyboard, mouse, microphone, touch screen display, camera, other input buttons and controls.

[0070] The memory 450 may be removable, non-removable, or a combination thereof. Exemplary hardware devices include solid-state memory, hard disk drives, optical disk drives, etc. The memory 450 may optionally include one or more storage devices that are physically remote from the processor 410.

[0071] The memory 450 includes a volatile memory or a non-volatile memory, and may also include both volatile and non-volatile memories. The non-volatile memory may be a read-only memory (ROM), and the volatile memory may be a random access memory (RAM). The memory 450 described in the embodiments of the present application is intended to include any suitable type of memory.

[0072] In some embodiments, memory 450 can store data to support various operations, examples of which include programs, modules, and data structures, or a subset or superset thereof, as exemplarily described below.

[0073] Operating system 451, including system programs for processing various basic system services and performing hardware-related tasks, such as a framework layer, a core library layer, a driver layer, etc., for implementing various basic services and processing hardware-based tasks;

[0074] A network communication module 452, for reaching other computing devices via one or more (wired or wireless) network interfaces 420, exemplary network interfaces 420 include: Bluetooth, Wireless Compatibility Authentication (WiFi), and Universal Serial Bus (USB);

[0075] a presentation module 453 for enabling presentation of information via one or more output devices 431 (e.g., display screen, speaker, etc.) associated with the user interface 430 (e.g., a user interface for operating peripherals and displaying content and information);

[0076] The input processing module 454 is used to detect one or more user inputs or interactions from one of the one or more input devices 432 and translate the detected inputs or interactions.

[0077] In some embodiments, the device provided in the embodiments of the present application can be implemented in software. Figure 2 The interactive strength evaluation device 455 of the virtual object stored in the memory 450 terminal is shown, which can be software in the form of a program and a plug-in, including the following software modules: an acquisition module 4551, a feature encoding module 4552, a first output module 4553 and a second output module 4554. These modules are logical, so they can be arbitrarily combined or further split according to the functions implemented. The functions of each module will be explained below.

[0078] In other embodiments, the device provided in the embodiments of the present application can be implemented in hardware. As an example, the evaluation device for the interaction strength of a virtual object provided in the embodiments of the present application can be a processor in the form of a hardware decoding processor, which is programmed to execute the image processing method in video editing provided in the embodiments of the present application. For example, the processor in the form of a hardware decoding processor can adopt one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs) or other electronic components.

[0079] In some embodiments, the terminal or server can implement the evaluation method of the interactive intensity of the virtual object provided in the embodiment of the present application by running a computer program. For example, the computer program can be a native program or software module in the operating system; it can be a native application (APP, Application), that is, a program that needs to be installed in the operating system to run, such as an instant messaging APP, a web browser APP; it can also be a small program, that is, a program that only needs to be downloaded to the browser environment to run; it can also be a small program that can be embedded in any APP. In short, the above-mentioned computer program can be an application, module or plug-in in any form.

[0080] Based on the above description of the evaluation system and electronic device for the interaction strength of virtual objects provided by the embodiment of the present application, the following describes the evaluation method for the interaction strength of virtual objects provided by the embodiment of the present application. In actual implementation, the evaluation method for the interaction strength of virtual objects provided by the embodiment of the present application can be implemented by a terminal or a server alone, or by a terminal and a server in collaboration, so that Figure 1 The server 200 in the embodiment of the present invention independently executes the evaluation method of the interaction strength of the virtual object provided by the embodiment of the present invention as an example for explanation. Figure 3 , Figure 3 is a flow chart of a method for evaluating the interactive strength of a virtual object provided in an embodiment of the present application, which will be combined with Figure 3 The steps shown are explained.

[0081] Step 101: The server obtains attribute data of a virtual object in a virtual scene.

[0082] Here, the virtual scene can be different types of games, such as dance competition games, racing competition games, or confrontation games, and the virtual objects can be virtual props or virtual characters in different types of games. As for the attribute data of the virtual objects, for example, when the virtual object is a virtual prop (such as a vehicle) in a racing competition game, the virtual data can be the speed, weight, gas collection speed, small spray, large spray, drifting, collision and other attribute values ​​of the racing car, and it can also be the player's racing level; when the virtual object is a virtual character in a confrontation game, the virtual data can be the player's level, the cooling time of the character's skills, the operation order of the skills and other attribute values.

[0083] Step 102 , feature encoding is performed on the attribute data of the virtual object through the feature encoding layer of the interaction strength evaluation model to obtain a corresponding feature vector.

[0084] Here, there are two different structures of the feature coding layer. For the two different structures of the feature coding layer, the process of feature encoding the attribute data of the virtual object and obtaining the corresponding feature vector is also different.

[0085] In some embodiments, see Figure 4A , Figure 4A It is an optional structural diagram of the feature coding layer provided in an embodiment of the present application. In actual implementation, the acquired attribute data of at least one attribute dimension is input into the feature coding layer for processing. The feature coding layer performs feature coding on the attribute data of different attribute dimensions one by one to obtain corresponding feature vectors.

[0086] As an example, see Figure 5A , Figure 5A : is an optional schematic diagram of a feature vector determination process provided by an embodiment of the present application. When the virtual scene is a racing game and the virtual object is a racing car in the game, the attribute data of at least two attribute dimensions obtained include the player's racing level, the racing car's flat running speed, the racing car's weight, the racing car's gas collection speed, the racing car's small spray attribute, the racing car's large spray attribute, the racing car's drifting attribute, and the racing car's collision attribute, etc. The attribute data of each attribute dimension is feature encoded one by one through a feature encoding layer to obtain feature data X1, X2, X3, ..., X corresponding to each attribute dimension. n , then based on the feature data corresponding to each attribute dimension, determine the feature vector [X1, X2, X3, …, X n ].

[0087] In other embodiments, see Figure 4B , Figure 4B This is an optional structural diagram of a feature coding layer provided in an embodiment of the present application. Here, the feature coding layer includes at least one coding sublayer. In actual implementation, feature coding is performed on the acquired attribute data of at least one attribute dimension through at least one coding sublayer. Here, each coding sublayer processes the attribute data of one attribute dimension, that is, each coding sublayer performs feature coding on the attribute data of the corresponding attribute dimension to obtain a corresponding feature vector.

[0088] As an example, see Figure 5B , Figure 5B : is an optional schematic diagram of a feature vector determination process provided by an embodiment of the present application. When the virtual scene is a racing game and the virtual object is a racing car in the game, the attribute data of at least two attribute dimensions obtained include the player's racing level, the racing car's flat running speed, the racing car's weight, the racing car's gas collection speed, the racing car's small spray attribute, the racing car's large spray attribute, the racing car's drifting attribute, and the racing car's collision attribute, etc. The attribute data of one attribute dimension is feature-encoded by each encoding sublayer included in the feature encoding layer to obtain feature data X1, X2, X3, ..., X corresponding to each attribute dimension. n, then based on the feature data corresponding to each attribute dimension, determine the feature vector [X1, X2, X3, …, X n ].

[0089] It should be noted that before the feature encoding layer of the interaction intensity assessment model performs feature encoding on the acquired attribute data and obtains the corresponding feature vector, it is necessary to first standardize the attribute data to the target interval, such as using the Min-Max normalization method to standardize it to the interval of [-1, 1].

[0090] In this way, by standardizing the attribute data, the impact of different attributes between different data can be eliminated, making the results more comparable.

[0091] Step 103 , based on the feature vector, the interaction strength of the virtual object is predicted respectively through at least two first sub-output layers included in the interaction strength evaluation model to obtain a probability distribution of the interaction strength of the virtual object.

[0092] Here, see Figure 6 , Figure 6 It is an optional structural diagram of the output layer provided in an embodiment of the present application, where the output layer includes a softmax layer and a second sub-output layer, wherein the softmax layer includes at least two first sub-output layers.

[0093] In actual implementation, for each first sub-output layer included in the interaction strength evaluation model, the mapping relationship between the probability distribution of the interaction strength and the feature vector of the attribute data is first obtained, and then the probability distribution of the interaction strength of the virtual object is determined based on the feature vector of the attribute data of the virtual object and the mapping relationship.

[0094] See also Figure 7 , Figure 7 is an optional flow chart of obtaining the mapping relationship between the probability distribution of interaction intensity and the feature vector of attribute data provided in the embodiment of the present application, see Figure 7 , by executing steps 201 to 203, a mapping relationship between the probability distribution of the interaction intensity and the feature vector of the attribute data is obtained.

[0095] Step 201: Obtain a first mapping relationship between a feature vector of attribute data and an interaction strength parameter, wherein the interaction strength parameter includes a mean value of the interaction strength and a standard deviation of the interaction strength.

[0096] In actual implementation, after determining the feature vector corresponding to the attribute data, for the feature data of different attribute dimensions, determine the first mapping relationship between the feature data of each attribute dimension and the mean and standard deviation of the interaction intensity, that is,

[0097]

[0098] Here, i in formula (1) represents different first sub-output layers, x represents the feature data of each attribute dimension in the input feature vector, σ represents the standard deviation node of each first sub-output layer for the interaction strength, and μ represents the mean node of each first sub-output layer for the interaction strength.

[0099] It should be noted that the first mapping relationship between the feature vector of the attribute data and the interaction strength parameter is a mapping relationship between the feature data of each attribute dimension in the feature vector and the interaction strength parameter.

[0100] Continuing with the above example, the feature vector of the attribute data here is [X1, X2, X3, …, X n ], feature vectors [X1, X2, X3, …, X n The characteristic data of each attribute dimension in ] is X1, X2, X3, ..., X n , and then for each attribute dimension, determine the first mapping relationship between it and the interaction intensity parameter.

[0101] Step 202: Determine a second mapping relationship between the probability distribution of interaction strength and the interaction strength parameter.

[0102] In actual implementation, for the interaction intensity of each attribute dimension in the feature vector, the second mapping relationship between the probability distribution of the interaction intensity and the interaction intensity parameter is determined, that is,

[0103]

[0104] Here, i in formula (2) represents different first sub-output layers, t represents the interaction strength, σ represents the standard deviation node of each first sub-output layer for the interaction strength, and μ represents the mean node of each first sub-output layer for the interaction strength.

[0105] Continuing with the above example, for the feature data of each attribute dimension, there is a corresponding interaction strength, for example, the feature vector of the attribute data [X1, X2, X3, ..., X n ] in each attribute dimension feature data X1, X2, X3, ..., X n , there are corresponding interaction intensities t1, t2, …, t n , then for each attribute dimension, determine the second mapping relationship between the probability distribution of the interaction intensity and the interaction intensity parameter.

[0106] Step 203 : determining a mapping relationship between the probability distribution of the interaction intensity and the feature vector of the attribute data based on the first mapping relationship and the second mapping relationship.

[0107] In actual implementation, the mapping relationship between the probability distribution of the interaction intensity and the feature vector of the attribute data is determined according to the first mapping relationship and the second mapping relationship. That is, according to the above formula (1) and formula (2), for each first sub-output layer, the mapping relationship between the probability distribution of the interaction intensity and the feature vector of the attribute data is determined, that is,

[0108]

[0109] Here, formula (3) is used to represent the Gaussian distribution for each attribute dimension in the feature vector. φ in formula (3) is the probability density function representing the interaction strength, i represents different first sub-output layers, t represents the interaction strength, x represents the feature data of each attribute dimension in the input feature vector, σ represents the standard deviation node of each first sub-output layer for the interaction strength, μ represents the mean node of each first sub-output layer for the interaction strength, and c is a constant.

[0110] It should be noted that, for different first sub-output layers, the standard deviation node and the mean node are different; the probability distribution of the interaction strength of the determined virtual object is the probability distribution of the interaction strength for different attribute dimensions.

[0111] See also Figure 8 , Figure 8 : is a schematic diagram of the probability distribution of the interaction strength of the virtual object provided in the embodiment of the present application. As an example, i is 3, μ is 1, -1 and 0, σ is 0.5, 1.0 and 1.5 respectively, then for each first sub-output layer, the mapping relationships obtained are respectively

[0112]

[0113]

[0114]

[0115] Here, formula (4) is the mapping relationship corresponding to a first sub-output layer when μ is 1 and σ is 0.5, formula (5) is the mapping relationship corresponding to a first sub-output layer when μ is -1 and σ is 1.0, and formula (6) is the mapping relationship corresponding to a first sub-output layer when μ is 0 and σ is 1.5. In this way, three different probability distributions of the interaction strength of the virtual object are determined.

[0116] Step 104 , superimposing the probability distributions of the interaction strength output by each first sub-output layer through the second sub-output layer of the interaction strength evaluation model, to obtain a target probability distribution of the interaction strength of the virtual object.

[0117] In actual implementation, the mapping relationship obtained for each first sub-output layer included in the interaction strength assessment model has different degrees of influence on the interaction strength, that is, the mapping relationship corresponding to each first sub-output layer has different weights of influence on the interaction strength. Therefore, while obtaining the mapping relationship between the probability distribution of the interaction strength and the characteristic vector of the attribute data through each first sub-output layer of the interaction strength assessment model, the weight corresponding to the first sub-output layer is also obtained, and then based on the obtained weight, the probability distribution of the interaction strength output by each first sub-output layer is weighted and summed to obtain the target probability distribution of the interaction strength of the virtual object.

[0118] In actual implementation, the probability distribution of the interaction intensity of the virtual object can be represented in different forms, such as a mapping relationship between probability and interaction intensity, a probability distribution diagram, and multiple interaction intensity probability pairs (i.e., the interaction intensity of the virtual object and the corresponding probability), etc. Correspondingly, the target probability distribution of the interaction intensity of the virtual object also includes different presentation methods, specifically, a mapping relationship between probability and interaction intensity, a probability distribution diagram of interaction intensity, and multiple interaction intensity probability pairs.

[0119] In actual implementation, when the probability distribution of interaction intensity is represented by a mapping relationship between probability and interaction intensity, the mapping relationships output by each of the first sub-output layers are weighted and summed based on the acquired weights to obtain the target mapping relationship of the interaction intensity of the virtual object, that is, the target probability distribution of the interaction intensity of the virtual object.

[0120] As an example, when the probability distribution of interaction intensity is represented by the mapping relationship between probability and interaction intensity, the mapping relationship between the probability distribution of interaction intensity output by each first sub-output layer and the feature vector of attribute data is weighted summed according to the weight corresponding to the first sub-output layer obtained, and the target mapping relationship is

[0121] F(t|x)=∑φ(t,μ i , σ i |x)*P i ……(7)

[0122] Here, P in formula (7) i is the weight corresponding to each first sub-output layer, that is, the weight of the feature data corresponding to each attribute dimension.

[0123] See also Fig. 9 , Fig. 9 : is a schematic diagram of the target probability distribution of the interaction intensity of the virtual object provided in the embodiment of the present application. Continuing with the above example, when i is 3, the weights corresponding to φ1, φ2 and φ3 are P1, P2 and P3 respectively, and the weighted sum of the three different mapping relationships is obtained to obtain the target mapping relationship:

[0124] F(t|x)=φ1*P1+φ2*P2+φ3*P3……(8)

[0125] Here, after obtaining the target mapping relationship, the target probability distribution of the interaction intensity of the virtual object is obtained.

[0126] It should be noted that, since the mapping relationship obtained by each first sub-output layer included in the interaction strength evaluation model is for different attribute dimensions of the input feature vector, the weight of the influence of the feature data of different attribute dimensions in the input feature vector on the interaction strength is the same as the weight corresponding to the mapping relationship obtained by each first sub-output layer, that is, the weight corresponding to each first sub-output layer. Here, after obtaining the weight corresponding to the first sub-output layer, the weight corresponding to each first sub-output layer can also be changed to adjust the influence of each attribute dimension in the feature vector on the interaction strength.

[0127] Continuing with the above example, by adjusting the weight of the attribute dimension "Player Racing Level", the influence of "Player Racing Level" on the time required for the car to complete a specific track can be enhanced to increase the player's upgrade speed; or by enhancing the influence of the car's own attributes on the time required for the car to complete a specific track, players can buy cars with high attribute values. In this way, by adjusting the weight of each attribute dimension, the influence of the corresponding attribute dimension on the interaction intensity can be changed, and then the application scenarios of various virtual objects can be coped with, and the applicability of the evaluation results can be enhanced.

[0128] In this way, by performing weighted summation on the mapping relationship between the probability distribution of the interaction strength output by each first sub-output layer and the characteristic vector of the attribute data, the probability distribution of the target value of the virtual object interaction strength is determined based on the importance of different attribute dimensions of the acquired attribute data to the virtual object interaction strength, thereby improving the accuracy of the evaluation results.

[0129] In some embodiments, see Fig.10 , Fig.10 is an optional flow chart of the interaction intensity evaluation method provided in the embodiment of the present application, based on Figure 3 , after step 104, the probability that the interaction intensity belongs to the target evaluation interval is determined by executing steps 301 to 303.

[0130] Step 301: Obtain a target probability distribution of the interaction strength of a virtual object.

[0131] See also Fig.11 , Fig.11: is a schematic diagram of the target probability distribution of the interaction intensity of the virtual object provided in the embodiment of the present application. As an example, when the virtual scene is a racing game, the virtual object is a vehicle in the game, and the interaction intensity includes the time taken by the vehicle to travel the target distance, the target probability distribution obtained is the probability distribution of the time taken by the vehicle to travel the target distance. Here, the horizontal axis is the predicted time t required for the new vehicle to complete the travel target distance, and the vertical axis is the probability density value p corresponding to t.

[0132] Step 302: Integrate the target probability distribution based on the target evaluation interval of the interaction intensity.

[0133] In actual implementation, the target evaluation interval is the expected interval of interaction intensity.

[0134] Continuing with the above example, when the target probability distribution obtained is the probability distribution of the time taken by the vehicle to travel the target distance, the target probability distribution is integrated based on the target evaluation interval [t1, t2]. As an example, the mapping relationship between the probability of the obtained interaction intensity and the interaction intensity is integrated to obtain the integral result, that is,

[0135]

[0136] Step 303: Using the integrated result as the probability that the interaction intensity belongs to the target evaluation interval.

[0137] Continuing with the above example, after determining the target evaluation interval [t1, t2], the target probability distribution is integrated based on the target evaluation interval [t1, t2] to predict the probability that the time taken by the vehicle to travel the target distance falls within the interval [t1, t2].

[0138] In this way, numerical planners can adjust the various numerical attributes of the new vehicle according to the target probability distribution curve until the strength of the new vehicle meets expectations.

[0139] In some embodiments, see Fig.12 , Fig.12 is an optional flow chart of the interaction intensity assessment model training method provided in the embodiment of the present application, based on Figure 3 Before step 101, it is necessary to train the interactive evaluation model of the virtual object, combined with Fig.12 The steps shown are explained.

[0140] Step 401, obtaining training sample data carrying labels.

[0141] The training sample data is the attribute data of the virtual object in the virtual scene, and the label is the real interaction strength of the virtual object.

[0142] In actual implementation, the training sample data collection and preprocessing is first performed. Specifically, the original data set is established, that is, the data attribute values ​​of the players and virtual objects in the existing virtual scene are obtained, and the real interaction intensity corresponding to the virtual objects is obtained; then the original data set is preprocessed. Specifically, the obtained attribute values ​​and real interaction intensity are standardized, and each attribute value and real interaction intensity are standardized to the target range respectively.

[0143] For example, when the virtual scene is a racing game, the virtual object is a vehicle in the racing game, and the label is the time required for the vehicle to complete a specific track, first, a data set of the numerical attribute data X of the vehicle or the player and the target data Y (i.e., the label) is obtained, as shown below: Fig.13 , Fig.13 This is a schematic diagram of the training sample data provided by the embodiment of the present application. Here, each row of data represents a data set, where the last column of data corresponds to the time required to complete a specific track, and the remaining columns correspond to the various attribute values ​​of the car, such as speed, weight, gas collection, collision, etc. Then, for each attribute value X, the Min-Max standardization method is used to standardize it to the interval of [-1, 1]; for the target value Y, the Min-Max standardization method is used to standardize it to the interval of [0, 1].

[0144] In actual implementation, after preprocessing the acquired training sample data with labels, the preprocessed training sample data is input into the interaction strength evaluation model, see Fig.14 , Fig.14 It is an optional structural diagram of the interaction strength evaluation model provided in the embodiment of the present application. Here, the pre-processed training sample data is input into the interaction strength evaluation model through the input layer of the interaction strength evaluation model.

[0145] Step 402, feature encoding is performed on the training sample data through the feature encoding layer of the interaction strength evaluation model to obtain a corresponding sample feature vector.

[0146] In actual implementation, continue to refer to Fig.14 After the preprocessed training sample data is input into the input layer, the feature encoding layer is used to perform feature encoding on the preprocessed training sample data to obtain the corresponding sample feature vector.

[0147] Step 403 , based on the sample feature vector, the interaction strength of the virtual object is predicted respectively through at least two first sub-output layers included in the interaction strength evaluation model to obtain a first probability distribution of the interaction strength of the virtual object.

[0148] In actual implementation, continue to refer to Fig.14, through at least two first sub-output layers included in the softmax layer in the output layer, the interaction strength of the virtual objects is predicted respectively, and three output nodes of each first sub-output layer, namely, the standard deviation node σ, the mean node μ and the weight node P, are determined.

[0149] It should be noted that the three output nodes of different first sub-output layers are different.

[0150] In actual implementation, after the output nodes σ and μ of each first sub-output layer are determined, the probability distribution of the interaction intensity of the virtual object is determined. As an example, when the probability distribution of the interaction intensity is represented by the mapping relationship between probability and interaction intensity, the mapping relationship between probability and interaction intensity, that is, the probability density function of interaction intensity, is:

[0151] N(t, μ, σ|x)=φ i (t|x)……(10)

[0152] Here, φ in formula (10) i (t|x) is formula (3).

[0153] Step 404 , superimposing the first probability distributions of the interaction strength output by each first sub-output layer through the second sub-output layer of the interaction strength evaluation model, to obtain a second probability distribution of the interaction strength of the virtual object.

[0154] In actual implementation, continue to refer to Fig.14 After determining the three output nodes of each first sub-output layer, the first probability distribution of the interaction strength output by each first sub-output layer is weighted and summed to obtain the second probability distribution of the interaction strength of the virtual object.

[0155] Continuing with the above example, when the probability distribution of the interaction intensity is represented by the mapping relationship between probability and interaction intensity, the mapping relationship output by each first sub-output layer is weighted and summed through the weight node P of each first sub-output layer to obtain the target mapping relationship of the interaction intensity of the virtual object, that is, the objective function

[0156] Loss = sum(∑N(t, μ i , σ i |x)*P i )……(11)

[0157] Here, x in formula (11) is the corresponding input feature value in the training data, μ is the mean node corresponding to each first sub-output layer, σ is the standard deviation node corresponding to each first sub-output layer, P is the weight node of each first sub-output layer, t is the true target value (i.e., label) corresponding to the training data, and φ is the value calculated according to the Gaussian distribution probability density function.

[0158] In actual implementation, after determining the target mapping relationship of the interaction strength of the virtual object, the second probability distribution of the interaction strength of the virtual object is determined.

[0159] Step 405 : determining the interaction strength of the virtual object when the probability is maximum based on the second probability distribution.

[0160] In actual implementation, when the probability distribution of interaction intensity is represented by the mapping relationship between probability and interaction intensity, the target mapping relationship of the interaction intensity of the obtained virtual object is integrated, that is,

[0161]

[0162] Here, the interaction strength t of the virtual object when the probability p is maximum is determined by formula (12).

[0163] It should be noted that when the true target value t is substituted into the objective function, the larger the value obtained, the closer the predicted probability density function is to the actual situation. Obtaining this maximum value is the goal of parameter optimization, that is, the meaning of the objective function Loss.

[0164] Step 406, obtaining the difference between the determined interaction strength and the actual interaction strength, and updating the model parameters of the interaction strength evaluation model based on the difference.

[0165] In actual implementation, the corresponding gradient information is determined based on the difference between the interaction intensity corresponding to the maximum probability and the actual interaction intensity, and the model parameters of the interaction intensity evaluation model are updated based on the gradient information.

[0166] In some embodiments, a general gradient optimization algorithm can be used to train the model parameters, such as using the Adam gradient optimization algorithm to train the model parameters, and finally obtaining the weights of the model parameters of the interaction intensity evaluation model. Based on the weights of the model parameters, the model parameters of the interaction intensity evaluation model are updated.

[0167] By using the above-mentioned embodiment of the present application, when evaluating the interaction strength of virtual objects in a virtual scene, the acquired attribute data of the virtual objects is predicted, and after obtaining at least two probability distributions of the interaction strength of the virtual objects, the at least two probability distributions are superimposed to obtain the target probability distribution of the interaction strength of the virtual objects. In this way, by obtaining the target probability distribution of the interaction strength of the virtual objects, the evaluation results are more consistent with the real data in actual applications, thereby improving the accuracy of the evaluation results.

[0168] The following describes exemplary applications of the embodiments of the present application in actual application scenarios.

[0169] Here, we take the application of the interactive intensity assessment model in racing games as an example. Here, the virtual scene is a racing game, the virtual object is a vehicle in the racing game, and the label is the time required for the vehicle to complete a specific track. Of course, this technical solution can also be applied to target value prediction and evaluation in other game products. In racing games, the time required for a car to complete a track is affected by the car's speed, weight, gas speed, small spray, large spray, drifting, collision and other attribute values, as well as the player's racing rank. In order to evaluate and predict the time required for a new car to complete a track, see Fig.15 , Fig.15 is a schematic diagram of the overall structure of the interaction intensity evaluation model provided in the embodiment of the present application, combined with Fig.15 The interaction strength evaluation model includes an input layer, a feature encoding layer and an output layer, wherein the output layer includes at least a first sub-output layer and a second sub-output layer.

[0170] In actual implementation, the acquired attribute data is input into the interaction intensity evaluation model through the input layer. After the attribute data is feature encoded by the feature encoding layer, the corresponding feature vector is obtained and then input into the output layer. The softmax layer in the output layer includes at least one first sub-output layer which determines the Gaussian distribution of the interaction intensity of the corresponding virtual object by respectively determining the output layer nodes corresponding to each output layer, and then inputs the determined Gaussian distribution into the second sub-output layer. The second sub-output layer obtains the target probability distribution of the interaction intensity of the virtual object by superimposing the Gaussian distribution of the interaction intensity of each virtual object.

[0171] here, Fig.16 is an optional flow chart of a method for evaluating the interactive strength of a virtual object provided in an embodiment of the present application, see Fig.16 The virtual object interaction strength evaluation method provided in the embodiment of the present application is implemented by the client and the server in collaboration. Fig.16 The interaction intensity evaluation method provided in the embodiment of the present application includes:

[0172] Step 501: The client obtains training sample data carrying labels.

[0173] Here, the training sample data is the attribute data of the virtual objects in the virtual scene, that is, the attribute data of the vehicles in the racing game, and the label is the real interaction intensity of the virtual objects, that is, the time required for the vehicle to complete a specific track.

[0174] In actual implementation, the client obtains training sample data carrying labels in response to an import operation on the attribute data of the virtual object.

[0175] Here, the client can be an interactive strength assessment client for virtual objects set on the terminal. The attribute data of the virtual object can be obtained by the user based on the human-computer interaction interface of the client, triggering the upload function item in the human-computer interaction interface so that the client presents a data import interface on the human-computer interaction interface. The user can then import the attribute data of the virtual object from the terminal locally or manually input the attribute data of the virtual object based on the data import interface, thereby enabling the client to obtain the attribute data of the virtual object.

[0176] In some embodiments, the attribute data of the virtual object may also be automatically collected during the game. After collecting the attribute data of the virtual object, the game client transmits the collected attribute data to the terminal, and the terminal automatically uploads it to the client.

[0177] Step 502: Send the acquired training sample data carrying labels to the server.

[0178] Step 503: The server preprocesses the received training sample data and inputs the preprocessed training sample data into the interaction intensity evaluation model.

[0179] In actual implementation, the training sample data collection preprocessing is first performed. Specifically, the original data set is established, that is, the data set of the numerical attribute data X and the target data Y (i.e., label) of the vehicle or player is obtained, wherein the numerical attribute data X of the vehicle or player includes various attribute values ​​such as speed, weight, gas, collision, and the player's racing level, and the target data Y is the time required to complete a specific track. Then, for each attribute value X, the Min-Max normalization method is used to normalize it to the interval of [-1, 1]; for the target value Y, the Min-Max normalization method is used to normalize it to the interval of [0, 1].

[0180] Here, the interactive strength evaluation model includes a hidden layer and an output layer. The hidden layer usually extracts features from the input data and can be adjusted and optimized according to the training effect. The output layer nodes contain K Gaussian distribution means μ, K Gaussian distribution standard deviations σ, and K Gaussian distribution probability weights P. The final probability distribution density function is generated by superimposing the determined K μ, σ, and P, which can be used to evaluate the strength of game characters or props.

[0181] It should be noted that the hidden layer is the feature encoding layer, and the output layer includes K first sub-output layers and one second sub-output layer. Here, the output layer nodes are determined on the first sub-output layer, and the output layer nodes of each first sub-output layer are different. K is a positive integer greater than 1.

[0182] Step 504 , feature encoding is performed on the training sample data through the feature encoding layer of the interaction strength evaluation model to obtain a corresponding sample feature vector.

[0183] Step 505 , based on the feature vector, the interaction strength of the virtual object is predicted respectively through at least two first sub-output layers included in the interaction strength evaluation model to obtain a probability distribution of the interaction strength of the virtual object.

[0184] In actual implementation, based on the feature vector, the interaction strength of the virtual object is predicted respectively through at least two first sub-output layers included in the interaction strength assessment model, and the probability distribution of the interaction strength of the virtual object is obtained. That is, based on the feature vector, the interaction strength assessment model is used to predict the time taken by the vehicle to travel the target distance respectively through at least two first sub-output layers, and the probability distribution of the time (that is, Gaussian distribution of the time) is obtained.

[0185] Here, the probability distribution of duration is the probability density function of duration; the softmax layer in the output layer includes at least two first sub-output layers, which predict the interaction strength of the vehicles respectively, and determine the three output nodes of each first sub-output layer, namely the standard deviation node σ, the mean μ and the weight node P.

[0186] In actual implementation, after determining the output layer nodes σ and μ, the probability density function of the time taken by the vehicle to travel the target distance is determined, that is,

[0187]

[0188] Here, the probability density functions output by different first sub-output layers correspond to different weight nodes P.

[0189] Step 506 , superimposing the probability distributions of the interaction strengths output by the first sub-output layers through the second sub-output layer of the interaction strength evaluation model, to obtain a target probability distribution of the interaction strength of the virtual object.

[0190] In actual implementation, the probability distributions of the interaction strength output by each first sub-output layer are superimposed to obtain the target probability distribution of the interaction strength of the virtual object. That is, the probability distributions of the duration output by each first sub-output layer are superimposed through the second sub-output layer of the interaction strength evaluation model to obtain the target probability distribution of the duration taken by the vehicle to travel the target distance.

[0191] Here, the target probability distribution of the time taken by the vehicle to travel the target distance is obtained, that is, the target probability density function of the time taken by the vehicle to travel the target distance is obtained, and the probability distributions of the time taken by each first sub-output layer are superimposed, that is, the probability density functions output by each first sub-output layer are weighted summed, that is, after determining the probability density function of the time taken by the vehicle to travel the target distance, the target probability density function is obtained by superimposing K probability density functions according to K weights P, that is,

[0192] Loss = sum(∑N(t, μ k , σ k |x)*P k )……(14)

[0193] Step 507: Determine the interaction strength of the virtual object when the probability is maximum based on the target probability density function.

[0194] In actual implementation, the interaction strength of the virtual object when the probability is the highest is determined by determining the time taken for the vehicle to travel the target distance when the probability is the highest.

[0195] Here, the true target value t is substituted into the target probability density function. The larger the value obtained, the closer the predicted target probability density function is to the actual situation. Obtaining this maximum value is the goal of parameter optimization, that is, the meaning of the target probability density function Loss.

[0196] Step 508, obtaining the difference between the determined duration and the actual duration, and updating the model parameters of the interaction intensity assessment model based on the difference to obtain a trained interaction intensity assessment model.

[0197] In actual implementation, a general gradient optimization algorithm can be used to train the model parameters, such as using the Adam gradient optimization algorithm to train the model parameters, and finally the weights of the model parameters of the interaction intensity evaluation model are obtained. Based on the weights of the model parameters, the model parameters of the interaction intensity evaluation model are updated.

[0198] Step 509: The server generates a prompt message indicating that the training of the interaction intensity assessment model is completed.

[0199] Step 510: Send a prompt message to the client.

[0200] Step 511: The client obtains attribute data of a virtual object in a virtual scene.

[0201] In actual implementation, the client obtains the attribute data of the virtual object in the virtual scene in response to the import operation on the attribute data of the virtual object.

[0202] Step 512: The client sends the acquired attribute data to the server.

[0203] In step 513, the server inputs the received attribute data of the vehicle to be evaluated into the interaction strength evaluation model, so that the interaction strength evaluation model predicts the time required for the vehicle to complete a specific track and obtains the corresponding target probability distribution.

[0204] Step 514: The server sends the target probability distribution to the client.

[0205] Step 515: The client outputs the target probability distribution.

[0206] In some embodiments, after outputting the target probability distribution, the client can also obtain the target interval [t1, t2] of the time required for the new vehicle to complete a specific track input by the user, and then the client integrates the target probability distribution based on the target interval [t1, t2], and uses the result of the integration as the predicted probability that the time required for the new vehicle to complete the specific track falls within the interval [t1, t2]. In this way, the numerical planner can adjust the various numerical attributes of the new vehicle according to the target probability distribution until the time required for the new vehicle to complete the specific track meets expectations.

[0207] In some embodiments, after outputting the target probability distribution, the client can also compare it with the distribution statistics of the time t required for a large number of real players to complete a specific track, see Fig.17 , Fig.17 This is a comparative diagram of the target probability distribution and the real distribution statistics provided by the embodiment of the present application. Here, the horizontal axis is the predicted time t required for a new vehicle to complete a specific track, and the vertical axis is the probability density value p corresponding to t. The curve is the target probability distribution output by the client, and the shaded part is the real distribution statistics. Fig.17 , we can see the consistency of the two distributions. It should be noted that when the two distributions are inconsistent, the consistency of the two distributions can be improved by adjusting the model parameters of the interaction intensity evaluation model until the target probability distribution conforms to the real distribution statistics.

[0208] By using the above-mentioned embodiment of the present application, when evaluating the interaction strength of virtual objects in a virtual scene, the acquired attribute data of the virtual objects is predicted, and after obtaining at least two probability distributions of the interaction strength of the virtual objects, the at least two probability distributions are superimposed to obtain the target probability distribution of the interaction strength of the virtual objects. In this way, by obtaining the target probability distribution of the interaction strength of the virtual objects, the evaluation results are more consistent with the real data in actual applications, thereby improving the accuracy of the evaluation results.

[0209] The following is a description of an exemplary structure of a virtual object interaction strength evaluation device 455 provided in an embodiment of the present application implemented as a software module. In some embodiments, Figure 2 As shown, the software modules stored in the virtual object interaction strength evaluation device 455 in the memory 440 may include:

[0210] An acquisition module 4551 is used to acquire attribute data of a virtual object in a virtual scene;

[0211] A feature coding module 4552 is used to perform feature coding on the attribute data of the virtual object through the feature coding layer of the interaction strength evaluation model to obtain a corresponding feature vector;

[0212] The first output module 4553 is used to predict the interaction strength of the virtual object based on the feature vector through at least two first sub-output layers included in the interaction strength evaluation model, so as to obtain the probability distribution of the interaction strength of the virtual object.

[0213] The second output module 4554 is used to superimpose the probability distributions of the interaction strength output by each of the first sub-output layers through the second sub-output layer of the interaction strength evaluation model to obtain a target probability distribution of the interaction strength of the virtual object.

[0214] In some embodiments, the first output module 4553 is also used to perform the following processing through each first sub-output layer of the interaction strength assessment model: obtaining the mapping relationship between the probability distribution of interaction strength and the feature vector of attribute data; and determining the probability distribution of the interaction strength of the virtual object based on the feature vector of the attribute data of the virtual object and the mapping relationship.

[0215] In some embodiments, the first output module 4553 is also used to obtain a first mapping relationship between a characteristic vector of attribute data and an interaction intensity parameter, wherein the interaction intensity parameter includes a mean of the interaction intensity and a standard deviation of the interaction intensity; determine a second mapping relationship between a probability distribution of the interaction intensity and the interaction intensity parameter; and determine a mapping relationship between the probability distribution of the interaction intensity and the characteristic vector of the attribute data based on the first mapping relationship and the second mapping relationship.

[0216] In some embodiments, the second output module 4554 is also used to obtain the weights corresponding to each of the first sub-output layers; based on the obtained weights, the probability distribution of the interaction intensity output by each of the first sub-output layers is weighted summed to obtain the target probability distribution of the interaction intensity of the virtual object.

[0217] In some embodiments, the second output module 4554 is also used to perform weighted summation processing on the mapping relationships output by each of the first sub-output layers based on the acquired weights when the probability distribution of the interaction intensity is represented by a mapping relationship between probability and interaction intensity, so as to obtain the target probability distribution of the interaction intensity of the virtual object.

[0218] In some embodiments, the device also includes a training module, which is used to obtain training sample data carrying labels; wherein the training sample data is attribute data of a virtual object in a virtual scene, and the label is the real interaction strength of the virtual object; the training sample data is feature encoded through the feature encoding layer of the interaction strength assessment model to obtain a corresponding sample feature vector; based on the sample feature vector, the interaction strength of the virtual object is predicted respectively through at least two first sub-output layers included in the interaction strength assessment model to obtain a first probability distribution of the interaction strength of the virtual object; through the second sub-output layer of the interaction strength assessment model, the first probability distribution of the interaction strength output by each first sub-output layer is superimposed to obtain a second probability distribution of the interaction strength of the virtual object; based on the second probability distribution, the interaction strength of the virtual object when the probability is maximum is determined; the difference between the determined interaction strength and the real interaction strength is obtained, and the model parameters of the interaction strength assessment model are updated based on the difference.

[0219] In some embodiments, the second output module 4554 is further used to output the target probability distribution of the interaction intensity of the virtual object in at least one of the following ways: a mapping relationship between probability and interaction intensity, a probability distribution diagram of interaction intensity, and a plurality of interaction intensity probability pairs; wherein the interaction intensity probability pair includes: the interaction intensity of the virtual object and the corresponding probability.

[0220] In some embodiments, the virtual object includes a vehicle in a virtual scene, and the interaction intensity includes the time taken by the vehicle to travel the target distance. The first output module 4553 is further used to predict the time taken by the vehicle to travel the target distance based on the feature vector through at least two first sub-output layers included in the interaction intensity assessment model to obtain a probability distribution of the time. The second output module 4554 is further used to superimpose the probability distributions of the time output by each of the first sub-output layers through the second sub-output layer of the interaction intensity assessment model to obtain a target probability distribution of the time taken by the vehicle to travel the target distance.

[0221] In some embodiments, the device also includes an integration module, which is used to obtain a target probability distribution of the interaction intensity of the virtual object; integrate the target probability distribution based on a target evaluation interval of the interaction intensity; and use the integrated result as the probability that the interaction intensity belongs to the target evaluation interval.

[0222] The embodiment of the present application provides a computer program product or a computer program, which includes a computer instruction stored in a computer-readable storage medium. The processor of the computer device reads the computer instruction from the computer-readable storage medium, and the processor executes the computer instruction, so that the computer device executes the method for evaluating the interactive strength of the virtual object described in the embodiment of the present application.

[0223] The present application embodiment provides a computer-readable storage medium storing executable instructions, wherein the executable instructions are stored. When the executable instructions are executed by a processor, the processor will execute the method for evaluating the interactive strength of a virtual object provided by the present application embodiment, for example, Figure 3 A method for evaluating the interaction strength of a virtual object is shown.

[0224] In some embodiments, the computer-readable storage medium may be a memory such as FRAM, ROM, PROM, EP ROM, EEPROM, flash memory, magnetic surface storage, optical disk, or CD-ROM; or it may be various devices including one or any combination of the above memories.

[0225] In some embodiments, executable instructions may be in the form of a program, software, software module, script or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including as a stand-alone program or as a module, component, subroutine or other unit suitable for use in a computing environment.

[0226] As an example, executable instructions may, but need not, correspond to a file in a file system, may be stored as part of a file that stores other programs or data, such as in one or more scripts in a HyperText Markup Language (HTML) document, in a single file dedicated to the program in question, or in multiple coordinated files (e.g., files storing one or more modules, subroutines, or code portions).

[0227] By way of example, executable instructions may be deployed to be executed on one computing device, or on multiple computing devices located at one site, or on multiple computing devices distributed across multiple sites and interconnected by a communication network.

[0228] In summary, the embodiments of the present application have the following beneficial effects:

[0229] (1) By obtaining the target probability distribution of the interaction strength of virtual objects, not only the target value of the interaction strength of virtual objects can be predicted, but also the probability distribution of the target value of the interaction strength of virtual objects can be predicted, which is more in line with the real data in practical applications and improves the accuracy of the evaluation results.

[0230] (2) The interactive strength evaluation model can be used to evaluate the strength of characters and props in the game, and provide a basis for numerical planners to adjust the attribute values ​​of characters or props.

[0231] (3) Compared with the traditional regression algorithm based on machine learning, this application increases the depth of the network model, and the model has a better fitting ability for complex data and can fit more complex data.

[0232] (4) By superimposing multiple Gaussian distributions through the interaction intensity evaluation model, any probability distribution model can be fitted in theory.

[0233] The above is only an embodiment of the present application and is not intended to limit the protection scope of the present application. Any modifications, equivalent substitutions and improvements made within the spirit and scope of the present application are included in the protection scope of the present application.

Claims

1. A method for evaluating the interaction strength of a virtual object, characterized in that: The method comprises: Obtaining attribute data of virtual objects in the virtual scene; Through the feature coding layer of the interaction strength evaluation model, feature coding is performed on the attribute data of the virtual object to obtain a corresponding feature vector; The following processing is performed through each first sub-output layer of the interaction strength evaluation model: Acquire a first mapping relationship between a feature vector of the attribute data and an interaction strength parameter, wherein the interaction strength parameter includes a mean value of the interaction strength and a standard deviation of the interaction strength; Determine a second mapping relationship between the probability distribution of interaction strength and the interaction strength parameter; Based on the first mapping relationship and the second mapping relationship, determining a mapping relationship between the probability distribution of the interaction intensity and a feature vector of the attribute data; Determining a probability distribution of interaction intensity of the virtual object based on a feature vector of the attribute data of the virtual object and the mapping relationship; The probability distributions of the interaction strength output by each of the first sub-output layers are superimposed through the second sub-output layer of the interaction strength evaluation model to obtain a target probability distribution of the interaction strength of the virtual object.

2. The method according to claim 1, characterized in that: The method of superimposing the probability distributions of the interaction strengths output by the first sub-output layers through the second sub-output layer of the interaction strength evaluation model to obtain a target probability distribution of the interaction strength of the virtual object includes: Obtaining weights corresponding to each of the first sub-output layers; Based on the obtained weights, a weighted summation process is performed on the probability distributions of the interaction strengths output by each of the first sub-output layers to obtain a target probability distribution of the interaction strengths of the virtual objects.

3. The method according to claim 2, characterized in that The method of performing weighted summation processing on the probability distributions of the interaction strengths outputted by each of the first sub-output layers based on the acquired weights to obtain a target probability distribution of the interaction strengths of the virtual objects includes: When the probability distribution of the interaction intensity is represented by a mapping relationship between probability and interaction intensity, Based on the obtained weights, a weighted summation process is performed on the mapping relationships output by each of the first sub-output layers to obtain a target probability distribution of the interaction intensity of the virtual object.

4. The method according to claim 1, characterized in that: The method further comprises: Acquire training sample data carrying labels; wherein the training sample data is attribute data of a virtual object in a virtual scene, and the label is the real interaction strength of the virtual object; Performing feature encoding on the training sample data through the feature encoding layer of the interaction intensity evaluation model to obtain a corresponding sample feature vector; Based on the sample feature vector, the interaction intensity of the virtual object is predicted respectively through at least two first sub-output layers included in the interaction intensity evaluation model to obtain a first probability distribution of the interaction intensity of the virtual object; The first probability distributions of the interaction strength output by each of the first sub-output layers are superimposed through the second sub-output layer of the interaction strength evaluation model to obtain a second probability distribution of the interaction strength of the virtual object; Based on the second probability distribution, determining the interaction strength of the virtual object when the probability is maximum; A difference between the determined interaction strength and the true interaction strength is obtained, and a model parameter of the interaction strength evaluation model is updated based on the difference.

5. The method according to claim 1, characterized in that The method further comprises: Outputting a target probability distribution of the interaction intensity of the virtual object in at least one of the following ways: The mapping relationship between probability and interaction intensity, the probability distribution diagram of interaction intensity, and the probability pairs of multiple interaction intensity; The interaction strength probability pair includes: the interaction strength of the virtual object and the corresponding probability.

6. The method according to claim 1, characterized in that The virtual object includes a vehicle in a virtual scene, and the interaction intensity includes the time taken by the vehicle to travel a target distance; The following processing is performed respectively through each first sub-output layer of the interaction strength evaluation model: obtaining a first mapping relationship between a feature vector of attribute data and an interaction strength parameter, wherein the interaction strength parameter includes a mean value of the interaction strength and a standard deviation of the interaction strength; determining a second mapping relationship between a probability distribution of the interaction strength and the interaction strength parameter; determining a mapping relationship between the probability distribution of the interaction strength and the feature vector of the attribute data based on the first mapping relationship and the second mapping relationship; determining a probability distribution of the interaction strength of the virtual object based on the feature vector of the attribute data of the virtual object and the mapping relationship, including: Based on the feature vector, the first mapping relationship between the feature vector of the attribute data and the time parameter used for the vehicle to travel the target distance is obtained respectively through at least two first sub-output layers included in the interaction intensity assessment model, wherein the time parameter used for the vehicle to travel the target distance includes the mean of the time used for the vehicle to travel the target distance and the standard deviation of the time used for the vehicle to travel the target distance; determine the second mapping relationship between the probability distribution of the time used for the vehicle to travel the target distance and the time parameter used for the vehicle to travel the target distance; based on the first mapping relationship and the second mapping relationship, determine the mapping relationship between the probability distribution of the time used for the vehicle to travel the target distance and the feature vector of the attribute data; based on the feature vector of the attribute data of the virtual object and the mapping relationship, determine the probability distribution of the time used for the vehicle to travel the target distance of the virtual object; The method of superimposing the probability distributions of the interaction strengths output by the first sub-output layers through the second sub-output layer of the interaction strength evaluation model to obtain a target probability distribution of the interaction strength of the virtual object includes: The probability distributions of the durations output by the first sub-output layers are superimposed through the second sub-output layer of the interaction intensity assessment model to obtain a target probability distribution of the duration taken by the vehicle to travel the target distance.

7. The method according to claim 1, characterized in that The method further comprises: Obtaining a target probability distribution of the interaction intensity of the virtual object; integrating the target probability distribution based on a target evaluation interval of the interaction intensity; The result obtained by integration is used as the probability that the interaction intensity belongs to the target evaluation interval.

8. A device for evaluating the interaction strength of a virtual object, characterized in that: The device comprises: An acquisition module, used to acquire attribute data of virtual objects in a virtual scene; A feature encoding module, used to perform feature encoding on the attribute data of the virtual object through a feature encoding layer of an interaction strength evaluation model to obtain a corresponding feature vector; The first output module is used to perform the following processing respectively through each first sub-output layer of the interaction intensity evaluation model: obtaining a first mapping relationship between a feature vector of attribute data and an interaction intensity parameter, wherein the interaction intensity parameter includes a mean value of the interaction intensity and a standard deviation of the interaction intensity; determining a second mapping relationship between a probability distribution of the interaction intensity and the interaction intensity parameter; determining a mapping relationship between the probability distribution of the interaction intensity and the feature vector of the attribute data based on the first mapping relationship and the second mapping relationship; determining a probability distribution of the interaction intensity of the virtual object based on the feature vector of the attribute data of the virtual object and the mapping relationship; The second output module is used to superimpose the probability distributions of the interaction strength output by each of the first sub-output layers through the second sub-output layer of the interaction strength evaluation model to obtain a target probability distribution of the interaction strength of the virtual object.

9. An electronic device, characterized in that: The electronic device comprises: A memory for storing computer executable instructions; A processor, configured to implement the method for evaluating the interactive strength of a virtual object as described in any one of claims 1 to 7 when executing the computer executable instructions stored in the memory.

10. A computer-readable storage medium storing computer-executable instructions or a computer program, characterized in that: When the computer executable instructions or computer program are executed by a processor, the method for evaluating the interaction strength of a virtual object as described in any one of claims 1 to 7 is implemented.

11. A computer program product comprising computer executable instructions or a computer program, characterized in that: When the computer executable instructions or computer program are executed by a processor, the method for evaluating the interaction strength of a virtual object as described in any one of claims 1 to 7 is implemented.

Citation Information

Patent Citations

  • User attribute influence assessment method and device

    CN110020211A

  • Method of determining virtual object behavior and hosting virtual object behavior

    CN111437605A

  • Game interaction behavior model generation method and device, server and storage medium

    CN111760291A

  • Population attribute tag prediction method and system, computer equipment and storage medium

    CN113052635A

  • Fuzzy users' attributes prediction based on users' behaviors

    US20110131166A1