A character scene modeling method, device, equipment and storage medium

Abstract

The application discloses a character scene modeling method and device, equipment and storage medium, and relates to the technical field of metaverse privacy protection, and comprises the following steps: acquiring initial data for generating a metaverse character scene model; the initial data comprises structured data, image data, voice data, video data and text description data; the initial data is converted into natural language text data by using a preset data conversion model; the natural language text data is desensitized by using a preset text data desensitization model and in combination with preset user privacy requirements, to obtain desensitized text data; and a corresponding metaverse character scene model is generated based on the desensitized text data. In this way, the application can convert cross-modal initial data into text data, then perform privacy desensitization processing, and then generate a corresponding metaverse character scene model by using the desensitized data, so that the data leakage problem of the metaverse character scene model can be eliminated, and the privacy requirements of users can be met.

Description

Technical Field

[0001] This invention relates to the field of metaverse privacy protection technology, and in particular to a method, apparatus, device and storage medium for character and scene modeling. Background Technology

[0002] The metaverse is a virtual, three-dimensional, interactive, multi-user, internet-based virtual space containing numerous virtual reality scenes and objects that can interact with the real world. Many of these scenes and objects are replicas of real-world counterparts. Building the metaverse requires extensive 3D scene modeling. Traditional 3D scene modeling methods involve significant manual model design and detail processing, which is time-consuming, labor-intensive, and costly. Secondly, much real-world scene information is difficult to directly convert into 3D scene models, resulting in low modeling efficiency. Finally, due to data privacy protection, much data cannot be directly used, limiting the generation of metaverse scenes.

[0003] As metaverse technology continues to develop and its applications expand, privacy protection has gradually become a focus of attention. When users interact, create, buy, sell, and entertain themselves in the metaverse using virtual identities, they are required to provide more personal information. This information includes basic personal information, virtual identity information, real-world photos, videos, and location information, all of which may be at risk of leakage. Furthermore, the use of VR headsets, sensors, and other devices may collect users' behavioral and biometric data, such as facial expressions, body postures, circadian rhythms, and voice information. This data could be misused, thereby affecting user privacy and security.

[0004] To address the privacy concerns of the metaverse, numerous technologies are being researched and applied. However, due to the characteristics of the metaverse, such as realism, interactivity, and immersive user experience, traditional encryption and authentication technologies have limitations. Therefore, how to generate metaverse character and scene models while protecting user privacy has become a pressing issue in this field. Summary of the Invention

[0005] In view of this, the purpose of this invention is to provide a method, apparatus, device, and storage medium for character and scene modeling, capable of converting multimodal data into text data and generating a corresponding metaverse character and scene model based on the anonymized text data, which can meet the user's privacy requirements. The specific solution is as follows:

[0006] Firstly, this application provides a method for character scene modeling, including:

[0007] Obtain initial data for generating metaverse character and scene models; the initial data includes structured data, image data, audio data, video data, and text description data;

[0008] The initial data is converted into natural language text data using a preset data conversion model;

[0009] By using a preset text data desensitization model and combining preset user privacy requirements, the natural language text data is desensitized to obtain desensitized text data.

[0010] Based on the anonymized text data, a corresponding metaverse character scene model is generated.

[0011] Optionally, before converting the initial data into natural language text data using a preset data conversion model, the method further includes:

[0012] The parameters of the preset data conversion model are adjusted based on information related to the business domain corresponding to the initial data, so that the initial data can be converted into natural language text data using the preset data conversion model with adjusted parameters.

[0013] Optionally, the step of converting the initial data into natural language text data using a preset data transformation model includes:

[0014] The data excluding the text description data in the initial data is transformed using the preset data transformation model to obtain the transformed data.

[0015] The converted data and the text description data are subjected to text fusion processing to obtain the natural language text data.

[0016] Optionally, after desensitizing the natural language text data using a preset text data desensitization model and in conjunction with preset user privacy requirements to obtain the desensitized text data, the process further includes:

[0017] The de-identified text data is judged using preset privacy level judgment rules;

[0018] If it is determined that the privacy level of the de-identified text data is less than the preset privacy and security level, then the relevant parameters of the preset text data de-identification model are adjusted based on deep reinforcement learning technology.

[0019] Using a preset text data anonymization model with adjusted parameters, and in conjunction with the preset user privacy requirements, the natural language text data is anonymized to obtain new anonymized data, so as to generate a corresponding metaverse character scene model based on the new anonymized data.

[0020] Optionally, before generating the corresponding metaverse character / scene model based on the desensitized text data, the method further includes:

[0021] Several model parameters are randomly selected from the dataset corresponding to the metaverse character scene model, and the model parameters are encoded to obtain the first feature vector;

[0022] The first feature vector is subjected to noise processing to obtain a second feature vector, so as to generate the metaverse character scene model based on the second feature vector and the desensitized text data.

[0023] Optionally, generating the metaverse character scene model based on the second feature vector and the desensitized text data includes:

[0024] Based on the preset vector generation model, a corresponding feature vector is generated according to the desensitized text data to obtain the third feature vector;

[0025] The metaverse character scene model is generated based on the second feature vector and the third feature vector.

[0026] Optionally, generating the metaverse character scene model based on the second feature vector and the third feature vector includes:

[0027] The second feature vector and the third feature vector are denoised together, and the denoised vector is decoded to obtain the metaverse character scene model.

[0028] Secondly, this application provides a character / scene modeling apparatus, comprising:

[0029] The initial data acquisition module is used to acquire initial data for generating the metaverse character and scene model; the initial data includes structured data, image data, audio data, video data, and text description data;

[0030] The data conversion module is used to convert the initial data into natural language text data using a preset data conversion model;

[0031] The data desensitization processing module is used to desensitize the natural language text data by using a preset text data desensitization model and combining preset user privacy requirements to obtain desensitized text data.

[0032] The model generation module is used to generate corresponding metaverse character scene models based on the desensitized text data.

[0033] Fourthly, this application provides an electronic device, comprising:

[0034] Memory, used to store computer programs;

[0035] A processor is used to execute the computer program to implement the character and scene modeling method described above.

[0036] Fifthly, this application provides a computer-readable storage medium for storing a computer program, which, when executed by a processor, implements the character and scene modeling method described above.

[0037] Therefore, this application first obtains initial data for generating metaverse character and scene models; the initial data includes structured data, image data, audio data, video data, and text description data; then, it uses a preset data conversion model to convert the initial data into natural language text data; next, it uses a preset text data anonymization model, combined with preset user privacy requirements, to anonymize the natural language text data, obtaining anonymized text data; and finally, it generates corresponding metaverse character and scene models based on the anonymized text data. In this way, this application can convert multimodal initial data into natural language text data, then anonymize the text data, and then generate corresponding metaverse character and scene models based on the anonymized text data. This fully explores the potential connections between various forms of data in multimodal initial data; and after anonymization, it can improve the protection of privacy data and meet users' privacy needs. Attached Figure Description

[0038] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0039] Figure 1 This is a flowchart of a character scene modeling method disclosed in this application;

[0040] Figure 2 This is a flowchart of a specific character scene modeling method disclosed in this application;

[0041] Figure 3 Here is a flowchart of a condition vector generation method disclosed in this application;

[0042] Figure 4 This is a schematic diagram of the structure of a character and scene modeling device disclosed in this application;

[0043] Figure 5 This is a structural diagram of an electronic device disclosed in this application. Detailed Implementation

[0044] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0045] This application can be effectively applied to modeling real and complex scenes in the metaverse, and fully considers users' personalized needs. By converting various cross-modal data into text data and constructing data anonymization prompts according to users' privacy requirements, it achieves text semantic-level privacy anonymization. Finally, it effectively integrates various forms of data such as structured data, text, voice, images, and videos to generate metaverse character and scene models that meet users' personalized needs, improves the efficiency of character and scene generation, and protects users' data privacy.

[0046] See Figure 1 As shown, this embodiment of the invention discloses a method for character scene modeling, including:

[0047] Step S11: Obtain initial data for generating metaverse character scene models; the initial data includes structured data, image data, audio data, video data, and text description data.

[0048] In this application, the initial data for generating the metaverse character and scene model is first obtained. It should be noted that the metaverse character and scene model is composed of 3D character and scene model parameters within the metaverse environment, including its 3D geometric shape model, material model, collision model, lighting model, motion model, and other model parameters. Accordingly, the initial data used to generate the metaverse character and scene model can include structured data, image data, audio data, video data, and related text description data. It is understood that through this cross-modal data, a metaverse character and scene model that meets the user's expectations can be obtained.

[0049] Step S12: Convert the initial data into natural language text data using a preset data conversion model.

[0050] Furthermore, after obtaining the initial data, it can be converted into corresponding natural language text data using a preset data conversion model. It should be noted that, in one specific embodiment, before converting the initial data into natural language text data using the preset data conversion model, the process may further include: adjusting the parameters of the preset data conversion model based on information related to the business domain corresponding to the initial data, so that the adjusted preset data conversion model can be used to convert the initial data into the natural language text data. Specifically, the parameters of the preset data conversion model can be adjusted based on data related to the business domain corresponding to the initial data; that is, the preset data conversion model can be reinforced and trained using data related to that business domain. This yields a preset data conversion model suitable for that business domain, and then the initial data can be converted into corresponding natural language text data using the preset data conversion model suitable for that business domain.

[0051] In a specific embodiment, converting the initial data into natural language text data using a preset data conversion model may include: converting the data in the initial data excluding the text description data using the preset data conversion model to obtain converted data; and performing text fusion processing on the converted data and the text description data to obtain the natural language text data. Specifically, the preset data conversion model can be used to convert structured data, image data, audio data, and video data in the initial data to obtain converted data; then, the obtained converted data is fused with the text description data in the initial data to obtain natural language text data in text form.

[0052] Step S13: Desensitize the natural language text data using a preset text data desensitization model and in accordance with preset user privacy requirements to obtain desensitized text data.

[0053] In this application, after converting initial data to obtain natural language text data, the obtained natural language text data can be further anonymized using a preset text data anonymization model and preset user privacy requirements to obtain anonymized text data. It should be noted that, in specific embodiments, after anonymizing the natural language text data using the preset text data anonymization model and preset user privacy requirements to obtain anonymized text data, the process may further include: judging the anonymized text data using preset privacy level judgment rules; if the privacy level of the anonymized text data is determined to be less than a preset privacy security level, then adjusting the relevant parameters of the preset text data anonymization model based on deep reinforcement learning technology; using the pre-adjusted preset text data anonymization model and the preset user privacy requirements to anonymize the natural language text data to obtain new anonymized data, so as to generate a corresponding metaverse character scene model based on the new anonymized data. Specifically, a privacy level determination rule can be preset to judge the privacy level of the de-identified data. If the privacy level of the de-identified data is lower than the preset privacy security level, then the de-identified data does not meet the user's privacy requirements. At this time, the preset text data de-identification model can be trained and adjusted based on deep reinforcement learning technology to ensure that the de-identified data obtained by using the adjusted preset text data de-identification model to de-identify natural language text data can meet the user's privacy requirements, that is, the privacy level of the new de-identified data is not lower than the preset privacy security level.

[0054] Step S14: Generate the corresponding metaverse character scene model based on the desensitized text data.

[0055] In this application, after obtaining anonymized text data corresponding to natural language text data through a preset text data anonymization model, a corresponding metaverse character scene model can be generated based on the anonymized text data. In a specific embodiment, before generating the corresponding metaverse character scene model based on the anonymized text data, the process may further include: randomly selecting several model parameters from the dataset corresponding to the metaverse character scene model and encoding the model parameters to obtain a first feature vector; adding noise to the first feature vector to obtain a second feature vector, so as to generate the metaverse character scene model based on the second feature vector and the anonymized text data. Specifically, several model parameters can be randomly selected from the dataset corresponding to the metaverse character scene model; further, the first feature vector corresponding to the model parameters can be added with noise, and after several processing steps, a Gaussian distributed second feature vector can be obtained, and subsequently, the corresponding metaverse character scene model can be generated based on the second feature vector and the anonymized text data.

[0056] In another specific embodiment, generating the metaverse character scene model based on the second feature vector and the desensitized text data may include: generating a corresponding feature vector based on the desensitized text data using a preset vector generation model to obtain a third feature vector; and generating the metaverse character scene model based on the second feature vector and the third feature vector. Specifically, the process of generating the metaverse character scene model based on the second feature vector and the desensitized text data firstly generates a third feature vector corresponding to the desensitized data based on the preset vector generation model, and then generates the corresponding metaverse character scene model based on the second feature vector and the third feature vector. Further, generating the metaverse character scene model based on the second feature vector and the third feature vector may include: performing denoising processing on the second feature vector and the third feature vector together, and decoding the denoised vector to obtain the metaverse character scene model. Specifically, the second feature vector obtained from the dataset based on the metaverse character scene model and the third feature vector obtained from the desensitized text data may be denoised together, and the denoised vector may be decoded, thus generating the metaverse character scene model corresponding to the desensitized text data.

[0057] Therefore, this application first obtains initial data for generating metaverse character and scene models; the initial data includes structured data, image data, audio data, video data, and text description data; then, it uses a preset data conversion model to convert the initial data into natural language text data; next, it uses a preset text data anonymization model, combined with preset user privacy requirements, to anonymize the natural language text data, obtaining anonymized text data; and then, it generates corresponding metaverse character and scene models based on the anonymized text data. Furthermore, this application can use data from the business domain corresponding to the initial data to specifically train the preset data conversion model, and then use the trained preset data conversion model to convert the initial data into natural language text data; then, it can judge the privacy of the anonymized text data according to preset privacy level judgment rules, ensuring that the obtained anonymized text data meets the user's privacy requirements; and then, it generates corresponding metaverse character and scene models based on the anonymized text data, which can eliminate the risk of privacy data leakage to a greater extent.

[0058] like Figure 2 As shown in the figure, this application discloses a method for character scene modeling, including:

[0059] In this embodiment, a feature vector V (first feature vector) can be obtained by encoding several model parameters in the dataset corresponding to the metaverse character scene model. After denoising, a feature vector Vt (second feature vector) can be obtained. Furthermore, the user-provided information (initial data) and user privacy requirements can be processed by a cross-modal conditional vector generator (ConVect-Gen) to obtain a conditional vector Vc (third feature vector). Then, the feature vector Vg can be obtained by denoising Vt and Vc. Decoding Vg can then yield the metaverse character scene model corresponding to the initial data.

[0060] Among them, such as Figure 3 As shown, the process by which the cross-modal conditional vector generator obtains conditional vectors based on initial data can include: processing the initial data (user-provided information) to obtain a third feature vector; the process can also include: converting the user-provided information into text; whereby the information provided by the user for generating the metaverse character scene model can include structured data, text description data (personalization requirements), image data, audio data, and video data. Different forms of data can be converted into corresponding text through corresponding modules; it is understood that the feature processing module can be based on a transformer structure; specifically, structured data can be converted into corresponding text using an S2T converter, image data can be converted into corresponding text using an I2TC converter, audio data can be converted into corresponding text using an A2T converter, and video data can be converted into corresponding text using a V2T converter; subsequently, the converted text and the text description data (personalization requirements) can be aggregated to obtain natural language text data. Furthermore, the user privacy requirements processed by the Prompt-Gen data generator can be input together with the obtained natural language text data into the Text-Desens module. This Text-Desens module can be based on the GPT structure of a large NLP model, and the Prompt-Gen data generator can convert the user privacy requirements into desensitized text prompt data. This allows for the desensitization of the obtained natural language text data based on the user privacy requirements, resulting in desensitized text (desensitized text data). Then, based on the desensitized text data, a corresponding conditional vector (third feature vector) can be obtained. Specifically, the Cross-Fusion desensitized data fusion module can transform the user-input cross-modal data and the desensitized text data, projecting them into a vector space to generate a fusion conditional vector, i.e., the third feature vector. This can then be combined with... Figure 2The feature vectors Vt and Vc are denoised, and the denoised feature vector Vg is decoded to obtain a metaverse character scene model that meets user privacy requirements. In one specific embodiment, user feedback data on the generated metaverse character scene model can be collected to adjust the parameters of the relevant preset model in this application.

[0061] Therefore, this application can perform anonymization processing on cross-modal data and generate corresponding condition vectors, i.e., the third feature vector, based on the cross-modal data of the user-generated metaverse character scene model in combination with user privacy requirements; at the same time, this application can obtain feature vectors, i.e., the second feature vector, based on several model parameters in the model dataset corresponding to the metaverse character scene model; then, by combining the second feature vector and the third feature vector, a metaverse character scene model that meets user privacy requirements and corresponds to the cross-modal data can be generated.

[0062] like Figure 4 As shown, this application discloses a character scene modeling device, comprising:

[0063] The initial data acquisition module 11 is used to acquire initial data for generating the metaverse character scene model; the initial data includes structured data, image data, audio data, video data, and text description data;

[0064] Data conversion module 12 is used to convert the initial data into natural language text data using a preset data conversion model;

[0065] Data desensitization processing module 13 is used to desensitize the natural language text data by using a preset text data desensitization model and combining preset user privacy requirements to obtain desensitized text data;

[0066] The model generation module 14 is used to generate corresponding metaverse character scene models based on the desensitized text data.

[0067] Therefore, this application can convert multimodal initial data into natural language text data, then perform anonymization processing on the text data, and then generate corresponding metaverse character scene models based on the anonymized text data. This can fully explore the potential connections between various forms of data in multimodal initial data; and after anonymization processing, it can improve the protection of privacy data and meet users' privacy needs.

[0068] In one specific embodiment, the device may further include:

[0069] The first model adjustment unit is used to adjust the parameters of the preset data conversion model according to information related to the business domain corresponding to the initial data, so as to use the preset data conversion model with adjusted parameters to convert the initial data into the natural language text data.

[0070] In one specific embodiment, the data conversion module 12 may include:

[0071] The data conversion unit is used to convert the data in the initial data, excluding the text description data, using the preset data conversion model to obtain the converted data.

[0072] The data fusion unit is used to perform text fusion processing on the converted data and the text description data to obtain the natural language text data.

[0073] In one specific embodiment, the device may further include:

[0074] A data privacy level determination unit is used to determine the de-identified text data using preset privacy level determination rules;

[0075] The second model adjustment unit is used to adjust the relevant parameters of the preset text data desensitization model based on deep reinforcement learning technology when it is determined that the privacy level corresponding to the desensitized text data is less than the preset privacy security level.

[0076] The data desensitization unit is used to desensitize the natural language text data using a preset text data desensitization model with adjusted parameters, combined with the preset user privacy requirements, to obtain new desensitized data, so as to generate a corresponding metaverse character scene model based on the new desensitized data.

[0077] In one specific embodiment, the device may further include:

[0078] The first vector generation unit is used to randomly select several model parameters from the dataset corresponding to the metaverse character scene model, and encode the model parameters to obtain the first feature vector.

[0079] The second vector generation unit is used to add noise to the first feature vector to obtain a second feature vector, so as to generate the metaverse character scene model based on the second feature vector and the desensitized text data.

[0080] In one specific embodiment, the model generation module 14 may include:

[0081] The third vector generation unit is used to generate a corresponding feature vector based on the desensitized text data according to a preset vector generation model, and obtain the third feature vector.

[0082] The model generation submodule is used to generate the metaverse character scene model based on the second feature vector and the third feature vector.

[0083] In another specific embodiment, the model generation submodule may include:

[0084] The model generation unit is used to denoise the second feature vector and the third feature vector together, and decode the vector obtained after denoising to obtain the metaverse character scene model.

[0085] Furthermore, embodiments of this application also disclose an electronic device, Figure 5 This is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content of the diagram should not be construed as limiting the scope of this application.

[0086] Figure 5 This is a schematic diagram of the structure of an electronic device 20 provided in an embodiment of this application. Specifically, the electronic device 20 may include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 stores a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the character scene modeling method disclosed in any of the foregoing embodiments. Furthermore, the electronic device 20 in this embodiment may specifically be an electronic computer.

[0087] In this embodiment, the power supply 23 is used to provide operating voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and external devices, and the communication protocol it follows can be any communication protocol applicable to the technical solution of this application, and is not specifically limited here; the input / output interface 25 is used to acquire external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs, and is not specifically limited here.

[0088] In addition, the memory 22, as a carrier for resource storage, can be a read-only memory, random access memory, disk or optical disk, etc. The resources stored thereon can include operating system 221, computer program 222, etc., and the storage method can be temporary storage or permanent storage.

[0089] The operating system 221 is used to manage and control the various hardware devices on the electronic device 20 and the computer program 222, which may be Windows Server, Netware, Unix, Linux, etc. In addition to including computer programs capable of performing the character and scene modeling method executed by the electronic device 20 as disclosed in any of the foregoing embodiments, the computer program 222 may further include computer programs capable of performing other specific tasks.

[0090] Furthermore, this application also discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, it implements the aforementioned character and scene modeling method. Specific steps of this method can be found in the corresponding content disclosed in the foregoing embodiments, and will not be repeated here.

[0091] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section.

[0092] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0093] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0094] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0095] The technical solutions provided in this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for character scene modeling, characterized in that, include: Obtain initial data for generating metaverse character and scene models; The initial data includes structured data, image data, audio data, video data, and text description data; The initial data is converted into natural language text data using a preset data conversion model; By using a preset text data desensitization model and combining preset user privacy requirements, the natural language text data is desensitized to obtain desensitized text data. Based on the anonymized text data, a corresponding metaverse character scene model is generated.

2. The character and scene modeling method according to claim 1, characterized in that, Before converting the initial data into natural language text data using a preset data conversion model, the process also includes: The parameters of the preset data conversion model are adjusted based on information related to the business domain corresponding to the initial data, so that the initial data can be converted into natural language text data using the preset data conversion model with adjusted parameters.

3. The character and scene modeling method according to claim 1, characterized in that, The process of converting the initial data into natural language text data using a preset data conversion model includes: The data excluding the text description data in the initial data is transformed using the preset data transformation model to obtain the transformed data. The converted data and the text description data are subjected to text fusion processing to obtain the natural language text data.

4. The character and scene modeling method according to claim 1, characterized in that, After desensitizing the natural language text data using a preset text data desensitization model and in accordance with preset user privacy requirements to obtain the desensitized text data, the process further includes: The de-identified text data is judged using preset privacy level judgment rules; If it is determined that the privacy level of the de-identified text data is less than the preset privacy and security level, then the relevant parameters of the preset text data de-identification model are adjusted based on deep reinforcement learning technology. Using a preset text data anonymization model with adjusted parameters, and in conjunction with the preset user privacy requirements, the natural language text data is anonymized to obtain new anonymized data, so as to generate a corresponding metaverse character scene model based on the new anonymized data.

5. The character scene modeling method according to any one of claims 1 to 4, characterized in that, Before generating the corresponding metaverse character scene model based on the desensitized text data, the process also includes: Several model parameters are randomly selected from the dataset corresponding to the metaverse character scene model, and the model parameters are encoded to obtain the first feature vector; The first feature vector is subjected to noise processing to obtain a second feature vector, so as to generate the metaverse character scene model based on the second feature vector and the desensitized text data.

6. The character and scene modeling method according to claim 5, characterized in that, The process of generating the metaverse character scene model based on the second feature vector and the desensitized text data includes: Based on the preset vector generation model, a corresponding feature vector is generated according to the desensitized text data to obtain the third feature vector; The metaverse character scene model is generated based on the second feature vector and the third feature vector.

7. The character and scene modeling method according to claim 6, characterized in that, The process of generating the metaverse character scene model based on the second feature vector and the third feature vector includes: The second feature vector and the third feature vector are denoised together, and the denoised vector is decoded to obtain the metaverse character scene model.

8. A character and scene modeling device, characterized in that, include: The initial data acquisition module is used to acquire the initial data for generating the metaverse character and scene models; The initial data includes structured data, image data, audio data, video data, and text description data; The data conversion module is used to convert the initial data into natural language text data using a preset data conversion model; The data desensitization processing module is used to desensitize the natural language text data by using a preset text data desensitization model and combining preset user privacy requirements to obtain desensitized text data. The model generation module is used to generate corresponding metaverse character scene models based on the desensitized text data.

9. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program to implement the character scene modeling method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, Used to store a computer program, which, when executed by a processor, implements the character and scene modeling method as described in any one of claims 1 to 7.

Images