WO2025194356A1 - Method for quickly generating multiple groups of customized user avatars - Google Patents

Abstract

Disclosed in the present invention is a method for quickly generating multiple groups of customized user avatars, comprising: a user selects a plurality of classification tags such as character tags, action/background tags and object/accessory tags according to his/her preferences; a computation processing unit combining the classification tags into a tag parameter group, extracts a model parameter list from a model database, and screens for a plurality of identical or similar model parameters from the model parameter list on the basis of the tag parameter group; the computation processing unit extracts a plurality of corresponding avatar models from the model database on the basis of these model parameters, then packs these avatar models, and sends same to an application program; and the application program receives these avatar models, and unpacks same for display, allowing for the user to select, and if the user selects one of these avatar models, the application program binds the avatar model with the user.

Description

Method for quickly generating multiple sets of customized user avatars Technical Field

The present invention relates to a method for generating a head portrait image, and in particular to a method for quickly generating multiple user head portraits.

Background Art

Existing social media apps often allow users to change their profile pictures. However, current methods for changing profile pictures are relatively fixed. For example, they provide a few default images for users to choose from, or use user-uploaded pictures as profile pictures. These methods lack personalization and have low recognition, failing to meet user experience requirements.

Some apps also generate user avatars based on user photos using drawing software, similar to the functionality of beauty cameras. However, this real-time image generation method can take 5 to 10 minutes to create, requiring user participation and is very time-consuming. Furthermore, it cannot meet the needs of users with specific requirements.

Summary of the Invention

In view of this, the present invention provides a method for quickly generating multiple sets of customized user avatars, which solves the shortcomings of user avatars in the existing technology, such as lack of personalization and low recognition. It can also quickly generate multiple sets of user avatars for users to choose from based on their preferences, reducing production time and increasing the specificity of the avatars.

A method for rapidly generating multiple sets of customized user avatars according to the present invention includes:

The administrator selects several classification tags including character tags, action/background tags, and object/accessory tags;

The processing unit obtains corresponding label parameters from the label database according to the classification labels, combines the label parameters into a plurality of label parameter groups according to a label combination method, and stores the combined labels into a label parameter group list;

The processing unit extracts a corresponding image file list from the multimedia database according to the classification tags involved in the tag parameter group list, and compiles a plurality of model parameters according to the tag parameter group list and the image file list and stores them into a model parameter list;

An avatar training unit extracts a list of model parameters from the model database, extracts several corresponding images from the multimedia database according to each set of model parameters, and uses a deep learning text-to-image generative model (Stable Diffusion) to generate several avatar models. These avatar models are respectively matched with these model parameters and stored in the model database.

The user executes an application on an electronic device to open a category tag selection interface. The user selects several category tags including character tags, action/background tags, and object/accessory tags according to his or her preferences. The application transmits these category tags to the processing unit; the processing unit combines these category tags into a tag parameter group, and extracts a model parameter list from a model database, and filters out several model parameters corresponding to the same or similar tag parameter groups from the model parameter list according to the tag parameter group; further, the processing unit extracts several corresponding avatar models from the model database according to these model parameters, and then packages these avatar models and transmits them to the application; the application receives these avatar models and unpacks and displays them for the user to select; if the user selects one of these avatar models, the application binds the selected avatar model to the user.

If there is no avatar model that the user is satisfied with, the user can click a regenerate indicator to make the application notify the processing unit to extract the corresponding several images from the multimedia database according to each group of model parameters corresponding to the tag parameter group. The data is sent to the avatar training unit, which then uses a deep learning text-to-image generation model (Stable Diffusion) to generate these avatar models in real time. These avatar models are then packaged and sent to the application for the user to select.

Classification labels can specify three ways of expressing user preferences so that AI can create images that meet user expectations.

The deep learning text-to-image generation model (StableDiffusion) is preset to generate avatar models in a less realistic direction, such as anthropomorphism and cartoonization, avoiding overly realistic styles.

ChatGPT generates highly relevant action descriptions based on the text selected or entered by the user in the category label selection interface. To ensure diversity in the images generated by the category labels, one category label will generate multiple action descriptions to help generate images. For example, if the action category label is dessert, the generated images include eating pie, cooking cookies, enjoying chocolate cake, etc.

Based on the user's preferences and the generated action description, a highly relevant image background is generated. For example, if the action classification label is cooking cookies, the generated background image includes a restaurant, a kitchen, etc.

Generate highly relevant accessories based on user preferences and generated action descriptions.

Each avatar model image is generated based on multiple preferences and user selections. In addition to generating images that match the three preferences as much as possible, these generated tags will be recorded in the database. When the user selects three category tags to be generated on the application, the system will search the database for the most similar avatar model and display it on the selection page.

The application will record the user's selection and non-selection when the user registers the avatar model, and will superimpose the user behavior.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a system diagram of a method for rapidly generating multiple sets of customized user avatars;

FIG2 is a schematic diagram of a method for generating image files corresponding to classification labels;

FIG3 is a flow chart of an avatar model training method;

FIG4 is a flow chart of a method for a user to generate a user avatar;

FIG5 is a schematic diagram of an embodiment of generating a user avatar according to a selected category tag;

FIG6 is a schematic diagram of an embodiment of generating a user avatar according to a selected category tag;

FIG7 is a schematic diagram of an embodiment of generating a user avatar based on input category labels;

FIG8 is a schematic diagram of an embodiment of generating a user avatar based on a text string input in a category tag.

Explanation of the accompanying symbols: 10-database server; 11-label database; 12-multimedia database; 13-model database; 14-user database; 20-model training server; 21-processing unit; 22-avatar training unit; 30-electronic device; 31-application; 32-display screen; 50-Internet; 60-classification label list; 70-deep learning text to generate image model; 100-system for quickly generating multiple sets of customized user avatars; A10~A50-avatar model training method process; A100~A180-user generation user avatar method process.

DETAILED DESCRIPTION

The present invention is described in detail below with reference to the accompanying drawings. Whenever possible, the same reference numerals are used in the drawings and the following description to refer to the same or similar parts. Although this specification describes several illustrative embodiments, modifications, adaptations, and other implementations are possible. For example, substitutions, additions, or modifications may be made to the components and steps shown in the drawings, and the illustrative methods described herein may be modified by substituting, reordering, removing, or adding steps to the disclosed methods. Therefore, the present invention is not limited to the disclosed embodiments, and the scope of protection of the present invention is defined by the claims.

In this specification, content is a concept that refers to information or individual information elements implemented in text, images, videos, audio files, or a combination thereof, and can be displayed.

In this specification, terms such as "unit," "device," "terminal," "server," or "system" generally refer to the combination of hardware and software executed by the corresponding hardware. For example, the hardware may be a data processing device, such as a mobile device or personal computer with a built-in central processing unit or other processor. Furthermore, the software executed by the hardware may refer to the executed program, object, executable file, thread, or program.

Those skilled in the art should have general knowledge of computer structure and computer organization, and will understand that the present invention is not limited to the form of these computers and the architecture of network connections. The database server and model training server of the present invention can be composed of a plurality of computers, as long as they provide corresponding functions. The computer includes a processor, a memory connected to the processor module, and a servo network interface. The processor can be used to execute operating systems and applications stored in the memory, including a database management system (DBMS), a web server, and/or a web application server, to implement the multiple steps mentioned in the embodiments of the present invention. The present invention as a whole can be implemented through a website platform, an application (APP), or other means.

FIG1 shows a system 100 for rapidly generating multiple customized user avatars according to the present invention. The system 100 includes:

A database server 10, comprising:

The tag database 11 includes a plurality of category tags and a category tag list 60. Each of the category tags corresponds to a tag parameter. The category tags include a clothing tag, an action tag, an object tag, and another tag.

The multimedia database 12 includes a plurality of images and corresponding image codes and an image list, wherein the images are associated with classification tags;

A model database 13, including a plurality of avatar models and corresponding plurality of model parameters, a label parameter group list, and a model parameter list;

User database 14, including user information and user behavior;

A model training server 20 connected to the database server 10, the model training server 20 comprising:

The processing unit 21 is used to combine the label parameters corresponding to the classification labels into a plurality of label parameter groups according to a label combination method, and store the label parameter group list in the model database 13;

The avatar training unit 22 is configured to extract the associated images from the multimedia database 12 based on the label parameters corresponding to the label parameter groups, generate the avatar models using a deep learning text-to-image model 70 (Stable Diffusion), and store the avatar models in the model database 13, where the model parameters correspond to the label parameter groups.

The user connects to the model training server 20 via the Internet 50 on the electronic device 30. The electronic device 30 includes:

Application 31 is used to provide a category tag selection interface, allowing the user to select a category tag for which the user avatar is desired to be generated on the category tag selection interface; and

A display screen 32 is used to display these avatar models for a user to select and/or register a user avatar.

Stable Diffusion is a deep learning text-to-image generation model. It is primarily used to generate detailed images from text descriptions and to generate image-to-image transformations guided by prompts. Stable Diffusion is a variant of the diffusion model, called the latent diffusion model (LDM).

The number of times each set of model parameters is generated by the deep learning text-to-image generation model 70 (Stable Diffusion) to generate these avatar models is set by an administrator.

The application 31 further includes a registration unit, through which the user registers the selected avatar model as the user's avatar. The registration unit sends a message to the model database 13 to lock the avatar model and marks a corresponding status code as registered.

The processing unit 21 further includes collecting statistics on the user's behavior based on the classification tags selected by the user and/or the model parameters corresponding to the user avatar registered by the user.

In one embodiment, the status code includes 0: removed from shelves, 1: listed, and 2: registered.

The application 31 further includes an editing unit, which allows the user to edit the user avatar.

The application 31 further includes a regeneration indicator. If the user is not satisfied with the batch of avatar models, the user can click the regeneration indicator to allow the avatar training unit 22 to generate the avatar models in real time according to the classification labels selected by the user.

Preferably, the tag combination method includes three classification tags, consisting of clothing tags, action tags, and object tags.

In one embodiment, these classification tags further include several main tags and several secondary tags, these main tags include clothing tags, action tags, object tags and other main tags, but are not limited to these, these secondary tags include a character tag, a background tag, an accessories tag, a style tag and other secondary tags, but are not limited to these.

In one embodiment, secondary tags are bound to primary tags, such as an action tag being bound to a background tag, and an object tag being bound to an accessory tag. For example, if the action tag is playing basketball, the background tag is bound to a basketball court, and if the object tag is a motorcycle, the accessory tag is bound to a helmet.

In one embodiment, the label combination method further includes randomly selecting from the primary label and the secondary label as a model training condition.

In one embodiment, the action tag is combined with the background tag, and the object tag is combined with the accessories tag. The classification tag selection interface provides users with selections based on clothing tags, action/background tags, and object/accessories tags.

In another embodiment, the category tag selection interface randomly displays the primary tags and/or the secondary tags for the user to select.

In another embodiment, the label combination method further includes the administrator setting these main labels and/or these secondary labels as model training conditions, and the administrator can also set the classification label selection interface to display these main labels and/or these secondary labels for user selection.

In one embodiment, the classification labels corresponding to the model parameters with an avatar generation success rate higher than a threshold are set as the training conditions of the avatar training unit 22 and/or the classification labels displayed on the classification label selection interface, and the avatar generation success rate calculation method includes the success rate of satisfying the user's needs in the first generation.

In one embodiment, the category tags displayed on the category tag selection interface are accumulated from all the category tags selected by the user on the application 31, and are then sorted and displayed according to the category statistics and proportions or the top three with the highest numbers.

Clothing labels include basketball uniforms, swimsuits, hip-hop clothing, etc., but are not limited to these.

Action tags include shooting, swimming, dancing, etc., but are not limited to these.

Object labels include, but are not limited to, basketballs, swimming goggles, and jazz drums.

Background labels include, but are not limited to, basketball court, beach, snow scene, etc.

Accessory labels include, but are not limited to, badges, headbands, swim rings, tattoos, etc.

Character tags include humans, anime characters, virtual characters, animals, etc., but are not limited thereto.

Style tags include Chinese retro style, Japanese style, American style, Disney style, personification, cartoon style, etc., but are not limited to these.

In one embodiment, the category tag selection interface further includes a category tag input field. The category tags are input by the user in the category tag input field and are classified into similar category tags after natural language analysis.

These images and their image codes are all associated with relevant classification labels. For example, the classification label is the clothing label (A), the image is the basketball uniform (a), and the image codes are basketball uniform image 1 (the image code is Aa001), basketball uniform image 2 (the image code is Aa002), and basketball uniform image 3 (the image code is Aa003).

The file size of the avatar model includes the large image (for example, resolution 512*512) and the thumbnail (for example, resolution 128*128).

User information includes name, account number, password, interests, gender, age, blood type, etc.

User behavior is further collected through APP tracking points to collect user interaction data on platforms associated with the application 31 server, including clicks on content, participation in competitions, joining teams, etc.; for example, in the user's past interaction data record statistics {'basketball':{'cnt':100,'pref':0.8}}, it means that the user has interacted with a total of 100 basketball-related categories of content, accounting for 80% of the user's preferences.

The model training server 20 further includes a similarity calculation unit for calculating a model similarity between the portrait models in the same group.

The system 100 of the present invention for quickly generating multiple sets of customized user avatars further includes an administrator unit. The administrator can set an avatar model similarity percentage and extract avatar models with a similarity percentage higher than or equal to the avatar model similarity percentage from the model database 13 in the form of groups. The administrator can further select which avatar models to retain or delete.

The manager unit also includes a head portrait model optimization record. The manager screens the head portrait models stored in the model database 13. The manager unit stores the screening record as a head portrait model optimization record and transmits it to the head portrait training unit 22 for learning and training.

The electronic device 30 includes a computer, a tablet, a smart watch, a personal computer (PC), a mobile terminal, or the like.

A method for rapidly generating multiple sets of customized user avatars according to the present invention is shown in FIG2 , wherein a method for generating images corresponding to classification labels includes an avatar training unit 22 extracting a classification label list 60 from a label database 11 , generating a plurality of images based on the textual content of the plurality of classification labels using a deep learning text-to-image generation model 70 (Stable Diffusion), and mapping these images to these classification labels and storing them in a multimedia database 12 .

FIG3 shows a method for quickly generating multiple sets of customized user avatars according to the present invention, including: step A10, the processing unit 21 extracts a plurality of classification tags and corresponding plurality of tag parameters stored in a tag database 11, combines the tag parameters corresponding to these classification tags into a plurality of tag parameter groups according to a tag combination method, and stores them as a tag parameter group list in a model database 13. An example of using 4 tags to form each group of 3 is described below, but this does not limit the present invention.

In step A20 , the processing unit 21 extracts a corresponding image list from a multimedia database 12 according to the classification tags involved in the tag parameter group list, as illustrated below but not limiting the present invention.

In step A30 , the processing unit 21 compiles a plurality of model parameters according to the label parameter list and the image file list and stores the model parameters into a model parameter list in the model database 13 . The following is an example but does not limit the present invention.

In step A40, these model parameters are analyzed by natural language to exclude unreasonable groups. For example, the group of basketball uniform-shooting-bat (a001-e001-k001) will be determined to be unreasonable after natural language analysis, and the group will be deleted from the model parameter list. Exceptionally, this step can be omitted to increase interest and creative space.

In step A50, the avatar training unit 22 extracts a list of model parameters from the model database 13, extracts several corresponding images from the multimedia database 12 according to each set of model parameters, and uses a deep learning text-to-image model 70 (Stable Diffusion) to generate several avatar models. These avatar models are respectively assigned to these model parameters and stored in the model database 13.

In one embodiment, the avatar training unit 22 feeds back the generated avatar models, based on user registration and/or behavior statistics selected by an administrator, to the avatar training unit 22 for training and learning.

In one embodiment, these images include multimedia images collected from other platforms, which are stored in the multimedia database 12 according to classification tags after being reviewed and approved by the administrator.

FIG4 shows a method for rapidly generating multiple sets of customized user avatars according to the present invention, comprising:

A100 - electronic device 30 transmits to processing unit 21 via Internet 50 a plurality of category tags selected by the user on the category tag selection interface of application 31 displayed on display screen 32 of electronic device 30;

A110, the processing unit 21 receives the classification labels and combines them into a label parameter group;

A120, the processing unit 21 extracts a model parameter list from the model database 13, and selects a plurality of model parameters corresponding to the same or similar tag parameter groups from the model parameter list according to the tag parameter groups;

A130, the processing unit 21 extracts corresponding avatar models from the model database 13 according to the model parameters;

A140, the processing unit 21 packages the avatar models and transmits them to the application 31;

A150, the application 31 receives and unpacks the avatar models and displays them on the display screen 32 for the user to select. The display screen 32 also displays a regeneration indicator;

A160, if the user selects one of these avatar models, the application 31 binds the selected avatar model to the user and sends a registration notification to the model database 13. The model database 13 changes the status code corresponding to the avatar model to registered according to the registration notification, and the process ends;

A170, if the user clicks the regeneration indicator, the application 31 sends a regeneration notification to the processing unit 21;

A180, the processing unit 21 receives the regeneration notification, extracts the model parameters corresponding to the label parameter group from the multimedia database 12 and transmits them to an avatar training unit 22 in each group. The avatar training unit 22 uses a deep learning text-to-image generation model 70 (Stable Diffusion) to generate these avatar models in real time, corresponds these avatar models to these model parameters respectively and stores them in the model database 13; repeat A140.

In one embodiment of a method for rapidly generating multiple customized user avatars, the method for rapidly extracting avatar models from the model database 13 based on user-selected category tags includes: the user launching an application 31 on an electronic device 30, displaying a category tag selection interface on screen 32; the user selecting "dog" in the "personality" tag, "listening to music" in the "action" tag, "park" in the "background" tag, "headphones" in the "accessories" tag, and "best quality," "high resolution," "simple background," and "ultra-detailed eyes" in the "other" tag; the application 31 transmitting these category tags to the processing unit 21, which combines the category tags into a tag parameter group and extracts a model parameter list from the model database 13. Based on the tag parameter group, the model parameters corresponding to the same or similar tag parameter groups are filtered from the model parameter list, and the corresponding avatar models are then extracted from the model database 13 based on these model parameters, as shown in FIG5 . This method can rapidly provide user avatars that meet user needs, resolving the drawback of prior art techniques that require 5 to 10 minutes to generate user avatars.

In one embodiment of the method for quickly generating multiple sets of customized user avatars, the method for real-time generation of user avatars in which the user selects these category tags includes: the user opens the application 31 on the electronic device 30, and the display screen 32 displays the category tag selection interface; the user selects "cat" in the character tag, "drums" in the object tag, "performance" in the action tag, "park" in the background tag, "headphones" in the accessories tag, and "best quality", "high resolution", "simple background", "super detailed eyes", and "no human" in other tags; the application 31 transmits these category tags to the processing unit 21, and the operation The processing unit 21 combines these classification labels into label parameter groups, and extracts the corresponding model parameters of the label parameter groups from the multimedia database 12 on a per-group basis, and transmits them to the avatar training unit 22. The avatar training unit 22 uses a deep learning text-to-image generative model 70 (Stable Diffusion) to generate these avatar models in real time, as shown in FIG6 .

In one embodiment, the user selects or enters - a rabbit in the character tag, - playing guitar in the action tag, - in the park in the background tag, - wearing headphones in the accessories tag, and - best quality, high resolution, simple background, ultra-detailed eyes, no humans in the category tag input fields of the category tag selection interface; the application 31 transmits the category tags to the processing unit 21, and the processing unit 21 combines the category tags into a tag parameter group and extracts a model parameter list from the model database 13, and filters out the model parameters corresponding to the same or similar tag parameter groups from the model parameter list according to the tag parameter group, and then extracts the corresponding avatar models from the model database 13 according to these model parameters, as shown in Figure 7.

In one embodiment, a user enters a string of text in the category label input field of the category label selection interface - dog's personality, anime style, 2D image, wearing clothes, playing basketball, park background, animal city character, ink painting, full-body photo, low contrast image; the application 31 transmits the text string of the category label input to the avatar training unit 22 of the processing unit 21, and the avatar training unit 22 uses the deep learning text to generative image model 70 (Stable Diffusion) to generate these avatar models in real time, as shown in Figure 8.

In one embodiment, a method of quickly generating multiple sets of customized user avatars of the present invention includes: an electronic device transmits several category tags selected by a user in a category tag selection interface displayed on the display screen of the electronic device to a processing unit via the Internet; the processing unit receives these category tags, combines these category tags into a tag parameter group and extracts a model parameter list from a database server, further filters out several model parameters corresponding to the same or similar tag parameter groups, and then extracts several corresponding avatar models from the database server based on these model parameters, packages these avatar models and transmits them to the electronic device, which receives these avatar models, unpacks them, and displays them on the display screen for the user to select.

The display screen displays these avatar models and also displays a regeneration identifier. If the user clicks on the regeneration identifier, the electronic device transmits a regeneration notification to the processing unit; the processing unit receives the regeneration notification and transmits these classification labels to the processing unit. The processing unit uses a deep learning text-to-image model to generate these avatar models in real time, packages these avatar models and transmits them to the electronic device. The electronic device receives these avatar models, unpacks them and displays them on the display screen for the user to select.

In one embodiment, the processing unit receives a regeneration notification, extracts several corresponding images from the database server according to each group of model parameters corresponding to the label parameter group, and uses a deep learning text-to-image generation model to generate these avatar models in real time, packages these avatar models and transmits them to the electronic device, which receives these avatar models and unpacks them and displays them on the display screen for the user to select; the method for generating these images includes the processing unit extracting a classification label list from the database server, generating these images according to the text content of these classification labels using the deep learning text-to-image generation model, and storing these images corresponding to these classification labels in the database server.

The mechanism for changing these classification labels includes: 1. When the number of times the avatar model corresponding to the classification label has been registered exceeds the limit, the avatar training unit will no longer generate the avatar model of the classification label; 2. When the avatar model corresponding to the classification label has been registered The more times, the lower the probability of generating a avatar model with a classification label. 3. When the avatar model corresponding to the classification label is selected not to be registered more times, the avatar training unit will regularly generate avatar models with the classification label.

While certain exemplary embodiments and implementations have been described above, other embodiments and modifications will be apparent from this description. Accordingly, the present invention is not limited to such exemplary embodiments, but rather to the broader scope of the claims set forth and various obvious modifications and equivalent arrangements.

Claims (10)

A method for quickly generating multiple sets of customized user avatars, including: The electronic device transmits a plurality of category tags selected by the user on a category tag selection interface displayed on a display screen of the electronic device to the processing unit via the Internet; The processing unit receives the classification tags, combines them into a tag parameter group, extracts a model parameter list from the database server, selects a plurality of model parameters corresponding to the same or similar tag parameter group, extracts a plurality of corresponding head portrait models from the database server based on the model parameters, packages the head portrait models, and transmits them to the electronic device; The electronic device receives these avatar models, unpacks them, and displays them on a display screen for the user to select. The method for rapidly generating multiple customized user avatars as claimed in claim 1, wherein when the user selects one of the avatar models, the electronic device transmits a registration notification to the database server to change the status code of the avatar model. The method for rapidly generating multiple customized user avatars as claimed in claim 1 , wherein the display screen displays a regeneration indicator simultaneously with the display of the avatar models, and when the user clicks on the regeneration indicator, the electronic device transmits a regeneration notification to the processing unit. The method for rapidly generating multiple sets of customized user avatars as described in claim 3, wherein the processing unit receives the regeneration notification, transmits the classification labels to the processing unit, and the processing unit generates the avatar models in real time using a deep learning text-to-image model. The avatar models are packaged and transmitted to the electronic device, which receives the avatar models, unpacks them, and displays them on a display screen for user selection. The method for rapidly generating multiple sets of customized user avatars as described in claim 3, wherein the processing unit receives a regeneration notification, extracts corresponding image files for each set of model parameters corresponding to the tag parameter groups from a database server, and utilizes a deep learning text-to-image model to generate these avatar models in real time. The avatar models are packaged and transmitted to the electronic device, which receives and unpacks these avatar models and displays them on a display screen for selection by the user. The method for rapidly generating multiple customized user avatars as described in claim 5, wherein the processing unit extracts a list of classification labels from a database server, generates the images based on the text content of the classification labels using the deep learning text-to-image model, and stores the images corresponding to the classification labels in the database server. The method for rapidly generating multiple customized user avatars as claimed in claim 1, wherein the category tags include clothing tags, action tags, object tags, character tags, background tags, accessories tags, style tags, other tags, and category tag input fields. The method for rapidly generating multiple customized user avatars as claimed in claim 1, wherein the model parameters are subjected to natural language analysis to exclude unreasonable groups. The method for rapidly generating multiple customized user portraits as claimed in claim 1 , wherein the processing unit further comprises a similarity calculation for calculating a model similarity between the portrait models in the same group. The method for rapidly generating multiple sets of customized user avatars as described in claim 1, wherein these classification labels are preset by the user's user behavior, and the user behavior collects interactive data on the user's related platforms through APP tracking.