Creation assisting method and device based on 3D visualization technology

Abstract

The invention discloses a creation assisting method and device based on a 3D visualization technology, and relates to the field of 3D visualization. In the method, when a target user starts a 3D creation project, behavior data of the target user is collected, and a user portrait is constructed according to basic information, historical work styles and historical evaluation data of the target user; determining a creation intention based on the behavior data and a label in a user portrait of the target user through a preset deep learning model; screening and recommending 3D elements from a 3D element library according to the creation intention; and adjusting the preset deep learning model according to the adoption condition of the target user on the 3D elements and the relevancy and satisfaction evaluation, and recommending new 3D elements according to the adjusted preset deep learning model. By implementing the technical scheme provided by the invention, the 3D creation efficiency and quality are improved, and the user experience is improved.

Description

Technical Field

[0001] This application relates to the technical field of 3D visualization, and specifically relates to a creation assistance method and device based on 3D visualization technology. Background Art

[0002] With the progress of technology, 3D creation platforms have gradually incorporated more intelligent functions, such as automatic modeling, scene synthesis, etc., which enables creators to focus more on the creativity itself without excessive concern for technical details. However, despite the many conveniences brought by these platforms, there are still some problems that need to be solved urgently in the actual use process.

[0003] In order to improve the quality and efficiency of 3D creation, existing technical solutions mainly focus on the following aspects: controlling the generation of models by manually inputting key parameters, such as specifying specific texture or lighting settings; or using a template library to allow creators to select appropriate objects from predefined elements for combination. The main drawback of the above methods is that they lack precise understanding and support for users' personalized needs. Specifically, these methods usually cannot accurately capture users' creative intentions and preferences, resulting in the recommended 3D elements often not meeting users' actual needs, thereby reducing the creation efficiency and the quality of the final work.

[0004] Therefore, a more intelligent and personalized creation assistance method is needed. Summary of the Invention

[0005] This application provides a creation assistance method and device based on 3D visualization technology, which improves the efficiency and quality of 3D creation and enhances the user experience.

[0006] In the first aspect of this application, a creation assistance method based on 3D visualization technology is provided, which is applied to a 3D creation platform. The method includes: When a target user starts a 3D creation project, collect the behavior data of the target user, where the behavior data includes the mouse movement trajectory, keyboard input records, operation frequency and sequence during the creation process, and construct a user profile based on the basic information, historical work style, and historical evaluation data of the target user; Determine the creative intention through a preset deep learning model based on the behavior data and the tags in the user profile of the target user; Screen and recommend 3D elements from a 3D element library, where the 3D elements include textures, objects, lighting settings, and material effects; Adjust the preset deep learning model according to the adoption situation, relevance, and satisfaction evaluation of the target user for the 3D elements, and recommend new 3D elements according to the adjusted preset deep learning model.

[0007] Optionally, constructing a user portrait based on the basic information, historical work style, and historical evaluation data of the target user includes: Extracting the basic information of the target user from the user database of the 3D creation platform, where the basic information includes age, gender, professional background, hobbies, registration time, active time period, and preference settings, and constructing a basic feature vector based on the basic information; Performing theme classification, color analysis, and composition recognition on the historical works of the target user to extract style features, and constructing a style feature vector based on the style features; Obtaining the historical evaluation data of the target user on the works created by himself / herself on the 3D creation platform and the works created by other users on the 3D creation platform, where the historical evaluation data includes a first score, a first comment content, a first like count, and a first share count, and constructing an evaluation feature vector based on the historical evaluation data; Fusing the basic feature vector, the style feature vector, and the evaluation feature vector to construct a user portrait.

[0008] Optionally, the first score includes a second score and a third score, and constructing an evaluation feature vector based on the historical evaluation data includes: Obtaining the second score of the target user on his / her own works, and constructing a first sub-evaluation vector of the target user on the works created by himself / herself on the 3D creation platform based on the second score; Obtaining the third score, the first comment content, the first like count, and the first share count of the target user on other works, constructing a score feature sub-vector based on the second score, performing text mining on the comment content, extracting keywords and sentiment tendencies to form a comment content feature sub-vector, analyzing the correlation between the first like count and the first share count and the work type, style, and creation time to form a social interaction feature sub-vector, and constructing a second sub-evaluation vector based on the score feature sub-vector, the comment content feature sub-vector, and the social interaction feature sub-vector; Performing weighted fusion on the first sub-evaluation vector and the second sub-evaluation vector to construct an evaluation feature vector.

[0009] Optionally, determining the creation intention by a preset deep learning model based on the behavior data and the tags in the user portrait of the target user includes: Constructing a behavior feature vector based on the behavior data; Extracting target tags related to the creation intention from the user portrait, where the target tags include creation theme, style preference, sentiment tendency, and creation purpose, and constructing a tag feature vector based on the target tags; Input the behavioral feature vector and the label feature vector into the preset deep learning model to capture the temporal features in the behavioral data and the static features in the user profile; Output the result with the highest probability through the output layer of the preset deep learning model as the creative intention.

[0010] Optionally, the screening and recommending 3D elements from the 3D element library according to the creative intention includes: Calculate the similarity between the feature vector of the 3D element and the feature vector of the creative intention, generate a recommendation list based on the 3D elements with a similarity higher than the threshold, and sort the 3D elements in the recommendation list according to the matching degree with the creative intention; Display the recommendation list to the target user and provide an interactive interface. The target user can preview, select, and adjust the target 3D elements in the recommendation list. The interactive interface includes a rendering view of the 3D model constructed by the target 3D elements, so that the target user can view the target 3D elements through rotation, scaling, and moving operations.

[0011] Optionally, the adjustment of the preset deep learning model according to the adoption situation, relevance, and satisfaction evaluation of the 3D elements by the target user includes: Collect the adoption situation data of the target user for the target 3D elements. The adoption situation data includes the adopted element types, quantities, and unadopted elements, and obtain the fourth score of the relevance and the fifth score of the satisfaction of the target user for the target 3D elements. Input the adoption situation data, the fourth score, and the fifth score into the preset deep learning model together with the behavioral data and the user profile; Update the parameters of the preset deep learning model through the backpropagation algorithm.

[0012] Optionally, the method further includes: Statistically analyze the interaction data of all users in the platform community of the 3D creation platform. The interaction data includes the download volume, the second comment content, the second like count, the second share count, the collection, and the sixth score. Regularly analyze the changes in the popularity trend according to the interaction data, and adjust the training data and feature weights of the deep learning model.

[0013] In the second aspect of the present application, a creation assistance system based on 3D visualization technology is provided, including an acquisition module, a prediction module, a recommendation module, and an execution module, where: The acquisition module is configured to collect the behavior data of the target user when the target user starts a 3D creation project. The behavior data includes the mouse movement trajectory, keyboard input records, operation frequency and order during the creation process, and constructs a user profile based on the basic information, historical work style, and historical evaluation data of the target user; The prediction module is configured to determine the creation intention based on the behavior data and the tags in the user profile of the target user through a preset deep learning model; The recommendation module is configured to screen and recommend 3D elements from a 3D element library according to the creation intention. The 3D elements include textures, objects, lighting settings, and material effects; The execution module is configured to adjust the preset deep learning model according to the adoption situation, relevance, and satisfaction evaluation of the target user for the 3D elements, and recommend new 3D elements according to the adjusted preset deep learning model.

[0014] In the third aspect of the present application, an electronic device is provided, including a processor, a memory, a user interface, and a network interface. The memory is used to store instructions, and both the user interface and the network interface are used to communicate with other devices. The processor is used to execute the instructions stored in the memory so that the electronic device executes the method described in any one of the above.

[0015] In the fourth aspect of the present application, a computer-readable storage medium is provided. The computer-readable storage medium stores instructions that, when executed, execute the method described in any one of the above.

[0016] In summary, one or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: 1. The relevant 3D elements are accurately recommended according to the creation intention through the deep learning model, enabling the creator to quickly find the elements that meet the creation requirements, thus greatly shortening the creation time. For example, when creating a 3D scene in a science fiction style, the model can directly recommend building textures full of a sense of the future, high-tech equipment objects, cold-toned lighting settings, and metallic texture material effects, etc. The creator does not need to search one by one among many non-science fiction style elements and can directly focus on the recommended elements for creation, with a significant improvement in efficiency; 2. The collection and analysis of behavioral data can help the system understand the operating habits and sequences of creators during the creation process. Based on this, the system can provide creators with a creation process guide that better suits their operation logic. For example, when it is detected that a creator usually performs material texturing operations after completing object modeling, the system will automatically pop up material recommendations that match the style of the object after modeling and guide the creator to perform the next operation such as lighting settings, making the creation process more coherent and smooth, avoiding situations where the creator's train of thought is interrupted or they need to repeatedly modify during the creation process, and further improving the creation efficiency; 3. By constructing user portraits, the system can fully understand the historical work styles of target users, including theme preferences, color usage, composition characteristics, etc. When recommending 3D elements, it will strictly follow the style characteristics in the user portrait to ensure that the recommended elements are consistent with the styles of the user's previous works. This helps creators continue their unique artistic styles in new creation projects and makes the works more visually unified and recognizable. For example, for a creator who is good at creating realistic-style portrait paintings, the textures recommended by the system will pay more attention to the realism of details, the shapes of objects will be closer to real-life proportions, the lighting settings will be more in line with the laws of natural lighting, and the material effects will also emphasize the authenticity of texture, thus helping the creator create high-quality 3D portrait works with consistent styles; 4. The deep learning model can not only recommend elements based on existing creative intentions but also inspire the creativity of creators to a certain extent. When the 3D elements recommended by the system are slightly different from the creator's initial creative ideas but are still somewhat relevant, it may trigger new inspirations for the creators and prompt them to optimize and expand their creative plans. At the same time, the model will continuously adjust the recommendation strategy according to the adoption and evaluation of the recommended elements by the creators, making the recommended elements more in line with the creative needs of the creators and helping the creators achieve the creation of higher-quality and more creative 3D works; 5. By fully considering the basic information, historical work styles, and historical evaluation data of target users, a unique user portrait is constructed for each user, thus realizing highly personalized creation assistance. When different users use the same 3D creation platform, they will receive different recommended elements and creation guides according to their own characteristics, making each user feel that the platform is customized for themselves, greatly enhancing the user's satisfaction and loyalty to the platform. For example, for a novice creator and a senior creator, the system will recommend 3D elements suitable for their skill levels and creation styles according to their user portraits. The novice creator may receive more basic and easy-to-operate element recommendations, while the senior creator will get more challenging and professional elements, meeting the needs of different user groups; 6. For novice 3D creators, complex 3D element libraries and professional creation software are often daunting. Through intelligent recommendation and guidance, the embodiments of this application enable novice creators to more easily find creation elements and operation processes suitable for themselves, reducing the entry threshold of 3D creation. Even novices without rich creation experience and professional skills can quickly get started with the help of the system, create 3D works of a certain quality, thereby attracting more people to participate in 3D creation, expanding the user group of the 3D creation platform, and promoting the development of the 3D creation field. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 is a schematic flowchart of a creation assistance method based on 3D visualization technology disclosed in the embodiments of this application; Figure 2 is a schematic block diagram of a creation assistance system based on 3D visualization technology disclosed in the embodiments of this application; Figure 3 is a schematic structural diagram of an electronic device disclosed in the embodiments of this application.

[0018] Description of Reference Numerals: 201, acquisition module; 202, prediction module; 203, recommendation module; 204, execution module; 301, processor; 302, communication bus; 303, user interface; 304, network interface; 305, memory. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of this specification. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all of the embodiments.

[0020] In the description of the embodiments of this application, words such as "for example" or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design solution described as "for example" or "for instance" in the embodiments of this application should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Rather, the use of words such as "for example" or "for instance" is intended to present relevant concepts in a specific manner.

[0021] In the description of the embodiments of the present application, the term "plurality" means two or more. For example, a plurality of systems means two or more systems, and a plurality of screen terminals means two or more screen terminals. In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance or implicitly specifying the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of such features. The terms "comprise", "include", "have" and their variants all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0022] This embodiment discloses a creation assistance method based on 3D visualization technology, which is applied to a 3D creation platform. Figure 1 It is a schematic flowchart of the creation assistance method based on 3D visualization technology disclosed in the embodiments of the present application. As Figure 1 shown, the method includes the following steps: S101. When the target user starts a 3D creation project, collect the behavior data of the target user. The behavior data includes the mouse movement trajectory, keyboard input records, operation frequency and order during the creation process, and construct a user profile based on the basic information, historical work style, and historical evaluation data of the target user; S102. Determine the creation intention based on the behavior data and the labels in the user profile of the target user through a preset deep learning model; S103. Screen and recommend 3D elements from the 3D element library according to the creation intention. The 3D elements include textures, objects, lighting settings, and material effects; S104. Adjust the preset deep learning model according to the adoption situation, relevance, and satisfaction evaluation of the target user for the 3D elements, and recommend new 3D elements according to the adjusted preset deep learning model.

[0023] When the target user starts a new 3D creation project on the 3D creation platform, the system will start collecting the behavior data of this user. These behavior data mainly include: Mouse movement trajectory: Record the movement path of the mouse on the screen during the creation process by the user. For example, when the user is performing object modeling, the movement trajectory of the mouse in the 3D view can reflect the process of the user shaping the object's shape; when selecting tools or elements, the trajectory of the mouse moving to different options can also reflect the user's operation habits and thinking process.

[0024] Keyboard input recording: Capture various instructions, parameter settings, and text information input by the user through the keyboard during the creation process. For example, when setting the size and position parameters of an object, the specific values input by the user through the keyboard; the text content input when adding annotations or descriptions to the scene, etc. These can provide clues for understanding the user's creative intentions and operation details.

[0025] Operation frequency and sequence during the creation process: Count the number of operations of the user on different functional modules, tools, and 3D elements, as well as the sequence of these operations. For example, record whether the user first builds the scene, then models the object, or first sets the lighting and then adds the object, etc.; at the same time, count the operation frequency of the user on a certain commonly used tool. For example, a user who focuses on detailed carving may frequently use the carving tool, while a user mainly responsible for scene layout may use the object movement and scaling tools more frequently, etc.

[0026] While collecting behavioral data, the system constructs a user profile based on the basic information of the target user, the historical work style, and the historical evaluation data. Specifically, it includes: Basic information: Extract the basic information of the target user from the user database of the 3D creation platform, such as age, gender, professional background, hobbies, registration time, active time period, and preference settings, etc. These information helps to understand the user's background and creative tendencies. For example, young users may be more inclined to create works with a modern and trendy style; users engaged in the field of architectural design may pay more attention to the accuracy of proportion and structure when creating 3D architectural scenes, etc.

[0027] Historical work style: Analyze the historical works created by the target user on the platform, including theme classification (such as science fiction, fantasy, realistic, etc.), color analysis (common color combinations, color style tendencies, etc.), and composition recognition (common composition methods, perspective selections, etc.), so as to extract the user's style characteristics. These style characteristics can help the system ensure consistency with the user's previous creative style when recommending 3D elements, making the works more visually coherent.

[0028] Historical evaluation data: Obtain the historical evaluation data of the target user on the works created by themselves and other users on the platform, such as ratings, comment content, number of likes, number of shares, and interaction with other users, etc. By analyzing these evaluation data, we can understand the user's satisfaction with their own works and the possible problems and improvement directions in the creation process; at the same time, we can also understand the user's preferences for other works, so as to provide a reference for recommending 3D elements that better meet the user's aesthetic and creative needs.

[0029] Using a pre-set deep learning model, the collected behavioral data and the tags in the constructed user profile are used as inputs. The deep learning model can perform complex analysis and processing on this data, capture the patterns and features therein, so as to determine the creative intent of the target user. For example, if a user frequently uses science fiction-style object elements during the creative process and the themes of their historical works shown in the user profile are mostly science fiction-themed, the model will comprehensively consider this information and judge that the user's current creative intent may be to create a science fiction-style 3D scene or character. The deep learning model can utilize its powerful learning ability and pattern recognition function to mine the potential creative intent of the user from a large amount of data, providing an accurate basis for subsequent element recommendations. According to the determined creative intent, the system will screen out relevant 3D elements from the 3D element library for recommendation. The 3D element library contains various elements, such as textures (texture maps of different materials and styles), objects (3D models of various shapes and functions), lighting settings (different lighting types and parameter configurations), and material effects (such as visual effects like metallic texture and glass texture), etc. The system will calculate the similarity between the feature vectors of these elements and the feature vectors of the creative intent, generate a recommendation list based on the elements with a similarity higher than the set threshold, and sort them according to the degree of matching with the creative intent. Then the recommendation list is presented to the target user, and an interactive interface is provided. The user can preview, select, and adjust the recommended 3D elements on the interactive interface. For example, the user can view the rendered view of the 3D model constructed by the target 3D element through rotation, scaling, and moving operations to determine whether the element meets their creative needs. According to the adoption situation of the recommended 3D elements by the target user (i.e., whether the user has selected the recommended elements and which elements have been selected, etc.) and the user's evaluation of the relevance and satisfaction of the elements (such as the score given by the user for the relevance of the element to the creative intent and the score for the satisfaction of the element), the pre-set deep learning model is adjusted. The system will use this feedback data as new training samples and update and optimize the parameters of the model through deep learning optimization methods such as backpropagation algorithm. The adjusted deep learning model can better capture the creative needs and preferences of the user, so as to recommend more accurate and more in line with the user's creative intent and satisfaction 3D elements in the subsequent creative process, realizing the continuous improvement and optimization of the model, and further enhancing the effect and quality of creative assistance.

[0030] Collecting the behavioral data of target users, such as mouse movement trajectories, keyboard input records, etc., can accurately capture the real-time operation habits and preferences of users during the creation process. Combining the basic information of users, historical work styles, and historical evaluation data to construct user portraits can comprehensively understand the creation background, aesthetic tendencies, and skill levels of users. This lays a solid foundation for accurately determining the creation intention and recommending 3D elements subsequently, making the recommendation more in line with the personalized needs of users. The tag information in the user portrait, such as creation theme preferences, style characteristics, etc., can help the system provide more targeted creation guidance for users. The preset deep learning model can comprehensively analyze the behavioral data and the tags in the user portrait, capturing the complex patterns and correlation relationships therein. Compared with traditional rule-based or simple statistical methods, the deep learning model can learn deeper feature representations, thus more accurately identifying the creation intention of users. Screening and recommending elements from the 3D element library according to the creation intention can ensure that the recommended textures, objects, lighting settings, and material effects, etc., are highly relevant to the creation needs of users. This not only saves the time and effort of users searching in the massive element library but also improves the usability of the elements, enabling users to quickly find high-quality elements that can be used in the current creation project. The recommended 3D elements may bring new creation inspirations and ideas to users.

[0031] Optionally, constructing the user portrait according to the basic information, historical work style, and historical evaluation data of the target user includes: Extracting the basic information of the target user from the user database of the 3D creation platform, where the basic information includes age, gender, professional background, hobbies, registration time, active time period, and preference settings, and constructing a basic feature vector according to the basic information; Conducting theme classification, color analysis, and composition recognition on the historical works of the target user to extract style features, and constructing a style feature vector according to the style features; Obtaining the historical evaluation data of the target user on the works created by himself / herself on the 3D creation platform and the works created by other users on the 3D creation platform, where the historical evaluation data includes the first score, the first comment content, the first like count, and the first share count, and constructing an evaluation feature vector according to the historical evaluation data; Fusing the basic feature vector, the style feature vector, and the evaluation feature vector to construct a user portrait.

[0032] Age can reflect the life stage of the user, and users of different age groups may have differences in creative styles, subject preferences, etc. For example, young users may be more inclined to create works full of vitality and innovation, while older users may pay more attention to the expression of depth and connotation in their works. Gender will affect the aesthetic tendency and creative perspective of users to a certain extent. For example, female users may prefer softer and warmer tones in color matching and may focus more on aspects such as human emotions in creative themes; male users may prefer a tougher and stronger style, and their creative themes mostly involve technology, machinery, etc. The user's occupation will have a profound impact on their creation. For example, when creating 3D building models, architects will demonstrate more professional designs that conform to actual building specifications by virtue of their professional knowledge; artists may incorporate more unique artistic concepts and expression techniques into their works. The professional background can also reflect the user's professional skill level, providing a basis for recommending creative elements of moderate difficulty. Hobbies are directly related to the creative themes and styles that users may be interested in. If a user is interested in science fiction, their creative works are likely to revolve around the future world, space exploration, etc. When recommending elements, textures, objects, etc. related to science fiction should be emphasized; for users who are interested in history and culture, elements such as ancient buildings and ancient costumes are suitable for recommendation. Registration time reflects the user's seniority on the platform. Newly registered users may not be very familiar with the platform functions and creative processes and need more basic guidance and recommendations; while users with a long registration time may already have certain creative experience and have higher requirements for the quality and uniqueness of the recommended elements. Active time period: Understand which time period of the day the user is more inclined to create. This helps the platform provide better recommendations and creative support during the user's active period. For example, during the night creation period that the user often uses, optimize the recommendation algorithm to respond to the user's needs more quickly. Preference settings: The user's personalized settings on the platform, such as interface style preferences, toolbar layout preferences, etc. These settings reflect the user's operating habits and aesthetic preferences, providing a reference for recommending creative elements that meet the user's operational convenience and visual comfort. Quantify and encode the above-extracted basic information to form a basic feature vector. For example, age can be encoded according to different age groups, such as 1 for 18 - 30 years old, 2 for 31 - 50 years old, etc.; gender is represented by 0 and 1 for male and female respectively; the professional background can adopt the one-hot encoding method, assigning different binary encodings to different occupations; hobbies can also be encoded in a similar way. In this way, unstructured personal information is transformed into a structured feature vector, facilitating subsequent calculations and analyses. Classify the historical works of the target user by theme, such as people, scenery, architecture, science fiction, fantasy, etc. Theme classification can clarify the user's preferences and areas of expertise in creative content. For example, if most of the user's works are portrait creations with people as the theme, it indicates that the user may have high skills and unique insights in aspects such as figure modeling and expression portrayal. When making subsequent recommendations, elements related to figure creation, such as figure textures, clothing objects, etc., should be emphasized.Analyze the use of colors in historical works, including common colors, color matching styles (such as warm colors, cool colors, contrasting colors), and color saturation, brightness, and other characteristics. Color is an important part of the visual effect of the work. Through color analysis, we can understand the user's color aesthetic tendency. For example, if the user often uses high-saturation bright colors, it means that the style of his work may be more lively and flamboyant; while the user who prefers low-saturation and soft colors may be more reserved and restrained. When recommending elements, the texture and material effects of the corresponding tones can be recommended according to the user's color style to keep the color of the work harmonious and unified. Identify the composition characteristics of historical works, such as whether symmetrical composition, golden ratio composition, and the layout of the elements of the picture (such as centralized, decentralized, etc.). The composition reflects the user's visual expression techniques and creative concepts. For example, users who are good at using symmetrical composition may pay more attention to balance and stability in their works; users who like to use golden ratio composition pursue harmony and beauty in the picture. After understanding the user's composition habits, when recommending 3D elements, the layout of the elements in the scene can be considered to provide users with creative suggestions that are more in line with their composition style, helping them to create works with more visual impact and artistic sense. The results of theme classification, color analysis and composition recognition are quantified to form a style feature vector. For example, the theme classification results are represented by one-hot encoding, the color analysis results are represented by color feature values ​​(such as the RGB value of the main color, the average value of saturation and brightness, etc.), and the composition characteristics are represented by specific numerical codes (such as symmetrical composition is 1, asymmetric composition is 0, etc.). By integrating these quantified style features, a style feature vector that can fully reflect the user's creative style is obtained, providing a key basis for accurately recommending 3D elements that match the user's style. Evaluation of one's own works: that is, the target user's rating of the works he created on the 3D creation platform. These data reflect the user's satisfaction and self-awareness of his own creative results. For example, if a user gives a high score to a certain work of his own, it means that he is relatively satisfied with the work and may have used his own skills or elements in the creation process. Evaluation of other people's works: including users' ratings, comments, likes (first likes) and shares (first shares) of other users' works on the platform. These data reflect users' aesthetic tendencies and preferences for different types of works. For example, users often like and share works with a certain unique style or theme, indicating that they have a high interest in this style or theme; and make detailed evaluations of a certain detail of the work (such as material texture) in the comments, indicating that users pay more attention to this aspect and have a certain appreciation ability. The historical evaluation data obtained is processed and quantified to construct the evaluation feature vector.For rating data, numerical representation can be used directly; the comment content can be extracted through text mining technology. Keywords are represented by word frequency-inverse document frequency (TF-IDF) values, and emotional tendencies are represented by emotional polarity values ​​(such as positive emotions are 1 and negative emotions are -1); the number of likes and shares can be represented by specific values ​​or normalized values. These quantified evaluation data are integrated together to form an evaluation feature vector, which can fully reflect the characteristics and preferences of users in creation evaluation and provide a reference for recommending 3D elements that meet the user's aesthetics and interests. The basic feature vector, style feature vector and evaluation feature vector are fused. The fusion method can be weighted summation, vector splicing, etc. For example, different weights are assigned according to the importance of each feature vector, and then weighted summation is performed to obtain a comprehensive user feature vector; or the three feature vectors can be directly spliced ​​together to form a higher-dimensional feature vector. The fused feature vector contains many aspects of the user's basic information, creation style, and evaluation preferences, which can fully and accurately characterize the user's creation characteristics and needs.

[0033] By integrating the basic information of the target user, the style of historical works and historical evaluation data, the user can be fully and deeply understood from multiple dimensions. The theme classification, color analysis and composition recognition of historical works can accurately capture the unique style and preference of the user in creation. The constructed user portrait provides a solid foundation for the personalized recommendation of 3D elements. Based on the basic feature vectors, style feature vectors and evaluation feature vectors in the user portrait, the system can accurately filter out elements that highly match the user's creative intention and style from the 3D element library. This personalized recommendation can significantly improve the efficiency of users in finding suitable elements in the creative process, reduce invalid searches and attempts, and make creation smoother. The detailed information in the user portrait can also help the system provide users with more accurate creative guidance and suggestions. At the same time, based on the style characteristics and evaluation data in the user portrait, the system can provide users with targeted creative suggestions. For example, when the composition is found to be insufficient in the evaluation of the user's work, the system can recommend relevant composition tutorials or excellent cases for users to learn and refer to, helping users to improve their creative level and optimize the quality of their works. By deeply analyzing the user portrait, the system can discover the potential and possible new directions that users have not yet fully utilized in their creation.

[0034] Optionally, the first score includes a second score and a third score, and constructing an evaluation feature vector according to the historical evaluation data includes: Obtain a second rating of the target user on his / her work, and construct a first sub-evaluation vector of the target user's work created on the 3D creation platform according to the second rating; Obtaining the third rating, first comment content, first number of likes, and first number of shares of the target user for other works, constructing a rating feature subvector according to the second rating, performing text mining on the comment content, extracting keywords and sentiment tendencies to form a comment content feature subvector, analyzing the correlation between the first number of likes and the first number of shares and the type, style, and creation time of the work to form a social interaction feature subvector, and constructing a second sub-evaluation vector according to the rating feature subvector, the comment content feature subvector, and the social interaction feature subvector; The first sub-evaluation vector and the second sub-evaluation vector are weightedly fused to construct an evaluation feature vector.

[0035] The second rating refers to the rating given by the target user to the work he created on the 3D creation platform. This reflects the user's self-evaluation and satisfaction with his own work. For example, after completing a 3D model, the user may give his work a score based on his original intention of creation, the presentation effect of the work, etc. This score is the second rating. The third rating refers to the rating given by the target user to the work created by other users on the 3D creation platform. This reflects the user's evaluation criteria and preference for other people's works. For example, when browsing other users' 3D works, users will score these works based on factors such as creativity, technical application, and visual effects. These scores are the third ratings. After obtaining the second rating of the target user's own work, the first sub-evaluation vector is constructed based on these rating data. This vector mainly reflects the user's overall evaluation level of his own work. For example, if a user gives 8 points, 7 points, 9 points, etc. to multiple of his own works, these rating values ​​can be converted into a feature vector to represent the distribution of the user's satisfaction with his own work. This vector can provide the system with a quantitative reference for user self-evaluation and help the system understand the user's self-expectations and standards in the creation process. Get the third ratings of other works by the target user, and construct the rating feature subvector based on these ratings. The rating feature subvector shows the user's rating preference for works of different types and styles. For example, users generally give higher ratings to realistic works, but lower ratings to cartoon-style works. These rating data can be organized into a vector to reflect the user's style tendency when evaluating other people's works. Comment content feature subvector: Text mining is performed on the first comment content of other works by the target user. By analyzing the keywords in the comments, we can understand the points that users are most concerned about. For example, in comments such as "the details are handled well" and "the creativity is unique", "details" and "creativity" are keywords. At the same time, extract the emotional tendency of the comments to determine whether the user has a positive, negative or neutral attitude towards the work. For example, through the sentiment analysis algorithm, the comment "This work is great, I like it very much!" is judged as positive sentiment. These keywords and emotional tendencies are converted into feature vectors to represent the focus and emotional attitude of the user's comments. Social interaction feature subvector: Analyze the first likes and first shares of other works by the target user, and study the correlation between these data and the type, style and creation time of the work. For example, it is found that users are more inclined to like and share works with a sci-fi style and a relatively recent creation time, while they are less inclined to like and share works with a classical style. Based on the results of this correlation analysis, a social interaction feature sub-vector is constructed to reflect the user's preferences and behavior patterns at the social interaction level. The above-mentioned rating feature sub-vector, comment content feature sub-vector, and social interaction feature sub-vector are integrated to construct the second sub-evaluation vector.This vector integrates information such as users' ratings, comments, and social interactions on other people's works, and can fully reflect the user's behavioral characteristics and preferences when evaluating other people's works. The first sub-evaluation vector and the second sub-evaluation vector are weighted and fused. The weighting process is to assign different weight coefficients according to the importance and relevance of the two sub-vectors. For example, assuming that the weight of the first sub-evaluation vector is 0.4 and the weight of the second sub-evaluation vector is 0.6, the final evaluation feature vector is obtained through weighted calculation (such as mathematical operations such as multiplication and addition of corresponding elements). This evaluation feature vector integrates the user's evaluation information on his own works and other people's works, and can more comprehensively and accurately reflect the user's evaluation behavior patterns and preference tendencies on the 3D creation platform, providing an important reference for subsequent creation assistance and recommendations.

[0036] By obtaining the target user's second rating of his own work and the third rating of other works, comment content, number of likes and number of shares, etc., the evaluation tendency of users on the 3D creation platform can be fully and accurately reflected. The second rating directly reflects the user's satisfaction and self-awareness of his own work, while the third rating, comment content, etc. show the user's aesthetic standards and preferences for other people's works. The number of likes and shares reflects the user's recognition and willingness to spread the work from the perspective of social interaction. Integrating these data to construct an evaluation feature vector can more accurately capture the evaluation characteristics of users in different aspects, and provide a basis that is more in line with the user's subjective intention for subsequent creation assistance. Text mining of the comment content, extraction of keywords and emotional tendencies, can deeply explore the user's emotional expression and potential needs in the evaluation process. Based on the constructed evaluation feature vector, the system can more accurately adjust the 3D element recommendation strategy and improve the relevance and practicality of the recommended elements.

[0037] Optionally, determining the creative intent based on the behavior data and the label in the user portrait of the target user by using a preset deep learning model includes: constructing a behavior feature vector according to the behavior data; Extracting target tags related to creative intent from the user portrait, the target tags including creative theme, style preference, emotional tendency and creative purpose, and constructing a tag feature vector based on the target tags; Inputting the behavior feature vector and the label feature vector into the preset deep learning model to capture the temporal features in the behavior data and the static features in the user portrait; The output layer of the preset deep learning model outputs the result with the highest probability as the creative intention.

[0038] Behavioral data includes mouse movement trajectory, keyboard input records, operation frequency and sequence in the creation process, etc. These data are collected when the user starts a 3D creation project. For example, the mouse movement trajectory can reflect the user's stay time and movement path in different functional areas (such as toolbars, property panels, view areas, etc.); keyboard input records can capture the commands and shortcut key operations entered by the user; the operation frequency and sequence can reflect the user's habitual operation mode in the creation process, such as modeling the object first, then adjusting the material, and finally setting the lighting. These behavioral data are preprocessed, such as smoothing the mouse movement trajectory and removing noise points; encoding the keyboard input records, converting different commands and shortcut keys into numerical or vector forms; statistically analyzing the operation frequency and sequence, extracting key operation sequences, etc. Then, the processed data is converted into a behavioral feature vector. This vector can quantitatively represent the user's behavioral pattern and the dynamic characteristics of the creation process, providing rich behavioral information input for the deep learning model. The target label is information closely related to the creative intention from the user portrait, including the creation theme, style preference, emotional tendency and creation purpose. The creative theme refers to the core content around which the user's work revolves, such as science fiction, nature, and characters; style preferences involve the user's unique style in terms of color, composition, texture, etc., such as realism, cartoon, abstract, etc.; emotional tendency reflects the emotional atmosphere conveyed by the user's work, such as joy, sadness, mystery, etc.; the creative purpose is the original intention of the user to create, such as commercial promotion, artistic expression, learning practice, etc. After extracting these target labels from the user portrait, they are converted into label feature vectors. For example, discrete labels such as creative themes and style preferences can be converted into numerical vectors using methods such as One-Hot Encoding or Word Embedding; for labels with a certain order or hierarchy such as emotional tendency and creative purpose, ordered encoding or other appropriate encoding methods can be used. The label feature vector can present the static information in the user portrait in a quantitative form, providing the deep learning model with prior knowledge of the user's creative intention. The preset deep learning model can be a recurrent neural network (RNN), a long short-term memory network (LSTM), a convolutional neural network (CNN), or a Transformer architecture. The specific choice depends on the characteristics of the behavioral data and the label feature vector and the requirements of the creative intention recognition task. For example, if the behavior data has obvious temporal dependencies, models such as LSTM or Transformer that can capture temporal features may be more appropriate. By inputting the behavior feature vector and the label feature vector into the deep learning model, the model can simultaneously capture the temporal features in the behavior data and the static features in the user portrait. The temporal features reflect the dynamic changes and operating habits of the user during the creation process, such as the sequence and time interval between different operations; the static features reflect the user's basic creative preferences and style tendencies.The deep learning model uses a complex neural network structure and learning algorithm to comprehensively analyze and integrate these features, and dig out the potential associations and patterns, thereby providing strong support for the accurate identification of creative intent. The output layer of the deep learning model usually uses activation functions such as softmax to convert the output of the model into a probability distribution. Each output node corresponds to a possible creative intent, such as creating a science fiction scene, drawing cartoon characters, and designing realistic buildings. The model calculates the probability value of each creative intent based on the input behavior feature vector and label feature vector. By comparing these probability values, the result with the highest probability is selected as the final creative intent. For example, if the probability of "creating a science fiction scene" in the creative intent probability distribution output by the model is 0.8, the probability of "drawing cartoon characters" is 0.1, and the probability of "designing realistic buildings" is 0.1, then "creating a science fiction scene" is determined to be the user's creative intent. This probability-based method can make full use of the prediction ability of the deep learning model to provide users with accurate and reliable creative intent identification results, and provide clear guidance for subsequent creative auxiliary links such as 3D element recommendation.

[0039] The behavior feature vector constructed from the behavior data is combined with the label feature vector in the user portrait and input into the deep learning model for creative intent recognition. The behavior data reflects the real-time operation habits and preferences of users in the creative process, while the labels in the user portrait cover static information such as the user's creative theme, style preference, emotional tendency and creative purpose. This fusion of multi-source data provides the model with more comprehensive and rich feature input, which helps the model to more accurately capture the user's creative intent. The deep learning model can effectively capture the complex patterns between the temporal features in the behavior data and the static features in the user portrait. Accurate creative intent recognition is the key to accurately recommending 3D elements. By using the result with the highest probability output by the deep learning model as the creative intent, the system can filter and recommend highly matching textures, objects, lighting settings, material effects and other elements from the 3D element library based on the intent. This not only saves users time and energy searching in the massive element library, but also improves the practicality of the elements, allowing users to quickly find high-quality elements that can be used in current creative projects.

[0040] Optionally, screening and recommending 3D elements from a 3D element library according to the creative intention includes: Calculating the similarity between the feature vector of the 3D element and the feature vector of the creative intent, generating a recommendation list based on the 3D elements with similarity higher than a threshold, and sorting the 3D elements in the recommendation list according to the matching degree with the creative intent; The recommendation list is displayed to the target user, and an interactive interface is provided, so that the target user can preview, select and adjust the target 3D elements in the recommendation list. The interactive interface includes a rendering view of the 3D model constructed by the target 3D elements, so that the target user can view the target 3D elements through rotation, scaling and movement operations.

[0041] Calculate feature vectors for each 3D element in the 3D element library. These feature vectors can contain various attributes of the element, such as the color distribution of the texture, the shape characteristics of the object, the intensity and direction of the light, the reflectivity and roughness of the material, etc. For example, for a 3D object element, its feature vector may include geometric features such as the size of the object, the number of vertices, the surface curvature distribution, and the encoding information of the category to which the object belongs (such as architecture, people, props, etc.). A feature vector is also generated according to the creative intent. The feature vector of the creative intent is constructed based on the creative intent results output by the deep learning model and the relevant information in the user portrait (such as style preferences, creative themes, etc.). For example, if the creative intent is to create a natural scenery scene, the feature vector may emphasize the attributes of natural elements, such as the color range of green vegetation, the contour characteristics of mountains, etc. Calculate the similarity between the feature vector of the 3D element and the feature vector of the creative intent. Commonly used similarity calculation methods include cosine similarity, Euclidean distance, etc. Taking cosine similarity as an example, by calculating the dot product of two vectors divided by their modulus, a value between -1 and 1 is obtained. The closer the value is to 1, the higher the similarity. According to the set similarity threshold, 3D elements with a similarity higher than the threshold are screened out. The setting of the threshold can be adjusted according to actual needs and experience, with the purpose of ensuring that the recommended elements have a high correlation with the creative intent. For example, if the threshold is set to 0.8, only when the similarity between the 3D element and the creative intent is higher than 0.8, the element will be included in the recommendation list. The 3D elements with a similarity higher than the threshold are screened out to generate a recommendation list. In the recommendation list, the 3D elements are sorted according to the matching degree between the 3D elements and the creative intent. The matching degree can be directly determined by the similarity value. The higher the similarity, the higher the matching degree, and the higher the ranking. In this way, when viewing the recommendation list, users can give priority to the elements that best match the creative intent, thereby improving the creative efficiency. The generated recommendation list is displayed to the target user. The recommendation list can be presented in a graphical way, and each recommended 3D element can be displayed with a thumbnail or a simplified model view to show its appearance and basic features. At the same time, a brief description information can be provided for each element, such as the name of the element, its category, main features, etc., to help users quickly understand the basic situation of the element. Provide users with an interactive interface, allowing them to preview, select and adjust the target 3D elements in the recommendation list. The interactive interface is an important tool for users to interact with 3D elements. It provides a variety of operating functions: users can enter the detailed preview mode by clicking on the elements in the recommendation list. In preview mode, users can see the rendering view of the 3D model constructed by the 3D element. The rendering view can show the effect of the element in the actual scene with high-quality images, including light and shadow effects, material texture, etc., to help users more accurately evaluate whether the element meets the creative needs. After previewing, users can select the 3D elements they need and add them to their own creative projects.The selection operation can be a simple click of the "Add" button, and the system will pass the relevant data and parameters of the selected element to the creation software for further use by the user. The interactive interface also allows users to adjust the selected 3D elements. Adjustment operations may include basic transformation operations such as rotation, scaling, and movement. Through these operations, users can place elements in the appropriate position and adjust their size and direction to better integrate them into the creative scene. For example, users can rotate a 3D object to face a specific direction; they can scale the size of the object to make it proportional to other elements in the scene; they can move the position of the object to find the best placement. These adjustment operations provide users with greater creative flexibility, allowing users to fine-tune elements according to their own creative ideas and create ideal 3D works.

[0042] By calculating the similarity between the feature vector of the 3D element and the feature vector of the creative intent, and generating a recommendation list based on the 3D elements with similarity higher than the threshold, it can ensure that the recommended elements are highly matched with the user's creative intent. This method avoids the situation in which the recommended elements that may appear in the traditional recommendation method do not match the user's needs, allowing users to quickly find the elements that best suit their current creative projects. For example, when a user intends to create an ancient courtyard scene, the system will accurately recommend architectural textures, flower objects, soft lighting settings, and wood material effects with ancient characteristics. These elements have a high degree of match with the creative intent and can directly help users build the required scene, greatly improving the creative efficiency. Sorting the 3D elements in the recommendation list according to the degree of match with the creative intent further improves the practicality and user experience of the recommendation. When viewing the recommendation list, users can see the elements that best meet their needs at the first time without having to repeatedly screen among many recommendations. This optimized sorting method enables users to locate key elements more quickly, saving time and energy, and also improves user satisfaction with the recommendation system.

[0043] Optionally, adjusting the preset deep learning model according to the target user's adoption of the 3D element and the relevance and satisfaction evaluation includes: collecting the target user's adoption data of the target 3D element, the adoption data including the type and quantity of the adopted elements and the elements not adopted, obtaining the fourth score of the relevance and the fifth score of the satisfaction of the target user to the target 3D element, Inputting the adoption data, the fourth score, and the fifth score into the preset deep learning model together with the behavior data and the user portrait; The parameters of the preset deep learning model are updated through the back propagation algorithm.

[0044] Record the element types of 3D elements adopted by the target user, such as texture, object, lighting setting, material effect, etc. For example, the user adopted 5 different texture elements for different parts of the scene, such as the ground and wall, and adopted 3 object elements as the main props in the scene. Count the number of various 3D elements adopted by the user, which helps to understand the degree of user demand for different types of elements. For example, if the user adopts more texture elements, it means that in the current creative project, texture is very important to the user, perhaps to create a specific atmosphere or style. It is equally important to pay attention to the elements that the user did not adopt. These unadopted elements may not be consistent with the user's creative intention, or the user may feel that the effect is not good after previewing. Analyzing the unadopted elements can help the model understand which recommendations are unsuccessful, so as to avoid similar situations in subsequent adjustments. Fourth scoring (relevance): Ask users to score the relevance of the recommended 3D elements to their creative intentions. The higher the score, the higher the relevance. For example, if a user gives a recommended lighting setting element a score of 4 (out of 5), it means that the user believes that the element is highly relevant to the scene atmosphere he wants to create. The fifth score (satisfaction): measures the user's satisfaction with the recommended elements. This includes not only the quality of the elements themselves, but also the application effect of the elements in actual creation. If the user is very satisfied with the shape, details, etc. of an object element, and it works well after being placed in the scene, the user may give a high satisfaction score. The collected adoption data (including element type, quantity, and unadopted elements), the fourth score (relevance), and the fifth score (satisfaction) are input into the preset deep learning model together with the previously collected behavior data (mouse movement trajectory, keyboard input records, operation frequency and sequence) and user portraits (feature vectors constructed including basic information, historical work styles, and historical evaluation data). These data provide comprehensive feedback information for the model, covering various performances of users in the creation process and their evaluation of the recommended results. The back propagation algorithm is a commonly used optimization algorithm in deep learning. It calculates the error between the model output result and the actual target (here, the user's evaluation and adoption of the recommended elements), and then backpropagates this error from the output layer to the input layer to update the model parameters (such as weights and biases in the neural network). During each iteration, the model adjusts parameters according to the error gradient so that the error gradually decreases and the model's prediction results gradually approach the actual target. In this scenario, the model calculates the error between the current recommendation result and the user's expectation based on the user's relevance and satisfaction ratings for the recommended 3D elements, as well as the adoption data. For example, if the user's satisfaction rating for a recommended element is low, it means that the model's recommendation result is not ideal. The model will adjust the parameters related to the recommendation of the element through the back-propagation algorithm. Specifically, the weights of the neural network layers related to element feature extraction and creative intent matching calculations may be adjusted to reduce the probability of the element being recommended under similar creative intents in the future.On the contrary, if the user gives a high evaluation to the recommended element, the model will strengthen the relevant parameters and increase the recommendation weight of similar elements, so that the model can more accurately recommend 3D elements that meet the user's needs in the future.

[0045] By collecting data on the target users' adoption of 3D elements, including the type and quantity of elements adopted and elements not adopted, as well as obtaining the user's relevance (fourth score) and satisfaction (fifth score) of the elements, real-time feedback information can be provided to the model. These data directly reflect the user's actual acceptance and preference for the recommendation results, allowing the model to promptly understand the accuracy and effectiveness of its own recommendations. The above data, together with the behavioral data and user portraits, are input into the preset deep learning model, and the model parameters are updated through the back propagation algorithm, which enables the model to dynamically adjust its own parameters based on the user's real-time feedback. This process is similar to the self-learning and optimization of the model, allowing the model to gradually adapt to the user's behavior patterns and preference changes, thereby improving the accuracy and adaptability of the model to future recommendations.

[0046] Optionally, the method further includes: Collect the interactive data of all users in the platform community of the 3D creation platform, wherein the interactive data includes the download volume, the second comment content, the second number of likes, the second number of shares, the collection and the sixth score; regularly analyze the changes in popular trends based on the interactive data, and adjust the training data and feature weights of the deep learning model.

[0047] In addition to collecting the behavioral data and evaluation data of individual target users, statistical analysis of the interaction data of all users in the platform community, such as download volume, comment content, number of likes, number of shares, favorites, and ratings, can provide a broader and richer data source for the deep learning model. These data cover various behaviors and feedback of different users on the platform, reflecting the creative dynamics and user preferences of the entire community. By analyzing these interaction data, the changing trends in the current 3D creation field can be understood regularly. For example, if the download volume, number of likes, and number of shares of a certain type of 3D element (such as a texture in a certain style, a specific type of object model) suddenly increase significantly, it indicates that this type of element may be becoming a popular trend. Incorporating the data related to these popular elements and styles into the training data of the model enables the model to better capture the latest dynamics of market and user demands, providing more trendy recommendations for subsequent creative assistance. Different interaction data metrics may vary in their importance for reflecting user preferences and popular trends. For example, in some cases, the download volume and number of likes may more directly reflect the degree of user preference for 3D elements, while in other cases, the keywords and sentiment tendencies in the comment content may be more valuable for reference. By regularly analyzing the interaction data, the weights of the corresponding features in the deep learning model can be dynamically adjusted according to the influence of each metric in the current popular trend. For example, if it is found that a certain new creative technique is frequently mentioned in the comments and the number of likes and shares of the related works are also high, the weight of the comment content feature in the model can be appropriately increased to make the model pay more attention to the recommendation of elements related to this new technique. The optimized model can provide more timely and diverse 3D element recommendations according to the latest popular trends. Users can not only obtain elements that match their personal creative intentions and styles but also timely learn about and try using the current popular elements and styles, thereby enhancing the fashion sense and attractiveness of their works. At the same time, this also encourages users to continuously explore new creative directions and element combinations, inspiring creative inspiration and promoting the innovation and development of the entire 3D creation community.

[0048] This embodiment also discloses a creative assistance system based on 3D visualization technology. Figure 2 It is a schematic diagram of the modules of the creative assistance system based on 3D visualization technology disclosed in the embodiments of the present application, as Figure 2 shown. The system includes a collection module 201, a prediction module 202, a recommendation module 203, and an execution module 204, where: The collection module 201 is configured to collect the behavioral data of the target user when the target user starts a 3D creation project. The behavioral data includes the mouse movement trajectory, keyboard input records, operation frequency and sequence during the creation process, and constructs a user profile based on the basic information, historical work style, and historical evaluation data of the target user. The prediction module 202 is configured to determine a creative intention based on the behavior data and the labels in the user profile of the target user through a preset deep learning model; The recommendation module 203 is configured to screen and recommend 3D elements from a 3D element library according to the creative intention, where the 3D elements include textures, objects, lighting settings, and material effects; The execution module 204 is configured to adjust the preset deep learning model according to the adoption situation, relevance, and satisfaction evaluation of the target user for the 3D elements, and recommend new 3D elements according to the adjusted preset deep learning model.

[0049] Optionally, the acquisition module 201 is configured to: Extract the basic information of the target user from the user database of the 3D creation platform, where the basic information includes age, gender, professional background, hobbies, registration time, active time period, and preference settings, and construct a basic feature vector according to the basic information; Perform theme classification, color analysis, and composition recognition on the historical works of the target user to extract style features, and construct a style feature vector according to the style features; Obtain the historical evaluation data of the target user for the works created by himself on the 3D creation platform and the works created by other users on the 3D creation platform, where the historical evaluation data includes a second score, first comment content, first like count, and first share count, and construct an evaluation feature vector according to the historical evaluation data; Fuse the basic feature vector, the style feature vector, and the evaluation feature vector to construct a user profile.

[0050] Optionally, the first score includes a second score and a third score, and the acquisition module 201 is configured to: Obtain the second score of the target user for his own work, and construct a first sub-evaluation vector of the target user for the works created by himself on the 3D creation platform according to the second score; Obtain the third score, first comment content, first like count, and first share count of the target user for other works, construct a score feature sub-vector according to the second score, perform text mining on the comment content, extract keywords and sentiment tendencies to form a comment content feature sub-vector, analyze the correlation between the first like count and the first share count and the work type, style, and creation time to form a social interaction feature sub-vector, and construct a second sub-evaluation vector according to the score feature sub-vector, the comment content feature sub-vector, and the social interaction feature sub-vector; Perform weighted fusion on the first sub-evaluation vector and the second sub-evaluation vector to construct an evaluation feature vector.

[0051] Optionally, the prediction module 202 is configured to: Construct a behavior feature vector based on the behavior data; Extract target tags related to the creative intention from the user profile, where the target tags include creative themes, style preferences, emotional tendencies, and creative purposes, and construct a tag feature vector based on the target tags; Input the behavior feature vector and the tag feature vector into the preset deep learning model to capture the temporal features in the behavior data and the static features in the user profile; Output the result with the highest probability through the output layer of the preset deep learning model as the creative intention.

[0052] Optionally, the recommendation module 203 is configured to: Calculate the similarity between the feature vector of the 3D element and the feature vector of the creative intention, generate a recommendation list based on the 3D elements with a similarity higher than the threshold, and sort the 3D elements in the recommendation list according to the degree of matching with the creative intention; Display the recommendation list to the target user and provide an interactive interface, where the target user can preview, select, and adjust the target 3D elements in the recommendation list, and the interactive interface includes a rendering view of the 3D model constructed by the target 3D elements, so that the target user can view the target 3D elements through rotation, scaling, and moving operations.

[0053] Optionally, the execution module 204 is configured to: Collect the adoption situation data of the target user for the target 3D elements, where the adoption situation data includes the adopted element types, quantities, and unadopted elements, and obtain the fourth score of the relevance and the fifth score of the satisfaction of the target user for the target 3D elements, Input the adoption situation data, the fourth score, and the fifth score together with the behavior data and the user profile into the preset deep learning model; Update the parameters of the preset deep learning model through the backpropagation algorithm.

[0054] Optionally, the system further includes an adjustment module, and the adjustment module is configured to: Statistically analyze the interaction data of all users in the platform community of the 3D creation platform, where the interaction data includes download volume, second comment content, second like count, second share count, collection, and sixth score, and regularly analyze the changes in the popular trend according to the interaction data, and adjust the training data and feature weights of the deep learning model.

[0055] It should be noted that: when the device provided in the above embodiments realizes its functions, only the division of the above function modules is used for illustration. In actual applications, the above functions can be allocated to different function modules according to needs, that is, the internal structure of the device is divided into different function modules to complete all or part of the functions described above. In addition, the device and method embodiments provided in the above embodiments belong to the same concept. For the specific implementation process, please refer to the method embodiments and will not be elaborated here.

[0056] This embodiment also discloses an electronic device. Referring to Figure 3 , the electronic device may include: at least one processor 301, at least one communication bus 302, a user interface 303, a network interface 304, and at least one memory 305.

[0057] Among them, the communication bus 302 is used to realize the connection and communication between these components.

[0058] Among them, the user interface 303 may include a display screen (Display) and a camera (Camera). Optionally, the user interface 303 may further include a standard wired interface and a wireless interface.

[0059] Among them, the network interface 304 may optionally include a standard wired interface and a wireless interface (such as a WI-FI interface).

[0060] Among them, the processor 301 may include one or more processing cores. The processor 301 connects various parts within the entire server through various interfaces and lines, and executes various functions of the server and processes data by running or executing instructions, programs, code sets, or instruction sets stored in the memory 305, and by calling data stored in the memory 305. Optionally, the processor 301 may be implemented in at least one hardware form of digital signal processing (DSP), field-programmable gate array (FPGA), or programmable logic array (PLA). The processor 301 may integrate one or several combinations of a central processing unit (CPU), a graphics processing unit (GPU), and a modem, etc. Among them, the CPU mainly processes the operating system, user interface, and application programs, etc.; the GPU is responsible for rendering and drawing the content to be displayed on the display screen; the modem is used to process wireless communication. It can be understood that the above modem may not be integrated into the processor 301 and may be implemented separately by a single chip.

[0061] Among them, the memory 305 may include a Random Access Memory (RAM), or may include a Read-Only Memory. Optionally, the memory 305 includes a non-transitory computer-readable storage medium. The memory 305 can be used to store instructions, programs, codes, code sets or instruction sets. The memory 305 may include a program storage area and a data storage area. Among them, the program storage area can store instructions for implementing the operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above method embodiments, etc.; the data storage area can store the data involved in the above method embodiments. Optionally, the memory 305 may also be at least one storage device located far from the aforementioned processor 301. As Figure 3 shown, the memory 305 as a computer storage medium may include an operating system, a network communication module, a user interface module, and an application program of the creation assistance method based on 3D visualization technology.

[0062] In Figure 3 the electronic device shown, the user interface 303 is mainly used to provide an input interface for the user to obtain the data input by the user; and the processor 301 can be used to call the application program of the creation assistance method based on 3D visualization technology stored in the memory 305. When executed by one or more processors 301, the electronic device executes the method of one or more of the above embodiments.

[0063] It should be noted that for the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that this application is not limited by the described action sequence, because according to this application, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0064] In the above embodiments, the descriptions of the various embodiments have their own emphases. For the parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0065] In several embodiments provided in this application, it should be understood that the disclosed device can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection between each other can be through some service interfaces. The indirect coupling or communication connection of the device or unit can be in electrical or other forms.

[0066] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place, or they can be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0067] In addition, each functional unit in various embodiments of this application can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.

[0068] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory 305 and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods in various embodiments of this application. And the aforementioned memory 305 includes: various media such as USB flash drives, mobile hard disks, magnetic disks, or optical discs that can store program codes.

[0069] The above are only exemplary embodiments of the present disclosure and should not be used to limit the scope of the present disclosure. That is, all equivalent changes and modifications made in accordance with the teachings of the present disclosure still fall within the scope covered by the present disclosure. Those skilled in the art will easily think of other implementation schemes of the present disclosure after considering the disclosure of the specification. This application aims to cover any variations, uses, or adaptive changes of the present disclosure, and these variations, uses, or adaptive changes follow the general principles of the present disclosure and include common general knowledge or conventional technical means in the technical field not recorded in the present disclosure. The specification and embodiments are only regarded as exemplary, and the scope and spirit of the present disclosure are defined by the claims.

Claims

1. A creation assistance method based on 3D visualization technology, characterized in that: Applied to a 3D creation platform, the method comprises: When the target user starts a 3D creation project, the target user's behavior data is collected, including mouse movement tracks, keyboard input records, and operation frequency and sequence during the creation process, and a user portrait is constructed based on the target user's basic information, historical work style, and historical evaluation data; Determine the creative intent based on the behavior data and the labels in the user profile of the target user through a preset deep learning model; Filtering and recommending 3D elements from a 3D element library according to the creative intent, the 3D elements including textures, objects, lighting settings, and material effects; The preset deep learning model is adjusted according to the target user's adoption of the 3D element as well as the relevance and satisfaction evaluation, and new 3D elements are recommended based on the adjusted preset deep learning model.

2. The creation assistance method based on 3D visualization technology according to claim 1, characterized in that: The constructing of a user portrait according to the basic information, historical work style, and historical evaluation data of the target user includes: Extracting basic information of the target user from a user database of the 3D creation platform, the basic information including age, gender, professional background, hobbies, registration time, active time period and preference settings, and constructing a basic feature vector based on the basic information; Performing theme classification, color analysis, and composition recognition on the historical works of the target user to extract style features, and constructing a style feature vector based on the style features; Acquire historical evaluation data of the target user on the works created by the target user on the 3D creation platform and on the works created by other users on the 3D creation platform, wherein the historical evaluation data includes a first score, a first comment content, a first number of likes, and a first number of shares, and construct an evaluation feature vector according to the historical evaluation data; The basic feature vector, the style feature vector and the evaluation feature vector are fused to construct a user portrait.

3. The creation assistance method based on 3D visualization technology according to claim 2, characterized in that: The first rating includes a second rating and a third rating, and constructing an evaluation feature vector according to the historical evaluation data includes: Obtain a second rating of the target user on his / her work, and construct a first sub-evaluation vector of the target user's work created on the 3D creation platform according to the second rating; Obtaining the third rating, first comment content, first number of likes, and first number of shares of the target user for other works, constructing a rating feature subvector according to the second rating, performing text mining on the comment content, extracting keywords and sentiment tendencies to form a comment content feature subvector, analyzing the correlation between the first number of likes and the first number of shares and the type, style, and creation time of the work to form a social interaction feature subvector, and constructing a second sub-evaluation vector according to the rating feature subvector, the comment content feature subvector, and the social interaction feature subvector; The first sub-evaluation vector and the second sub-evaluation vector are weightedly fused to construct an evaluation feature vector.

4. The creation assistance method based on 3D visualization technology according to claim 1, characterized in that: Determining the creative intent based on the behavior data and the labels in the user portrait of the target user by using a preset deep learning model includes: constructing a behavior feature vector according to the behavior data; Extracting target tags related to creative intent from the user portrait, the target tags including creative theme, style preference, emotional tendency and creative purpose, and constructing a tag feature vector based on the target tags; Inputting the behavior feature vector and the label feature vector into the preset deep learning model to capture the temporal features in the behavior data and the static features in the user portrait; The output layer of the preset deep learning model outputs the result with the highest probability as the creative intention.

5. The creation assistance method based on 3D visualization technology according to claim 1, characterized in that: The screening and recommending of 3D elements from the 3D element library according to the creative intention includes: Calculating the similarity between the feature vector of the 3D element and the feature vector of the creative intent, generating a recommendation list based on the 3D elements with similarity higher than a threshold, and sorting the 3D elements in the recommendation list according to the matching degree with the creative intent; The recommendation list is displayed to the target user, and an interactive interface is provided, so that the target user can preview, select and adjust the target 3D elements in the recommendation list. The interactive interface includes a rendering view of the 3D model constructed by the target 3D elements, so that the target user can view the target 3D elements through rotation, scaling and movement operations.

6. The creation assistance method based on 3D visualization technology according to claim 5, characterized in that: The adjusting the preset deep learning model according to the target user's adoption of the 3D element and the relevance and satisfaction evaluation includes: collecting the target user's adoption data of the target 3D element, the adoption data including the type and quantity of the adopted elements and the elements not adopted, obtaining the fourth score of the relevance and the fifth score of the satisfaction of the target user to the target 3D element, Inputting the adoption data, the fourth score, and the fifth score into the preset deep learning model together with the behavior data and the user portrait; The parameters of the preset deep learning model are updated through the back propagation algorithm.

7. The creation assistance method based on 3D visualization technology according to claim 1, characterized in that: The method further comprises: Collect the interactive data of all users in the platform community of the 3D creation platform, wherein the interactive data includes the download volume, the second comment content, the second number of likes, the second number of shares, the collection and the sixth score; regularly analyze the changes in popular trends based on the interactive data, and adjust the training data and feature weights of the deep learning model.

8. A creation assistance system based on 3D visualization technology, characterized in that: It includes acquisition module, prediction module, recommendation module and execution module, among which: A collection module is configured to collect the target user's behavior data when the target user starts a 3D creation project, wherein the behavior data includes mouse movement trajectory, keyboard input record, operation frequency and sequence during the creation process, and construct a user portrait according to the target user's basic information, historical work style, and historical evaluation data; A prediction module configured to determine the creation intention based on the behavior data and the label in the user portrait of the target user through a preset deep learning model; A recommendation module configured to filter and recommend 3D elements from a 3D element library according to the creative intent, wherein the 3D elements include textures, objects, lighting settings, and material effects; An execution module is configured to adjust the preset deep learning model according to the target user's adoption of the 3D element and the relevance and satisfaction evaluation, and recommend new 3D elements according to the adjusted preset deep learning model.

9. An electronic device, characterized in that: It includes a processor, a memory, a user interface and a network interface, the memory is used to store instructions, the user interface and the network interface are both used to communicate with other devices, and the processor is used to execute the instructions stored in the memory so that the electronic device executes the method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores instructions, and when the instructions are executed, the method according to any one of claims 1 to 7 is executed.

Images