An animation engine resource asset management system and method based on artificial intelligence

Abstract

The application provides an animation engine resource asset management system and method based on artificial intelligence, which constructs a unified artificial intelligence driven asset management framework by utilizing multi-modal processing, knowledge base construction and natural language processing technology; an AI model is trained by utilizing animation engine data, 3D animation assets are generated and added to an asset library; the model is optimized by combining meta-learning, small sample learning and fine-tuning technology to give explainability; a distributed asset management platform is constructed by fusing Internet of Things, edge computing and cloud computing technology to optimize distribution and version control; the asset library is automatically expanded by utilizing deep learning; and unified data standards and development specifications are formulated to integrate artificial intelligence into animation production tools and asset management platforms. The application scheme is more intelligent, automated and humanized, provides highly explainable and controllable experience, and ensures high efficiency and scalability of the system.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, in particular to an animation engine resource asset management system and method based on artificial intelligence. BACKGROUND

[0002] Animation production is a complex and huge process, which requires a large number of 3D models, texture maps, animation clips, sound effects and various asset materials. These assets are usually stored in animation engines or special asset management systems to support subsequent animation creation.

[0003] With the increasing demand for high-quality animation in the fields of film, television, games and other fields, the number of animation assets is growing exponentially. Traditional manual management methods have been difficult to cope with the effective management and reuse of massive assets. Therefore, it is urgent to establish an intelligent animation asset management system to improve asset utilization efficiency and reduce production costs. SUMMARY

[0004] The present application is based on the above problems, and proposes an animation engine resource asset management system and method based on artificial intelligence. The present application scheme is more intelligent and automated, providing end-to-end one-stop service. At the same time, it is more humanized, deeply cooperates with users, and provides highly interpretable and controllable experience. Distributed architecture ensures the efficiency and scalability of the system.

[0005] Therefore, one aspect of the present application proposes an animation engine resource asset management system based on artificial intelligence, comprising: a management server and a content server.

[0006] The management server is configured to:

[0007] A unified artificial intelligence driven first asset management framework is constructed by using multi-modal data processing technology, knowledge base construction technology and natural language processing technology;

[0008] Obtain existing animation engine resource asset data, and use the animation engine resource asset data and a preset first neural network to train an AI model to obtain a first model;

[0009] According to the animation engine resource asset data, 3D animation assets are generated using 3D animation modeling technology and added to the animation asset library;

[0010] The first model is adjusted to improve the understanding, generation and processing ability of the AI model for the 3D animation assets by combining meta-learning technology, small sample learning technology and fine-tuning technology to obtain a second model;

[0011] An interpretable artificial intelligence technology is used to give the second model interpretability;

[0012] According to the first asset management framework and the second model, Internet of Things technology, edge computing technology and cloud computing technology are fused to build a distributed asset management platform;

[0013] Optimize asset distribution and version control process;

[0014] Using deep learning technology, new animation assets are generated based on the animation engine resource asset data, and the animation asset library is automatically expanded;

[0015] Unified data standards, interface standards and development specifications are formulated for the asset management platform;

[0016] Design human-computer interaction interface, integrate artificial intelligence technology into animation production tool and the asset management platform.

[0017] Another aspect of the application provides an animation engine resource asset management method based on artificial intelligence, comprising:

[0018] By using multi-modal data processing technology, knowledge base construction technology and natural language processing technology, a unified first asset management framework driven by artificial intelligence is constructed;

[0019] Obtain existing animation engine resource asset data, and use the animation engine resource asset data and a preset first neural network to train an AI model to obtain a first model;

[0020] According to the animation engine resource asset data, 3D animation assets are generated using 3D animation modeling technology and added to the animation asset library;

[0021] Combined with meta-learning technology, small sample learning technology and fine-tuning technology, the first model is adjusted to improve the understanding, generation and processing ability of the AI model for the 3D animation assets, and a second model is obtained;

[0022] Using explainable artificial intelligence technology, the second model is given explainability;

[0023] According to the first asset management framework and the second model, Internet of Things technology, edge computing technology and cloud computing technology are fused to build a distributed asset management platform;

[0024] Optimize asset distribution and version control process;

[0025] Using deep learning technology, new animation assets are generated based on the animation engine resource asset data, and the animation asset library is automatically expanded;

[0026] Unified data standards, interface standards and development specifications are formulated for the asset management platform;

[0027] designing human-computer interaction interfaces to integrate artificial intelligence technology into animation production tools and the asset management platform.

[0028] Optionally, the step of constructing a unified artificial intelligence-driven first asset management framework by utilizing multi-modal data processing technology, knowledge base construction technology, and natural language processing technology comprises:

[0029] Defining a unified data format standard for different types of first animation asset data;

[0030] Using computer vision technology and natural language processing technology to preprocess and standardize the first animation asset data to obtain basic asset data;

[0031] Establishing metadata standards for animation assets to describe various attributes and semantic information of animation assets;

[0032] Based on the basic asset data and metadata standards, automatically constructing a knowledge graph of animation assets;

[0033] Using graph embedding technology to represent the structured knowledge graph as a low-dimensional knowledge graph vector;

[0034] Inputting the basic asset data of different modalities (such as images, text, 3D, etc.) into a deep neural network for feature extraction;

[0035] Combining the extracted features with the knowledge graph vector to obtain a multi-modal fusion encoding vector;

[0036] Based on the multi-modal fusion encoding vector, developing AI-enhanced functional modules;

[0037] Integrating and deploying each functional module into the asset management platform;

[0038] Collecting feedback information through active learning and human-computer interaction, and continuously optimizing the performance of the asset management platform according to the feedback information;

[0039] Introducing interpretable AI and interactive learning mechanisms to enhance the interpretability and controllability of the asset management platform.

[0040] Optionally, the step of obtaining existing animation engine resource asset data and training an AI model using the animation engine resource asset data and a pre-set first neural network to obtain a first model comprises:

[0041] Collecting various resource asset data from existing animation engines and animation projects;

[0042] After classifying, labeling, and auditing the asset data, performing cleaning, format unification, and standardization processing;

[0043] augmenting a data volume of the asset data using a data augmentation technique to obtain first asset data;

[0044] converting the first asset data in different formats into a unified tensor representation format to obtain the animation engine resource asset data;

[0045] designing a network architecture of a corresponding first neural network for different types of asset data;

[0046] fusing data in different modalities in the animation engine resource asset data using a multi-head attention mechanism;

[0047] training the first neural network using the animation engine resource asset data and learning a multi-modal representation of asset data using a self-supervised learning method to obtain a general animation asset representation base model;

[0048] fine-tuning and transfer learning of the animation asset representation base model on corresponding labeled data for a specific task to generate a dedicated first model;

[0049] designing evaluation indicators and test sets to measure model performance, continuously optimizing through parameter fine-tuning and data increment based on evaluation feedback, realizing closed-loop model improvement to improve the quality of the first model.

[0050] Optionally, the step of generating 3D animation assets using 3D animation modeling technology based on the animation engine resource asset data and adding them to the animation asset library comprises:

[0051] analyzing specific asset requirements for 3D assets in historical animation projects;

[0052] determining the type, quantity, and detail level of the required 3D assets according to the specific asset requirements;

[0053] conducting creative design according to the requirement indicators to output concept design drafts and sketches;

[0054] creating an initial 3D model based on the concept design drafts and sketches;

[0055] generating 3D rendering assets using real-time or offline rendering tools based on the initial 3D model;

[0056] inputting the 3D rendering assets into animation production software to produce keyframe animations according to the shot and action design;

[0057] rendering output animation sequence frames based on the produced keyframes to form complete 3D animation segments;

[0058] Classify and organize 3D static and dynamic assets, label metadata information, and combine with other 2D and audio asset collections to build a complete animation asset library;

[0059] Deploy the animation asset library to the asset management platform for creation and retrieval calls.

[0060] Optionally, the step of adjusting the first model to improve the AI model's understanding, generation, and processing capabilities of the 3D animation assets to obtain a second model by combining meta-learning techniques, few-shot learning techniques, and fine-tuning techniques includes:

[0061] Divide the 3D animation assets into multiple first tasks, each corresponding to a 3D asset type;

[0062] Based on the first model, an optimized meta-learning algorithm or a metric-based meta-learning algorithm is used to train a meta-model that can understand, generate, and process 3D animation assets;

[0063] Collect annotated new type 3D animation asset data, use the new type 3D animation asset data as input for small sample learning, and fine-tune the meta-model to adapt to new type 3D animation assets;

[0064] Collect corresponding first 3D animation asset data for a predetermined specific application scenario;

[0065] Use fine-tuning techniques to optimize and train the meta-model using the first 3D animation asset data while preserving previously learned knowledge to obtain the second model.

[0066] Optionally, the step of giving the second model interpretability using explainable artificial intelligence techniques includes:

[0067] Embedding explainable artificial intelligence techniques into the architecture of the second model to enable the second model to generate a decision-making process that users can understand and clarify its internal reasoning logic;

[0068] Introduce an active learning strategy, where the second model actively asks users for key information to narrow down the model's uncertainty;

[0069] The second model learns the user's preferences and feedback through human-computer interaction and adjusts its behavior and decision-making accordingly;

[0070] Design an interactive learning interface for users to modify and adjust the second model's knowledge base and parameters in real time.

[0071] Optionally, the step of constructing a distributed asset management platform according to the first asset management framework and the second model, integrating Internet of Things technology, edge computing technology and cloud computing technology, comprises:

[0072] Distributed file system and distributed database are used to store asset data;

[0073] Microservice architecture and containerization technology are used to build the asset management platform, realizing service orchestration and elastic scaling;

[0074] Service mesh technology is introduced to manage cross-platform service calls and data interactions;

[0075] An edge node integrating the second model is deployed in a collaborative field to support edge-side asset processing and local caching;

[0076] Based on the second model, an edge AI model with local rendering and retrieval functions is generated;

[0077] Content distribution network technology is introduced to distribute and update cached animation assets;

[0078] An AI-based asset intelligent management engine and analysis module are deployed in the cloud;

[0079] An identity authentication and access control system is deployed to manage asset access permissions for multiple users;

[0080] Encryption and digital signature are used to protect important assets and prevent leaks and data tampering.

[0081] Optionally, the step of optimizing asset distribution and version control process comprises:

[0082] Multiple CDN nodes are deployed to realize edge caching and distribution of assets;

[0083] According to the location and router address of the CDN nodes, in combination with a preset router intelligent control strategy and a dynamic routing allocation scheme, multiple distribution paths are generated;

[0084] File system snapshot technology and data deduplication technology are used to calculate the first difference data between asset versions;

[0085] The first difference data is distributed through the distribution paths;

[0086] The CDN nodes are connected to a preset first blockchain network;

[0087] The version management data, propagation path data and equity transaction data of animation assets involved in each CDN node are all chained;

[0088] Automated copyright protection, usage authorization and revenue distribution are realized by using smart contracts of the first blockchain network;

[0089] A version control system is introduced to support distributed collaborative editing of animation assets.

[0090] Optionally, the step of automatically expanding the animation asset library by using deep learning technology to generate new animation assets based on the animation engine resource asset data comprises:

[0091] According to a preset data compression and enhancement strategy, the data volume of the first animation asset data is expanded to obtain first training data;

[0092] A generation model architecture suitable for different asset types is designed;

[0093] The first training data is used to train the generation model architecture to obtain a first asset generation model;

[0094] For a specific asset production task, the first asset generation model is fine-tuned to obtain a second asset generation model;

[0095] New animation assets are generated according to the animation engine resource asset data and the second asset generation model;

[0096] The generated new animation assets are quality evaluated, and unqualified results are filtered out.

[0097] The technical scheme of the present application is based on an animation engine resource asset management method based on artificial intelligence, which comprises the following steps: constructing a unified first asset management framework driven by artificial intelligence by using multi-modal data processing technology, knowledge base construction technology and natural language processing technology; obtaining existing animation engine resource asset data, using the animation engine resource asset data and a preset first neural network to train an AI model to obtain a first model; using 3D animation modeling technology to generate 3D animation assets according to the animation engine resource asset data and adding the 3D animation assets to an animation asset library; combining meta-learning technology, small sample learning technology and fine-tuning technology to adjust the first model to improve the understanding, generation and processing capabilities of the AI model for the 3D animation assets to obtain a second model; using explainable artificial intelligence technology to give the second model explainability; constructing a distributed asset management platform by combining the first asset management framework and the second model, and combining Internet of Things technology, edge computing technology and cloud computing technology; optimizing asset distribution and version control processes; using deep learning technology to generate new animation assets based on the animation engine resource asset data to automatically expand the animation asset library; formulating unified data standards, interface standards and development specifications for the asset management platform; designing a human-computer interaction interface and integrating artificial intelligence technology into animation production tools and the asset management platform. The present application is more intelligent and automated, provides end-to-end one-stop services, is more user-friendly, deeply cooperates with users, provides highly explainable and controllable experiences, and the distributed architecture ensures the efficiency and scalability of the system. BRIEF DESCRIPTION OF DRAWINGS

[0098] Figure 1 is a schematic block diagram of an animation engine resource asset management system based on artificial intelligence provided by an embodiment of the present application;

[0099] Figure 2 is a flowchart of an animation engine resource asset management method based on artificial intelligence provided by an embodiment of the present application. DETAILED DESCRIPTION

[0100] In order to more clearly understand the above-mentioned purposes, features and advantages of the present application, the present application will be further described in detail below with reference to the drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0101] In the following description, many specific details are set forth in order to provide a thorough understanding of the present application, however, the present application can also be implemented in other ways different from those described herein, therefore, the scope of protection of the present application is not limited by the specific embodiments disclosed below.

[0102] The terms "first", "second", and the like in the description and in the claims of the present application and above-described drawings are used to distinguish different objects, and are not used to describe a particular order. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the listed steps or units, but can optionally include steps or units not listed, or can optionally include other steps or units inherent to the process, method, product, or device.

[0103] Reference herein to "embodiments" means that a particular feature, structure, or characteristic described in connection with an embodiment can be included in at least one embodiment of the application. The appearance of the phrase in various places in the specification does not necessarily all refer to the same embodiment, nor does it necessarily exclude other embodiments that are not explicitly identified as such. It is explicitly and implicitly understood that the embodiments described herein can be combined with other embodiments.

[0104] Some embodiments of the application will now be described with reference to the following drawings: Figures 1 to 2 An artificial intelligence-based animation engine resource asset management system and method according to some embodiments of the application will be described below with reference to

[0105] As shown in Figure 1 , one embodiment of the application provides an artificial intelligence-based animation engine resource asset management system, comprising: a management server and a content server;

[0106] The management server is configured to:

[0107] By utilizing multi-modal data processing technology, knowledge base construction technology, and natural language processing technology, a unified artificial intelligence-driven first asset management framework is constructed;

[0108] Obtain existing animation engine resource asset data, and use the animation engine resource asset data and a preset first neural network to train an AI model to obtain a first model;

[0109] According to the animation engine resource asset data, 3D animation assets are generated using 3D animation modeling technology and added to an animation asset library;

[0110] In combination with meta-learning technology, small sample learning technology, and fine-tuning technology, the first model is adjusted to improve the understanding, generation, and processing capabilities of the AI model for the 3D animation assets, to obtain a second model;

[0111] An interpretable artificial intelligence technology is adopted to give the second model interpretability;

[0112] According to the first asset management framework and the second model, Internet of Things technology, edge computing technology and cloud computing technology are fused to build a distributed asset management platform.

[0113] Optimize asset distribution and version control processes.

[0114] Using deep learning technology, new animation assets are generated based on the animation engine resource asset data to automatically expand the animation asset library.

[0115] A unified data standard, interface standard and development specification are formulated for the asset management platform.

[0116] A human-computer interaction interface is designed, and artificial intelligence technology is integrated into the animation production tool and the asset management platform.

[0117] It should be understood that, Figure 1 It should be understood that,

[0118] Referring to Figure 2 The present application provides another embodiment of an animation engine resource asset management method based on artificial intelligence, comprising:

[0119] A unified first asset management framework driven by artificial intelligence is constructed by using multi-modal data processing technology, knowledge base construction technology and natural language processing technology (to cover the whole life cycle of assets, realize data standardization, knowledge fusion and closed-loop optimization);

[0120] Obtain existing animation engine resource asset data, and use the animation engine resource asset data and a preset first neural network training AI model to obtain a first model;

[0121] According to the animation engine resource asset data, 3D animation assets are generated using 3D animation modeling technology and added to the animation asset library;

[0122] In combination with meta-learning technology, small sample learning technology and fine-tuning technology, the first model is adjusted to improve the understanding, generation and processing capabilities of the AI model for the 3D animation assets, and a second model is obtained;

[0123] An interpretable artificial intelligence technology is used to give the second model interpretability (to enable users to understand the system decision-making process; through active learning, interactive learning and other technologies, the controllability and individualization capabilities of the system are improved);

[0124] According to the first asset management framework and the second model, Internet of Things technology, edge computing technology and cloud computing technology are fused to build a distributed asset management platform (realizing asset management across platforms and in multiple scenarios, supporting real-time collaboration);

[0125] Optimizing asset distribution and version control processes (communication technology can support efficient asset distribution and version control; new distributed technologies such as blockchain can provide reliable asset traceability and permission management mechanisms);

[0126] Using deep learning technology (such as generative adversarial networks), new animation assets are generated based on the animation engine resource asset data, automatically expanding the animation asset library;

[0127] Developing unified data standards, interface standards and development specifications for the asset management platform (to standardize animation asset management);

[0128] In this step, the data structure and attributes of each asset can be defined by sorting out common asset types in the animation industry (such as models, textures, actions, etc.); a unified file format standard (such as FBX, Alembic, USD, etc.) is developed to ensure the interoperability of assets between different software; asset metadata standards are established, including asset description, version information, creator, etc., to facilitate asset management and retrieval; API interface standards between asset management systems / platforms and animation production software, asset libraries, etc. are defined; interface specifications for common operations such as asset upload, download, version control, etc. are standardized; seamless integration and data interaction between different vendor systems are ensured; standardized specifications for asset creation, modification, submission, etc. development process are developed; unified asset naming, organizational structure, metadata filling, etc. development practice standards are established; code specifications, testing specifications, documentation specifications, etc. are established to improve development efficiency and maintainability. Through this step, the unified data standards ensure seamless import and export of assets between different software; standardized asset attribute information facilitates retrieval and cross-project reuse; standardized development processes and specifications reduce the complexity of asset management; standardized metadata information facilitates asset classification, searching and version tracing; standardized interface specifications enable seamless integration between different systems; the asset management system can be seamlessly integrated with animation production software, asset libraries, etc. In summary, by developing unified data standards, interface standards and development specifications for the asset management platform, the standardization level of animation asset management can be greatly improved.

[0129] Designing human-computer interaction interfaces and integrating artificial intelligence technology into animation production tools and the asset management platform (using natural language processing technology, etc. to realize human-computer semantic interaction and improve the intelligent level of the asset management platform).

[0130] By utilizing multi-modal data processing technology, knowledge base construction technology and natural language processing technology, a unified artificial intelligence driven first asset management framework is constructed; existing animation engine resource asset data is acquired, and a first model is obtained by utilizing the animation engine resource asset data and a preset first neural network training AI model; according to the animation engine resource asset data, 3D animation assets are generated by utilizing 3D animation modeling technology and are added to an animation asset library; in combination with meta-learning technology, small sample learning technology and fine-tuning technology, the first model is adjusted to improve the understanding, generation and processing capability of the AI model on the 3D animation assets, and a second model is obtained; by utilizing explainable artificial intelligence technology, the second model is endowed with explainability; according to the first asset management framework and the second model, a distributed asset management platform is constructed by utilizing Internet of Things technology, edge computing technology and cloud computing technology; asset distribution and version control processes are optimized; by utilizing deep learning technology, new animation assets are generated based on the animation engine resource asset data, and the animation asset library is automatically expanded; unified data standards, interface standards and development specifications are formulated for the asset management platform; a man-machine interaction interface is designed, and artificial intelligence technology is integrated into animation production tools and the asset management platform. The AI-based animation resource asset management scheme is more intelligent and automated, provides end-to-end one-stop service, is more humanized, deeply cooperates with users and provides highly explainable and controllable experience, and the distributed architecture guarantees the efficiency and scalability of the system.

[0131] In some possible embodiments of the present application, the step of constructing a unified artificial intelligence driven first asset management framework by utilizing multi-modal data processing technology, knowledge base construction technology and natural language processing technology comprises:

[0132] A unified data format standard is defined for different types of first animation asset data (such as 2D images, videos, 3D models, etc.);

[0133] Computer vision technology and natural language processing technology are used to preprocess and standardize the first animation asset data, and basic asset data is obtained;

[0134] A metadata standard of animation assets is established to describe various attributes and semantic information of the animation assets;

[0135] In this step, the animation production process is thoroughly understood, and the requirements of each link for asset metadata are sorted out; different roles of users (such as animators, artists, producers, etc.) are collected for asset metadata expectations; industry mature metadata standards (such as ACES, MDEC, etc.) are referred to, and excellent practices are learned; according to the aforementioned collection of asset metadata requirements, expectations and industry mature metadata standards, the metadata model of animation assets is established, including general attributes and specific attributes; general attributes such as asset ID, name, version, creation time, etc. Basic information; specific attributes such as model geometry information, mapping materials, action data, etc. Professional attributes; determine the relationship between metadata, such as the relationship between assets and materials, assets and projects; make detailed definitions, data types, value ranges, etc. Standards for each metadata attribute; establish an asset classification system and customize appropriate metadata attributes for different types of assets; develop metadata filling specifications, clearly define mandatory items, optional items, unified naming, format constraints, etc. Based on the metadata model, develop an asset metadata management software system la Provide a user-friendly metadata editing interface, support batch import and export; realize the version control, permission management, search query, etc. of metadata. Through the above steps, detailed metadata information can be obtained, which facilitates the classification, retrieval and tracing of assets; users can quickly locate the required assets, improve the asset reuse rate; realize standardized metadata, ensure the interoperability of asset information between different systems and platforms; assets can be efficiently shared and circulated across projects and teams; the metadata management system provides a visual asset management interface, simplifying operations; realize the version control and permission management of metadata, support multi-person collaborative development; standardized metadata description ensures the completeness and accuracy of asset information; accumulated rich metadata forms a reference knowledge base. In short, by establishing the metadata standard of animation assets, not only can the manageability and discoverability of assets be greatly improved, but also the efficient circulation and team collaboration of assets can be promoted, bringing significant technical benefits to the animation production industry.

[0136] Based on the basic asset data and metadata standards, automatically build a knowledge graph of animation assets (graph nodes represent different types of asset entities, and edges represent semantic relationships between entities);

[0137] In this step, natural language processing techniques are applied to identify various entities embedded in the metadata standards, and corresponding key entity information such as asset type, name, attributes, etc. is extracted from the basic asset data. Machine learning models are used to analyze the semantic relationships between entities such as inclusion, use, association, etc. Entity-relation triples are constructed to form a preliminary knowledge graph structure. The extracted entities and relationships are organized and stored according to the standard knowledge graph data model, and each entity node is given rich attribute information such as asset ID, name, type, etc. The types and attributes of semantic relationships between entities are defined, such as "has_texture", "part_of", etc. Graph algorithms are applied to analyze the structural characteristics of the knowledge graph and discover potential errors and conflicts. Domain expert knowledge is used to manually correct errors in the knowledge graph. The coverage and accuracy of the knowledge graph are continuously optimized to improve its overall quality. A graph visualization interface is developed to visually display the knowledge graph of animation assets. Asset search, recommendation, and correlation analysis functions based on the knowledge graph are provided. The knowledge graph supports intelligent management and innovative applications of animation assets. Through the above steps, the knowledge graph can capture asset entities and their complex semantic relationships, forming a more detailed knowledge representation. Compared to traditional metadata, the knowledge graph can better reflect the internal relationships between animation assets. Graph structure-based algorithms can enable more intelligent asset search and recommendation functions. Users can quickly discover associated asset information based on specific needs. The knowledge graph provides semantic relationships between assets, supporting more advanced asset classification, organization, and management. Knowledge graph analysis can also uncover novel combinations of assets, promoting the generation of innovative applications. The knowledge graph provides a foundation for the systematic accumulation and effective use of animation asset knowledge. Cross-team and cross-project asset knowledge sharing becomes more convenient and intelligent. In summary, building a knowledge graph based on animation asset metadata not only enables more comprehensive asset knowledge representation, but also significantly improves asset management and innovative application capabilities, providing significant technical advantages for the animation industry.

[0138] Using graph embedding techniques to represent structured knowledge graphs as low-dimensional knowledge graph vectors.

[0139] In this step, first, clean and standardize the entities and relationships in the knowledge graph to ensure data quality. Based on requirements, selectively retain or delete some entities and relationships to focus on specific domains.

[0140] According to the characteristics and application scenarios of the knowledge graph, select a suitable graph embedding algorithm, such as Node2Vec, DeepWalk, TransE, etc. (considering factors such as algorithm complexity, embedding dimension, whether to consider relationships, etc.); input the preprocessed knowledge graph into the selected graph embedding algorithm for training, adjust algorithm hyperparameters such as random walk length, embedding dimension, etc., and optimize model performance; ensure that the training process converges to obtain stable knowledge graph vector representation; design appropriate evaluation indicators such as similarity, link prediction, etc. to evaluate the quality of graph embedding vectors; according to the evaluation results, further adjust the parameters and settings of the graph embedding technology; apply the trained knowledge graph vector to downstream tasks such as asset association analysis, asset recommendation, etc.; according to specific application requirements, explore how to best utilize graph embedding vectors. Through the above steps, the original complex knowledge graph can be compressed into low-dimensional vector form, greatly improving storage and computing efficiency; vector form of knowledge representation is more convenient for subsequent machine learning and data analysis applications; graph embedding technology can capture the semantic information contained in the knowledge graph, which is reflected in the topological structure of the vector space; similar entities and relationships are closer in the vector space, preserving the semantic association of the original graph; based on graph embedding vectors, knowledge graphs can be seamlessly integrated with other data modalities (such as text, images, etc.); cross-modal semantic association analysis and application become more efficient and accurate; graph embedding vectors can serve as a general feature representation, providing strong support for downstream asset analysis, recommendation, etc.; compared to the original graph, vector-based calculations are usually more efficient. In summary, representing structured knowledge graphs as low-dimensional graph embedding vectors can effectively compress knowledge representation, preserve semantic information, and provide strong support for cross-modal fusion and downstream applications, playing an important role in animation asset management.

[0141] Input the basic asset data of different modalities (such as images, text, 3D, etc.) into a deep neural network for feature extraction;

[0142] In this step, according to specific application scenarios, extract relevant visual, semantic, and other multi-modal features; use mature feature extraction models such as CNN, BERT, etc. to ensure feature quality.

[0143] Merge the extracted features with the knowledge graph vector to obtain a multi-modal fusion encoding vector;

[0144] In this step, the extracted multi-modal features (such as visual, semantic, etc.) are spliced with the knowledge graph vector to form a multi-modal fusion encoding vector (different splicing methods can be tried, such as simple splicing, weighted average, etc., choose the appropriate method according to the needs); introduce attention mechanism, dynamically allocate weights for different modalities, highlight important information and suppress secondary information; through the attention mechanism, adaptively aggregate multi-modal features to improve the fusion effect; apply the multi-modal fusion encoding vector to downstream tasks such as classification, retrieval, etc.; through end-to-end training, optimize the performance of the multi-modal fusion encoding vector; adjust the model structure and hyperparameters to improve the fusion effect. Through the above steps, multi-modal information such as visual, semantic, and knowledge is fused, and the feature representation is more comprehensive and semantic; compared with single-modal features, multi-modal fusion features can better capture various attributes of assets; multi-modal fusion encoding vectors provide more effective feature representations for downstream classification, retrieval, and other tasks; the fusion of knowledge graph information can significantly improve task performance, such as more accurate asset correlation analysis; multi-modal feature fusion can improve the generalization ability of the model and cope with more complex scenarios and data distribution; especially in the case of insufficient data, knowledge graph information can effectively supplement the role; multi-modal fusion features contain rich semantic information, enhancing the model's interpretability; it helps to understand the model's decision-making process and provides a basis for subsequent manual intervention and optimization. In summary, by fusing the extracted visual, semantic, and other features with the knowledge graph vector, and dynamically aggregating each modal information using the attention mechanism, we can obtain information-rich, high-performance, and strong generalization ability feature representations, which play an important role in animation asset management.

[0145] Based on the multi-modal fusion encoding vector, develop AI-enhanced function modules (such as intelligent retrieval: multi-modal retrieval based on semantic similarity; asset generation: generate new assets through generative adversarial networks; asset archiving: automated asset archiving and organization based on clustering).

[0146] Integrate and deploy each function module into the asset management platform;

[0147] Collect feedback information through active learning and human-computer interaction, and continuously optimize the performance of the asset management platform according to the feedback information;

[0148] Introduce interpretable AI and interactive learning mechanisms to enhance the interpretability and controllability of the asset management platform.

[0149] In the embodiment, intelligent management of the whole life cycle of assets is realized, covering various links such as asset creation, use, retrieval and distribution; through multi-modal fusion coding, basic asset data of different modalities are fully integrated, deep semantic associations are mined, and data utilization efficiency is improved; AI enhancement function greatly improves the automation capability of asset management, reduces the labor intensity, and improves the production efficiency; semantic understanding and reasoning based on knowledge graph are carried out, and more intelligent and accurate asset retrieval and decision are supported; continuous model optimization and man-machine cooperation mechanism are used to continuously improve the performance and user experience of the system.

[0150] In some possible embodiments of the present application, the step of obtaining existing animation engine resource asset data and training an AI model using the animation engine resource asset data and a preset first neural network to obtain a first model comprises:

[0151] Collect various resource asset data (including 2D image materials, 3D models, animation clips, material maps, etc.) from existing animation engines and animation projects;

[0152] After classifying, labeling and auditing the asset data, clean up, format uniformity and standardization processing are performed;

[0153] Use data enhancement techniques (such as rotation, scaling, noise addition, etc.) to expand the data volume of the asset data to obtain first asset data;

[0154] Convert the first asset data in different formats into a unified tensor representation format to obtain the animation engine resource asset data;

[0155] Design a corresponding network architecture of the first neural network for different types of asset data (such as using a 3D convolutional network to process 3D models and using a Transformer to process sequence data);

[0156] Use a multi-head attention mechanism to fuse data of different modalities in the animation engine resource asset data;

[0157] Train the first neural network using the animation engine resource asset data and learn the multi-modal representation of asset data using a self-supervised learning method to obtain a general animation asset representation basic model;

[0158] In this step, multi-modal features are extracted from the animation engine resource asset data, including visual, semantic and other types of features; a self-supervised learning task is designed to enable the first neural network to learn the latent representation of asset data autonomously; common self-supervised tasks include asset attribute prediction, asset relationship inference, cross-modal reconstruction, etc.; through these self-supervised learning tasks, the first neural network can learn a general and semantically rich asset representation; through large-scale asset data training, a general and robust asset representation basic model is learned; the trained basic model is applied to downstream specific tasks such as asset classification and retrieval; according to the task requirements, the basic model is appropriately fine-tuned to improve the performance of the model on specific tasks. Through the above steps, the model has learned a general representation of animation asset data, covering visual, semantic and other aspects; the basic model can be widely used in different animation asset management tasks and has strong transferability; based on the general representation model, key tasks such as asset classification and retrieval can be quickly completed; the efficiency and accuracy of animation asset management are greatly improved; the self-supervised learning method enables the model to learn the internal laws and general features of asset data; self-supervised learning does not require a large amount of manually labeled data, greatly reducing the cost of model training. In summary, training a general animation asset representation basic model using self-supervised learning can effectively enhance the ability of animation asset management and provide strong support for animation production.

[0159] For specific tasks such as asset identification, retrieval, generation, etc., the animation asset representation basic model is fine-tuned and transferred on the corresponding labeled data to generate a specialized first model (improving performance on the task);

[0160] Design evaluation indicators and test sets to measure model performance, based on evaluation feedback, continuously optimize through parameter fine-tuning and data increment, realize closed-loop model improvement to improve the quality of the first model.

[0161] Through the steps of this embodiment, a powerful asset representation capability can be built to efficiently encode multi-modal asset data; through the pre-training + fine-tuning paradigm, the first model is suitable for a variety of downstream tasks; the limited labeled data in the industry is maximally utilized to achieve optimal performance; continuous optimization enables the model to continuously adapt to new requirements and has good generalization ability, laying a solid foundation for subsequent intelligent asset management, generation and other tasks. In summary, the pre-training + transfer learning paradigm can efficiently build AI models with excellent performance, opening up new possibilities for animation asset management enhanced by artificial intelligence.

[0162] In some possible embodiments of the present application, the step of generating 3D animation assets from the animation engine resource asset data using 3D animation modeling technology and adding them to the animation asset library includes:

[0163] analyzing specific asset requirements (such as character models, scene models, props, etc.) for 3D assets in the historical animation project;

[0164] determining the type, quantity, and level of detail of the required 3D assets according to the specific asset requirements;

[0165] conducting creative design according to the requirements, and outputting concept design drafts and sketches (including character settings, modeling designs, action sketches, etc.);

[0166] based on the concept design drafts and sketches, creating initial 3D models (such as bone binding for characters and creatures to prepare for animation production; importing reference pictures and videos to refine model details; applying physical-based rendering (PBR) principles to design materials; giving the model reasonable physical properties such as light, shadow, and reflection; etc.);

[0167] generating 3D rendering assets using real-time or offline rendering tools based on the initial 3D models;

[0168] inputting the 3D rendering assets into animation production software and creating keyframe animations according to the shot and action design;

[0169] rendering output animation sequence frames based on the created keyframes to form complete 3D animation segments;

[0170] sorting and organizing 3D static and dynamic assets, labeling metadata information, and integrating with other 2D and audio assets to build a complete animation asset library;

[0171] deploying the animation asset library to the asset management platform for creation and retrieval.

[0172] Through the above process of the embodiment, the following significant technical effects can be achieved: high-quality and high-fidelity 3D assets are generated using professional 3D modeling tools; realistic materials and lighting effects are achieved through physical rendering; 3D modeling and animation production are integrated to efficiently output dynamic 3D animation assets; standardized processes ensure the consistency and traceability of 3D assets; integration with other asset types builds a rich and complete asset library; by incorporating 3D animation modeling technology into the entire workflow, not only does it enrich the asset library capacity, but more importantly, it improves the quality and immersion of the entire animation asset; moreover, high-quality 3D assets lay a solid foundation for subsequent intelligent management, retrieval, and generation of innovative applications, contributing to the higher-dimensional innovation and development of the entire animation industry.

[0173] In some possible embodiments of the present application, the step of adjusting the first model to improve the understanding, generation and processing capability of the AI model for the 3D animation assets to obtain a second model by using the meta-learning technology, the few-shot learning technology and the fine-tuning technology comprises:

[0174] dividing the 3D animation assets into a plurality of first tasks, each of which corresponds to a 3D asset type;

[0175] training a meta-model capable of understanding, generating and processing 3D animation assets based on the first model by using an optimization-based meta-learning algorithm (such as MAML) or a metric-based meta-learning algorithm (such as Prototypical Networks);

[0176] collecting annotated new type 3D animation asset data, taking the new type 3D animation asset data as the input of few-shot learning, and fine-tuning the meta-model to adapt to new type 3D animation assets (i.e., capable of understanding, generating and processing new type 3D animation assets);

[0177] collecting corresponding first 3D animation asset data for a preset specific application scenario;

[0178] using the fine-tuning technology to optimize and train the meta-model based on the previously learned knowledge to obtain the second model.

[0179] The model obtained by the present embodiment has stronger generalization capability and can quickly adapt to new types of 3D assets; has higher data efficiency and can be optimized for specific purposes with only a small amount of labeled data; has better 3D asset understanding, generation and processing capability, and meets the demand of intelligent management of animation engines. This method fully utilizes the advantages of meta-learning, few-shot learning and fine-tuning, and can effectively improve the performance of AI models in 3D animation asset management. The scheme of the present embodiment can greatly improve the applicability and robustness of AI models in complex animation asset fields, support multi-modal animation asset generation, reduce the difficulty and workload of artists' creation, realize intelligent asset creation, editing and processing, and improve the automation level of animation production; the 3D asset creation and physical simulation capability lays a foundation for new animation creation methods and immersive experience; with the help of transfer learning and meta-learning technologies, the present embodiment can continuously adapt to emerging animation creation needs.

[0180] In some possible embodiments of the present application, the step of giving the second model interpretability by using the explainable artificial intelligence technology comprises:

[0181] An explainable artificial intelligence technique (such as attention mechanism, visualizable explanation, causal reasoning, etc.) is selected to be embedded into the architecture of the second model, so that the second model can generate a decision-making process that can be understood by the user, and the internal reasoning logic is elucidated;

[0182] An active learning strategy is introduced, and the second model actively asks the user for key information to narrow down the uncertainty of the model;

[0183] The second model learns the user's preferences and feedback through the human-computer interaction process and adjusts its own behavior and decision-making accordingly;

[0184] An interactive learning interface is designed, and the user can modify and adjust the knowledge base and parameters of the second model in real time.

[0185] In this embodiment, through the above steps, the second model obtained in this way is explainable, controllable and personalized, which greatly improves the intelligent level and user experience of the 3D animation asset management system. Users are no longer passive in accepting the system's decisions, but can actively participate and guide the system to obtain better customized services. This human-machine collaborative approach helps to enhance user trust and reliance on the system, significantly improves the explainability of AI systems in the field of animation asset management, and wins the trust of users. Human-computer interactive explanation enables users to better understand the system and provide feedback on incorrect decisions to promote continuous improvement of the system. In summary, through a series of innovative technical means, future AI-based animation asset management systems will be more reliable, explainable, controllable, and form a deep collaboration with users to maximize the potential of artificial intelligence and achieve true intelligent assistance.

[0186] In some possible embodiments of the present application, the step of constructing a distributed asset management platform according to the first asset management framework and the second model, integrating Internet of Things technology, edge computing technology and cloud computing technology, comprises:

[0187] Distributed file systems and distributed databases are used to store asset data;

[0188] Microservice architecture and containerization technology are used to build the asset management platform to realize service orchestration and elastic scaling;

[0189] Service mesh technology is introduced to manage cross-platform service calls and data interactions;

[0190] An edge node integrating the second model is deployed on the collaboration site to support edge-side asset processing and local caching;

[0191] Based on the second model, an edge AI model with local rendering and retrieval functions is generated;

[0192] Introduce content distribution network technology to distribute and update animation assets;

[0193] Deploy AI-based asset intelligent management engine and analysis module in the cloud;

[0194] Deploy identity authentication and access control system to manage multi-user asset access rights;

[0195] Encrypt and digitally sign important assets to prevent leaks and data tampering.

[0196] In this embodiment, through the above distributed architecture, unified management and high-availability storage of asset data can be realized, eliminating data silos; support for distributed collaboration in different places, breaking away from the limitations of traditional single-machine production mode; use the powerful computing power of the cloud to support high-performance asset processing and analysis; edge local processing and caching reduce data transmission delay and improve response speed; Internet of Things and augmented reality technology gives animation creation a new immersive experience. By innovatively integrating emerging technologies such as cloud, edge, and end, this distributed asset management platform can significantly improve the intelligence level, collaboration efficiency, and security of animation asset management, and inject new momentum into the digital and intelligent transformation of the industry.

[0197] In some possible embodiments of the present application, the step of optimizing asset distribution and version control process includes:

[0198] Deploy multiple CDN nodes to realize edge caching and distribution of assets;

[0199] According to the location and router address of the CDN node, combined with the preset router intelligent control strategy and dynamic routing allocation scheme, a plurality of distribution paths are generated (a P2P distribution network is established, each node is both a downloader and an upload source; high-efficiency multi-point transmission is realized based on code tower distributed coding technology; dynamic load balancing and fault switching are provided to provide high-availability distribution services; combined with new transmission protocols such as HTTP / 2 and QUIC, distribution efficiency is improved);

[0200] Use file system snapshot technology and data deduplication technology to calculate the first difference data between asset versions;

[0201] Distribute the first difference data through the distribution path (only synchronize and transmit the difference part, greatly saving bandwidth and synchronization time; support incremental synchronization and breakpoint resume, improve synchronization reliability);

[0202] Connect the CDN node to a preset first blockchain network;

[0203] Chain all the version management data, propagation path data, and equity transaction data of animation assets involved in each CDN node;

[0204] Automated copyright protection, usage authorization, and revenue distribution are realized by smart contracts of the first blockchain network;

[0205] A version control system is introduced to support distributed collaborative editing of animation assets.

[0206] In this embodiment, end-to-end encryption and integrity verification are used during asset transmission; distributed storage realizes fault-tolerant backup to prevent data loss caused by single-point failure; and an external security defense system prevents various attacks to ensure secure and reliable distribution.

[0207] The scheme of this embodiment can greatly improve the efficiency, reliability, and bandwidth utilization of asset distribution; the asset traceability and copyright management mechanism based on blockchain ensures that asset transactions are trustworthy and traceable; and the automated interest distribution based on smart contracts solves the problems of animation asset transaction and revenue distribution. In summary, by integrating innovative technologies such as new communication networks, distributed storage, and blockchain, the efficiency, security, and copyright issues in traditional asset distribution and version control can be fundamentally addressed, injecting new momentum into the digital and intelligent transformation of the animation industry.

[0208] In some possible embodiments of the present application, the step of generating new animation assets based on the animation engine resource asset data using deep learning technology to automatically expand the animation asset library includes:

[0209] According to a preset data compression and enhancement strategy, the data volume of the first animation asset data is expanded to obtain first training data;

[0210] Design a generation model architecture suitable for different asset types (such as StyleGAN, etc.);

[0211] Train the generation model architecture using the first training data to obtain a first asset generation model (supporting asset generation based on text description, sketch, and small amount of examples, etc.);

[0212] Fine-tune the first asset generation model for specific asset production tasks (such as generating cartoon characters, 3D models), to obtain a second asset generation model (combining visual, semantic, and other modal condition information to guide the generation process; attention mechanism gives the model the ability to selectively focus on and integrate different condition information);

[0213] Generate new animation assets according to the animation engine resource asset data and the second asset generation model;

[0214] Perform quality assessment on the generated new animation assets to filter out unqualified results;

[0215] Through the method of the embodiment, the time and workload for creating animation assets can be greatly reduced; high-quality and diversified animation assets are generated, and the capacity of the asset library is automatically expanded.

[0216] It should be noted that, for the foregoing method embodiments, in order to simply describe, they are all expressed as a series of action combinations, but those skilled in the art should know that the present application is not limited to the action sequence described, because according to the present application, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should know that the embodiments described in the specification all belong to preferred embodiments, and the actions and modules involved are not necessarily necessary for the present application.

[0217] In the above embodiments, the description of each embodiment has its own focus, and the parts not described in detail in a certain embodiment can be referred to the related description of other embodiments.

[0218] In several embodiments provided in the present application, it should be understood that the disclosed device can be implemented by other ways. For example, the device embodiments described above are only schematic, for example, the division of the above units is only a logical function division, and actual implementation can be another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interface, device or unit, and can be electrical or other forms.

[0219] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, that is, they can be located in one place, or can be distributed on multiple network units. According to actual needs, part or all of the units can be selected to achieve the purpose of the embodiment scheme.

[0220] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0221] The integrated unit described above, if realized in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable memory. Based on this understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a memory and includes a plurality of instructions for causing 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 above-mentioned method of each embodiment of the present application. The aforementioned memory includes: a U disk, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk, and various media that can store program codes.

[0222] A person of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by a program instructing relevant hardware, and the program can be stored in a computer readable memory, which can include a flash disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0223] The embodiments of the present application are described in detail above, and the principles and implementation manners of the present application are described by applying specific examples. The above description of the embodiments is only used to help understand the method of the present application and its core idea; at the same time, for those of ordinary skill in the art, according to the idea of the present application, the specific implementation manner and application range can be changed, and the above description of the embodiments should not be understood as a limitation of the present application.

[0224] Although the present application is disclosed as above, the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements without departing from the spirit and scope of the present application, and can make various changes and modifications, including the combination of different functions and implementation steps, including software and hardware implementation manners, which are all within the protection scope of the present application.

Claims

1. An AI-based animation engine resource asset management system, characterized in that, include: Management server and content server; The management server is configured as follows: By leveraging multimodal data processing, knowledge base construction, and natural language processing technologies, a unified AI-driven first asset management framework is constructed. Obtain existing animation engine resource asset data, and use the animation engine resource asset data and a preset first neural network to train an AI model to obtain a first model; Based on the animation engine resource asset data, 3D animation assets are generated using 3D animation modeling technology and added to the animation asset library; By combining meta-learning techniques, few-shot learning techniques, and fine-tuning techniques, the first model is adjusted to improve the AI ​​model's ability to understand, generate, and process the 3D animation assets, resulting in the second model. The second model is given interpretability by employing interpretable artificial intelligence technology. Based on the first asset management framework and the second model, a distributed asset management platform is constructed by integrating Internet of Things (IoT) technology, edge computing technology, and cloud computing technology. Optimize asset distribution and version control processes; Using deep learning technology, new animation assets are generated based on the animation engine resource asset data, automatically expanding the animation asset library; Establish unified data standards, interface standards, and development specifications for the asset management platform; Design a human-computer interaction interface and integrate artificial intelligence technology into the animation production tool and the asset management platform.

2. A method for managing resource assets of an animation engine based on artificial intelligence, characterized in that, include: By leveraging multimodal data processing, knowledge base construction, and natural language processing technologies, a unified AI-driven first asset management framework is constructed. Obtain existing animation engine resource asset data, and use the animation engine resource asset data and a preset first neural network to train an AI model to obtain a first model; Based on the animation engine resource asset data, 3D animation assets are generated using 3D animation modeling technology and added to the animation asset library; By combining meta-learning techniques, few-shot learning techniques, and fine-tuning techniques, the first model is adjusted to improve the AI ​​model's ability to understand, generate, and process the 3D animation assets, resulting in the second model. The second model is given interpretability by employing interpretable artificial intelligence technology. Based on the first asset management framework and the second model, a distributed asset management platform is constructed by integrating Internet of Things (IoT) technology, edge computing technology, and cloud computing technology. Optimize asset distribution and version control processes; Using deep learning technology, new animation assets are generated based on the animation engine resource asset data, automatically expanding the animation asset library; Establish unified data standards, interface standards, and development specifications for the asset management platform; Design a human-computer interaction interface and integrate artificial intelligence technology into the animation production tool and the asset management platform.

3. The method for managing animation engine resource assets based on artificial intelligence according to claim 2, characterized in that, The steps involved in constructing a unified AI-driven first asset management framework by utilizing multimodal data processing technology, knowledge base construction technology, and natural language processing technology include: Define a unified data format standard for different types of first-generation animation asset data; The first animation asset data was preprocessed and standardized using computer vision and natural language processing technologies to obtain basic asset data. Establish metadata standards for animation assets to describe the various attributes and semantic information of animation assets; Based on the aforementioned basic asset data and metadata standards, a knowledge graph of the animation assets is automatically constructed. Graph embedding techniques are used to represent structured knowledge graphs as low-dimensional knowledge graph vectors; The basic asset data of different modalities (such as images, text, 3D, etc.) are input into a deep neural network for feature extraction; The extracted features are merged with the knowledge graph vector to obtain a multimodal fusion encoding vector; Based on the aforementioned multimodal fusion coding vector, an AI enhancement function module was developed; The various functional modules are integrated and deployed into the asset management platform; The asset management platform collects feedback information through active learning and human-computer interaction, and continuously optimizes the performance of the asset management platform based on the feedback information. The asset management platform is enhanced by introducing explainable AI and interactive learning mechanisms.

4. The method for managing animation engine resource assets based on artificial intelligence according to claim 3, characterized in that, The step of acquiring existing animation engine resource asset data and using the animation engine resource asset data and a preset first neural network to train an AI model to obtain a first model includes: Collect various resource and asset data from existing animation engines and animation projects; After classifying, labeling, and reviewing the asset data, it is then cleaned, formatted, and standardized. The data volume of the asset data is expanded using data augmentation techniques to obtain the first asset data; The first asset data in different formats is converted into a unified tensor representation format to obtain the animation engine resource asset data; Design the corresponding network architecture of the first neural network for different types of asset data; A multi-head attention mechanism is employed to fuse data from different modalities within the animation engine's resource asset data; The first neural network is trained using the animation engine resource asset data, and a self-supervised learning method is used to learn the multimodal representation of the asset data to obtain a general animation asset representation basic model. For a specific task, the basic model representing the animation assets is fine-tuned and transfer-learned on the corresponding labeled data to generate a dedicated first model; Design evaluation metrics and test sets to measure model performance. Based on evaluation feedback, continuously optimize the model through parameter fine-tuning and data increment to achieve closed-loop model improvement and enhance the quality of the first model.

5. The AI-based animation engine resource asset management method according to claim 4, characterized in that, The step of generating 3D animation assets using 3D animation modeling technology based on the animation engine resource asset data and adding them to the animation asset library includes: Analyze the specific asset requirements for 3D assets in historical animation projects; Based on the specific asset requirements, determine the required 3D assets, including their type, quantity, and level of detail. Based on the aforementioned requirements, creative designs are developed, and conceptual design drafts and sketches are output. Based on the conceptual design draft and the sketches, create an initial 3D model; Based on the initial 3D model, generate 3D rendered assets using real-time or offline rendering tools; The 3D rendered assets are input into animation production software, and keyframe animation is produced based on the storyboard and action design. Based on the keyframes created, render and output the animation sequence frames to form a complete 3D animation clip; The 3D static and dynamic assets are classified and organized, metadata information is labeled, and they are combined with other 2D and audio assets to build a complete animation asset library; The animation asset library is deployed to the asset management platform for creation and retrieval.

6. The method for managing animation engine resource assets based on artificial intelligence according to claim 5, characterized in that, The step of adjusting the first model by combining meta-learning techniques, few-shot learning techniques, and fine-tuning techniques to improve the AI ​​model's ability to understand, generate, and process the 3D animation assets, thereby obtaining the second model, includes: The 3D animation assets are divided into multiple first tasks, and each first task corresponds to a type of 3D asset. Based on the first model, an optimization-based meta-learning algorithm or a metric-based meta-learning algorithm is used to train a meta-model that can understand, generate, and process 3D animation assets. Collect labeled new types of 3D animation asset data, use the new types of 3D animation asset data as input for few-sample learning, and fine-tune the meta-model to adapt to the new types of 3D animation assets; Collect corresponding first 3D animation asset data for specific pre-defined application scenarios; By employing fine-tuning techniques, while retaining the previously learned knowledge, the meta-model is optimized and trained using the first 3D animation asset data to obtain the second model.

7. The method for managing animation engine resource assets based on artificial intelligence according to claim 6, characterized in that, The step of employing interpretable artificial intelligence technology to endow the second model with interpretability includes: The second model is designed to incorporate interpretable artificial intelligence techniques into its architecture, enabling it to generate decision-making processes that are understandable to users and to elucidate its internal reasoning logic. An active learning strategy is introduced, in which the second model actively asks the user for key information, thereby reducing the uncertainty of the model; The second model learns user preferences and feedback through human-computer interaction and adjusts its own behavior and decisions accordingly. Design an interactive learning interface that allows users to modify and adjust the knowledge base and parameters of the second model in real time.

8. The method for managing animation engine resource assets based on artificial intelligence according to claim 7, characterized in that, The steps of constructing a distributed asset management platform based on the first asset management framework and the second model, integrating IoT technology, edge computing technology, and cloud computing technology, include: Asset data is stored using a distributed file system and a distributed database; An asset management platform is built based on microservice architecture and containerization technology to achieve service orchestration and elastic scaling. Introducing service mesh technology to manage cross-platform service calls and data interactions; Deploy edge nodes that integrate the second model at the collaborative site to support edge-side asset processing and local caching; Based on the second model, an edge AI model with local rendering and retrieval functions is generated; Introducing content delivery network technology for the distribution and caching of animation assets; Deploy an AI-based asset intelligent management engine and analysis module in the cloud; Deploy an identity authentication and access control system to manage asset access permissions for multiple users; Encryption and digital signatures protect critical assets from leaks and data tampering.

9. The method for managing animation engine resource assets based on artificial intelligence according to claim 8, characterized in that, The steps for optimizing asset distribution and version control include: Deploy multiple CDN nodes to achieve edge caching and distribution of assets; Based on the location of the CDN node and the router address, and combined with the preset router intelligent control strategy and dynamic routing allocation scheme, multiple distribution paths are generated; Using file system snapshot technology and data deduplication technology, the first difference data between asset versions is calculated; The first difference data is distributed through the distribution path; Connect the CDN node to the preset first blockchain network; All version management data, propagation path data, and equity transaction data of the animation assets involved in each CDN node will be uploaded to the blockchain; Utilize smart contracts on the first blockchain network to automate copyright protection, licensing, and revenue distribution; A version control system was introduced to support distributed collaborative editing of animation assets.

10. The method for managing animation engine resource assets based on artificial intelligence according to claim 9, characterized in that, The step of using deep learning technology to generate new animation assets based on the animation engine resource asset data and automatically expanding the animation asset library includes: The data volume of the first animation asset data is increased according to a preset data compression and enhancement strategy to obtain the first training data; Design generative model architectures suitable for different asset types; The first asset generation model is obtained by training the generative model architecture using the first training data. For a specific asset production task, the first asset generation model is fine-tuned to obtain a second asset generation model; New animation assets are generated based on the animation engine resource asset data and the second asset generation model; Perform a quality assessment on the newly generated animation assets and filter out unqualified results.

Images