A method and system for intelligent indexing of a movie 3D digital asset

Abstract

The application discloses a kind of movie 3D digital asset intelligent indexing method and system.The method constructs the indexing system covering asset type, scene style, scene purpose and precision level;Heterogeneous format is uniformly converted into OpenUSD format and output six-view rendering image;Through API, the visual semantic analysis of rendering image is carried out to extract asset type and scene style, while extracting metadata and geometric information as basic attributes, and then the regularized indexing of scene purpose and precision level is completed based on map resolution and grid face number.The application realizes the efficient, accurate, unified and multi-dimensional fusion intelligent indexing of movie 3D digital asset, and improves the asset retrieval accuracy and reuse efficiency.

Description

Technical Field

[0001] This application relates to the field of digital asset processing technology, and in particular to a smart indexing method and system for 3D digital assets of movies. Background Technology

[0002] With the digital development of the film industry, 3D digital assets (including character models, scene models, prop models, etc.) have become a core production element in filmmaking. Throughout the entire film production lifecycle, the efficient management, rapid retrieval, and reuse of massive amounts of 3D digital assets heavily rely on a precise and unified asset indexing system. Indexing is the process of semantically describing, classifying, and structurally organizing 3D digital assets; its quality directly affects the retrieval efficiency, reuse rate, and cross-project, cross-platform asset interoperability of the asset library.

[0003] Currently, the indexing of 3D digital assets mainly relies on manual annotation. Operators manually classify and tag assets based on their personal experience. However, manual annotation has significant technical drawbacks: First, it is inefficient and costly, making it difficult to meet the needs of rapid data entry and real-time management of massive amounts of 3D movie digital assets. Second, due to differences in subjective understanding among annotators, annotation standards are inconsistent and tag semantics are ambiguous, resulting in low asset retrieval accuracy and poor reusability. Third, manual annotation typically only focuses on the surface category information of assets, lacking the ability to integrate and index multi-dimensional metadata information (such as asset format, creation information, file size, texture resolution, etc.) and geometric information (such as vertex count, mesh face count, etc.), thus failing to form a comprehensive and objective technical description of the assets.

[0004] In recent years, several intelligent indexing technologies have been proposed to replace or assist manual annotation. Some solutions convert 3D assets into 2D images for visual feature analysis, which reduces the difficulty of indexing to some extent, but still suffers from problems such as inconsistent 3D format processing, incomplete indexing systems, and insufficient multi-dimensional feature extraction. Other solutions use keyword-based or simple rule-based matching for automatic indexing, which cannot handle complex semantic and multimodal information, resulting in limited accuracy and generalization capabilities. Furthermore, while some digital content retrieval methods incorporate cross-modal technologies, their solutions primarily focus on the retrieval process, neglecting the indexing of 3D film digital assets, thus failing to meet the specific needs of high-precision, multi-dimensional, and intelligent indexing of 3D film digital assets. Simultaneously, the current lack of a unified indexing system and indexing field specifications in the 3D film digital asset field leads to incompatibility between asset indexing results, severely restricting cross-platform interoperability and shared reuse of assets.

[0005] In summary, existing technologies for indexing 3D digital assets of movies generally suffer from technical problems such as low indexing efficiency, inconsistent standards, insufficient integration of multi-dimensional features, lack of a unified indexing system, and low level of intelligence. Summary of the Invention

[0006] Based on this, the embodiments of this application provide a method and system for intelligent indexing of 3D digital assets of movies, which can improve the indexing efficiency of 3D digital assets of movies and effectively meet the needs of high-precision, multi-dimensional and intelligent indexing of 3D digital assets of movies.

[0007] Firstly, a smart indexing method for 3D digital assets of movies is provided, the method comprising:

[0008] Constructing a 3D digital asset indexing system for films;

[0009] Various heterogeneous 3D digital assets are uniformly converted into the OpenUSD format to obtain standardized digital assets;

[0010] The standardized digital asset is rendered from multiple perspectives using rendering software, resulting in six different viewpoints: front view, rear view, left view, right view, top view, and isometric view.

[0011] Construct system prompts and user prompts, call the large model via API, send the system prompts, the rendered images from six different perspectives, and the user prompts to the large model, and receive the structured indexing results returned by the large model;

[0012] Parse the structured indexing results, extract the asset type and asset scene style of the 3D digital assets, and complete semantic indexing;

[0013] The metadata and geometric information of the standardized digital assets are extracted and indexed as basic attributes of the 3D digital assets; wherein, the metadata information includes asset format, creation information, file size, and texture resolution, and the geometric information includes the number of vertices and the number of mesh faces;

[0014] Based on the texture resolution and mesh count in the basic attributes, the 3D digital assets are indexed according to their scene usage and accuracy level, resulting in multi-dimensional fusion indexing results.

[0015] Optionally, the construction of the film 3D digital asset indexing system specifically includes:

[0016] The asset indexing system is divided into four dimensions: asset type, asset scenario style, asset scenario purpose, and asset accuracy level. Under different dimensions, first-level, second-level, and third-level subcategories of assets are established to construct a hierarchical asset indexing system.

[0017] Optionally, the system prompts are used to guide the large model to index the type and scene style of the digital assets according to the movie 3D digital asset indexing system, and the user prompts include the large model output requirements and output format requirements.

[0018] Optionally, the step of calling the large model via API to send the system prompt, the rendered images from the six different perspectives, and the user prompt to the large model includes:

[0019] The rendered images from the six different perspectives are Base64 encoded;

[0020] Set the request body, and organize the system prompt, the Base64 encoded rendered image, and the user prompt into request parameters;

[0021] Send the request body to the API interface of the large model, requesting the large model to output indexed results in structured JSON format.

[0022] Optionally, the step of indexing the 3D digital asset according to the texture resolution and mesh count in the basic attributes, including:

[0023] The geometric complexity and texture accuracy of the 3D digital asset are calculated based on the texture resolution and the number of mesh faces.

[0024] Based on the geometric complexity and texture precision, and combined with the classification standards for asset scene use and asset precision level in the asset indexing classification system, the asset scene use and asset precision level of the 3D digital asset are determined.

[0025] Optionally, the asset types include character assets, scene assets, prop assets, special effects assets, texture assets, Gaussian assets, and animation assets; the scene styles include modern / modern, ancient, and science fiction / fantasy styles; the use of the scene is determined according to the applicable stage of the 3D digital asset in the film technology process, including the creation stage, production stage, post-production stage, distribution and marketing stage, and operation and derivative stage; the asset precision level is divided into high precision level, medium precision level, and low precision level according to the geometric complexity and texture precision.

[0026] Optionally, the method further includes:

[0027] Based on a multimodal visual understanding model, visual semantic analysis is performed on the rendered images from six different perspectives to identify the subject type, style features, and emotional atmosphere in the 3D digital assets and generate multimodal indexing tags.

[0028] By fusing the multimodal indexing tags with the semantic indexing, the basic attribute indexing, and the asset scenario usage and asset precision level indexing, a full-dimensional indexing result covering visual content, technical parameters, and application scenarios is obtained.

[0029] Secondly, a smart indexing system for 3D digital assets of movies is provided, the system comprising:

[0030] The building module is used to construct a 3D digital asset indexing system for movies;

[0031] The format conversion module is used to convert 3D digital assets in various heterogeneous formats into the OpenUSD format to obtain standardized digital assets.

[0032] The rendering module is used to call rendering software to perform multi-view rendering of the standardized digital assets and output rendered images from six different perspectives; wherein, the six different perspectives include front view, rear view, left view, right view, top view, and isometric view.

[0033] The large model indexing module is used to construct system prompts and user prompts. It calls the large model through API, sends the system prompts, the rendered images from six different perspectives, and the user prompts to the large model, and receives the structured indexing results returned by the large model.

[0034] The semantic indexing module is used to parse the structured indexing results, extract the asset type and asset scene style of the 3D digital assets, and complete the semantic indexing.

[0035] The attribute extraction module is used to extract the metadata and geometric information of the standardized digital assets, and index them as the basic attributes of the 3D digital assets; wherein, the metadata information includes asset format, creation information, file size, texture resolution, and the geometric information includes the number of vertices and the number of mesh faces;

[0036] The rule indexing module is used to index the 3D digital assets according to the texture resolution and mesh count in the basic attributes, and to obtain multi-dimensional fusion indexing results.

[0037] Thirdly, an electronic device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement any of the methods described in the first aspect above.

[0038] Fourthly, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements any of the methods described in the first aspect above.

[0039] The beneficial effects of the technical solutions provided in this application include at least the following:

[0040] (1) By constructing a unified four-dimensional indexing system and combining it with the visual semantic analysis of multi-view rendered images using a large model, the traditional manual annotation is replaced, which solves the problems of inconsistent indexing standards and ambiguous label semantics. This enables the rapid entry and efficient indexing of massive amounts of movie 3D digital assets, and significantly reduces manual costs.

[0041] (2) By unifying heterogeneous formats into the OpenUSD format and integrating large model semantic indexing with regularized indexing based on geometric information (number of vertices, number of mesh faces) and metadata information (texture resolution, file format, etc.), multi-dimensional integrated indexing of asset type, scene style, scene use and accuracy level is realized, which effectively improves the accuracy of asset retrieval and cross-project reuse rate.

[0042] (3) Based on OpenUSD standardized conversion and multi-view rendering, it takes into account the compatibility of assets produced by different methods such as traditional CG assets and Gaussian assets. Combined with structured JSON index output, it provides a unified data foundation for the full life cycle management, version traceability and cross-platform interoperability of the asset library. Attached Figure Description

[0043] To more clearly illustrate the embodiments of this application or the technical solutions in the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0044] Figure 1 A flowchart illustrating the steps of a smart indexing method for 3D digital assets of a film provided in this application embodiment;

[0045] Figure 2 This is a system block diagram of a smart indexing method for 3D digital assets of a movie, provided as an embodiment of this application. Detailed Implementation

[0046] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0047] In the description of this application, the terms "comprising," "having," and any variations thereof are intended to cover non-exclusive inclusion, such as a process, method, system, product, or apparatus that includes a series of steps or units, not necessarily limited to those steps or units that are expressly listed, but may also include other steps or units that are not expressly listed but are inherent to these processes, methods, products, or apparatuses, or steps or units added based on further optimizations conceived in this application.

[0048] Traditional 3D digital asset indexing relies on manual annotation, which has the following technical problems:

[0049] 1. Manual annotation is inefficient and costly, making it difficult to meet the rapid data entry needs of massive amounts of 3D digital movie assets.

[0050] 2. Inconsistent labeling standards and ambiguous tag semantics lead to low asset retrieval accuracy and poor reusability;

[0051] 3. Lack of multi-dimensional indexing capabilities for 3D digital asset metadata information (asset format, creation information, file size, texture resolution, etc.) and geometric information (number of vertices, number of mesh faces, etc.);

[0052] The existing intelligent indexing technology has the following technical problems:

[0053] 1. Currently, there is a lack of a unified indexing system and indexing fields for 3D digital assets of movies. Each platform or company uses its own technical solutions, which affects asset interoperability.

[0054] 2. Some 3D digital asset indexing methods convert 3D assets into 2D images for feature analysis. Although this reduces the difficulty of indexing to some extent, there are still problems such as inconsistent processing of 3D formats, incomplete indexing systems, and insufficient extraction of multi-dimensional features.

[0055] 3. Some solutions use simple rule matching, which cannot handle complex semantic information and multimodal information.

[0056] 4. Some other digital content retrieval methods incorporate cross-modal technology, but do not focus on the indexing of 3D digital assets in movies, thus failing to meet the needs of high-precision, multi-dimensional, and intelligent indexing of 3D digital assets in movies.

[0057] Therefore, there is an urgent need for a method that can solve the above problems and achieve efficient, accurate, and unified indexing of 3D digital assets for movies.

[0058] Please refer to Figure 1 The document illustrates a flowchart of a smart indexing method for 3D digital assets of a film provided in an embodiment of this application. The method may include the following steps:

[0059] S1, constructing a 3D digital asset indexing system for movies.

[0060] In this step, a film 3D digital asset indexing system is constructed, hierarchically divided according to four dimensions: asset type, asset scene style, asset scene purpose, and asset precision level. The asset type dimension, referencing the production methods and morphological characteristics of film digital assets, sets up primary categories such as character assets, scene assets, prop assets, special effects assets, texture assets, Gaussian assets, and animation assets, along with their secondary and tertiary subcategories. The scene style dimension sets up primary categories such as modern / modern style, ancient style, and science fiction / fantasy style. The asset scene purpose dimension is divided according to the film's technical process into creation stage, production stage, post-production stage, distribution and marketing stage, and operation and derivative stage. The asset precision level dimension is divided into high precision level, medium precision level, and low precision level based on geometric complexity and texture precision. This establishes a unified indexing and classification framework covering the application needs of the entire lifecycle of 3D digital assets, providing standardized classification criteria and a controlled vocabulary for subsequent large-scale model semantic indexing and rule-based indexing.

[0061] S2 converts various heterogeneous 3D digital assets into the OpenUSD format to obtain standardized digital assets.

[0062] In this step, after the indexing system is built, a standardized conversion is performed on the various heterogeneous 3D digital assets in the database. 3D digital assets generated from different sources and software, such as FBX, OBJ, PLY, and GLB / GLTF, are uniformly converted into OpenUSD format to eliminate data parsing barriers caused by format differences and form standardized digital assets with consistent structure. This standardized format not only provides a unified data input basis for subsequent calls to Blender software for multi-view automated rendering, but also facilitates the system to directly read the metadata and geometric information of the assets, ensuring the continuity and compatibility of the indexing process at the data level.

[0063] S3 calls rendering software to perform multi-view rendering of standardized digital assets, outputting rendered images from six different perspectives.

[0064] In this step, based on the converted OpenUSD standardized digital asset, the Blender rendering software is called to configure six-view rendering parameters, and six different 2D rendered images from different perspectives are output sequentially: front view, rear view, left view, right view, top view, and isometric view. This is to comprehensively capture the appearance texture features and spatial structure information of the 3D digital asset. The rendered images are then Base64 encoded to provide multi-view input data for the next step of calling the large model via API for visual semantic analysis, ensuring that the large model can comprehensively understand the three-dimensional shape and surface details of the asset from different angles.

[0065] S4 constructs system prompts and user prompts, calls the large model via API, sends the system prompts, six rendered images from different perspectives, and user prompts to the large model, and receives the structured indexing results returned by the large model.

[0066] In this step, six different perspectives are included: front view, rear view, left view, right view, top view, and isometric view. For the 3D digital assets to be indexed, system prompts containing indexing rules, classification constraints, and examples are constructed to guide the large model to output standardized classification results according to the preset movie 3D digital asset indexing system. Simultaneously, user prompts containing output format requirements, quality constraints, and Base64 encoded image reception methods are constructed. Then, via API, the large model service is called, and the system prompts, the six Base64-encoded rendered images from different perspectives, and the user prompts are organized into a request body and sent to the large model interface. The large model is required to output structured JSON format indexing results covering asset type and scene style based on multi-view visual information, achieving automated semantic indexing based on artificial intelligence visual understanding.

[0067] S5 parses the structured indexing results, extracts the asset type and asset scene style of 3D digital assets, and completes semantic indexing.

[0068] In this step, after receiving the structured JSON response returned by the large model, the response data is parsed to extract the asset type and asset scene style indexing information of the 3D digital assets. The visual semantic understanding results of the large model are mapped to the classification nodes of the indexing system constructed in step S1 to complete the semantic indexing of the core asset classification. This enables the assets to obtain semantic tags that conform to the unified indexing framework, laying the foundation for accurate asset retrieval, rapid location, and subsequent integration with the basic attribute indexing results.

[0069] S6 extracts the metadata and geometric information of standardized digital assets and indexes them as basic attributes of 3D digital assets.

[0070] In this step, metadata information includes asset format, creation information, file size, and texture resolution, while geometric information includes the number of vertices and the number of mesh faces. During or after the OpenUSD format conversion, the metadata and geometric information of the standardized digital asset are automatically extracted. The metadata information includes asset format, creation information, file size, and texture resolution, while the geometric information includes the number of vertices and the number of mesh faces. These objective technical parameters are indexed as basic attributes of the 3D digital asset, supplementing the asset's physical and technical indicators. This constructs a multi-dimensional descriptive foundation that complements semantic indexing and provides quantitative data support for subsequent rule-based calculations to determine the asset's scenario usage and accuracy level.

[0071] S7 indexes 3D digital assets based on texture resolution and mesh count in the basic attributes, and indexes them according to asset scene usage and asset precision level, to obtain multi-dimensional fusion indexing results.

[0072] In this step, based on the basic attribute information extracted in step S6, the texture resolution and mesh facet count of the 3D digital asset are quantitatively analyzed using preset calculation rules to determine its geometric complexity and texture accuracy. Combined with the classification standards for asset scene usage and asset accuracy level in the indexing system constructed in step S1, the applicable film technology process stage and its accuracy level for the 3D digital asset are automatically determined. This yields a multi-dimensional fusion indexing result covering semantic indexing, basic attribute indexing, and rule-based indexing, achieving a comprehensive structured description of the film's 3D digital asset from visual content to technical parameters and application scenarios. In an optional embodiment of this step, the geometric complexity can be calculated by combining the number of vertices and mesh facet counts.

[0073] In summary, this application proposes a film 3D digital asset indexing system, involving four indexing dimensions: asset type, asset scene style, asset scene purpose, and asset precision level. By calling large model APIs, it indexes the asset type and scene style of rendered images of 3D digital assets from different perspectives, and obtains metadata information such as the 3D digital asset's format, creation information, file size, texture resolution, vertex count, and mesh face count, as well as geometric information. This information serves as the basic attribute information of the digital asset, and the system indexes these basic attributes. This improves the indexing efficiency of film 3D digital assets and effectively meets the needs for high-precision, multi-dimensional, and intelligent indexing of film 3D digital assets.

[0074] like Figure 2 This application also provides an intelligent indexing system for 3D digital assets of movies, which may include:

[0075] The building module is used to construct a 3D digital asset indexing system for movies;

[0076] The format conversion module is used to convert 3D digital assets in various heterogeneous formats into the OpenUSD format to obtain standardized digital assets.

[0077] The rendering module is used to call rendering software to perform multi-view rendering of the standardized digital assets and output rendered images from six different perspectives; wherein, the six different perspectives include front view, rear view, left view, right view, top view, and isometric view.

[0078] The large model indexing module is used to construct system prompts and user prompts. It calls the large model through API, sends the system prompts, the rendered images from six different perspectives, and the user prompts to the large model, and receives the structured indexing results returned by the large model.

[0079] The semantic indexing module is used to parse the structured indexing results, extract the asset type and asset scene style of the 3D digital assets, and complete the semantic indexing.

[0080] The attribute extraction module is used to extract the metadata and geometric information of the standardized digital assets, and index them as the basic attributes of the 3D digital assets; wherein, the metadata information includes asset format, creation information, file size, texture resolution, and the geometric information includes the number of vertices and the number of mesh faces;

[0081] The rule indexing module is used to index the 3D digital assets according to the texture resolution and mesh count in the basic attributes, and to obtain multi-dimensional fusion indexing results.

[0082] Specific limitations regarding the intelligent indexing system for 3D digital assets of films can be found in the limitations of the intelligent indexing method for 3D digital assets of films mentioned above, and will not be repeated here. Each module in the aforementioned intelligent indexing system for 3D digital assets of films can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0083] In one embodiment, an electronic device, which may be a computer, is provided. The electronic device includes a processor, memory, and a network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database of the computer device is used for smart indexing data. The network interface of the computer device is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a smart indexing method.

[0084] In one embodiment of this application, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the steps of the above-described intelligent indexing method for 3D digital assets of movies.

[0085] In one embodiment of this application, a computer program product is provided, including a computer program / instructions, which, when executed by a processor, implements the steps of the above-described intelligent indexing method for 3D digital assets of movies.

[0086] The computer-readable storage medium and computer program product provided in this embodiment are similar in implementation principle and technical effect to the above method embodiments, and will not be repeated here.

[0087] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods.

[0088] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0089] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for intelligent indexing of 3D digital assets in film, characterized in that, The method includes: Constructing a 3D digital asset indexing system for films; Various heterogeneous 3D digital assets are uniformly converted into the OpenUSD format to obtain standardized digital assets; The standardized digital asset is rendered from multiple perspectives using rendering software, resulting in six different viewpoints: front view, rear view, left view, right view, top view, and isometric view. Construct system prompts and user prompts, call the large model via API, send the system prompts, the rendered images from six different perspectives, and the user prompts to the large model, and receive the structured indexing results returned by the large model; Parse the structured indexing results, extract the asset type and asset scene style of the 3D digital assets, and complete semantic indexing; The metadata and geometric information of the standardized digital assets are extracted and indexed as basic attributes of the 3D digital assets; wherein, the metadata information includes asset format, creation information, file size, and texture resolution, and the geometric information includes the number of vertices and the number of mesh faces; Based on the texture resolution and mesh count in the basic attributes, the 3D digital assets are indexed according to their scene usage and accuracy level, resulting in multi-dimensional fusion indexing results.

2. The method according to claim 1, characterized in that, The construction of the film 3D digital asset indexing system specifically includes: The asset indexing system is divided into four dimensions: asset type, asset scenario style, asset scenario purpose, and asset accuracy level. Under different dimensions, first-level, second-level, and third-level subcategories of assets are established to construct a hierarchical asset indexing system.

3. The method according to claim 1, characterized in that, The system prompts are used to guide the large model to index the type and scene style of the digital assets according to the movie 3D digital asset indexing system. The user prompts include the large model output requirements and output format requirements.

4. The method according to claim 1, characterized in that, The step of calling the large model via API to send the system prompts, the rendered images from six different perspectives, and the user prompts to the large model includes: The rendered images from the six different perspectives are Base64 encoded; Set the request body, and organize the system prompt, the Base64 encoded rendered image, and the user prompt into request parameters; Send the request body to the API interface of the large model, requesting the large model to output indexed results in structured JSON format.

5. The method according to claim 1, characterized in that, The step of indexing the 3D digital asset according to the texture resolution and mesh count in the basic attributes, including: The geometric complexity and texture accuracy of the 3D digital asset are calculated based on the texture resolution and the number of mesh faces. Based on the geometric complexity and texture precision, and combined with the classification standards for asset scene use and asset precision level in the asset indexing classification system, the asset scene use and asset precision level of the 3D digital asset are determined.

6. The method according to claim 1, characterized in that, The asset types include character assets, scene assets, prop assets, special effects assets, texture assets, Gaussian assets, and animation assets. The scene styles include modern / modern, ancient, and science fiction / fantasy styles. The purpose of the scene is determined according to the applicable stage of the 3D digital asset in the film technology process, including the creation stage, production stage, post-production stage, distribution and marketing stage, and operation and derivative stage. The asset precision level is divided into high precision level, medium precision level, and low precision level according to the geometric complexity and texture precision.

7. The method according to claim 1, characterized in that, The method further includes: Based on a multimodal visual understanding model, visual semantic analysis is performed on the rendered images from six different perspectives to identify the subject type, style features, and emotional atmosphere in the 3D digital assets and generate multimodal indexing tags. By fusing the multimodal indexing tags with the semantic indexing, the basic attribute indexing, and the asset scenario usage and asset precision level indexing, a full-dimensional indexing result covering visual content, technical parameters, and application scenarios is obtained.

8. A smart indexing system for 3D digital assets of movies, characterized in that, The system includes: The building module is used to construct a 3D digital asset indexing system for movies; The format conversion module is used to convert 3D digital assets in various heterogeneous formats into the OpenUSD format to obtain standardized digital assets. The rendering module is used to call rendering software to perform multi-view rendering of the standardized digital assets and output rendered images from six different perspectives; wherein, the six different perspectives include front view, rear view, left view, right view, top view, and isometric view. The large model indexing module is used to construct system prompts and user prompts. It calls the large model through API, sends the system prompts, the rendered images from six different perspectives, and the user prompts to the large model, and receives the structured indexing results returned by the large model. The semantic indexing module is used to parse the structured indexing results, extract the asset type and asset scene style of the 3D digital assets, and complete the semantic indexing. The attribute extraction module is used to extract the metadata and geometric information of the standardized digital assets, and index them as the basic attributes of the 3D digital assets; wherein, the metadata information includes asset format, creation information, file size, texture resolution, and the geometric information includes the number of vertices and the number of mesh faces; The rule indexing module is used to index the 3D digital assets according to the texture resolution and mesh count in the basic attributes, and to obtain multi-dimensional fusion indexing results.

9. An electronic device, characterized in that, It includes a memory and a processor, the memory storing a computer program that, when executed by the processor, implements the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the method as described in any one of claims 1 to 7.

Images