Art product design system based on Internet

Abstract

The invention relates to the technical field of art product design, in particular to an Internet-based art product design system. The system comprises a material acquisition management unit which obtains art product materials from a material platform through an internet interface protocol and generates a standardized material package; the semantic drive design generation unit converts user input into structured scene data based on a cross-modal semantic analysis technology, and constructs a semantic mapping matrix and an element confidence distribution mechanism to match the structured scene data with the standardized material package; the collaborative version fusion unit monitors and solves multi-version conflicts by using a timestamp consistency branch merging algorithm, and generates final art product work data; the cross-platform publishing rendering unit carries out adaptation processing on the final art product work data according to the requirements of the target publishing platform; according to the method, generation and release of an art product scene are realized in combination with a high-precision scene generation and timestamp consistency branch merging algorithm driven by cross-modal semantic analysis.

Description

Technical Field

[0001] This invention relates to the field of art product design technology, and more specifically to an Internet-based art product design system. Background Technology

[0002] With the deep integration of internet technology and the digital creative industry, art product design systems are gradually evolving from localized standalone software to cloud-based and collaborative approaches. Internet-based art product design platforms can integrate various design resources, creative tools, and rendering services into a single online environment, providing users with cross-platform and cross-regional collaborative creation capabilities. The application of technologies such as natural language processing, cross-modal content generation, and online collaborative management has significantly improved the interactive experience and production efficiency of the art product creation process. However, existing systems still have bottlenecks in areas such as the conversion efficiency from high-precision semantic parsing to scene generation and multi-user collaborative version management, which affect the generation quality of complex scenes and the stability of collaborative creation.

[0003] In existing internet art product design systems, when users describe their creative needs in natural language, the lack of a sophisticated cross-modal semantic parsing mechanism and dynamic precision adjustment strategy often makes it difficult for the system to accurately match elements in the material library, resulting in the generated scene's details not conforming to the creative intent. In scenarios where multiple users are editing the same art product project in parallel, another type of problem exists: due to the lack of an effective timestamp consistency branch merging and event-driven differential synchronization mechanism, version conflicts, data redundancy, and transmission delays are easily generated, leading to decreased collaboration efficiency and incomplete work data. These problems are particularly prominent in art product creation projects involving a large number of material references, high-concurrency collaboration, and cross-platform publishing, urgently requiring a technical solution that achieves breakthroughs in both semantic-to-scene generation precision and multi-user collaborative version management. Summary of the Invention

[0004] The purpose of this invention is to provide an Internet-based art product design system to solve the problems mentioned in the background art, which are prone to version conflicts, data redundancy and transmission delays due to the lack of an effective timestamp consistency branch merging and event-driven differential synchronization mechanism, resulting in decreased collaboration efficiency and incomplete work data.

[0005] To achieve the above objectives, the invention aims to provide an internet-based art product design system, comprising:

[0006] The material acquisition and management unit is used to acquire art product materials from the material platform through the Internet interface protocol, organize and index them according to preset classification rules, and generate standardized material packages with version identifiers and copyright labels.

[0007] The semantic-driven design generation unit converts user input into structured scene data based on cross-modal semantic parsing technology. It constructs a semantic mapping matrix and an element confidence distribution mechanism to match the structured scene data with standardized material packages and dynamically adjusts the scene generation accuracy according to the confidence distribution.

[0008] The collaborative version fusion unit uses a timestamp consistency branch merging algorithm to monitor and resolve multi-version conflicts, and adopts an event-driven differential synchronization mechanism to transmit edited change data blocks and generate the final artwork data.

[0009] The cross-platform publishing and rendering unit is used to adapt the final artwork data to the requirements of the target publishing platform and generate the target format file using a progressive resolution rendering strategy.

[0010] Preferably, in the material acquisition and management unit, the standardized material package is a collection of materials stored with multi-dimensional tags based on material type, purpose and resolution level, which is used for automatic retrieval and matching in the semantic-driven design generation unit;

[0011] The version identifier of the standardized material package adopts a globally unique identifier generation algorithm and is bound to the hash verification value of the material content; the copyright label of the standardized material package is generated based on a digital signature mechanism and includes material source information, copyright authorization scope and validity period data.

[0012] Preferably, in the semantic-driven design generation unit, cross-modal semantic parsing technology is used to receive user input and map it to the semantic feature space;

[0013] User input includes three modalities: text input, voice input, and sketch input.

[0014] The cross-modal semantic parsing technology establishes a semantic feature alignment network to align the feature vectors of the three modalities in the same semantic coordinate system, thereby enabling the collaborative participation and information complementarity of the three modalities in the process of generating structured scene data.

[0015] Preferably, in the semantic-driven design generation unit, the semantic mapping matrix is ​​a two-dimensional weighted matrix with semantic feature vectors as row indices and standardized material package element feature vectors as column indices, used to represent the matching strength between semantic units and material elements;

[0016] The semantic mapping matrix construction method is as follows: obtain the semantic feature vector of the user input processed by cross-modal semantic parsing technology; obtain the material feature vector in the standardized material package; calculate the matching similarity value between the semantic feature vector and the material feature vector; assign weights to the two-dimensional weight matrix according to the matching similarity value, and perform regularization processing on the two-dimensional weight matrix to obtain the semantic mapping matrix.

[0017] Preferably, in the semantic-driven design generation unit, the element confidence distribution mechanism is used to perform probabilistic evaluation of the matching results of each material element in the semantic mapping matrix, and output the results to the scene generation engine in the form of a probability distribution.

[0018] The calculation method for the element confidence distribution mechanism is as follows:

[0019] The weight value corresponding to each material element in the semantic mapping matrix is ​​input into the normalization function to obtain the initial confidence value; the initial confidence value is then adjusted by weighting based on the matching stability of the material element in the historical generation records and the frequency of user corrections to form the final confidence distribution;

[0020] The final confidence distribution is used to determine the priority and replacement strategy of material elements in scene generation during dynamic precision adjustment.

[0021] Preferably, in the semantic-driven design generation unit, the scene generation accuracy is dynamically adjusted according to the confidence distribution. The specific method is as follows:

[0022] When the variance of the confidence distribution is less than the first preset threshold, a detail downsampling strategy is executed to reduce the rendering overhead of low-difference material elements; when the variance of the confidence distribution is greater than the second preset threshold, a detail enhancement strategy is executed to improve the scene expressiveness of high-difference material elements through high-precision rendering and material replacement operations.

[0023] The first preset threshold and the second preset threshold are dynamically calculated based on the complexity of the target scene and the performance parameters of the user equipment.

[0024] Preferably, in the collaborative version fusion unit, the timestamp consistency branch merging algorithm is used to merge branch versions based on the timestamp order and dependency relationship of each editing operation when multiple users are editing the same art product project in parallel.

[0025] The specific operation of the timestamp consistency branch merging algorithm is as follows:

[0026] For each editing operation, a timestamp accurate to milliseconds is generated, and the identifiers of the material elements and operation types that the operation depends on are recorded. A directed acyclic graph of timestamps is constructed, and all editing operations are sorted according to time order and dependencies. In the merge conflict detection phase, when two or more editing operations involve the same material element and the timestamp difference is less than a preset conflict threshold, they are marked as potential conflicts. In the merge execution phase, based on the topological sequence of the directed acyclic graph of timestamps, the editing operations with newer timestamps are preferentially retained, while the overwritten operations are recorded.

[0027] Preferably, in the collaborative version fusion unit, the event-driven differential synchronization mechanism is used to transmit only the data differential blocks related to the change when a version change event is detected;

[0028] The event-driven differential synchronization mechanism includes:

[0029] The differential synchronization process is triggered by listening to user operation events, material replacement events, and scene structure adjustment events; the differential generation module is called to compare the current version data with the previous version data to generate a differential data block containing the change location index, change type, and change data content; the differential data block is compressed and encoded and sent to the target client, and differential application operations are performed on the target client.

[0030] Preferably, in the collaborative version fusion unit, the final art product data includes a material reference index table, a design structure description file, and a rendering parameter configuration file;

[0031] The material reference index table records the unique identifier of the material in the standardized material package and its version information; the design structure description file defines the spatial position and relationship of each element in the scene in a hierarchical node manner; the rendering parameter configuration file contains rendering control parameters for color, lighting, shadow and resolution.

[0032] Preferably, in the cross-platform publishing rendering unit, the progressive resolution rendering strategy is used to gradually increase the rendering resolution in stages after the target publishing platform receives the final art product data.

[0033] The progressive resolution rendering strategy specifically includes:

[0034] In the first stage, scene elements are quickly pre-rendered at a low resolution to provide an instant preview effect; in the second stage, the rendering resolution is increased and detailed texture maps are applied based on the user's device processing power and network bandwidth; in the third stage, when the available resources of the user's device are detected to be higher than a preset threshold, full-resolution rendering is performed and the target format file is generated.

[0035] Compared with the prior art, the above-mentioned technical solution of the present invention has the following beneficial technical effects:

[0036] 1. In this invention, based on cross-modal semantic parsing, semantic mapping matrix construction and element confidence distribution mechanism, user natural language input can be accurately converted into structured scene data and adaptively matched with standardized material packages, thereby significantly improving the accuracy and detail of semantic-to-scene generation;

[0037] 2. In this invention, the synergistic application of the timestamp consistency branch merging algorithm and the event-driven differential synchronization mechanism enables automated processing of version conflicts and efficient differential data transmission during multi-user collaboration, effectively improving collaboration efficiency and ensuring the integrity of the final artwork data. Attached Figure Description

[0038] Figure 1 This is a schematic diagram of one embodiment of the present invention;

[0039] Attached label: 1. Material acquisition and management unit; 2. Semantic-driven design generation unit; 3. Collaborative version fusion unit; 4. Cross-platform publishing and rendering unit. Detailed Implementation

[0040] like Figure 1 As shown, an Internet-based art product design system is provided, including: material acquisition and management unit 1, semantic-driven design generation unit 2, collaborative version fusion unit 3, and cross-platform publishing and rendering unit 4.

[0041] In the material acquisition and management unit 1 of this embodiment, the material acquisition and management unit 1 is used to obtain art product materials from the material platform through the Internet interface protocol, and organize and index them according to the preset classification rules to generate a standardized material package with version identification and copyright label.

[0042] In this embodiment, the standardized material package is a collection of materials stored with multi-dimensional tags based on material type, purpose and resolution level, which is used for automatic retrieval and matching in the semantic-driven design generation unit 2;

[0043] The version identifier of the standardized material package adopts a globally unique identifier generation algorithm and is bound to the hash verification value of the material content; the copyright label of the standardized material package is generated based on a digital signature mechanism and includes material source information, copyright authorization scope and validity period data.

[0044] In this embodiment, the standardized material package consists of original art product material files, material metadata files, and version control information. The original art product material files include various file types such as images, audio, video, and models. The material metadata files record information such as material type, resolution, color space, file size, and usage restrictions. The version control information includes the material generation timestamp, version number, and change history. To ensure the global uniqueness of version identifiers between different material packages, the system uses a combination of timestamp and UUID for generation, and after generation, hash verification ensures that the identifier is not duplicated in the material database. The copyright label consists of a copyright statement field, creator information, authorization scope, and an encrypted signature. The encrypted signature can be generated using RSA or ECC algorithms and is used for copyright verification during material calls or cross-platform transmission. During the copyright label generation process, the system can also automatically generate corresponding label fields based on authorization data returned by internet material platforms to ensure the matching and traceability of copyright information with the material source.

[0045] In the semantic-driven design generation unit 2 of this embodiment, the semantic-driven design generation unit 2 converts user input into structured scene data based on cross-modal semantic parsing technology, constructs a semantic mapping matrix and element confidence distribution mechanism to match the structured scene data with standardized material packages, and dynamically adjusts the scene generation accuracy according to the confidence distribution.

[0046] In this embodiment, cross-modal semantic parsing technology is used to receive user input and map it to a semantic feature space;

[0047] User input includes three modalities: text input, voice input, and sketch input.

[0048] The cross-modal semantic parsing technology establishes a semantic feature alignment network to align the feature vectors of the three modalities in the same semantic coordinate system, thereby enabling the collaborative participation and information complementarity of the three modalities in the process of generating structured scene data.

[0049] In this embodiment, the cross-modal semantic parsing technology first performs word segmentation, part-of-speech tagging, and dependency parsing on the user-input text data to extract keywords and semantic relationships. The speech data is converted into text using an end-to-end speech recognition model and acoustic sentiment features are extracted. The sketch data is extracted using a convolutional neural network to extract geometric shape features and spatial layout information. Then, the feature vectors of the three modalities are embedded into a pre-trained multimodal semantic embedding space. Vector alignment and weight normalization are performed through a semantic feature alignment network, enabling high-dimensional feature mapping of different modalities in the same semantic coordinate system. In the process of generating structured scene data, the key information of each modality is fused based on a modal complementarity strategy to improve the completeness and semantic consistency of the generated results.

[0050] In this embodiment, the semantic mapping matrix is ​​a two-dimensional weighted matrix with semantic feature vectors as row indices and standardized material package element feature vectors as column indices, used to represent the matching strength between semantic units and material elements;

[0051] The semantic mapping matrix construction method is as follows: obtain the semantic feature vector of the user input processed by cross-modal semantic parsing technology; obtain the material feature vector in the standardized material package; calculate the matching similarity value between the semantic feature vector and the material feature vector; assign weights to the two-dimensional weight matrix according to the matching similarity value, and perform regularization processing on the two-dimensional weight matrix to obtain the semantic mapping matrix.

[0052] In this embodiment, during the construction of the semantic mapping matrix, cross-modal semantic parsing technology is first invoked to obtain the semantic feature vector input by the user, and corresponding feature vectors are generated for each element in the standardized material package based on the material feature extraction model. Then, similarity calculation methods such as cosine similarity, Euclidean distance, or Mahalanobis distance are used to calculate the matching similarity value between the semantic feature vector and the material feature vector, and the similarity value is mapped to the initial weight in the interval [0,1] and filled into a two-dimensional weight matrix with the semantic feature vector as the row index and the material feature vector as the column index. After all weights are filled, L2 regularization and weight smoothing are performed on the matrix to suppress noise interference and improve the robustness of semantic matching. Finally, a semantic mapping matrix that can be directly used for scene generation and dynamic precision adjustment is generated.

[0053] In this embodiment, the element confidence distribution mechanism is used to perform probabilistic evaluation of the matching results of each material element in the semantic mapping matrix and output them to the scene generation engine in the form of a probability distribution.

[0054] The calculation method for the element confidence distribution mechanism is as follows:

[0055] The weight value corresponding to each material element in the semantic mapping matrix is ​​input into the normalization function to obtain the initial confidence value; the initial confidence value is then adjusted by weighting based on the matching stability of the material element in the historical generation records and the frequency of user corrections to form the final confidence distribution;

[0056] The final confidence distribution is used to determine the priority and replacement strategy of material elements in scene generation during dynamic precision adjustment.

[0057] In this embodiment, the element confidence distribution mechanism first converts the weight value of each material element in the semantic mapping matrix into an initial confidence value through a Softmax or Min-Max normalization function to ensure that the sum of the confidence values ​​of each material element is 1. Then, based on the historical generation records recorded in the material management module, the matching stability index of the material element is extracted, and the correction frequency of the element by the user in the subsequent editing process is counted. The two are used as positive weighting factors and negative weighting factors, respectively, to weight and correct the initial confidence value, forming the final confidence distribution. This distribution is output to the scene generation engine in the form of a probability vector. During the dynamic precision adjustment process, it is used to prioritize the use of high-confidence material elements and execute an automatic replacement strategy when low-confidence material elements trigger the replacement threshold, thereby improving the stability of the generated scene and user satisfaction.

[0058] In this embodiment, the scene generation accuracy is dynamically adjusted based on the confidence level distribution. The specific method is as follows:

[0059] When the variance of the confidence distribution is less than the first preset threshold, a detail downsampling strategy is executed to reduce the rendering overhead of low-difference material elements; when the variance of the confidence distribution is greater than the second preset threshold, a detail enhancement strategy is executed to improve the scene expressiveness of high-difference material elements through high-precision rendering and material replacement operations.

[0060] The first preset threshold and the second preset threshold are dynamically calculated based on the complexity of the target scene and the performance parameters of the user equipment.

[0061] In this embodiment, the process of dynamically adjusting the scene generation accuracy first calculates the variance value based on the element confidence distribution, and then calculates a first preset threshold and a second preset threshold in real time by combining the complexity parameters of the scene generation task and the performance indicators of the current user terminal. When the variance value is lower than the first preset threshold, the system determines that the matching difference between material elements is small, triggers a detail downsampling strategy to reduce unnecessary rendering resolution and texture details, thereby reducing rendering overhead and speeding up the generation speed. When the variance value is higher than the second preset threshold, the system determines that the matching difference between material elements is large, triggers a detail enhancement strategy, and enhances the scene expressiveness by calling a high-precision rendering mode, loading higher resolution material files, and replacing low-matching material elements. When the variance value is between the two thresholds, the system maintains the standard rendering mode to balance rendering performance and scene quality.

[0062] In the collaborative version fusion unit 3 of this embodiment, the collaborative version fusion unit 3 uses the timestamp consistency branch merging algorithm to monitor and resolve multi-version conflicts, and adopts an event-driven differential synchronization mechanism to transmit edit change data blocks and generate the final art product data.

[0063] In this embodiment, the timestamp consistency branch merging algorithm is used to merge branch versions based on the timestamp order and dependency of each editing operation when multiple users are editing the same art product project in parallel.

[0064] The specific operation of the timestamp consistency branch merging algorithm is as follows:

[0065] For each editing operation, a timestamp accurate to milliseconds is generated, and the identifiers of the material elements and operation types that the operation depends on are recorded. A directed acyclic graph of timestamps is constructed, and all editing operations are sorted according to time order and dependencies. In the merge conflict detection phase, when two or more editing operations involve the same material element and the timestamp difference is less than a preset conflict threshold, they are marked as potential conflicts. In the merge execution phase, based on the topological sequence of the directed acyclic graph of timestamps, the editing operations with newer timestamps are preferentially retained, while the overwritten operations are recorded.

[0066] In this embodiment, the timestamp consistency branch merging algorithm first generates a timestamp accurate to the millisecond level for each multi-user editing operation, and simultaneously records the material element identifiers, operation types, and editor identity identifiers that the operation depends on. Then, it constructs a directed acyclic graph of timestamps with timestamps as nodes and dependencies as edges, and performs topological sorting in the graph to determine the global execution order of each operation. In the conflict detection phase, when two or more editing operations are detected acting on the same material element and their timestamp difference is less than a preset conflict threshold, the group of operations is marked as potential conflict operations, and further pre-classification of conflicts is performed according to the priority strategy of operation type. In the merging execution phase, the system prioritizes the operation with the latest timestamp and the highest priority according to the topological sequence, and records the overwritten operations in the operation backtracking log to support subsequent version rollback or difference analysis, thereby achieving high-precision version fusion in a high-concurrency collaborative environment.

[0067] In this embodiment, the event-driven differential synchronization mechanism is used to transmit only the data differential blocks related to the version change when a version change event is detected;

[0068] The event-driven differential synchronization mechanism includes:

[0069] The differential synchronization process is triggered by listening to user operation events, material replacement events, and scene structure adjustment events; the differential generation module is called to compare the current version data with the previous version data to generate a differential data block containing the change location index, change type, and change data content; the differential data block is compressed and encoded and sent to the target client, and differential application operations are performed on the target client.

[0070] In this embodiment, the event-driven differential synchronization mechanism first captures user operation events, material replacement events, and scene structure adjustment events in real time through event listeners. Once these events are detected, the differential synchronization process is immediately triggered. The system calls the differential generation module to compare the current version data with the previous version data, and uses a fast comparison algorithm based on content hashing to generate differential data blocks. These data blocks contain the change location index, change type, and specific changed data content. The differential data blocks are then compressed to reduce network transmission overhead. During data transmission, the system prioritizes low-latency transmission channels and sends the encoded differential data blocks to the target client. After receiving the data, the target client calls the differential application module to accurately update the local version data based on the change location index and change type, thereby avoiding the redundant overhead caused by full data transmission and significantly improving the real-time performance of collaborative editing.

[0071] In this embodiment, the final artwork data includes a material reference index table, a design structure description file, and a rendering parameter configuration file;

[0072] The material reference index table records the unique identifier of the material in the standardized material package and its version information; the design structure description file defines the spatial position and relationship of each element in the scene in a hierarchical node manner; the rendering parameter configuration file contains rendering control parameters for color, lighting, shadow and resolution.

[0073] In this embodiment, the process of generating the final artwork data includes the structured integration of collaborative editing results. First, the version fusion engine generates a material reference index table based on the merged scene data. This index table records the reference position and version information of each material in the standardized material package in the form of a combination of unique material identifiers and version numbers, ensuring the accuracy of subsequent rendering and version backtracking processes. Then, a design structure description file is generated. This file uses a hierarchical node tree structure to describe the spatial coordinates, hierarchical relationships, and interactive dependencies of each element in the scene, so as to maintain the consistency of scene layout in different rendering environments. Finally, a rendering parameter configuration file is generated. This file records color model parameters, lighting source configuration, shadow calculation methods, and resolution control parameters, supporting the rendering module to perform personalized rendering output according to the performance requirements of different target platforms, thereby achieving cross-platform consistency and high-quality presentation of the final artwork data.

[0074] In this embodiment, the cross-platform publishing rendering unit 4 is used to adapt the final art product data according to the requirements of the target publishing platform, and to generate the target format file using a progressive resolution rendering strategy.

[0075] In this embodiment, the progressive resolution rendering strategy is used to gradually increase the rendering resolution in stages after the target publishing platform receives the final artwork data.

[0076] The progressive resolution rendering strategy specifically includes:

[0077] In the first stage, scene elements are quickly pre-rendered at a low resolution to provide an instant preview effect; in the second stage, the rendering resolution is increased and detailed texture maps are applied based on the user's device processing power and network bandwidth; in the third stage, when the available resources of the user's device are detected to be higher than a preset threshold, full-resolution rendering is performed and the target format file is generated.

[0078] In this embodiment, the progressive resolution rendering strategy, during execution, firstly, after receiving the final artwork data in the first stage, rapidly pre-renders scene elements in a low-resolution mode, prioritizing the loading of design structure description files and basic lighting information to provide users with an instant preview effect in a very short time. In the second stage, based on real-time detection of the user device's processor utilization, video memory usage, and network bandwidth, the system gradually increases the rendering resolution and dynamically loads and applies high-precision detail texture maps and shadow effects during the rendering process, thereby enhancing the image's detail performance. In the third stage, when it is detected that the user device's available computing resources are continuously higher than a preset threshold, a full-resolution rendering task is initiated, performing high-precision sampling and global illumination calculations on all scene elements, and finally generating a target format file that meets the requirements of the target publishing platform, thereby achieving the optimal balance between image quality and performance.

[0079] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited thereto. Various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention.

Claims

1. An internet-based art product design system, characterized in that, include: Material Acquisition and Management Unit (1) is used to acquire art product materials from the material platform through the Internet interface protocol, and organize and index them according to the preset classification rules to generate a standardized material package with version identification and copyright label; The semantic-driven design generation unit (2) converts user input into structured scene data based on cross-modal semantic parsing technology, constructs a semantic mapping matrix and element confidence distribution mechanism to match structured scene data with standardized material packages, and dynamically adjusts scene generation accuracy according to confidence distribution. Collaborative version fusion unit (3) uses a timestamp consistency branch merging algorithm to monitor and resolve multi-version conflicts, and adopts an event-driven differential synchronization mechanism to transmit edit change data blocks and generate final art product data; The cross-platform publishing rendering unit (4) is used to adapt the final art product data according to the requirements of the target publishing platform and generate the target format file using a progressive resolution rendering strategy.

2. The Internet-based art product design system according to claim 1, characterized in that, In the material acquisition and management unit (1), the standardized material package is a collection of materials stored in a multi-dimensional tagged manner according to material type, purpose and resolution level, which is used for automatic retrieval and matching in the semantic-driven design generation unit (2); The version identifier of the standardized material package adopts a globally unique identifier generation algorithm and is bound to the hash verification value of the material content; the copyright label of the standardized material package is generated based on a digital signature mechanism and includes material source information, copyright authorization scope and validity period data.

3. The Internet-based art product design system according to claim 2, characterized in that, In the semantic-driven design generation unit (2), cross-modal semantic parsing technology is used to receive user input and map it to the semantic feature space; User input includes three modalities: text input, voice input, and sketch input. The cross-modal semantic parsing technology establishes a semantic feature alignment network to align the feature vectors of the three modalities in the same semantic coordinate system, thereby enabling the collaborative participation and information complementarity of the three modalities in the process of generating structured scene data.

4. The Internet-based art product design system according to claim 3, characterized in that, In the semantic-driven design generation unit (2), the semantic mapping matrix is ​​a two-dimensional weight matrix with semantic feature vectors as row indices and standardized material package element feature vectors as column indices, used to represent the matching strength between semantic units and material elements; The semantic mapping matrix construction method is as follows: obtain the semantic feature vector of the user input processed by cross-modal semantic parsing technology; obtain the material feature vector in the standardized material package; calculate the matching similarity value between the semantic feature vector and the material feature vector; assign weights to the two-dimensional weight matrix according to the matching similarity value, and perform regularization processing on the two-dimensional weight matrix to obtain the semantic mapping matrix.

5. The Internet-based art product design system according to claim 4, characterized in that, In the semantic-driven design generation unit (2), the element confidence distribution mechanism is used to perform probabilistic evaluation of the matching results of each material element in the semantic mapping matrix and output them to the scene generation engine in the form of a probability distribution. The calculation method for the element confidence distribution mechanism is as follows: The weight value corresponding to each material element in the semantic mapping matrix is ​​input into the normalization function to obtain the initial confidence value; the initial confidence value is then adjusted by weighting based on the matching stability of the material element in the historical generation records and the frequency of user corrections to form the final confidence distribution; The final confidence distribution is used to determine the priority and replacement strategy of material elements in scene generation during dynamic precision adjustment.

6. The Internet-based art product design system according to claim 5, characterized in that, In the semantic-driven design generation unit (2), the scene generation accuracy is dynamically adjusted according to the confidence distribution. The specific method is as follows: When the variance of the confidence distribution is less than the first preset threshold, a detail downsampling strategy is executed to reduce the rendering overhead of low-difference material elements; when the variance of the confidence distribution is greater than the second preset threshold, a detail enhancement strategy is executed to improve the scene expressiveness of high-difference material elements through high-precision rendering and material replacement operations. The first preset threshold and the second preset threshold are dynamically calculated based on the complexity of the target scene and the performance parameters of the user equipment.

7. The Internet-based art product design system according to claim 6, characterized in that, In the collaborative version fusion unit (3), the timestamp consistency branch merging algorithm is used to merge branch versions based on the timestamp order and dependency relationship of each editing operation when multiple users are editing the same art product project in parallel. The specific operation of the timestamp consistency branch merging algorithm is as follows: For each editing operation, generate a timestamp accurate to the millisecond level and record the material element identifiers and operation types that the operation depends on; construct a directed acyclic graph of timestamps and sort all editing operations according to time order and dependencies; During the merge conflict detection phase, when two or more editing operations involve the same material element and the timestamp difference is less than the preset conflict threshold, they are marked as potential conflicts. During the merge execution phase, based on the topological sequence of the directed acyclic graph according to timestamps, the edit operations with newer timestamps are retained first, while the overwritten operations are recorded.

8. The Internet-based art product design system according to claim 7, characterized in that, In the collaborative version fusion unit (3), the event-driven differential synchronization mechanism is used to transmit only the data differential blocks related to the change when a version change event is detected. The event-driven differential synchronization mechanism includes: The differential synchronization process is triggered by listening to user operation events, material replacement events, and scene structure adjustment events; the differential generation module is called to compare the current version data with the previous version data to generate a differential data block containing the change location index, change type, and change data content; the differential data block is compressed and encoded and sent to the target client, and differential application operations are performed on the target client.

9. The Internet-based art product design system according to claim 8, characterized in that, In the collaborative version fusion unit (3), the final art product data includes a material reference index table, a design structure description file, and a rendering parameter configuration file; The material reference index table records the unique identifier of the material in the standardized material package and its version information; the design structure description file defines the spatial position and relationship of each element in the scene in a hierarchical node manner; the rendering parameter configuration file contains rendering control parameters for color, lighting, shadow and resolution.

10. The Internet-based art product design system according to claim 9, characterized in that, In the cross-platform publishing rendering unit (4), the progressive resolution rendering strategy is used to gradually increase the rendering resolution in stages after the target publishing platform receives the final art product data. The progressive resolution rendering strategy specifically includes: In the first stage, scene elements are quickly pre-rendered at a low resolution to provide an instant preview effect; in the second stage, the rendering resolution is increased and detailed texture maps are applied based on the user's device processing power and network bandwidth; in the third stage, when the available resources of the user's device are detected to be higher than a preset threshold, full-resolution rendering is performed and the target format file is generated.

Images