AI model production method and system based on meta universe data marking
By double-marking on the blockchain and training and testing on the metacosmic platform, data privacy, security and quality issues in AI model production are solved, efficient and secure AI model production is achieved, and the model is improved performance and adaptability.
Patent Information
- Application Number
- CN202411393751.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-08
- Publication Date
- 2025-01-10
AI Technical Summary
The existing AI model production methods have data privacy and security risks, data quality and authenticity issues, and the lack of user control over data.
By storing and performing double-labeled data sets on the blockchain, combining the metacosmic platform for training and testing of AI models, ensuring data security and credibility, and optimizing model performance through data tagging and virtual reality environments.
It improves the security and credibility of data, ensures the authenticity and integrity of data, enhances the performance and adaptability of AI models, and provides efficient and intelligent solutions for various application scenarios.
Smart Images
Figure CN119294558A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of AI model production technology, and in particular to an AI model production method and system based on metaverse data labeling. Background Art
[0002] An AI model is a mathematical model based on artificial intelligence technology that can receive input data, automatically process and analyze it, and then return the corresponding results. AI models are usually composed of multiple algorithms and programs, and can improve their accuracy and efficiency through training and learning. AI models can be divided into different types, including deep learning-based models and large language models (LLMs), which can process massive amounts of data and learn complex patterns and relationships. AI models are widely used in various fields, such as medicine, education, entertainment, and business. AI models have entered daily life in a wild way, and the speed of progress is rapid. From question-and-answer dialogues to assisted programming, from picture analysis to autonomous creation, all walks of life are increasingly inseparable from AI models.
[0003] There are some drawbacks in the current AI model production methods, including: 1) Data privacy and security: AI model training usually requires a large amount of data, which may involve the user's privacy information. Under the traditional centralized data set storage and processing method, there is a risk of data leakage and abuse. 2) Data quality and authenticity: The data used to train AI models may have quality problems, including missing data, incorrect labeling, etc., which affect the accuracy and reliability of the model. 3) Data ownership and control: Under the traditional centralized data set management, the ownership and control of data are usually concentrated in the data provider or data platform, and users often lack control over their own data.
[0004] By adopting new technologies such as Metaverse and data labeling, these drawbacks can be alleviated to a certain extent, the security, credibility and transparency of data can be improved, and more advantages can be brought to the production and application of AI models. Therefore, the present invention proposes an AI model production method and system based on Metaverse data labeling. Summary of the invention
[0005] In order to overcome the shortcomings of the prior art, the purpose of the present invention is to provide an AI model production method and system based on Metaverse data labeling. By utilizing data labeling and the Metaverse platform, the quality, security and credibility of data can be improved, and efficient, high-performance and highly adaptable AI models can be produced to provide intelligent solutions for various application scenarios.
[0006] To achieve the above objectives, the present invention adopts an AI model production method based on metaverse data labeling, comprising the following steps:
[0007] Collecting and organizing data sets, and storing said data sets on a blockchain;
[0008] Select an AI model framework according to the type of the data set, and use the data set for training to obtain an initial AI model;
[0009] The initial AI model is deployed to the Metaverse platform for testing and optimization to obtain an AI model.
[0010] Optionally, the specific process of collecting and organizing the data set is:
[0011] Collect data and handle data outliers, missing values, and duplicate values;
[0012] Use data labeling tools to label the processed data according to the data type to complete the initial labeling;
[0013] Using sampling inspection and cross-validation to evaluate whether the data after the initial labeling is accurate, if accurate, exporting the data after the initial labeling, if inaccurate, re-labeling using the data labeling tool;
[0014] Establishing data labels according to the exported data after the initial labeling to obtain a type of data labels, wherein the type of data labels is used for rapid labeling of subsequent data;
[0015] Select the format for data export to match the AI model framework and obtain the data set after initial labeling.
[0016] Optionally, the specific process of storing the data set on the blockchain is:
[0017] Using blockchain technology, the data set after the initial marking is re-marked, and the content of the secondary marking includes the data source, timestamp and data owner. According to the data after the secondary marking, a second type of data label is obtained, and the second type of data label is used for the rapid marking of subsequent data;
[0018] By using blockchain smart contracts and granting user permissions, only authorized users can access and use data, thus achieving data access control;
[0019] Utilize data transaction and authorization mechanisms to record and store every step of data transaction;
[0020] The secondary marked data set and data transaction information are stored on the blockchain for real-time supervision. If any violation is found, a warning message is automatically sent.
[0021] Optionally, the specific process of selecting an AI model framework according to the type of the data set and using the data set for training to obtain the initial AI model is as follows:
[0022] Based on the second labeled data set, the AI model framework is trained using a loss function to obtain a first sample model and a first loss value, and the first sample model is trained using the loss function and the first loss value to obtain a second sample model and a second loss value, and finally the second sample model is trained using the loss function, the second loss value and an optimizer to obtain the initial AI model;
[0023] During the training process, the model parameters are updated in real time through the back-propagation algorithm, and the updated model parameters are synchronized to the blockchain.
[0024] Optionally, the initial AI model is deployed to the Metaverse platform for testing and optimization. The specific process of obtaining the AI model is as follows:
[0025] Select and determine the test objectives and indicators of the initial AI model in the Metaverse, where the test objectives include the performance, stability, and scalability of the initial AI model, and the indicators are corresponding test thresholds set based on different test objectives of the initial AI model;
[0026] Use virtual reality and augmented reality technologies to build a simulated metaverse environment;
[0027] Based on the secondary labeled data set, deploy the initial AI model in the metaverse environment, and perform testing and evaluation;
[0028] Collect and analyze the output and behavior data of the initial AI model in the metaverse to obtain test results;
[0029] The test result is compared with the corresponding test threshold to obtain a comparison result, and then based on the comparison result, the initial AI model is optimized and adjusted by adjusting model parameters, increasing training data and improving the algorithm to obtain the AI model.
[0030] The present invention also adopts an AI model production system based on metaverse data labeling, including:
[0031] A data labeling module, used to collect and organize data sets and store the data sets on the blockchain;
[0032] A model training module, used to select an AI model framework according to the type of the data set, and use the data set for training to obtain an initial AI model;
[0033] The model optimization module is used to deploy the initial AI model to the Metaverse platform for testing and optimization to obtain an AI model.
[0034] The present invention discloses the following technical effects by providing an AI model production method and system based on metaverse data labeling:
[0035] 1. Through data labeling, the present invention can provide high-quality labeled data sets for machine learning model training, so that the model can achieve efficient learning; and through double labeling and establishing data labels, the data labeling process is made more accurate and faster, ensuring the consistency of the data, ensuring the quality of the labeled data, and meeting the requirements of model training.
[0036] 2. By using blockchain to store and mark data in AI models, the present invention can improve the security and credibility of data, ensure the authenticity and integrity of data, and provide more reliable data support for the training and application of AI models.
[0037] 3. The present invention performs training through multiple training sessions to reduce model loss and model error, ensure that the model gradually converges during the training process, improve the performance and generalization ability of the model, and enable the model to better solve practical problems.
[0038] 4. By building a metaverse platform, the present invention can continuously test and optimize AI models, improve the performance and adaptability of the models in virtual environments, and provide better support for solving practical problems.
[0039] The technical solution of the utility model is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS
[0040] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0041] Figure 1 A flow chart of a method provided by an embodiment of the present invention;
[0042] Figure 2 A system architecture diagram provided for an embodiment of the present invention. DETAILED DESCRIPTION
[0043] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0044] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.
[0045] like Figure 1 As shown, the present invention provides an AI model production method based on metaverse data labeling, comprising the following steps:
[0046] 1. Collect and organize data sets and store them on the blockchain.
[0047] 1) The specific process of collecting and organizing the data set is as follows:
[0048] Collect data and handle data outliers, missing values, and duplicate values;
[0049] Use data labeling tools to label the processed data according to the data type to complete the initial labeling;
[0050] Using sampling inspection and cross-validation to evaluate whether the data after the initial labeling is accurate, if accurate, exporting the data after the initial labeling, if inaccurate, re-labeling using the data labeling tool;
[0051] Establishing data labels according to the exported data after the initial labeling to obtain a type of data labels, wherein the type of data labels is used for rapid labeling of subsequent data;
[0052] Select the format for data export to match the AI model framework and obtain the data set after initial labeling.
[0053] 2) The specific process of storing the data set on the blockchain is:
[0054] Using blockchain technology, the data set after the initial marking is re-marked, and the content of the secondary marking includes the data source, timestamp and data owner. According to the data after the secondary marking, a second type of data label is obtained, and the second type of data label is used for the rapid marking of subsequent data;
[0055] By using blockchain smart contracts and granting user permissions, only authorized users can access and use data, thus achieving data access control;
[0056] Utilize data transaction and authorization mechanisms to record and store every step of data transaction;
[0057] The secondary marked data set and data transaction information are stored on the blockchain for real-time supervision. If any violation is found, a warning message is automatically sent.
[0058] 3) Through data labeling, the present invention can provide high-quality labeled data sets for machine learning model training, so that the model can achieve efficient learning; and through double labeling and establishing data labels, the data labeling process is made more accurate and faster, ensuring data consistency, ensuring the quality of labeled data, and meeting the requirements of model training.
[0059] By using blockchain to store and label data in AI models, the security and credibility of data can be improved, the authenticity and integrity of data can be ensured, and more reliable data support can be provided for the training and application of AI models.
[0060] 2. Select an AI model framework based on the type of the data set, and use the data set for training to obtain an initial AI model. The specific process is:
[0061] Based on the twice labeled data set, the AI model framework is trained using a loss function to obtain a first sample model and a first loss value, and then the first sample model is trained using the loss function and the first loss value to obtain a second sample model and a second loss value, and finally the second sample model is trained using the loss function, the second loss value and the optimizer to obtain the initial AI model. During the training process, the model parameters are updated in real time through the back propagation algorithm, and the updated model parameters are synchronized to the blockchain.
[0062] The present invention performs training by means of multiple trainings, thereby reducing model loss and model error, ensuring that the model gradually converges during the training process, improving the performance and generalization ability of the model, and enabling the model to better solve practical problems.
[0063] 3. Deploy the initial AI model to the Metaverse platform for testing and optimization to obtain an AI model, which improves the performance and adaptability of the model in the virtual environment and provides better support for solving practical problems. The specific process is:
[0064] Select and determine the test objectives and indicators of the initial AI model in the Metaverse, where the test objectives include the performance, stability, and scalability of the initial AI model, and the indicators are corresponding test thresholds set based on different test objectives of the initial AI model;
[0065] Use virtual reality and augmented reality technologies to build a simulated metaverse environment;
[0066] Based on the secondary labeled data set, deploy the initial AI model in the metaverse environment, and perform testing and evaluation;
[0067] Collect and analyze the output and behavior data of the initial AI model in the metaverse to obtain test results;
[0068] The test result is compared with the corresponding test threshold to obtain a comparison result, and then based on the comparison result, the initial AI model is optimized and adjusted by adjusting model parameters, increasing training data and improving the algorithm to obtain the AI model.
[0069] like Figure 2 As shown, the present invention also adopts an AI model production system based on metaverse data labeling, including:
[0070] The data labeling module is used to collect and organize data sets and store the data sets on the blockchain.
[0071] The model training module is used to select an AI model framework according to the type of the data set, and use the data set for training to obtain an initial AI model.
[0072] The model optimization module is used to deploy the initial AI model to the Metaverse platform for testing and optimization to obtain an AI model.
[0073] Therefore, the present invention provides an AI model production method and system based on Metaverse data labeling. By utilizing data labeling and the Metaverse platform, it is possible to improve the quality, security, and credibility of data, produce efficient, high-performance, and highly adaptable AI models, and provide intelligent solutions for various application scenarios.
[0074] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of the present invention. At the same time, for those skilled in the art, according to the idea of the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.
Claims
1. A method for producing an AI model based on metaverse data labeling, characterized in that: The following steps are involved: Collecting and organizing data sets, and storing said data sets on a blockchain; Select an AI model framework according to the type of the data set, and use the data set for training to obtain an initial AI model; The initial AI model is deployed to the Metaverse platform for testing and optimization to obtain an AI model.
2. The method for producing an AI model based on metaverse data labeling according to claim 1, characterized in that: The specific process of collecting and organizing the data set is as follows: Collect data and handle data outliers, missing values, and duplicate values; Use data labeling tools to label the processed data according to the data type to complete the initial labeling; Using sampling inspection and cross-validation to evaluate whether the data after the initial labeling is accurate, if accurate, exporting the data after the initial labeling, if inaccurate, re-labeling using the data labeling tool; Establishing data labels according to the exported data after the initial labeling to obtain a type of data labels, wherein the type of data labels is used for rapid labeling of subsequent data; Select the format for data export to match the AI model framework and obtain the data set after initial labeling.
3. The method for producing an AI model based on metaverse data labeling according to claim 2, characterized in that: The specific process of storing the dataset on the blockchain is as follows: Using blockchain technology, the data set after the initial marking is re-marked, and the content of the secondary marking includes the data source, timestamp and data owner. According to the data after the secondary marking, a second type of data label is obtained, and the second type of data label is used for the rapid marking of subsequent data; By using blockchain smart contracts and granting user permissions, only authorized users can access and use data, thus achieving data access control; Utilize data transaction and authorization mechanisms to record and store every step of data transaction; The secondary marked data set and data transaction information are stored on the blockchain for real-time supervision. If any violation is found, a warning message is automatically sent.
4. The method for producing an AI model based on metaverse data labeling according to claim 3, characterized in that: The specific process of selecting an AI model framework according to the type of the data set and using the data set for training to obtain the initial AI model is as follows: Based on the second labeled data set, the AI model framework is trained using a loss function to obtain a first sample model and a first loss value, and the first sample model is trained using the loss function and the first loss value to obtain a second sample model and a second loss value, and finally the second sample model is trained using the loss function, the second loss value and an optimizer to obtain the initial AI model; During the training process, the model parameters are updated in real time through the back-propagation algorithm, and the updated model parameters are synchronized to the blockchain.
5. The method for producing an AI model based on metaverse data labeling according to claim 4, characterized in that: The initial AI model is deployed to the Metaverse platform for testing and optimization. The specific process of obtaining the AI model is as follows: Select and determine the test objectives and indicators of the initial AI model in the Metaverse, where the test objectives include the performance, stability, and scalability of the initial AI model, and the indicators are corresponding test thresholds set based on different test objectives of the initial AI model; Use virtual reality and augmented reality technologies to build a simulated metaverse environment; Based on the secondary labeled data set, deploy the initial AI model in the metaverse environment, and perform testing and evaluation; Collect and analyze the output and behavior data of the initial AI model in the metaverse to obtain test results; The test result is compared with the corresponding test threshold to obtain a comparison result, and then based on the comparison result, the initial AI model is optimized and adjusted by adjusting model parameters, increasing training data and improving the algorithm to obtain the AI model.
6. An AI model production system based on metaverse data labeling, characterized in that: include: A data labeling module, used to collect and organize data sets and store the data sets on the blockchain; A model training module, used to select an AI model framework according to the type of the data set, and use the data set for training to obtain an initial AI model; The model optimization module is used to deploy the initial AI model to the Metaverse platform for testing and optimization to obtain an AI model.