Deep learning sample management and training automation platform

Abstract

The invention relates to the technical field of deep learning, in particular to a deep learning sample management and training automation platform which comprises a project initialization module, a data storage synchronization module, a picture sample labeling module, a data preprocessing module, an automation training module, a training result analysis module and a model management and meta-information module. According to the method, the model training process is simplified through user-friendly interface design; the distributed file storage system is used for effectively managing data, so that the safety is ensured; the integrated YOLOv5 algorithm optimizes the training process, and improves the efficiency and accuracy of model evaluation. According to the invention, a cooperative work environment is jointly constructed through efficient interaction among the modules, the development speed and quality of a deep learning project are remarkably improved, the operation complexity is reduced, a one-stop deep learning solution is provided for a user, and the method is suitable for various industries and application scenes.

Description

Technical Field

[0001] The present invention relates to the field of deep learning technology, and in particular to a deep learning sample management and training automation platform. Background Art

[0002] The field of deep learning technology focuses on parsing data by simulating the processing methods of the human brain to perform complex tasks such as image recognition and language processing. In the field of modern science and technology, especially in computer vision and natural language processing, the application of deep learning has become increasingly important. Effective sample management and automated training are the key to improving the efficiency and accuracy of deep learning. Efficiently organizing, storing, processing and training a large number of data samples is crucial to improving model performance, shortening training time and performing effective model iteration.

[0003] Among them, a deep learning sample management and training automation platform aims to provide efficient data management and training process automation for deep learning. The main purpose of this platform is to optimize the management of image samples and models, improve the efficiency of data processing, simplify the configuration of model training, and conduct real-time monitoring of the training process. Through this platform, researchers and developers can more conveniently manage training data, adjust and optimize the training process, and effectively evaluate model performance. The ultimate goal is to improve the accuracy and efficiency of model training, reduce resource consumption, and provide support for further research and application.

[0004] Traditional deep learning training methods face many challenges in terms of sample management, training configuration, and integration of operational processes. The lack of an efficient sample management strategy often leads to data redundancy and inconsistency, which directly affects the accuracy of training results. The cumbersome configuration of training tasks, operational fragmentation, and long training cycles are usually time-consuming and error-prone. In addition, traditional methods are not efficient enough in terms of processing and training efficiency of large-scale data sets, which becomes an obvious bottleneck when faced with growing data volumes. Therefore, the emergence of a deep learning sample management and training automation platform is of great significance for optimizing the overall process of deep learning projects and improving training efficiency and accuracy. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and propose a deep learning sample management and training automation platform.

[0006] In order to achieve the above-mentioned purpose, the present invention adopts the following technical solutions: a deep learning sample management and training automation platform includes a project initialization module, a data storage synchronization module, a picture sample annotation module, a data preprocessing module, an automated training module, a training result analysis module, and a model management and meta-information module; The project initialization module, based on user demand selection and original data set, adopts project framework setting method and data parsing and integration method to perform initialization configuration of the project, build a standardized empty project and generate a picture sample data set for subsequent processing; The data storage synchronization module uses cloud synchronization technology to perform cloud storage and backup based on the image sample data set to generate a synchronized cloud data copy; The image sample annotation module uses an image annotation tool to perform annotation processing on the image samples based on the image sample data, and generates corresponding annotation information that is uploaded to the cloud database; The data preprocessing module processes the cloud image samples and their annotation information based on the sample requirements of the automated training module, and generates the data sets and annotation files required for training; The automated training module uses the YOLOv5 deep learning algorithm to train the preprocessed data set based on the training parameters set by the user, and generates training progress monitoring data and models; The training result analysis module performs effect testing on the imported sample images based on the model generated by the automated training module and adopts a model testing method to generate corresponding detection results; The model management and metadata module manages and backs up the model and metadata and generates management records based on the model generated by training and the project-related metadata using cloud synchronization technology.

[0007] As a further solution of the present invention, the user demand selection includes options such as project name, project type, test file generation, etc. provided on the platform interface; the original data set includes uncleaned, unnamed, and image samples with unqualified resolution or size; the image sample data set includes image files of different resolutions and formats; the image sample data includes images, related metadata and their classification labels; the sample requirements of the automated training module include data set size, image format and its corresponding annotation information; the training parameters set by the user include pre-training weights, number of training rounds, rectangular training, multi-scale training, hyperparameter evolution, optimizer, cross-card synchronized BN, etc.; the model generated by the automated training includes the model's weight file, training log, performance indicators and its configuration file; the project-related metadata includes basic project information, image details, model and sample descriptions.

[0008] As a further solution of the present invention, the project initialization module includes a project creation submodule and a data import submodule; The project creation submodule uses a project framework setting method to establish a new project based on user demand selection and generates a standardized empty project; The data import submodule, based on the standardized empty project and the original data set, uses the data parsing and integration method in the project to import and integrate the data, and generates a picture sample data set for subsequent processing.

[0009] As a further solution of the present invention, the data storage synchronization module includes a cloud storage submodule and a data synchronization submodule; The cloud storage submodule uses distributed file storage technology to store images in the cloud based on the sample data set, generating a secure and efficient file storage system. The data synchronization submodule adopts a real-time synchronization strategy to update data based on the data differences between the local and cloud file storage systems, and generates a synchronization mechanism to maintain data consistency.

[0010] As a further solution of the present invention, the picture sample annotation module includes an annotation tool submodule and an annotation data management submodule; The annotation tool submodule uses a multifunctional image annotation tool to annotate the image based on the unannotated image sample data to generate high-precision annotation information; The annotation data management submodule uses a database management system to store and manage data based on the annotation information generated by annotation, and generates an efficient and accessible annotation information database.

[0011] As a further solution of the present invention, the data preprocessing module includes a data set processing submodule and a label information processing submodule; The data set processing submodule generates data sets suitable for training requirements at different stages by using data set diversity technology based on the sample requirements of the automated training module; The annotation information processing submodule uses the annotation information generation technology to process the annotation information of the image samples and generates the annotation file required by the training module.

[0012] As a further solution of the present invention, the automated training module includes a training configuration submodule and a training execution submodule; The training configuration submodule sets the training environment based on the training parameters specified by the user and adopts the parameter configuration process to generate the training configuration for the specific requirements; The training execution submodule performs model training based on the configured parameters and data sets using an automated training process to generate a training model and related training data.

[0013] As a further solution of the present invention, the training result analysis module includes a model detection submodule and a result visualization submodule; The model detection submodule analyzes the accuracy and performance of the model based on the model generated by the automatic training module using a model testing method to generate detailed test results; The result visualization submodule uses visualization tools to present the performance indicators and detection effects of the model based on the test results, and generates intuitive performance charts and analysis reports.

[0014] As a further solution of the present invention, the model management and meta-information module includes a model storage sub-module and a metadata management sub-module; The model storage submodule uses a cloud storage strategy to save and back up the model based on the trained model, generating a safe and reliable model storage solution; The metadata management submodule manages and synchronizes metadata based on project-related metadata using database technology to generate complete and updated metadata records.

[0015] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects: 1. The deep learning sample management and training automation platform provided by the present invention significantly reduces the operational complexity and time consumption while effectively improving the sample management efficiency and the automation of the training process through its innovative integrated workflow. The introduction of the automated training module simplifies the data preprocessing, model training and detection links, allowing researchers and developers to focus more on model optimization and innovation. In addition, the real-time monitoring and data display functions of the platform provide detailed feedback on the training process, allowing users to adjust the training strategy in a timely manner and optimize model performance, thereby significantly shortening the model development cycle while improving the model accuracy.

[0016] 2. The efficient data processing and management mechanism in the platform provides solutions to the big data processing challenges faced by the modern deep learning field. The efficient data upload and download mechanism ensures fast and secure data processing, while the data management technology optimizes the storage and retrieval of large-scale data sets. It improves the ability to process large data sets and ensures the integrity and consistency of the data, thereby improving the stability and reliability of the entire training process.

[0017] 3. This platform has lowered the technical threshold through its user-friendly interface design and simple operation process, making it easy for even non-professional users to get started. In addition, the model management and version control functions of the platform facilitate the tracking, storage and recovery of historical models, greatly enhancing the flexibility and efficiency of research and development. These features make this platform not only suitable for advanced users and professional researchers, but also provide strong support for beginners and non-technical users. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 is a flow chart of the platform of the present invention; Figure 2 It is the overall framework diagram of the platform of the present invention; Figure 3 It is a flow chart of the project initialization module of the present invention; Figure 4 It is a flow chart of the data storage synchronization module of the present invention; Figure 5 This is a flow chart of the image sample annotation module of the present invention; Figure 6 It is a flow chart of the data preprocessing module of the present invention; Figure 7 This is a flow chart of the automated training module of the present invention; Figure 8 It is a flow chart of the training result analysis module of the present invention; Fig. 9 This is a flow chart of the model management and meta-information module of the present invention. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0020] Example 1: Please refer to Figure 1 to Figure 2 , a deep learning sample management and training automation platform includes a project initialization module, a data storage synchronization module, an image sample annotation module, a data preprocessing module, an automated training module, a training result analysis module, and a model management and meta information module; The project initialization module uses the project framework setting method to perform initial configuration of the project based on user needs and build a standardized empty project. The data storage synchronization module uses cloud synchronization technology to perform cloud storage and backup based on the image sample data set, and generates a synchronized cloud data copy; The image sample annotation module uses image annotation tools to annotate the image samples based on the image sample data, and generates corresponding annotation information that is uploaded to the cloud database; The data preprocessing module processes the cloud image samples and their annotation information based on the sample requirements of the automated training module, and generates the data sets and annotation files required for training. The automated training module uses the YOLOv5 deep learning algorithm to train the preprocessed data set based on the training parameters set by the user, and generates training progress monitoring data and models; The training result analysis module performs effect testing on the imported image sample images based on the model generated by the automated training module and adopts the model testing method to generate corresponding detection results; The model management and metadata module uses cloud synchronization technology to manage and back up models and metadata and generate management records based on the models and project-related metadata generated by training.

[0021] User demand selections include options such as project name, project type, sample dataset format, etc. provided on the platform interface; image sample datasets include image files of different resolutions and formats; image sample data include images, related metadata and their classification labels; sample requirements for the automated training module include dataset size, image format and its corresponding annotation information; user-set training parameters include pre-training weights, number of training rounds, rectangular training, multi-scale training, hyperparameter evolution, optimizer, cross-card synchronized BN, etc.; the model generated by automated training includes the model's weight file, training log, performance indicators and its configuration file; project-related metadata includes basic project information, image details, model performance overview and sample feature description.

[0022] In this platform, the project initialization module provides customized configuration according to user needs, ensures the standardized construction of the project, and performs preliminary processing of the original data, laying a solid foundation for subsequent steps. The data storage synchronization module ensures data security and instant access through distributed cloud storage and backup mechanisms, while the generation of cloud copies further speeds up data access and backup efficiency. The image sample annotation module uses efficient annotation tools to not only speed up the sample preparation process and ensure data accuracy, but also greatly improves the reusability of annotation information by uploading to the cloud. The data preprocessing module significantly improves the data processing efficiency before training by optimizing the data set preparation process and simplifies the complex steps of set and sample file conversion. The graphical interface configuration of the automated training module simplifies the training operation process. The real-time monitoring function not only visualizes the training progress, but also effectively improves the efficiency of model training and guarantees the model quality to a certain extent. The training result analysis module displays the training data and model detection results through an intuitive interface, which is convenient for users to accurately evaluate the model performance. The model management and metadata module provides continuous support throughout the process to ensure the smooth progress of the project and the maintenance of data integrity.

[0023] See also Figure 3 ,The project initialization module includes a project creation submodule and a data import submodule; The project creation submodule uses the project framework setting method to establish a new project based on user needs and generate a standardized empty project; The data import submodule, based on the standardized empty project and original data set, uses data parsing and integration methods in the project to import and integrate data and generate image sample data sets for subsequent processing.

[0024] The project framework setting method specifically includes template selection, parameter configuration and structure definition. The data parsing and integration method specifically includes file format recognition, data extraction and merging processing.

[0025] In the project creation submodule, project settings are customized through user selection, and users select the type, framework and parameters of the project according to their needs. The module uses template selection technology to allow users to select from preset project templates to speed up the project startup process. The parameter configuration function enables users to customize parameter settings according to project requirements, while the structure definition provides users with a way to define and adjust the project structure to ensure that the project proceeds as expected, thereby generating a standard empty project.

[0026] In the data import submodule, file format recognition technology is used to process the original data set to ensure the compatibility and accuracy of the data format. Through the data extraction step, the necessary information and features are extracted from the data set. Next, the merging process integrates the extracted data to form a unified image sample data set that can be used in subsequent steps. In addition, the module automatically renames the imported image samples according to the serial number, thereby improving the efficiency of subsequent data processing and making the data organization more orderly and easy to manage. The comprehensive use of these steps not only improves the efficiency of data processing, but also ensures the quality and consistency of the data, generating a stable and reliable image sample data set for the platform for subsequent operations.

[0027] See also Figure 4 ,The data storage and synchronization module includes a cloud storage submodule and a data synchronization submodule; The cloud storage submodule uses distributed file storage technology to save the image sample data set in the cloud, generating a secure and efficient file storage system. The data synchronization submodule uses a real-time synchronization strategy to update data based on the data differences between the local and cloud file storage systems, and generates a synchronization mechanism to maintain data consistency.

[0028] Distributed file storage technology uses multiple server nodes to store data to improve access speed and data reliability. Real-time synchronization strategy monitors file changes and version control to maintain real-time data consistency.

[0029] In the cloud storage submodule, based on the image sample dataset imported into the project, the data is stored on different server nodes through distributed file storage technology to enhance storage capacity and data security. This method enhances data backup and recovery capabilities and reduces the risk of data loss or damage by replicating and distributing data storage among multiple nodes.

[0030] In the data synchronization submodule, a real-time synchronization strategy is adopted to perform incremental updates based on the differences in data between the local and cloud file storage systems to ensure the latest status of the data.

[0031] See also Figure 5 ,The image sample annotation module includes an annotation tool submodule and an annotation data management submodule; The annotation tool submodule uses a multifunctional image annotation tool to annotate images based on unannotated image sample data and generate high-precision annotation information; The annotation data management submodule uses a database management system to store and manage data based on the annotation information generated by annotation, and generates an efficient and accessible annotation information database.

[0032] The multifunctional image annotation tool is a customizable image annotation tool, including rectangle, polygon, key point and free-form annotation functions, supporting batch operations and automatic annotation suggestions. The annotation information generated by this tool exists in the form of key-value pairs, which is easy to store in the database. The database management system is MySQL, which has the characteristics of high performance, high reliability and easy scalability. It can efficiently process and manage a large amount of annotation data and support fast data retrieval and secure backup.

[0033] In the annotation tool submodule, based on the unlabeled image sample data, the image samples are annotated in detail and accurately by using multifunctional image annotation tools. Specifically, these tools include customized image annotation functions such as box selection, classification label assignment, and annotation addition, making the annotation process more flexible and accurate, providing high-quality training data for deep learning models.

[0034] In the annotation data management submodule, the annotation information generated by the annotation tool is efficiently stored and managed through the database management system. Specifically, this submodule uses MySQL database technology to organize and maintain the annotation data, ensuring the accessibility, security and sustainable management of the annotation information, and providing reliable data support for subsequent data training and analysis.

[0035] See also Figure 6 ,The data preprocessing module includes a data set processing submodule and a label information processing submodule; The data set processing submodule uses data set diversity technology to generate data sets suitable for training requirements at different stages based on the sample requirements of the automated training module; The annotation information processing submodule uses annotation information generation technology to process the annotation information of the image samples and generate the annotation files required by the training module.

[0036] Dataset diversity technology is specifically automatic data classification and subset partitioning technology, which divides large data sets into training, validation and test sets to ensure the diversity and balance of data sets. Annotation information generation technology includes format conversion and verification to ensure that the annotation information meets the format requirements of the training module and the data quality meets the standards.

[0037] In the data set processing submodule, based on training requirements, image samples are classified and subdivided to effectively manage and use large amounts of data, while providing suitable data sets for model training at different stages. The automatic classification and subdivision technology involved in this process not only improves the efficiency of data preprocessing, but also provides a solid foundation for subsequent model training and verification.

[0038] In the annotation information processing submodule, based on the training requirements, the annotation information is processed and converted to generate annotation files that meet the requirements of the training module. This process ensures the accuracy and completeness of the annotation information, which is crucial to maintaining the high quality of the data set and ensuring the efficient operation of the training module.

[0039] See also Figure 7 ,The automated training module includes a training configuration submodule and a training execution submodule; The training configuration submodule uses the parameter configuration process to set up the training environment and generates training configurations for specific needs based on the training parameters specified by the user; The training execution submodule uses an automated training process to perform model training based on the configured parameters and data sets, and generates training models and related training data.

[0040] The parameter configuration process is specifically a combination of simplified user interface operations and advanced configuration options, and the automated training process is specifically an integration of batch processing, model iterative optimization, and real-time monitoring.

[0041] In the training configuration submodule, the training environment is set up based on the training parameters specified by the user through the parameter configuration process. In this process, a combination of simplified user interface and advanced configuration options is used to ensure that the training environment not only meets the specific needs of the user, but also maintains the flexibility and customization of the configuration. The parameter optimization step is crucial to adapt to different training needs and data characteristics.

[0042] In the training execution submodule, the model is trained based on the configured parameters and data sets through an automated training process. This process includes batch processing technology to efficiently process large amounts of data, model iterative optimization to ensure continuous improvement of model performance, and real-time monitoring functions to provide users with instant feedback on the training process. These steps work together to ensure the efficiency of training and the effectiveness of the generated model.

[0043] See also Figure 8 ,The training result analysis module includes a model detection submodule and a result visualization submodule; The model detection submodule uses the model testing method to analyze the model accuracy and performance based on the model generated by the automated training module and generates detailed test results; The result visualization submodule uses visualization tools to present the performance indicators and detection effects of the model based on the test results, and generates intuitive performance charts and analysis reports.

[0044] The model detection method specifically imports test images, performs code testing on the trained model, and returns the test results. Visualization tools include chart generation and interactive interfaces to clearly display the results of model testing and other visual data.

[0045] In the model detection submodule, the accuracy and performance of the model are analyzed based on the model generated by the automated training module through the model testing method. In this process, the model code is actually tested by importing test images to evaluate the effectiveness and accuracy of the model. Through these tests, detailed model performance test results are generated, which provide an important basis for subsequent analysis and optimization.

[0046] In the result visualization submodule, the performance indicators and detection effects of the model are intuitively displayed based on the test results through visualization tools. Through these visualizations, users can clearly and intuitively understand the performance of the model, more easily interpret and analyze the model results, and effectively evaluate the overall effect of the model.

[0047] See also Fig. 9 ,The model management and meta information module includes a model storage submodule and a metadata management submodule; The model storage submodule uses cloud storage strategies to save and back up the model based on the training model, generating a secure and reliable model storage solution. The metadata management submodule uses database technology to manage and synchronize metadata based on project-related meta-information and generates complete and updated metadata records.

[0048] The cloud storage strategy specifically adopts encryption and redundant backup technology to ensure the security and recoverability of model data. The database technology specifically applies advanced query optimization and indexing technology to support efficient data processing and fast retrieval.

[0049] In the model storage submodule, the cloud storage strategy is used to securely store and back up the model data based on the trained model. In this process, encryption technology is used to ensure the security of the model data, while redundant backup provides data recovery capabilities to prevent data loss or damage. Through these measures, a safe and reliable model storage solution is provided to ensure the long-term storage and management of model data.

[0050] In the metadata management submodule, metadata is effectively managed and optimized based on project-related meta-information through database technology. The application of advanced query optimization and indexing technology not only improves the efficiency of data processing, but also enables fast data retrieval and update. This process not only ensures the integrity and timeliness of metadata, but also improves the overall performance of data processing, which is crucial for maintaining the data quality of the project and providing accurate information analysis.

[0051] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A deep learning sample management and training automation platform, characterized by: The platform includes a project initialization module, a data storage synchronization module, an image sample annotation module, a data preprocessing module, an automated training module, a training result analysis module, and a model management and meta-information module; The project initialization module, based on user demand selection and original data set, adopts project framework setting method and data parsing and integration method to perform initialization configuration of the project, build a standardized empty project and generate a picture sample data set for subsequent processing; The data storage synchronization module uses cloud synchronization technology to perform cloud storage and backup based on the image sample data set to generate a synchronized cloud data copy; The image sample annotation module uses an image annotation tool to perform annotation processing on the image samples based on the image sample data, and generates corresponding annotation information that is uploaded to the cloud database; The data preprocessing module processes the cloud image samples and their annotation information based on the sample requirements of the automated training module, and generates the data sets and annotation files required for training; The automated training module uses the YOLOv5 deep learning algorithm to train the preprocessed data set based on the training parameters set by the user, and generates training progress monitoring data and models; The training result analysis module performs effect testing on the imported sample images based on the model generated by the automated training module and adopts a model testing method to generate corresponding detection results; The model management and metadata module manages and backs up the model and metadata and generates management records based on the model generated by training and the project-related metadata using cloud synchronization technology.

2. A deep learning sample management and training automation platform according to claim 1, characterized in that: The user demand selection includes options such as project name, project type, test file generation, etc. provided on the platform interface; the original data set includes image samples that are not cleaned, not renamed, and have unqualified resolution or size; The picture sample data set includes image files of different resolutions and formats; the picture sample data includes pictures, related meta information and their classification labels; the sample requirements of the automated training module include data set size, picture format and its corresponding annotation information; the training parameters set by the user include pre-training weights, number of training rounds, rectangular training, multi-scale training, hyperparameter evolution, optimizer, cross-card synchronous BN, etc.; the model generated by the automated training includes the model's weight file, training log, performance indicators and its configuration file; the project-related metadata includes basic project information, picture details, model and sample description.

3. A deep learning sample management and training automation platform according to claim 1, characterized in that: The project initialization module includes a project creation submodule and a data import submodule; The project creation submodule uses a project framework setting method to establish a new project based on user demand selection and generates a standardized empty project; The data import submodule, based on the standardized empty project and the original data set, uses the data parsing and integration method in the project to import and integrate the data, and generates a picture sample data set for subsequent processing.

4. A deep learning sample management and training automation platform according to claim 1, characterized in that: The data storage synchronization module includes a cloud storage submodule and a data synchronization submodule; The cloud storage submodule uses distributed file storage technology to store images in the cloud based on the sample data set, generating a secure and efficient file storage system. The data synchronization submodule adopts a real-time synchronization strategy to update data based on the data differences between the local and cloud file storage systems, and generates a synchronization mechanism to maintain data consistency.

5. The deep learning sample management and training automation platform according to claim 1, characterized in that: The image sample annotation module includes an annotation tool submodule and an annotation data management submodule; The annotation tool submodule uses a multifunctional image annotation tool to annotate the image based on the unannotated image sample data to generate high-precision annotation information; The annotation data management submodule uses a database management system to store and manage data based on the annotation information generated by annotation, and generates an efficient and accessible annotation information database.

6. A deep learning sample management and training automation platform according to claim 1, characterized in that: The data preprocessing module includes a data set processing submodule and a label information processing submodule; The data set processing submodule generates data sets suitable for training requirements at different stages by using data set diversity technology based on the sample requirements of the automated training module; The annotation information processing submodule uses the annotation information generation technology to process the annotation information of the image samples and generates the annotation file required by the training module.

7. The deep learning sample management and training automation platform according to claim 1, characterized in that: The automated training module includes a training configuration submodule and a training execution submodule; The training configuration submodule sets the training environment based on the training parameters specified by the user and adopts the parameter configuration process to generate the training configuration for the specific requirements; The training execution submodule performs model training based on the configured parameters and data sets using an automated training process to generate a training model and related training data.

8. The deep learning sample management and training automation platform according to claim 1, characterized in that: The training result analysis module includes a model detection submodule and a result visualization submodule; The model detection submodule analyzes the accuracy and performance of the model based on the model generated by the automatic training module using a model testing method to generate detailed test results; The result visualization submodule uses visualization tools to present the performance indicators and detection effects of the model based on the test results, and generates intuitive performance charts and analysis reports.

9. The deep learning sample management and training automation platform according to claim 1, characterized in that: The model management and meta-information module includes a model storage submodule and a metadata management submodule; The model storage submodule uses a cloud storage strategy to save and back up the model based on the trained model, generating a safe and reliable model storage solution; The metadata management submodule manages and synchronizes metadata based on project-related metadata using database technology to generate complete and updated metadata records.

Images