Quality inspection model management method, system, electronic device and storage medium
Through the method of automatically updating the industrial quality inspection model, using database screening and training data, the problem of low quality inspection model update complexity and efficiency is solved, and the continuous adaptation and efficient update of model accuracy is achieved.
Patent Information
- Application Number
- CN202311190419.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-09-14
- Publication Date
- 2024-11-12
- Estimated Expiration
- 2043-09-14
AI Technical Summary
In actual use, the industrial quality inspection model does not match the requirements due to the difference in quality inspection output results and the model does not meet the accuracy requirements. The traditional update method is complex and inefficient, which affects the factory's work efficiency.
By obtaining the initial quality inspection model, monitoring whether its parameters are within the preset range. If not, filter sample training and test data from the database, train and test the model until the standard value is met, and automatically update the model to adapt to changes in demand.
It reduces the difficulty and time of updating the quality inspection model, improves the update efficiency, ensures that the model always meets the quality inspection accuracy requirements, and simplifies the update process of factory operators.
Smart Images

Figure HDA0004450314540000011 
Figure HDA0004450314540000012 
Figure HDA0004450314540000021
Abstract
Description
Technical Field
[0001] The present application relates to the field of model management technology, and in particular to a quality inspection model management method, system, electronic device and storage medium. Background Art
[0002] At present, quality inspection models are often used to improve the processing efficiency of quality inspection work. For example, in industrial quality inspection, industrial quality inspection models are often used to analyze and process relevant data of a large number of industrial products. In addition, the use of industrial quality inspection models improves the quality inspection efficiency of industrial products.
[0003] However, in the actual application of industrial quality inspection models, as time goes by, the quality inspection output results will not match the ever-changing quality inspection requirements, which will eventually lead to the industrial quality inspection model not meeting the accuracy requirements of industrial quality inspection. In related technologies, the adjustment and update of industrial quality inspection models depends on the development iteration of professional engineers. However, the actual users of industrial quality inspection models are usually factory personnel. The traditional industrial quality inspection model update and iteration methods have many steps, strong professionalism, and long update time. The update complexity of the industrial quality inspection model is relatively high, resulting in a low update efficiency of the industrial quality inspection model, which in turn affects the work efficiency of industrial quality inspection related work. Summary of the invention
[0004] The main purpose of the embodiments of the present application is to propose a quality inspection model management method, system, electronic device and storage medium, which can reduce the difficulty of updating the quality inspection model and improve the updating efficiency of the quality inspection model.
[0005] To achieve the above-mentioned purpose, the first aspect of an embodiment of the present application proposes a quality inspection model management method, the method comprising: obtaining an initial quality inspection model, performing a quality inspection operation on a target object according to the initial quality inspection model, and obtaining initial quality inspection parameters; when it is monitored that the initial quality inspection parameters are not within a preset quality inspection parameter range, screening and obtaining sample training data from a database, and training the initial quality inspection model according to the sample training data; screening and obtaining sample test data from a database, testing the trained initial quality inspection model according to the sample test data, and obtaining model test parameters; when the model test parameters match the preset model compliance values, determining that the tested initial quality inspection model is a target quality inspection model, and the target quality inspection model is used to re-perform quality inspection operations on the target object.
[0006] In some embodiments, after the trained initial quality inspection model is tested according to the sample test data and the model test parameters are obtained, the method further includes: when the model test parameters do not match the preset model compliance values, re-screening and obtaining sample training data and sample test data from the database; re-updating the initial quality inspection model according to the re-determined sample training data and sample test data, and obtaining updated model test parameters; when the model test parameters match the preset model compliance values, determining the updated initial quality inspection model as the target quality inspection model
[0007] In some embodiments, when it is monitored that the initial quality inspection parameters are not within the preset quality inspection parameter range, sample training data is screened and obtained from the database, including: parsing the initial quality inspection parameters to obtain multiple parameter categories of the initial quality inspection parameters, each parameter category including a corresponding confidence level, and the confidence level is used to characterize the degree of credibility that the initial quality inspection model belongs to the parameter category; when the confidence level is not within the preset quality inspection parameter range, determining the screening category of the sample training data screened from the database according to the confidence level, and selecting corresponding sample training data from the database according to the screening category.
[0008] In some embodiments, the sample training data includes difficult example data and regular data; the training of the initial quality inspection model based on the sample training data includes: when the sample training data is determined from a database, activating a training trigger of the initial quality inspection model; according to the training trigger in an activated state, inputting the difficult example data into the initial quality inspection model to obtain a difficult example training result, and inputting the regular data into the initial quality inspection model to obtain a regular training result; and completing the training of the initial quality inspection model based on the difficult example training result and the regular training result.
[0009] In some embodiments, the step of screening and obtaining sample test data from a database, testing the trained initial quality inspection model according to the sample test data, and obtaining model test parameters includes: activating a test trigger of the initial quality inspection model when the training of the initial quality inspection model is completed; screening and obtaining sample test data from a database according to the test trigger in an activated state; and testing the trained initial quality inspection model according to the sample test data, and obtaining model test parameters.
[0010] In some embodiments, after determining that the trained initial quality inspection model is the target quality inspection model, it also includes: displaying a quality inspection model management interface, displaying the configuration components of the initial quality inspection model on the quality inspection model management interface, wherein the configuration components include a data configuration component, a model training component, a model testing component and a model publishing component; in response to a click operation on the data configuration component, displaying the configured sample data on the quality inspection model management interface, wherein the sample data includes sample training data and sample testing data; in response to a click operation on the model training component, displaying the model training parameter values on the quality inspection model management interface, wherein the model training parameter values are obtained after the initial quality inspection model is trained according to the sample training data; in response to a click operation on the model testing component, displaying the model testing parameter values on the quality inspection model management interface, wherein the model testing parameter values are obtained after the initial quality inspection model is tested according to the sample test data, and determining that the initial quality inspection model after the test passes is the target quality inspection model; in response to a click operation on the model publishing component, displaying the published target quality inspection model on the quality inspection model management interface.
[0011] In some embodiments, the displaying of the configured sample data on the quality inspection model management interface includes: displaying the quality inspection model configuration interface, displaying the data generation component and the data annotation component of the sample data on the quality inspection model configuration interface; in response to a click operation on the data generation component, displaying the sample training data and the sample test data configured according to the initial quality inspection parameters in the data configuration area of the quality inspection model configuration interface, wherein the sample training data includes difficult example data and regular data; in response to a click operation on the data annotation component, displaying the annotation results of the sample training data in the data annotation area of the quality inspection model configuration interface, wherein the annotation results include difficult example data annotation results and regular data annotation results; in response to a data configuration completion operation, displaying the configured sample data on the quality inspection model management interface according to the sample test data and the sample training data including the annotation results.
[0012] To achieve the above-mentioned purpose, the second aspect of an embodiment of the present application proposes a quality inspection model management system, which includes: an acquisition module, which is used to acquire an initial quality inspection model, perform quality inspection operations on a target object according to the initial quality inspection model, and obtain initial quality inspection parameters; a model training module, which is used to screen and obtain sample training data from a database when it is monitored that the initial quality inspection parameters are not within a preset quality inspection parameter range, and train the initial quality inspection model according to the sample training data; a model testing module, which is used to screen and obtain sample test data from a database, test the trained initial quality inspection model according to the sample test data, and obtain model test parameters; a target quality inspection model module, which is used to determine that the tested initial quality inspection model is a target quality inspection model when the model test parameters match the preset model compliance values, and the target quality inspection model is used to re-perform quality inspection operations on the target object.
[0013] To achieve the above-mentioned purpose, the third aspect of the embodiments of the present application proposes an electronic device, which includes a memory and a processor, the memory stores a computer program, and the processor implements the method described in the above-mentioned first aspect embodiment when executing the computer program.
[0014] To achieve the above-mentioned purpose, the fourth aspect of an embodiment of the present application proposes a storage medium, which is a computer-readable storage medium, and the storage medium stores a computer program. When the computer program is executed by a processor, the method described in the embodiment of the first aspect is implemented.
[0015] The quality inspection model management method, system, electronic device and storage medium proposed in the present application first obtain an initial quality inspection model, and perform quality inspection operations on the target object according to the initial quality inspection model to obtain initial quality inspection parameters; then, when it is monitored that the initial quality inspection parameters are not within the preset quality inspection parameter range, sample training data is screened and obtained from the database, and the initial quality inspection model is trained according to the sample training data. In this way, the initial quality inspection model can be trained according to the sample training data to adapt to the ever-changing quality inspection needs; thereafter, sample test data is screened and obtained from the database, and the trained initial quality inspection model is tested according to the sample test data to obtain model test parameters; when the model test parameters match the preset model compliance values, the tested initial quality inspection model is determined to be the target quality inspection model, and the target quality inspection model is used to re-perform quality inspection operations on the target object. It can be understood that the initial quality inspection model after the test indicates that it can meet the current quality inspection needs after training and testing. Therefore, the initial quality inspection model after the test can be determined as the target quality inspection model, and the target quality inspection model can be used to re-quality inspect the target object. In the actual quality inspection model update process, it is only necessary to filter and obtain sample training data and sample test data from the preset database to automatically perform iterative updates of the quality inspection model, so as to reduce the difficulty of updating the quality inspection model, improve the updating efficiency of the quality inspection model, and thus improve the efficiency of quality inspection work. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] Figure 1 It is a schematic diagram of an application scenario of the quality inspection model management system provided in an embodiment of the present application;
[0017] Figure 2 This is an optional flow chart of the quality inspection model management method provided in the embodiment of the present application;
[0018] Figure 3 yes Figure 2 An implementation flow chart after step S104;
[0019] Figure 4 yes Figure 2 An implementation flow chart of step S102;
[0020] Figure 5 yes Figure 2 Another implementation flow chart of step S102;
[0021] Figure 6 yes Figure 2 An implementation flow chart of step S103;
[0022] Figure 7 This is an optional workflow configuration diagram of the quality inspection model management method provided in the embodiment of the present application;
[0023] Figure 8 This is another optional workflow configuration diagram of the quality inspection model management method provided in the embodiment of the present application;
[0024] Fig. 9 It is an optional workflow execution diagram of the quality inspection model management provided in the embodiment of the present application;
[0025] Fig.10 This is an optional flow chart of the quality inspection model management method provided in the embodiment of the present application;
[0026] Fig.11 is another optional flow chart of the quality inspection model management method provided in the embodiment of the present application;
[0027] Fig.12 yes Fig.11 An implementation flow chart of step S602;
[0028] Fig.13 It is a schematic diagram of the functional modules of the quality inspection model management provided in the embodiment of the present application;
[0029] Fig.14 It is a schematic diagram of the hardware structure of the electronic device provided in the embodiment of the present application. DETAILED DESCRIPTION
[0030] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application 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 application and are not used to limit the present application.
[0031] It should be noted that, although the functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first", "second", etc. in the specification, claims and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.
[0032] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.
[0033] First, some nouns involved in this application are analyzed:
[0034] Artificial intelligence (AI) is a new technical science that studies and develops theories, methods, technologies and application systems for simulating, extending and expanding human intelligence. AI is a branch of computer science that attempts to understand the essence of intelligence and produce a new type of intelligent machine that can respond in a similar way to human intelligence. Research in this field includes robotics, speech recognition, image recognition, natural language processing and expert systems. AI can simulate the information process of human consciousness and thinking. AI is also a theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.
[0035] Machine Learning Operations (MLOps) refers to applying software engineering best practices to the development, deployment, and maintenance of machine learning models to improve the reliability, scalability, and repeatability of machine learning systems. It is the continuous delivery and automated pipeline in machine learning.
[0036] Data extraction, transformation, and loading (ETL) is a data integration process, and its name stands for "Extract", "Transform", and "Load". ETL is usually used to extract data from different data sources (such as databases, files, or APIs), then transform it, and finally load it into the target data storage location (such as a data warehouse or data lake).
[0037] Message Queue (MQ) is used to transmit messages between different services.
[0038] A workflow is a specific pipeline for MLOps operations. A workflow consists of many steps, which have dependencies. By following the steps of the workflow, one model update iteration is completed.
[0039] A workflow server is a server software used to manage and execute workflows. It provides a way to centrally manage and automate various business processes, which can help organizations improve work efficiency, optimize business processes, and achieve task collaboration and monitoring.
[0040] Based on this, the embodiments of the present application provide a quality inspection model management method, system, electronic device and storage medium, which can reduce the difficulty of updating the quality inspection model and improve the updating efficiency of the quality inspection model.
[0041] The quality inspection model management method, system, electronic device and storage medium provided in the embodiments of the present application are specifically illustrated through the following embodiments. First, the quality inspection model management system in the embodiments of the present application is described.
[0042] At present, quality inspection models are often used to improve the processing efficiency of quality inspection work. For example, in industrial quality inspection, industrial quality inspection models are often used to analyze and process relevant data of a large number of industrial products. In addition, the use of industrial quality inspection models improves the quality inspection efficiency of industrial products.
[0043] However, in the actual application of industrial quality inspection models, as time goes by, the quality inspection output results will not match the ever-changing quality inspection requirements, which will eventually lead to the industrial quality inspection model not meeting the accuracy requirements of industrial quality inspection. In related technologies, the adjustment and update of industrial quality inspection models depends on the development iteration of professional engineers. However, the actual users of industrial quality inspection models are usually factory personnel. The traditional industrial quality inspection model update and iteration methods have many steps, strong professionalism, and long update time. The update complexity of the industrial quality inspection model is relatively high, resulting in a low update efficiency of the industrial quality inspection model, which in turn affects the work efficiency of industrial quality inspection related work.
[0044] Based on this, the quality inspection model management method, system, electronic device and storage medium proposed in the present application first obtain an initial quality inspection model, and perform quality inspection operations on the target object according to the initial quality inspection model to obtain initial quality inspection parameters; then, when it is monitored that the initial quality inspection parameters are not within the preset quality inspection parameter range, sample training data is screened and obtained from the database, and the initial quality inspection model is trained according to the sample training data. In this way, the initial quality inspection model can be trained according to the sample training data to adapt to the ever-changing quality inspection needs; thereafter, sample test data is screened and obtained from the database, and the trained initial quality inspection model is tested according to the sample test data to obtain model test parameters; when the model test parameters match the preset model compliance values, the tested initial quality inspection model is determined to be the target quality inspection model, and the target quality inspection model is used to re-perform quality inspection operations on the target object. It can be understood that the initial quality inspection model after the test indicates that it can meet the current quality inspection needs after training and testing. Therefore, the initial quality inspection model after the test can be determined as the target quality inspection model, and the target quality inspection model can be used to re-quality inspect the target object. In the actual quality inspection model update process, it is only necessary to filter and obtain sample training data and sample test data from the preset database to automatically perform iterative updates of the quality inspection model, so as to reduce the difficulty of updating the quality inspection model, improve the updating efficiency of the quality inspection model, and thus improve the efficiency of quality inspection work.
[0045] The quality inspection model management method, system, electronic device and storage medium provided in the embodiments of the present application are specifically described through the following embodiments.
[0046] For example, Figure 1 As shown, Figure 1 1 is a schematic diagram of an application scenario of the quality inspection model management system provided by the embodiment of the present application. In the quality inspection model management system (hereinafter referred to as the "system"), a user terminal 11 and a management terminal 12 may be included, wherein the quality inspection model management method and the quality inspection model management system of the embodiment of the present application may both be run on the user terminal 11, or the quality inspection model management system provided by the embodiment of the present application may be run on the user terminal 11, and the quality inspection model management method provided by the embodiment of the present application may be run on the management terminal 12, the user terminal 11 and the management terminal 12 may be connected by remote connection, and the quality inspection model management method in the management terminal 12 may be applied to the quality inspection model management system of the user terminal 11 by remote control. The specific setting is made according to the actual needs of the operator, and the embodiment of the present application does not make specific restrictions here.
[0047] The quality inspection model management method in the embodiments of the present application can be illustrated by the following embodiments.
[0048] In some embodiments, the embodiments of the present application can also acquire and process relevant data based on artificial intelligence technology. For example, the acquired data can be intelligently divided by artificial intelligence to obtain sample difficult case data and sample routine data, and the sample difficult case data and sample routine data are stored in the database as sample training data, reducing the manual screening and division of large amounts of data, and further improving the updating efficiency of the quality inspection model. Among them, artificial intelligence (AI) is a theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.
[0049] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.
[0050] The quality inspection model management method provided in the embodiment of the present application relates to the field of artificial intelligence technology. The quality inspection model management method provided in the embodiment of the present application can be applied to the terminal, can also be applied to the server side, and can also be software running in the terminal or the server side. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc.; the server side can be configured as an independent physical server, or it can be configured as a server cluster or distributed system composed of multiple physical servers, and can also be configured to provide cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms and other basic cloud computing services. Cloud servers; software can be applications that implement quality inspection model management methods, etc., but are not limited to the above forms.
[0051] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.
[0052] It should be noted that in each specific implementation of the present application, when it comes to the need to perform relevant processing based on data related to user identity or characteristics such as user information, user behavior data, user historical data, and user location information, the user's permission or consent will be obtained first. Moreover, the collection, use, and processing of these data will comply with relevant laws, regulations, and standards. In addition, when the embodiment of the present application needs to obtain the user's sensitive personal information, the user's separate permission or separate consent will be obtained through a pop-up window or by jumping to a confirmation page. After clearly obtaining the user's separate permission or separate consent, the necessary user-related data for enabling the normal operation of the embodiment of the present application will be obtained.
[0053] Based on this, the quality inspection model management method in the embodiment of the present application can be illustrated by the following embodiment.
[0054] like Figure 2 As shown, Figure 2is an optional flowchart of the quality inspection model management method provided in the embodiment of the present application. Figure 2 The method may include but is not limited to steps S101 to S104.
[0055] Step S101, obtaining an initial quality inspection model, performing a quality inspection operation on the target object according to the initial quality inspection model, and obtaining initial quality inspection parameters;
[0056] In some embodiments, the initial quality inspection model may be used in the industrial field, and the target object may be a product or component in industrial manufacturing, for example, an automobile component in automobile industry production.
[0057] In some embodiments, the initial quality inspection model can be generated according to the initial quality inspection requirements. For example, a batch of automobile engine parts needs to be quality inspected, that is, the initial quality inspection model is generated according to the first quality inspection requirement. However, as time goes by, a second quality inspection requirement arises. At this time, if the initial quality inspection model generated by the first quality inspection requirement is used to quality inspect the automobile engine parts, the expected quality inspection results cannot be achieved. Therefore, the initial quality inspection model needs to be updated.
[0058] In some embodiments, the quality inspection of the target object may include quality inspection of appearance defects, such as cracks, breakages, and deformations, and may also include quality inspection of size requirements, such as diameter, length, and flatness.
[0059] In some embodiments, the initial quality inspection parameters can be obtained by performing quality inspection operations on the target object using an initial quality inspection model. For example, when performing quality inspection on the size of automobile engine parts, the initial quality inspection parameters are size data records of the automobile engine parts. Specifically, they may also include diameter data, length data, and flatness data.
[0060] It should be noted that the initial quality inspection model can also be used in other product manufacturing industries that require quality inspection operations, and the target object can also be adjusted according to actual conditions. The embodiments of the present application are only described as preferred embodiments and are not specifically limited.
[0061] In some embodiments, the iterative update of the initial quality inspection model can be performed on a platform equipped with MLOps capabilities (hereinafter referred to as the "platform"), such as Kubeflow, MLflow, TensorFlow Extended, or DataRobot. Since platforms equipped with MLOps capabilities usually have a series of built-in tools and components suitable for machine learning development and management, such as model version control systems, automated deployment tools, model performance monitoring systems, etc., these tools and components enable developers to develop, deploy, and manage models more conveniently. By performing the update of the initial quality inspection model on a platform equipped with MLOps capabilities, the update and iteration efficiency of the quality inspection model can be improved.
[0062] Step S102, when it is monitored that the initial quality inspection parameters are not within the preset quality inspection parameter range, sample training data is screened and obtained from the database, and the initial quality inspection model is trained according to the sample training data;
[0063] In some embodiments, the preset quality inspection parameter range is actually the quality inspection requirement, and the quality inspection parameter range can be adjusted according to actual needs, and the preset quality inspection parameter range can be input into the platform equipped with MLOps capabilities. Among them, the preset quality inspection parameter range can be updated regularly by relevant operators.
[0064] In some embodiments, it is necessary to filter the corresponding sample training data in the database according to the category of the initial quality inspection parameter. For example, when the initial quality inspection model is used to perform dimensional quality inspection on automobile engine parts, the output is the diameter quality inspection result. At this time, it is necessary to filter the data of the "size-diameter" category in the database to obtain sample training data for training the initial quality inspection model.
[0065] In some embodiments, a database may be deployed in the platform, or sample data may be acquired through a communication connection to a server deployed with the database, wherein the sample data includes sample training data and sample test data.
[0066] In some embodiments, the platform may include a controller for real-time monitoring of generated initial quality inspection parameters. When the initial quality inspection parameters are not within the preset quality inspection parameter range, the acquisition of sample training data can be automatically triggered, or a prompt alarm can be issued to allow relevant operators to confirm whether the initial quality inspection model needs to be updated.
[0067] In some embodiments, the data in the database may be collected and updated regularly. For example, relevant operators may regularly update the data in the database based on experience, or the data may be obtained from a designated data source through an update script.
[0068] In some embodiments, the initial quality inspection model may be trained separately by acquiring multiple groups of sample training data to obtain multiple different trained initial quality inspection models.
[0069] In some embodiments, after completing the training of the initial quality inspection model, the trained initial quality inspection model can also be preliminarily evaluated. Common evaluation methods include using confusion matrix, characteristic curve (Receiver Operating Characteristic Curve, ROC) and area under the characteristic curve (Area Under the Curve, AUC) for evaluation, and the evaluation metrics usually include accuracy, precision, recall rate, F1 score, etc.
[0070] Step S103, screening and obtaining sample test data from the database, testing the trained initial quality inspection model according to the sample test data, and obtaining model test parameters;
[0071] In some embodiments, after the training of the initial quality inspection model is completed, the trained initial quality inspection model needs to be tested, and the model test parameters can be obtained to determine whether the trained initial quality inspection model meets the expected quality inspection requirements.
[0072] In some embodiments, if different training versions of the initial quality inspection models are generated under multiple sets of sample training data when training the initial quality inspection model, the multiple sets of different initial quality inspection models obtained after training can be tested separately.
[0073] In some embodiments, testing the trained initial quality inspection model may include A / B testing and online indicator observation. Among them, A / B testing refers to comparing the trained initial quality inspection models to determine which trained initial quality inspection model performs better. Online indicator observation refers to running the initial quality inspection model under A / B testing and comparing the indicators between A / B tests, such as model accuracy, recall rate, etc., and comprehensively judging which version of the model is better through online real-time operation.
[0074] In some embodiments, to ensure the authenticity of the sample test data, the sample test data obtained from the database may be randomly generated, so that the sample test data can be more representative to enhance the robustness and stability of the quality inspection model.
[0075] Step S104, when the model test parameters match the preset model compliance values, the initial quality inspection model after the test is determined to be the target quality inspection model, and the target quality inspection model is used to re-perform quality inspection operations on the target object.
[0076] In some embodiments, a model compliance value is preset, wherein the model compliance value is used to indicate whether the updated initial quality inspection model achieves the expected quality inspection effect.
[0077] In some embodiments, the model test parameters obtained by testing the initial quality inspection model can be compared with the preset model standard values. If the model test parameters fall within the preset model standard value range, it means that the initial quality inspection model after training and testing meets the model expected value, and the updated initial quality inspection model is determined as the target quality inspection model. The target quality inspection model is used to re-inspect the target object to complete the accurate quality inspection of the target object.
[0078] It should be noted that in the process of updating the initial quality inspection model, the decrease in the quality inspection accuracy of the initial quality inspection model can be automatically monitored by the platform, and the corresponding sample training data and sample test data are also pre-configured in the database, so there is no need for professionals to manually mine and screen the data. After testing, the quality inspection model with the best test effect is selected as the target quality inspection model. It is understandable that even factory operators with low professionalism can independently complete the update of the quality inspection model, which improves the updating efficiency of the quality inspection model. Moreover, the target quality inspection model after the update can be immediately applied to the quality inspection work of the target object, which greatly improves the efficiency of the quality inspection work. If the problem of decreased quality inspection accuracy occurs again in actual use, the quality inspection model can still be updated again.
[0079] like Figure 3 As shown, Figure 3 yes Figure 2 An implementation flow chart after step S104. In some embodiments, step S104 may further include steps S201 to S202:
[0080] Step S201, when the model test parameter does not match the preset model standard value, re-screen and obtain sample training data and sample test data from the database;
[0081] In some embodiments, if the model test parameters do not match the preset model compliance values, it means that the training of the initial quality inspection model has not achieved the expected results, and it is necessary to re-screen and obtain sample training data and sample test data from the database to re-train and test the initial quality inspection model. Among them, the sample training data screened during re-training should be different from the sample training data used in the previous training, and the sample test data can be the same data to better detect the training effect of the initial quality inspection model after training.
[0082] In some embodiments, when the model test parameters do not match the preset model compliance values, sample training data can also be screened based on the deviation values between the model test parameters and the preset model compliance values. For example, when the deviation value range between the model test parameters and the preset model compliance values is small, it indicates that the updated quality inspection model may have an overfitting problem. In this case, the first training data is selected, and the first training data includes diverse sample training data. When the deviation value range between the model test parameters and the preset model compliance values is large, it indicates that the updated quality inspection model may not be of high quality. In this case, the second training data is selected, and the second training data includes high-quality sample training data. In this way, the initial quality inspection model can be trained more targeted.
[0083] Step S202: re-train and re-test the initial quality inspection model based on the re-determined sample training data and sample test data, and obtain updated model test parameters. When the model test parameters match the preset model compliance values, the updated initial quality inspection model is determined as the target quality inspection model.
[0084] In some embodiments, after re-determining the sample training data and the sample test data, the initial quality inspection model can be retrained and tested using the sample training data and the sample test data. If the updated model test parameters obtained after retraining and testing still do not match the preset model compliance values, steps S201 to S202 need to be repeated until the model test parameters match the preset model compliance values to obtain the target quality inspection model.
[0085] like Figure 4 As shown, Figure 4 yes Figure 2 An implementation flow chart of step S102. In some embodiments, step S102 may include steps S301 to S302:
[0086] Step S301, parsing the initial quality inspection parameters to obtain multiple parameter categories of the initial quality inspection parameters, each parameter category including a corresponding confidence level, the confidence level being used to characterize the degree of credibility of the initial quality inspection model belonging to the parameter category;
[0087] In some embodiments, when the initial quality inspection parameter is not within the preset quality inspection parameter range, the initial quality inspection parameter can be parsed to obtain the parameter category of the initial quality inspection parameter. Exemplarily, when the initial quality inspection model is an image classification model, a result list is output, which includes the parameter category of each quality inspection target object and the confidence level corresponding to the parameter category.
[0088] In some embodiments, if the obtained initial quality inspection parameter is a parameter code, the code can be parsed to obtain the corresponding parameter category.
[0089] Step S302, when the confidence level is not within the preset quality inspection parameter range, determining the screening category of the sample training data screened from the database according to the confidence level, and selecting corresponding sample training data from the database according to the screening category.
[0090] In some embodiments, when the confidence level obtained by analysis is not within the preset quality inspection parameter range, it can be determined that the quality inspection accuracy of the initial quality inspection model has decreased, and the initial quality inspection model needs to be trained and updated. The basis for selecting sample training data from the database is the initial quality inspection parameters. Specifically, the sample training data to be screened from the database is determined according to the parameter category, and the number of samples of the screened sample training data is determined according to the confidence level.
[0091] In some embodiments, the sample training data includes difficult example data, wherein the difficult example data can be filtered using SQL or ETL, or a filtering program can be provided, and which strategy to use is related to the user's specific scenario.
[0092] like Figure 5 As shown, Figure 5 yes Figure 2 Another implementation flow chart of step S102. In some embodiments, step S102 may further include steps S401 to S403:
[0093] Step S401, when sample training data is determined from the database, a training trigger of the initial quality inspection model is activated;
[0094] In some embodiments, the sample training data includes difficult data and regular data, wherein regular data is data that the initial quality inspection model often encounters during the actual quality inspection process, and difficult data is data that the initial quality inspection model rarely encounters during the actual quality inspection process.
[0095] In some embodiments, a training trigger is also provided. When the sample training data is determined from the database, the training trigger of the initial quality inspection model can be activated.
[0096] Step S402, according to the activated training trigger, input the difficult case data into the initial quality inspection model to obtain the difficult case training result, and input the normal data into the initial quality inspection model to obtain the normal training result;
[0097] In some embodiments, when the training trigger is in an activated state, it can be determined that the initial quality inspection model needs to be trained based on the acquired sample training data. First, the difficult example data can be input into the initial quality inspection model. Because the difficult example data usually represents data in extreme or boundary cases, it plays a key role in the training of the initial quality inspection model and can improve the accuracy and generalization ability of the model.
[0098] In some embodiments, the hard-case data may include boundary data, special format data, abnormal data, extreme data, and the like.
[0099] In some embodiments, after completing the test on the difficult data, it is usually necessary to test the initial quality inspection model based on conventional data to observe the general data processing capability of the trained initial quality inspection model.
[0100] In some embodiments, after the routine data is input into the initial quality inspection model, the initial quality inspection model can be trained according to the input routine data and obtain routine training results. Usually, the difficult example training results are similar to the routine training results, including both correct training results and incorrect training results. The difficult example training results and the routine training results can be used to determine whether additional sample training data is needed.
[0101] Step S403, completing the training of the initial quality inspection model according to the difficult example training results and the conventional training results.
[0102] In some embodiments, the difficult example data may be a difficult example data set. Training the initial quality inspection model based on the difficult example data set may obtain multiple difficult example training results, and a target percentage of the difficult example training results may be set. For example, setting the difficult example data set to achieve a 70% accuracy rate means that the difficult example training passes.
[0103] In some embodiments, if the difficult case training fails, there is no need to perform regular data training, because the initial quality inspection model does not meet the extreme data training requirements, indicating that there are still quality inspection errors in the quality inspection of the target object corresponding to the difficult case data. In quality inspection, especially in industrial quality inspection activities, the training of difficult case data is very critical. Therefore, if the difficult case training fails, the sample training data and sample training data can be directly obtained from the database again, and there is no need to perform the training of the current sample training data. In this way, the updating efficiency of the quality inspection model can be further improved.
[0104] like Figure 6 As shown, Figure 6 yes Figure 2 An implementation flow chart of step S103. In some embodiments, step S103 may also include steps S501 to S503:
[0105] Step S501, when the training of the initial quality inspection model is completed, activating the test trigger of the initial quality inspection model;
[0106] In some embodiments, a test trigger is provided, and when the sample training data is determined from the database, the test trigger of the initial quality inspection model can be activated.
[0107] Step S502, according to the test trigger in the activated state, filter and obtain sample test data from the database;
[0108] In some embodiments, when the test trigger is in an activated state, it can be determined that the initial quality inspection model needs to be tested. Similar to the training process of the initial quality inspection model, it is necessary to first determine the sample test data of the initial quality inspection model from the database. Through the test trigger, the acquisition of sample test data is triggered, which avoids the tedious steps of manually waiting and monitoring until the training of the initial quality inspection model is completed, and manually starting the initial quality inspection model test in the traditional quality inspection model update, further improving the update efficiency of the quality inspection model.
[0109] In some embodiments, the sample test data may also be input in real time by relevant operators, and may be set specifically according to actual conditions, and this application does not impose any specific restrictions.
[0110] Step S503: Test the trained initial quality inspection model according to the sample test data and obtain model test parameters.
[0111] In some embodiments, the acquired sample test data is input into the trained initial quality inspection model to perform a test on the trained initial quality inspection model. After the test is completed, the model test parameters of the initial quality inspection model can be obtained.
[0112] In some embodiments, the model test parameters are used to compare with preset model compliance values. The model compliance value can be a range. If the obtained model test parameter value does not fall within the range, it means that the initial quality inspection model test fails, and it is necessary to re-acquire sample training data and sample test data to perform model update operations.
[0113] like Figure 7 As shown, Figure 7 It is an optional workflow configuration diagram of the quality inspection model management method provided in the embodiment of the present application. In order to improve the updating efficiency of the quality inspection model, the relevant update steps can be logically connected in the relevant platform equipped with MLOps to form a visual workflow step. In addition, the platform includes a management end and a user end. The management end is responsible for configuring the workflow, and the user end is responsible for displaying and running the workflow. The management end includes a management interface, and the user end includes a user interface.
[0114] For example, Figure 7 The left area is the visualization area, and the right area is the configuration area. The workflow in the visualization area can include multiple sub-workflows, and each sub-workflow includes the following steps: difficult case screening-difficult case annotation-data set-training evaluation-testing. The key step of each sub-workflow is the selection of difficult case data. Therefore, the configuration area is used to screen and annotate difficult case data for each sub-workflow. Specifically, the visualization table in the configuration area is used to select a specific visualization display method; the screening range is used to indicate the screening range of difficult case data; the model name is used to indicate the selection of the initial quality inspection model that needs to be updated and iterated; the annotation type is used to select the type that needs to be annotated; the label template is used to adjust the label style of the visualization area. Among them, each workflow node in the visualization area is followed by a judgment mark. If the node configuration is passed, the judgment mark displays the configuration success status. If the node configuration is not passed, the judgment mark displays the configuration failure status. It should be noted that the workflow configured by the management end requires the user end to set specific parameters before it can be executed.
[0115] like Figure 8 As shown, Figure 8 This is another optional workflow configuration diagram of the quality inspection model management method provided in the embodiment of the present application. After the workflow configuration of the quality inspection model is completed on the management end of the relevant platform equipped with MLOps, the corresponding workflow configuration of the quality inspection model management can be displayed in the user interface, wherein the training, evaluation and testing of the model are all in Figure 8 The "Model Evaluation" area in the QC model is displayed. Through the workflow logic pre-configured on the management side, the relevant operators only need to click the component buttons corresponding to each step on the user side to complete the automatic update operation of the QC model. When the quality inspection accuracy of the initial QC model decreases, there is no need to manually obtain the relevant data and input them one by one into the initial QC model to be updated, which reduces the professionalism of the operation and improves the updating efficiency of the QC model. In this way, the operator can update the QC model immediately and put it into use directly after obtaining the updated QC model, thereby improving the efficiency of related quality inspection work.
[0116] like Fig. 9 As shown, Fig. 9 This is an optional workflow execution diagram for quality inspection model management provided by the embodiment of the present application. Since the configuration of the management end and the operation and display of the user end are provided by the backend service workflow-server, after the workflow is configured on the platform management end, specific parameters can be configured on the user end (user end-web), and then specific operation steps are executed through the backend service workflow-server. The detailed process is as follows:
[0117] 1. Configure the workflow in the management interface of the platform management end. The configuration work includes creating an MLOps workflow template, configuring a virtual data set and inference service (that is, configuring a sample training set and the training method of the sample training set), instantiating the MLOps work, and generating a workflow configuration;
[0118] 2. Start an update iteration on the user side - web and run a step of the workflow, including obtaining the MLOps workflow configuration configured on the platform management side, displaying the instantiation of the MLOps steps according to the workflow configuration, and running the MLOps data set;
[0119] 3. According to the operation of the user-side web, find the corresponding trigger in the workflow-server backend service and trigger the corresponding Component logic;
[0120] 4. Component will call the corresponding service to complete the relevant logic, and after completion, send the event to MessageQueue (MQ);
[0121] 5. The workflow-controller obtains events from MQ and stores the execution results of the current step, where the running results can be persisted and the workflow advances to the next workflow step.
[0122] Among them, Trigger is the service that is initiated by Trigger when the user clicks on a workflow step component; Component is the service that specifically executes a step, such as difficult example screening, training, evaluation, etc.; Controller is used to execute the flow of workflow status. When a step is completed, it is responsible for storing the output result of this step, reversing the status of the current step, and moving on to the next step.
[0123] In addition, in order to more clearly understand the above embodiments of the present application, further examples will be given below.
[0124] like Fig.10 As shown, Fig.10 This is an optional flow chart of the quality inspection model management method provided in the embodiment of the present application. The following is a description of the specific steps of this example:
[0125] Step 1: Service deployment: First, deploy the initial quality inspection model to the service scenario. The initial quality inspection model will infer the initial quality inspection parameter data and generate feedback data.
[0126] Step 2: Data analysis. From the returned data, relevant model indicators can be statistically analyzed to determine whether the model effect has declined. If so, model iteration is required. Specifically, the decline in model effect can be judged through model indicator statistics. For the industrial quality inspection field, by collecting production line data in real time, the model accuracy, defect detection rate, false point removal rate, first pass rate, and other indicators can be statistically analyzed, and whether the indicators have declined can be judged through visualization.
[0127] Step 3: Model iteration is divided into the following steps: difficult example screening, data labeling, training, evaluation, A / B testing, online indicator observation, grayscale release, etc. Among them, the model iteration can be run multiple times, and the optimal target quality inspection model is selected to replace the initial quality inspection model running online.
[0128] like Fig.11 As shown, Fig.11 is another optional flow chart of the quality inspection model management method provided in the embodiment of the present application. Fig.11 The method may include but is not limited to steps S601 to S605.
[0129] Step S601, displaying a quality inspection model management interface, displaying configuration components of the initial quality inspection model on the quality inspection model management interface, wherein the configuration components include a data configuration component, a model training component, a model testing component, and a model publishing component;
[0130] In some embodiments, in order to further improve the updating efficiency of the quality inspection model, the updating operation steps of the quality inspection model may be displayed in a visual manner.
[0131] like Fig. 9 As shown, the data configuration component includes a difficult example screening component, a difficult example annotation component and a data set component. Clicking the difficult example screening component can display the start time and completion time of the selected difficult example screening, and different start times and end times correspond to multiple groups of different difficult example data; clicking the difficult example annotation component can display the difficult example annotation corresponding to the selected difficult example data, and the difficult example annotation can be annotated based on the artificial intelligence model, or manually annotated; clicking the data set component can display one or more selected groups of difficult example data, and can also include common conventional data.
[0132] In some embodiments, since the relevant sample data is pre-set and stored in the database, there is no need for operators to manually select and integrate them. At the same time, the relevant platform can perform automated testing based on the selected sample data, that is, the platform equipped with MLOps capabilities can perform automatic data collection and automatic model monitoring. Through preset difficult case screening methods and preset model training and evaluation methods, relevant operators only need to simply click on the relevant components and choose to start iterating the components to automatically complete the update of the initial quality inspection model, which greatly simplifies the operational steps of model updating and improves the updating efficiency of the quality inspection model.
[0133] Step S602, in response to a click operation on the data configuration component, displaying the configured sample data on the quality inspection model management interface, wherein the sample data includes sample training data and sample test data;
[0134] In some embodiments, after clicking the data configuration component, the configured sample data can be displayed in the quality inspection model management interface. Figure 8 The area below the Dataset component displays sample training data and sample test data.
[0135] Step S603, in response to a click operation on the model training component, the model training parameter values are displayed on the quality inspection model management interface, where the model training parameter values are obtained after the initial quality inspection model is trained according to the sample training data;
[0136] In some embodiments, after clicking the model training component, the model training parameter values after the initial quality inspection model training can be displayed in the quality inspection model management interface. Exemplarily, the model training parameter values may include information such as model weights and loss functions.
[0137] Step S604, in response to the click operation of the model test component, the model test parameter value is displayed in the quality inspection model management interface, the model test parameter value is obtained after the initial quality inspection model is tested according to the sample test data, and the initial quality inspection model after the test is determined to be the target quality inspection model;
[0138] In some embodiments, the model test component can be clicked to perform the test of the initial quality inspection model, and the model test parameter values related to the test are displayed in the quality inspection model management interface. Exemplarily, the test parameter values may include the model test values obtained after the model test, and the preset model compliance values, so as to facilitate intuitive data comparison with the model test values.
[0139] Step S605, in response to a click operation on the model publishing component, the published target quality inspection model is displayed on the quality inspection model management interface.
[0140] In some embodiments, when the test of the initial quality inspection model passes, the model release component can be clicked to indicate that the target quality inspection model is officially launched. Alternatively, a simulated launch component can be provided. When the simulated launch component is clicked, the launch of the target quality inspection model can be simulated on some terminals to ensure the reliability and actual availability of the target quality inspection model before the real launch.
[0141] like Fig.12 As shown, Fig.12 yes Fig.11 An implementation flow chart of step S602. In some embodiments, step S602 may include steps S701 to S704:
[0142] Step S701, displaying a quality inspection model configuration interface, displaying a data generation component and a data annotation component of sample data on the quality inspection model configuration interface;
[0143] In some embodiments, the quality inspection model configuration interface may be a pop-up interface, that is, when sample data is not being acquired, it may be hidden in the quality inspection model management interface to enhance the simplicity of the interface.
[0144] In some embodiments, to implement the configuration of sample data, a quality inspection model configuration interface can be further displayed in the quality inspection model management interface, wherein the quality inspection model configuration interface includes a data generation component and a data annotation component. The data generation component corresponds to Figure 8 The "Generate Hard Examples" component in the data annotation component corresponds to Figure 8 The Submit Annotation component in .
[0145] Step S702, in response to a click operation on the data generation component, the sample training data and sample test data obtained according to the initial quality inspection parameter configuration are displayed in the data configuration area of the quality inspection model configuration interface, wherein the sample training data includes difficult case data and regular data;
[0146] In some embodiments, a data generation component is clicked to perform generation of hard example data, and Figure 8 The area below the "Dataset" component displays the sample training data and sample test data obtained by configuration. Since the acquisition of difficult data is more important in the quality inspection of the target object, the data generation component is actually the generation of difficult data, and the generation component of regular data is not displayed, but is obtained from fixed storage in the database.
[0147] Step S703, in response to a click operation on the data annotation component, the annotation results of the sample training data are displayed in the data annotation area of the quality inspection model configuration interface, wherein the annotation results include the annotation results of the difficult data and the annotation results of the regular data;
[0148] In some embodiments, click on the data annotation component to Figure 8 The area below the "Submit Annotation" component displays the annotation results of the sample training data, where the annotation of the sample training data can be obtained through artificial intelligence annotation, or manually annotated by relevant staff.
[0149] Step S704, in response to the data configuration completion operation, the configured sample data is displayed on the quality inspection model management interface according to the sample test data and the sample training data including the annotation results.
[0150] In some embodiments, when sample training data, sample test data, and corresponding annotation results are generated, the acquisition of sample data is completed.
[0151] like Fig.13 As shown, Fig.13 : is a functional module diagram of the quality inspection model management provided in the embodiment of the present application. The embodiment of the present application also provides a quality inspection model management system, which can implement the above-mentioned quality inspection model management method. The quality inspection model management system includes:
[0152] The acquisition module 801 is used to acquire an initial quality inspection model, perform a quality inspection operation on the target object according to the initial quality inspection model, and obtain initial quality inspection parameters;
[0153] The model training module 802 is used to filter and obtain sample training data from the database when the initial quality inspection parameters are not within the preset quality inspection parameter range, and train the initial quality inspection model according to the sample training data;
[0154] The model testing module 803 is used to filter and obtain sample test data from the database, test the trained initial quality inspection model according to the sample test data, and obtain model test parameters;
[0155] The target quality inspection model module 804 is used to determine the initial quality inspection model after the test as the target quality inspection model when the model test parameters match the preset model compliance values. The target quality inspection model is used to re-perform quality inspection operations on the target object.
[0156] The specific implementation of the quality inspection model management system is basically the same as the specific implementation of the quality inspection model management method described above, and will not be repeated here. Under the premise of meeting the requirements of the embodiment of this application, the quality inspection model management system can also set other functional modules to implement the quality inspection model management method in the above embodiment.
[0157] It should be noted that in the process of updating the initial quality inspection model, the decrease in the quality inspection accuracy of the initial quality inspection model can be automatically monitored by the platform, and the corresponding sample training data and sample test data are also pre-configured in the database, so there is no need for professionals to manually mine and screen the data. After testing, the quality inspection model with the best test effect is selected as the target quality inspection model. It is understandable that even factory operators with low professionalism can independently complete the update of the quality inspection model, which improves the updating efficiency of the quality inspection model. Moreover, the target quality inspection model after the update can be immediately applied to the quality inspection work of the target object, which greatly improves the efficiency of the quality inspection work. If the problem of decreased quality inspection accuracy occurs again in actual use, the quality inspection model can still be updated again.
[0158] The embodiment of the present application also provides an electronic device, the electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the above-mentioned quality inspection model management method when executing the computer program. The electronic device can be any intelligent terminal including a tablet computer, a car computer, etc.
[0159] like Fig.14 As shown, Fig.14 : is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present application, the electronic device includes:
[0160] The processor 901 may be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application;
[0161] The memory 902 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 902 can store an operating system and other applications. When the technical solution provided in the embodiment of this specification is implemented by software or firmware, the relevant program code is stored in the memory 902, and the processor 901 calls and executes the quality inspection model management method of the embodiment of this application;
[0162] Input / output interface 903, used to implement information input and output;
[0163] Communication interface 904, used to realize communication interaction between the device and other devices, which can be realized by wired mode (such as USB, network cable, etc.) or wireless mode (such as mobile network, WIFI, Bluetooth, etc.);
[0164] A bus 905 that transmits information between various components of the device (e.g., the processor 901, the memory 902, the input / output interface 903, and the communication interface 904);
[0165] The processor 901 , the memory 902 , the input / output interface 903 and the communication interface 904 are connected to each other in communication within the device via a bus 905 .
[0166] An embodiment of the present application also provides a computer-readable storage medium, which stores a computer program, and when the computer program is executed by a processor, the above-mentioned quality inspection model management method is implemented.
[0167] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.
[0168] The embodiments described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.
[0169] Those skilled in the art will appreciate that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.
[0170] The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
[0171] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices may be implemented as software, firmware, hardware, or a suitable combination thereof.
[0172] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.
[0173] It should be understood that in the present application, "at least one (item)" and "several" refer to one or more, and "plurality" refers to two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the previous and subsequent associated objects are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.
[0174] In the several embodiments provided in the present application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely schematic. For example, the division of the above units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.
[0175] The units described above as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
[0176] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.
[0177] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including multiple instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM), disk or optical disk and other media that can store programs.
[0178] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but the scope of the rights of the present invention is not limited thereto. Any modification, equivalent substitution and improvement made by a person skilled in the art without departing from the scope and essence of the present invention should be within the scope of the rights of the present invention.
Claims
1. A quality inspection model management method, characterized in that: The method comprises: Acquire an initial quality inspection model, perform a quality inspection operation on the target object according to the initial quality inspection model, and obtain initial quality inspection parameters; When it is monitored that the initial quality inspection parameter is not within the preset quality inspection parameter range, sample training data is screened and obtained from the database, and the initial quality inspection model is trained according to the sample training data; Screening and obtaining sample test data from a database, testing the trained initial quality inspection model according to the sample test data, and obtaining model test parameters; When the model test parameter matches the preset model compliance value, the initial quality inspection model after the test is determined to be a target quality inspection model, and the target quality inspection model is used to re-perform a quality inspection operation on the target object; The training of the initial quality inspection model according to the sample training data includes: When the sample training data is determined from the database, a training trigger of the initial quality inspection model is activated; the sample training data includes difficult case data and regular data; According to the training trigger in an activated state, the difficult example data is input into the initial quality inspection model to obtain a difficult example training result; When the difficult example training result is not lower than the preset standard percentage, the conventional data is input into the initial quality inspection model to obtain the conventional training result; otherwise, the sample training data is re-acquired to train the initial quality inspection model; The training of the initial quality inspection model is completed according to the difficult example training results and the conventional training results.
2. The quality inspection model management method according to claim 1, characterized in that: After testing the trained initial quality inspection model according to the sample test data and obtaining the model test parameters, the method further includes: When the model test parameter does not match the preset model standard value, re-screening and obtaining sample training data and sample test data from the database; According to the re-determined sample training data and the sample test data, the initial quality inspection model is re-updated and re-tested to obtain updated model test parameters. When the model test parameters match the preset model compliance values, the updated initial quality inspection model is determined to be the target quality inspection model.
3. The quality inspection model management method according to claim 2, characterized in that: When it is monitored that the initial quality inspection parameter is not within the preset quality inspection parameter range, the sample training data is screened and obtained from the database, including: Parsing the initial quality inspection parameters to obtain a plurality of parameter categories of the initial quality inspection parameters, each parameter category including a corresponding confidence level, the confidence level being used to characterize the degree of credibility of the initial quality inspection model belonging to the parameter category; When the confidence level is not within a preset quality inspection parameter range, a screening category of sample training data screened from a database is determined according to the confidence level, and corresponding sample training data is selected from the database according to the screening category.
4. The quality inspection model management method according to claim 1, characterized in that: The method of screening and obtaining sample test data from the database, testing the trained initial quality inspection model according to the sample test data, and obtaining model test parameters includes: When the training of the initial quality inspection model is completed, activating a test trigger of the initial quality inspection model; According to the test trigger in an activated state, sample test data is screened and obtained from a database; The trained initial quality inspection model is tested according to the sample test data, and model test parameters are obtained.
5. The quality inspection model management method according to claim 1, characterized in that: After determining that the trained initial quality inspection model is the target quality inspection model, the method further includes: Display a quality inspection model management interface, and display configuration components of the initial quality inspection model on the quality inspection model management interface, wherein the configuration components include a data configuration component, a model training component, a model testing component, and a model publishing component; In response to a click operation on the data configuration component, the configured sample data is displayed on the quality inspection model management interface, wherein the sample data includes sample training data and sample test data; In response to a click operation on the model training component, a model training parameter value is displayed on the quality inspection model management interface, wherein the model training parameter value is obtained after the initial quality inspection model is trained according to the sample training data; In response to a click operation on the model test component, a model test parameter value is displayed on the quality inspection model management interface, the model test parameter value is obtained after the initial quality inspection model is tested according to the sample test data, and the initial quality inspection model after the test is determined to be the target quality inspection model; In response to a click operation on the model publishing component, the published target quality inspection model is displayed on the quality inspection model management interface.
6. The quality inspection model management method according to claim 5, characterized in that: The sample data obtained by configuration is displayed on the quality inspection model management interface, including: Displaying a quality inspection model configuration interface, wherein the data generation component and the data annotation component of the sample data are displayed on the quality inspection model configuration interface; In response to a click operation of the data generation component, sample training data and sample test data obtained according to the initial quality inspection parameter configuration are displayed in the data configuration area of the quality inspection model configuration interface, wherein the sample training data includes difficult case data and regular data; In response to a click operation on the data annotation component, the annotation results of the sample training data are displayed in the data annotation area of the quality inspection model configuration interface, wherein the annotation results include the annotation results of the difficult data and the annotation results of the conventional data; In response to the data configuration completion operation, the configured sample data is displayed on the quality inspection model management interface according to the sample test data and the sample training data including the annotation results.
7. A quality inspection model management system, characterized in that: The system comprises: An acquisition module is used to acquire an initial quality inspection model, perform a quality inspection operation on the target object according to the initial quality inspection model, and obtain initial quality inspection parameters; A model training module is used to filter and obtain sample training data from a database when it is monitored that the initial quality inspection parameters are not within a preset quality inspection parameter range, and train the initial quality inspection model according to the sample training data; A model testing module is used to filter and obtain sample test data from a database, test the trained initial quality inspection model according to the sample test data, and obtain model test parameters; A target quality inspection model module, used for determining the initial quality inspection model after the test as a target quality inspection model when the model test parameter matches the preset model compliance value, and the target quality inspection model is used for re-performing a quality inspection operation on the target object; The training of the initial quality inspection model according to the sample training data includes: When the sample training data is determined from the database, a training trigger of the initial quality inspection model is activated; the sample training data includes difficult case data and regular data; According to the training trigger in an activated state, the difficult example data is input into the initial quality inspection model to obtain a difficult example training result; When the difficult example training result is not lower than the preset standard percentage, the conventional data is input into the initial quality inspection model to obtain the conventional training result; otherwise, the sample training data is re-acquired to train the initial quality inspection model; The training of the initial quality inspection model is completed according to the difficult example training results and the conventional training results.
8. An electronic device, characterized in that: The electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the quality inspection model management method according to any one of claims 1 to 6 when executing the computer program.
9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the quality inspection model management method according to any one of claims 1 to 6 is implemented.
