A model management method, device and equipment and computer readable storage medium

Abstract

This application provides a model management method, apparatus, device, and computer-readable storage medium, relating to the field of vision technology. The method is applied to a vision algorithm development platform, which includes a vision task module for building vision task schemes. The method includes: receiving a global model management instruction; and displaying a global model management interface. The global model management interface displays model information and deployment status of the models used in the vision task scheme. This facilitates model management for users and improves model management efficiency.

Description

Technical Field

[0001] This application relates to the field of vision technology, and in particular to a model management method, apparatus, device, and computer-readable storage medium. Background Technology

[0002] With the rapid development of computer technology, visual task solutions can be executed based on visual algorithm development platforms. For example, a visual task workflow can be built within the interface of the visual algorithm development platform, and the built visual task workflow can be run to process images. The visual task workflow includes an inference module for image detection, and a model is deployed in the inference module.

[0003] Currently, for an inference module, users need to determine the visual task flow to which the inference module belongs within the visual algorithm development platform, and then click on the inference module within that visual task flow to view the model information of the model deployed in that inference module. This method is inconvenient for users to manage models. Summary of the Invention

[0004] The purpose of this application is to provide a model management method, apparatus, device, and computer-readable storage medium to facilitate user management of models and improve model management efficiency. The specific technical solution is as follows:

[0005] The first aspect of this application provides a model management method applied to a visual algorithm development platform, wherein the visual algorithm development platform includes a visual task module for building visual task schemes; the method includes:

[0006] Receive global model management instructions;

[0007] The global model management interface is displayed; wherein, the global model management interface is used to display: the model information and deployment status of the models carried in the vision task solution.

[0008] In an optional embodiment, the method further includes:

[0009] In response to the add model operation, retrieve the second model;

[0010] The global model management interface now displays the model identifier of the second model, as well as the model information of the second model.

[0011] In an optional embodiment, obtaining the second model in response to the add model operation includes:

[0012] In response to receiving a model creation command triggered in the global model management interface, a first model is created; the first model is trained using sample images to obtain a second model;

[0013] or,

[0014] In response to receiving a model import command triggered in the global model management interface, the second model indicated by the model import command is obtained.

[0015] In an optional embodiment, the step of obtaining the second model indicated by the model import instruction in response to receiving a model import instruction triggered in the global model management interface includes:

[0016] In response to receiving a local model import command triggered in the global model management interface, the first model is retrieved from the local machine;

[0017] or,

[0018] In response to receiving a cloud model import command triggered in the global model management interface, the first model is retrieved from the cloud.

[0019] In one optional embodiment, the first model is obtained in the following manner:

[0020] Models are categorized and displayed according to their type.

[0021] In response to receiving the first selection instruction, the first model is retrieved from the displayed models.

[0022] In an optional embodiment, the step of obtaining the first model from the cloud in response to receiving a cloud model import instruction triggered in the global model management interface includes:

[0023] In response to receiving a cloud model import command triggered in the global model management interface, the cloud login interface is displayed;

[0024] After successful login, the model in the cloud will be displayed;

[0025] In response to receiving a second selection instruction, the first model is retrieved from the displayed models.

[0026] In an optional embodiment, training the first model using sample images to obtain the second model includes:

[0027] In response to receiving a model training instruction triggered in the global model management interface, the first model is trained using sample images to obtain a second model.

[0028] In an optional embodiment, before training the first model using sample images to obtain the second model, the method further includes:

[0029] Images used to train the first model are determined from the real-time acquired images and used as sample images.

[0030] In an optional embodiment, determining the images used for training the first model from the real-time acquired images, as sample images, includes:

[0031] In response to receiving a single image addition instruction, the currently acquired single image is determined as a sample image;

[0032] or,

[0033] In response to receiving a batch image addition instruction, the acquired multiple images are identified as sample images.

[0034] In an optional embodiment, training the first model using sample images to obtain the second model includes:

[0035] The training settings interface is displayed; wherein, the training settings interface displays: sample images and label setting areas;

[0036] Obtain the labels that are used to annotate the sample image in the label setting area;

[0037] The first model is trained using sample images and acquired labels to obtain the second model.

[0038] In an optional embodiment, the method further includes:

[0039] The training status of the model is displayed in the global model management interface.

[0040] In an optional embodiment, the training state includes at least one of the following:

[0041] Model training progress;

[0042] This indicates the training result, whether the training was successful or unsuccessful.

[0043] Reasons for training failure.

[0044] In an optional embodiment, the method further includes:

[0045] In response to a detailed display instruction for the training status of any model triggered in the global model management interface, the training details of the model indicated by the detailed display instruction are displayed; wherein, the training details include at least one of the following: basic training information, training parameters, and loss curve.

[0046] In an optional embodiment, the vision task module includes an inference module that uses a model for inference; the method further includes:

[0047] The third model is deployed in batches to at least one inference module included in the vision task scheme.

[0048] In an optional embodiment, the detection type of the at least one inference module is consistent with the detection type of the third model.

[0049] In an alternative embodiment, before deploying the third model in batches to at least one inference module included in the vision task scheme, the method further includes:

[0050] Display the inference module in the vision task scheme that has the same detection type as the third model;

[0051] In response to receiving a third selection instruction, determine from the displayed inference modules the inference modules for which the third model needs to be deployed in batches.

[0052] In an optional embodiment, the model information displayed in the global model management interface includes at least one of the following: the model's source, version number, detection type, testing status, and annotation status;

[0053] And / or,

[0054] In the global model management interface, models are displayed according to model type;

[0055] And / or,

[0056] The method further includes:

[0057] In response to receiving a path modification instruction for the fourth model triggered in the global model management interface, the fourth model is saved according to the save path indicated by the path modification instruction.

[0058] In one alternative embodiment, the model carried in the vision task scheme includes a model with adjustable parameters.

[0059] A second aspect of this application provides a model management device applied to a visual algorithm development platform, the visual algorithm development platform including a visual task module for building visual task schemes; the device includes:

[0060] The instruction receiving unit is used to receive global model management instructions;

[0061] The display unit is used to display the global model management interface; wherein, the global model management interface is used to display: the model information and deployment status of the models carried in the vision task solution.

[0062] In an optional embodiment, the apparatus further includes:

[0063] The model acquisition unit is used to acquire a second model in response to the add model operation;

[0064] The refresh module is used to add and display the model identifier of the second model and the model information of the second model in the global model management interface.

[0065] In an optional embodiment, the model acquisition unit is specifically configured to, in response to receiving a model creation instruction triggered in the global model management interface, create a first model; and train the first model using sample images to obtain a second model;

[0066] or,

[0067] In response to receiving a model import command triggered in the global model management interface, the second model indicated by the model import command is obtained.

[0068] In an optional embodiment, the model acquisition unit is specifically used to acquire a first model from the local source in response to receiving a local model import instruction triggered in the global model management interface;

[0069] or,

[0070] In response to receiving a cloud model import command triggered in the global model management interface, the first model is retrieved from the cloud.

[0071] In an optional embodiment, the model acquisition unit is specifically used to acquire the first model in the following manner:

[0072] Models are categorized and displayed according to their type.

[0073] In response to receiving the first selection instruction, the first model is retrieved from the displayed models.

[0074] In an optional embodiment, the model acquisition unit is specifically used to display the cloud login interface in response to receiving a cloud model import instruction triggered in the global model management interface;

[0075] After successful login, the model in the cloud will be displayed;

[0076] In response to receiving a second selection instruction, the first model is retrieved from the displayed models.

[0077] In an optional embodiment, the model acquisition unit is specifically used to train the first model using sample images to obtain a second model in response to receiving a model training instruction triggered in the global model management interface.

[0078] In an optional embodiment, the apparatus further includes: a sample image determination unit, configured to determine, from real-time acquired images, images for training the first model as sample images before training the first model using the sample images to obtain the second model.

[0079] In an optional embodiment, the sample image determination unit is specifically configured to determine the currently acquired single image as a sample image in response to receiving a single image addition instruction;

[0080] or,

[0081] In response to receiving a batch image addition instruction, the acquired multiple images are identified as sample images.

[0082] In an optional embodiment, the model acquisition unit is specifically used to display a training settings interface; wherein the training settings interface displays: sample images and a label setting area;

[0083] Obtain the labels that are used to annotate the sample image in the label setting area;

[0084] The first model is trained using sample images and acquired labels to obtain the second model.

[0085] In an optional embodiment, the apparatus further includes:

[0086] The training status display unit is used to display the training status of the model in the global model management interface.

[0087] In an optional embodiment, the training state includes at least one of the following:

[0088] Model training progress;

[0089] This indicates the training result, whether the training was successful or unsuccessful.

[0090] Reasons for training failure.

[0091] In an optional embodiment, the device further includes: a training details display unit, configured to display the training details of the model indicated by the details display instruction in response to a details display instruction for the training status of any model triggered in the global model management interface; wherein the training details include at least one of the following: basic training information, training parameters, and loss curve.

[0092] In an optional embodiment, the vision task module includes an inference module that uses a model for inference;

[0093] The device further includes a batch deployment unit for batch deploying the third model to at least one inference module included in the vision task scheme.

[0094] In an optional embodiment, the detection type of the at least one inference module is consistent with the detection type of the third model.

[0095] In an optional embodiment, the apparatus further includes: an inference module determination unit, configured to display, before deploying the third model in batches to at least one inference module included in the vision task scheme, an inference module in the vision task scheme that is consistent with the detection type of the third model;

[0096] In response to receiving a third selection instruction, determine from the displayed inference modules the inference modules for which the third model needs to be deployed in batches.

[0097] In an optional embodiment, the model information displayed in the global model management interface includes at least one of the following: the model's source, version number, detection type, testing status, and annotation status;

[0098] And / or,

[0099] In the global model management interface, models are displayed according to model type;

[0100] And / or,

[0101] The device further includes:

[0102] The saving unit is used to respond to a path modification instruction for the fourth model triggered in the global model management interface, and save the fourth model according to the saving path indicated by the path modification instruction.

[0103] In one alternative embodiment, the model carried in the vision task scheme includes a model with adjustable parameters.

[0104] A third aspect of this application provides an electronic device, comprising:

[0105] Memory, used to store computer programs;

[0106] When a processor executes a program stored in memory, it implements any of the methods described in the first aspect above.

[0107] A fourth aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements any of the methods described in the first aspect above.

[0108] The fifth aspect of this application provides a computer program product containing instructions that, when run on a computer, cause the computer to perform any of the methods described in the first aspect above.

[0109] Beneficial effects of the embodiments in this application:

[0110] The model management method provided in this application embodiment can be applied to a visual algorithm development platform. The visual algorithm development platform includes a visual task module for building a visual task scheme. The method includes: receiving a global model management instruction; and displaying a global model management interface. The global model management interface is used to display the model information and deployment status of the models carried in the visual task scheme.

[0111] Based on the above processing, the model information and deployment status of the models carried in the vision task solution are uniformly displayed in the global model management interface. This allows users to view the model information and deployment status of any model in the global model management interface, making it convenient for users to manage models and improving model management efficiency.

[0112] Of course, implementing any product or method of this application does not necessarily require achieving all of the advantages described above at the same time. Attached Figure Description

[0113] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other embodiments can be obtained based on these drawings.

[0114] Figure 1 A flowchart illustrating the first model management method provided in this application embodiment;

[0115] Figure 2 A schematic diagram of a first global model management interface provided in an embodiment of this application;

[0116] Figure 3 A schematic diagram of a process management interface provided in an embodiment of this application;

[0117] Figure 4 A schematic diagram illustrating a global model management interface provided in an embodiment of this application;

[0118] Figure 5 A schematic diagram illustrating a second type of global model management interface provided in an embodiment of this application;

[0119] Figure 6 A schematic diagram illustrating a login interface provided in an embodiment of this application;

[0120] Figure 7 A schematic diagram illustrating a model for displaying cloud storage, provided as an embodiment of this application;

[0121] Figure 8 A schematic diagram of an import model provided in an embodiment of this application;

[0122] Figure 9 A schematic diagram of a training settings interface provided in an embodiment of this application;

[0123] Figure 10 A schematic diagram illustrating a third type of global model management interface provided in an embodiment of this application;

[0124] Figure 11 A schematic diagram illustrating a detailed information display interface provided in an embodiment of this application;

[0125] Figure 12 A schematic diagram illustrating a batch deployment interface provided in an embodiment of this application;

[0126] Figure 13 A flowchart illustrating the second model management method provided in this application embodiment;

[0127] Figure 14 This is a schematic diagram of the structure of a model management device provided in an embodiment of this application;

[0128] Figure 15 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0129] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art based on this application are within the scope of protection of this application.

[0130] Currently, users need to determine the visual task flow to which the inference module belongs within the visual algorithm development platform, and then click on the inference module within that visual task flow to view the model information of the model deployed in that inference module. This method is inconvenient for users to manage models.

[0131] To address the aforementioned technical problems, embodiments of this application provide a model management method, apparatus, device, and computer-readable storage medium.

[0132] The model management method provided in this application embodiment will be described below. The model management method provided in this application embodiment can be applied to a visual algorithm development platform in an electronic device. For example, the visual algorithm development platform is a software client installed in an electronic device. Users can build visual task flows in the interface (which can be called the process management interface) of the visual algorithm development platform according to business needs, and run the built visual task flows to process images. The visual task flow includes an inference module for image detection, and a model is deployed in the inference module.

[0133] In one embodiment, the model management method is applied to a visual algorithm development platform, which includes a visual task module for building visual task solutions. The method includes: receiving a global model management instruction; and displaying a global model management interface. The global model management interface displays the model information and deployment status of the models used in the visual task solutions. Based on the above processing, the global model management interface uniformly displays the model information and deployment status of the models used in the visual task solutions, allowing users to view the model information and deployment status of any model within the global model management interface. This facilitates model management and improves management efficiency.

[0134] In one embodiment, see Figure 1 The model management method provided in this application includes the following steps:

[0135] S101: Receive global model management instructions.

[0136] S102: Displays the global model management interface.

[0137] The global model management interface displays model information and deployment status of the models used in the vision task solution. For ease of description, model information and deployment status will be collectively referred to as the information to be displayed.

[0138] In one embodiment, the global model management interface displays model identifiers for all models included in the vision task solution. The interface only displays the information to be displayed for the model represented by the currently selected model identifier; other models only display their model identifiers. Alternatively, if the global model management interface displays a large amount of information to be displayed, a sliding button can be included. When the user triggers this button, the global model management interface can be flipped through pages, allowing the user to browse all the information to be displayed.

[0139] In one embodiment, the global model management interface displays models categorized by model type. Model type includes: detection type and / or model version. For example, in the models displayed on the global model management interface, multiple models of the same detection type are displayed in the same area, and multiple versions of the same model are displayed in the same area. Alternatively, in the models displayed on the global model management interface, multiple models of the same version are displayed in the same area. This facilitates model management for users.

[0140] In one embodiment, the process management interface is equipped with a component (referred to as the global management component) that triggers global model management commands. After the visual algorithm development platform is started, when the user triggers the global management component, the visual algorithm development platform can display the global model management interface.

[0141] In one embodiment, the visual algorithm development platform automatically displays a global model management interface upon startup. In this case, the global model management command can be understood as the user's startup command for the visual algorithm development platform.

[0142] In one embodiment, after the visual algorithm development platform is started, when the user selects to load a visual task process, the visual algorithm development platform automatically displays the global model management interface while displaying the visual task process; or, when the user selects to create a visual task process, the visual algorithm development platform automatically displays the global model management interface.

[0143] In one embodiment, the visual task flow includes inference modules that use models for reasoning. Each inference module has a corresponding visual detection function, and each inference module indicates a detection type. For example, a user creates an inference module for the desired detection type within the visual task flow. If an inference module is used to segment objects in an image, the detection type indicated by the inference module is image segmentation, and the model deployed in the inference module is an image segmentation model; if an inference module is used to recognize characters in an image, the detection type indicated by the inference module is character recognition, and the model deployed in the inference module is a character recognition model; if an inference module is used to recognize targets in an image, the detection type indicated by the inference module is target detection, and the model deployed in the inference module is a target detection model.

[0144] A visual task flow represents a complete process of visual processing of an image. Besides the modules for the visual task itself, a visual task flow can also include other modules, which may also contain corresponding models. For example, a visual task flow might include: a module for acquiring images, a module for visually inspecting the images (i.e., an inference module), and a module for outputting the detection results. It's understandable that multiple visual task flows can exist simultaneously in the same scenario, constituting a visual task scheme. For instance, in a scenario involving the inspection of printed circuit boards (PCBs), the visual task scheme might include multiple visual task flows, each inspecting the PCB from different perspectives. One visual task flow might be used to detect whether components are correctly mounted on the PCB, another to detect soldering defects, and a third to detect anomalies in the information codes on the PCB.

[0145] The models carried in the vision task solution include models that are currently deployed in the module, as well as models that have been acquired but not yet deployed in the module (e.g., newly created models that have not yet been deployed to the module).

[0146] Understandably, if the user has not yet created or loaded a visual task workflow in the workflow management interface—that is, the model included in the visual task solution is empty—then the global model management interface displayed by the visual algorithm development platform will be empty. However, in this case, the global model management interface will still display a component for obtaining models; the user can trigger the model addition operation in subsequent embodiments by interacting with this component. See also... Figure 2 The global model management interface includes a "Create New Model" component, which allows users to create new models; an "Import Local Model" component, which allows users to import locally stored models; and an "Import Cloud Model" component, which allows users to retrieve models from the cloud. Detailed descriptions of how these three components retrieve models can be found in the subsequent embodiments regarding creating new models, importing models from local electronic devices, and importing models from the cloud.

[0147] In one embodiment, the model used in the vision task solution includes a model with adjustable parameters. That is, the model has a model structure and model parameters. The process of training the model is also the process of adjusting the model parameters.

[0148] The model in this application can be an artificial intelligence (AI) model, such as a deep learning model. Alternatively, the model in this application can also be a traditional machine learning model, such as a random forest model or a logistic regression model. In different application scenarios, the model in this application can be an object detection model, a semantic segmentation model, an instance segmentation model, or a character recognition model.

[0149] It is understandable that some useful models in existing technologies may involve a "model training" process. However, the "model training" they refer to essentially does not involve adjusting model parameters; it merely increases the mapping relationship between the features extracted by the model and the inference results. For example, the character recognition model used previously was essentially a mapping library of extracted graphic features and character recognition results. The "training process" of this model included: extracting graphic features using conventional methods such as extracting grayscale features and texture features, and then having the user label the characters corresponding to those graphic features, thus completing the "training." This training did not involve optimizing the feature extraction process; the extracted features were fixed, and only the extracted features and characters were mapped. This approach itself does not have a model with model parameters, and correspondingly, it does not adjust the model parameters.

[0150] The model in this application is trained using sample images, meaning that the model parameters are adjusted using sample images. For example, gradients are calculated using backpropagation, and model parameters are optimized using gradient descent. In other words, the global model management interface in this application allows for the processing of each stage of adjusting the model parameters.

[0151] In one embodiment, in the process management interface, each module in the visual task process can be represented in the form of a block diagram, and the data flow direction of the visual task process can be represented by arrows between the block diagrams. See also Figure 3 The interface of the visual algorithm development platform includes a task flow setup area 301 and a module setting area 302. Figure 3 In the task flow setup area 301, four visual task flows (flow 1, flow 2, flow 3, and flow 4) have been created. Currently, the modules in flow 1 are displayed, including the image acquisition module, image matching module, and image segmentation module. For any module currently displayed in task flow setup area 301, if the module is selected, its information is displayed in module settings area 302. If a model has been deployed for this module, its model information is displayed in module settings area 302. For example, model information includes at least one of the following: model source, detection type, model identifier, version number, creation time (i.e., training completion time), test results, test parameters, annotation information, label information, deployment status, and remarks. Test results include at least one of the following: precision, recall, and F1 score. Test parameters include at least one of the following: number of images, confidence score, maximum overlap rate, and intersection over union (IOU). Annotation information includes the ratio of labeled to unlabeled images. The labeling information includes the number of images with different labels. The deployment information records the inference modules to which the model is deployed. Figure 3 In the image segmentation module, the selected module displays the model information of the image segmentation model 3 deployed in this module in the module settings area 302. "XXXX" represents the specific model information.

[0152] If the model used in the vision task solution is not empty, see [link to relevant documentation]. Figure 4 , Figure 4 In the process management interface, a global model management interface 303 is displayed on top of the overall interface. This interface includes a model list display area 3031 and an information display area 3032. The model list display area 3031 displays the model identifiers (image segmentation model 1, image segmentation model 2, object detection model 1, object detection model 2, object detection model 3, and character recognition model 1) of the models used in the visual task solution. The information display area 3032 displays the information to be shown for the model currently selected in the model list display area 3031, representing the model's identifier. Figure 4 The currently selected model is Image Segmentation Model 1. The right side displays information about different versions of Image Segmentation Model 1 to be displayed (including version number, testing status (test results and test parameters), annotation status, label status, deployment status, and remarks). In other words, information display area 3032 can display multiple pieces of information to be displayed. Based on user actions, the visual algorithm development platform can determine the selected model within information display area 3032. Additionally, the source of the model can be marked in the global model management interface 303. For example... Figure 4 In the diagram, some models are labeled "cloud," indicating that the model was obtained from the cloud; some models are labeled "local," indicating that the model was imported locally from the electronic device; and unlabeled models were created in the visual algorithm development platform.

[0153] In one embodiment, the method further includes: Step 1; in response to the add model operation, obtaining a second model; Step 2: adding and displaying the model identifier of the second model and the model information of the second model in the global model management interface. This facilitates users in obtaining new models in the global model management interface, further improving model management efficiency.

[0154] In this embodiment, users can operate in the global model management interface to obtain new models, which can then be deployed in modules of the created vision task process.

[0155] In one embodiment, the model identifier of the second model is selected in the refreshed global model management interface. This facilitates user management of the second model.

[0156] For example, the global model management interface includes a unified component for retrieving models. When a user triggers this component, the visual algorithm development platform can acquire a second model. The second model can be acquired through at least one of the following methods:

[0157] One approach (referred to as Approach 1): In response to a model creation command triggered in the global model management interface, a first model is created; the first model is then trained using sample images to obtain a second model. In this way, models can be created and trained within the global model management interface to meet different user needs.

[0158] In one embodiment, the first model and the second model are models with adjustable parameters, such as deep learning models.

[0159] Since the second model is trained based on the first model, the detection types of the second model are the same as those of the first model. The second model and the first model can be understood as different versions of the same model.

[0160] like Figure 4 As shown, the global model management interface includes a "Model Creation" component. When a user triggers this "Model Creation" component, the visual algorithm development platform can create a new model.

[0161] In one embodiment, when a user triggers the "model creation" component, the visual algorithm development platform can also display a setting option for the detection type, which the user can then trigger to set the detection type of the first model to be created.

[0162] In one embodiment, when a user triggers the "model creation" component, the visual algorithm development platform can automatically create a first model with a default detection type. The default detection type can be determined based on the current visual task scheme.

[0163] The sample images used to train the first model can be images local to the electronic device or images acquired in real time.

[0164] In one embodiment, in response to receiving a model import command triggered in the global model management interface, the second model indicated by the model import command is obtained. This supports importing existing models, further improving model management efficiency.

[0165] like Figure 4 As shown, the global model management interface includes a "Model Import" component. When a user triggers this component, the visual algorithm development platform can acquire existing models. The visual algorithm development platform can import models using at least one of the following methods:

[0166] Method 2: Responding to a local model import command triggered in the global model management interface, retrieve the first model from the local machine. This supports importing models from the local machine, meeting different business needs of users.

[0167] When a user triggers a local model import command, the visual algorithm development platform can display models stored locally on the electronic device. The user can then select at least one model from the displayed options as a second model. For example, the platform can pop up a new window displaying the models stored locally on the electronic device; alternatively, it can display the models stored locally on the electronic device in a designated area of ​​the global model management interface. For each model, its information can be displayed.

[0168] Method 3: Responding to the cloud model import command triggered in the global model management interface, the first model is retrieved from the cloud. This supports importing models from the cloud, meeting different business needs of users.

[0169] When a user triggers the command to import a model from the cloud, the visual algorithm development platform can display the model stored in the cloud. The user can then select at least one model from the displayed options as a secondary model. For example, the platform can pop up a new window displaying the cloud-stored model; alternatively, it can display the cloud-stored model in a designated area of ​​the global model management interface. For each model, its information can be displayed.

[0170] For example, see Figure 5 ,exist Figure 4 Based on this, when a user triggers the "Model Import" component, the visual algorithm development platform can display import model options, including "Local Import" and "Cloud Import" options. Users can import models from local or cloud sources by triggering different options.

[0171] In one embodiment, the above method three includes the following steps: in response to receiving a cloud model import instruction triggered in the global model management interface, displaying the cloud login interface; after successful login, displaying the cloud models; and in response to receiving a second selection instruction, retrieving the first model from the displayed models. This verifies the request to retrieve a model from the cloud, improving the security of models stored in the cloud.

[0172] In this embodiment, when a user triggers a cloud-based model import command, the visual algorithm development platform displays the cloud-based login interface. See also... Figure 6 When a user triggers the cloud model import command, the visual algorithm development platform no longer displays the global model management interface, but instead displays the cloud login interface (304 error). For example, the user can enter their username and password on this login interface (304 error).

[0173] Then, after successfully logging in, see Figure 7 The visual algorithm development platform displays a new window (called the cloud model display window) 305, which displays the model information of the model stored in the cloud. Figure 7 The information displayed is the model information of each model (including image segmentation model 1, image segmentation model 2, character recognition model 1, and character recognition model 2) stored in the cloud, including: detection type (i.e., image segmentation, character recognition), version number, training error, precision, recall, training start time, and training duration. Figure 7 The "XXX" indicates the specific information displayed. Window 305 displays information in sections based on the model's detection type. That is, information about models with the same detection type is displayed in the same area, making it easier for users to find the model they need.

[0174] In one embodiment, for both methods two and three described above, the first model can be obtained by: classifying and displaying models according to model type; and in response to receiving a first selection instruction, retrieving the first model from the displayed models. This facilitates the user's selection of the first model.

[0175] For either Method 2 or Method 3, models can be categorized and displayed according to their type. For example, multiple models of the same detection type can be displayed in the same area, and multiple versions of the same model can be displayed in the same area. Alternatively, multiple models of the same version can be displayed in the same area.

[0176] In one embodiment, for either method two or method three, the model can also be displayed in other ways. For example, the models can be displayed in chronological order of their creation time, or in a random order, etc.

[0177] See Figure 8 , Figure 8 The model import process shown includes the following steps:

[0178] S801: Determine the import method.

[0179] If a local model import instruction is received, the import method is confirmed as local import; if a cloud model import instruction is received, the import method is confirmed as cloud import.

[0180] S802: If the import method is local import, select the local model.

[0181] S803: If the import method is cloud import, log in to the cloud and select the model from the cloud.

[0182] S804: Deploy the imported model to the inference module.

[0183] Based on the user's actions, the imported model is deployed to the inference module. For example, the batch deployment method described in subsequent embodiments can be used.

[0184] In one embodiment, the step of training the first model includes: in response to receiving a model training instruction triggered in the global model management interface, training the first model using sample images to obtain a second model. Thus, triggering model training within the global model management interface supports this process, further improving model management efficiency.

[0185] For example, see Figure 4 The global model management interface 303 includes a "Label Training" component. After the first model is created, it can be displayed on the global model management interface (including the model's identifier and information to be displayed). Correspondingly, if the model identifier of the first model is currently selected, when the user triggers the "Label Training" component, the visual algorithm development platform can receive a model training instruction for the first model. The visual algorithm development platform then trains the first model. For example, the visual algorithm development platform can display the training settings interface as shown in subsequent embodiments, and use this training settings interface to determine sample images to train the first model in conjunction with the sample images, thereby obtaining the second model.

[0186] In one embodiment, a first model is trained using sample images by: displaying a training settings interface; wherein the training settings interface displays: sample images and a label setting area; obtaining labels that annotate the sample images in the label setting area; and training the first model using the sample images and the obtained labels to obtain a second model.

[0187] In this embodiment of the application, when training of the first model is triggered, the visual algorithm development platform can display a training settings interface. For example, see... Figure 9 , Figure 9 The training settings interface shown has a thumbnail display area 901 on the left, used to display thumbnails of sample images; a large image display area 902 in the middle, used to display the currently selected sample image; and a label setting area 903 on the right. Accordingly, users can operate within the image displayed in the large image display area 902 to select image content and enter labels for the selected image content in the label setting area 903, where labels are entered. Figure 9 In the image, the diagonal rectangle represents both the thumbnail and the full-size image.

[0188] Regarding the above-described embodiments for training the model, if the visual algorithm development platform itself has the capability to train the model, it can directly use the sample images to train the model after acquiring them. Alternatively, after acquiring the sample images, the visual algorithm development platform can export them to a platform used for model training (referred to as the training platform). Correspondingly, after the training platform completes model training, the trained model can be imported back into the visual algorithm development platform.

[0189] In one embodiment, the user can also set training parameters in the training settings interface, such as the learning rate and batch size.

[0190] In one embodiment, images used for training the first model are determined from real-time acquired images and used as sample images. This eliminates the need for the user to save the images acquired in real-time by the image acquisition device and import them into the model for training, further improving model management efficiency.

[0191] In this embodiment of the application, in a scenario where a visual task process is used to detect images acquired in real time, the visual algorithm development platform can communicate with the image acquisition device to obtain the images acquired in real time by the image acquisition device. Accordingly, the visual algorithm development platform can train a first model based on the images acquired in real time.

[0192] In one embodiment, the sample image is determined by: in response to receiving a single image addition instruction, determining the currently acquired single image as the sample image; or, in response to receiving a batch image addition instruction, determining multiple acquired images as sample images.

[0193] In this embodiment, a single image can be marked as a sample image using a single image addition command, while multiple images can be marked as sample images using a batch image addition command, thus meeting different user needs. For example, in response to receiving a batch image addition command, multiple consecutively acquired images are identified as sample images.

[0194] For example, see Figure 9 , Figure 9 The training settings interface, shown in the upper left corner, displays a "Single Run" component for triggering single image addition commands and a "Continuous Run" component for triggering batch image addition commands. In other words, the sample images used are determined through the training settings interface before training the model.

[0195] Alternatively, a separate component for setting sample images can be set in the global model management interface. When the user triggers this component, the visual algorithm development platform can display an interface for setting sample images, and then select the sample image to use based on the user's operation in this interface.

[0196] In one embodiment, the visual algorithm development platform can also display the training status of the model in the global model management interface to help users understand the current training status of the model.

[0197] Since the visual task process may include multiple inference modules, and correspondingly, in response to user operations, the visual algorithm development platform may train multiple models. To facilitate user management of the models, the training status of the models can be displayed.

[0198] In one approach, the visual algorithm development platform displays the training status (including detailed training status as described below) of all models (including those that have not yet completed training and those that have completed training). Alternatively, the visual algorithm development platform may also display detailed training status based on user actions.

[0199] In one embodiment, the training state includes at least one of the following: the training progress of the model, the training result indicating successful or unsuccessful training, and the reason for training failure.

[0200] In one embodiment, displaying the training status of a model in the global model management interface includes: displaying the number of models currently being trained, the number of models waiting to be trained, and the number of models that have completed training in the global model management interface; when a command to display details of incomplete training is received from the global model management interface, displaying the detailed training status of the models currently being trained and the detailed training status of the models waiting to be trained; and when a command to display details of completed training is received from the global model management interface, displaying the detailed training status of the models that have completed training. This allows users to easily understand the specific training details of the model.

[0201] For example, a visual algorithm development platform can display detailed training status in a region of the global model management interface, or it can display a new interface and show detailed training status in that interface.

[0202] The detailed training status of an incomplete training model can include at least one of the following: model identifier, detection type, version number, and training progress. Additionally, for incomplete training models, a "Stop Training" component can be displayed; when the user triggers this component, the visual algorithm development platform cancels training for that model. Furthermore, for models that have failed to train, a "Retrain" component can be displayed; when the user triggers this component, the visual algorithm development platform retrains the model.

[0203] For example, see Figure 10 ,exist Figure 4 Based on this, the global model management interface includes a number display area 3033, which displays the number of models that have not yet completed training and the number of models that have completed training. Figure 10 The "Task in progress 3 / 5" displayed means that there are currently 5 model training tasks, and 3 are still incomplete, that is, 2 models have been trained.

[0204] Correspondingly, when the user triggers the display area 3033, the visual algorithm development platform can display a window 306, showing both "In Progress" and "Completed" options. When the user triggers the "In Progress" option, the visual algorithm development platform displays the detailed training status of the currently training model and the detailed training status of models awaiting training in window 306; when the user triggers the "Completed" option, the visual algorithm development platform displays the detailed training status of the completed models in window 306. Figure 10 In the text, "3 in progress" indicates 3 models that have not yet completed training, and "2 completed" indicates 2 models that have completed training. Figure 10 When the user triggers the "In Progress" option, it displays the detailed training status of the three models that have not yet completed training (Image Segmentation Model 1, Object Detection Model 2, and Character Recognition Model 1). Image Segmentation Model 1 is the model that is currently being trained, displaying its training progress (i.e., the remaining training time). Object Detection Model 2 is the model waiting to be trained (i.e., the task is in the queue), and Character Recognition Model 1 is the model that has failed to train. When the user triggers the "Retry" component, the visual algorithm development platform retrains Character Recognition Model 1. Figure 10 In the middle, for each model, there is also an "×" component, which indicates the "End Training" component.

[0205] The detailed training status of a trained model can include at least one of the following: the model identifier, detection type, version number, training progress, and training duration.

[0206] In one embodiment, in response to a detailed display command for the training status of any model triggered in the global model management interface, the training details of the model indicated by the detailed display command are displayed. The training details include at least one of the following: basic training information, training parameters, and loss curve.

[0207] against Figure 10 For each model displayed in window 306, the visual algorithm development platform can also display the model's training details upon receiving a user's instruction to display details for that model. See also... Figure 11 The visual algorithm development platform can display a detailed information interface 307, which shows basic information (i.e., basic training information) and training parameters. Basic information includes: version number, task identifier, training status (success / failure), training duration, number of training images, and training image size. Training parameters include: training version (e.g., Graphics Processing Unit (GPU) or Tensor Processing Unit (TPU)), model capabilities, training type (e.g., single image segmentation / image object detection), iteration rounds, defect type, batch size, target platform (the platform on which the trained model is applied), sampling coefficients, and data augmentation strategies. The detailed information interface 307 can also display the model's loss curve. Figure 11 (Not shown in the image).

[0208] In one embodiment, the global model management interface also enables the export and saving path modification of any model. For example, the global model management interface includes a "Model Export" component. When the user triggers this component, the interface displays export options for the currently selected model, such as displaying a file explorer page and exporting the model to the local machine based on the user's actions on that page.

[0209] In one embodiment, in response to receiving a path modification instruction for the fourth model triggered in the global model management interface, the fourth model is saved according to the save path indicated by the path modification instruction. For example, the global model management interface has a "Change Path" component. When the user triggers the "Change Path" component, the global model management interface can change the save path of the model according to the user's operation.

[0210] In one embodiment, the vision task module includes an inference module that uses a model for inference. The method further includes: batch deploying a third model to at least one inference module included in the vision task solution. This supports batch deployment of models, meeting users' batch deployment needs and improving model management efficiency.

[0211] In this embodiment, the third model can be any model displayed in the global model management interface. Based on the user's selection, any model displayed in the global model management interface can be deployed to the inference module in batches.

[0212] In one embodiment, the detection type of at least one inference module is consistent with the detection type of the third model.

[0213] In one embodiment, the inference modules that need to be deployed in batches for the third model are determined by the following steps: displaying inference modules in the visual task scheme that have the same detection type as the third model; and in response to receiving a third selection instruction, determining the inference modules that need to be deployed in batches for the third model from the displayed inference modules.

[0214] In this way, the inference modules that match the detection type of the third model (referred to as the third detection type) are selected and displayed, making it easier for users to select inference modules and improving the efficiency of model deployment.

[0215] In one embodiment, the global model management interface also includes a "batch deployment" component, for example, see [link to relevant documentation]. Figure 4 , Figure 4 The black triangle button is displayed in the "Deployment Status" column. Correspondingly, when the user triggers this button, the visual algorithm development platform displays a batch deployment interface for the currently selected model (i.e., the third model). See also... Figure 12 The batch deployment interface 308 includes a module selection list area 3081 and a selected list area 3082. The module selection list area 3081 is used to display the module identifier of the inference module indicating the third detection type in the created visual task flow, as well as the flow identifier of the visual task flow to which each inference module belongs. The selected list area 3082 is used to display the module identifier of the inference module selected by the user through the command.

[0216] Figure 12 In the module selection list area 3081, the module identifiers of the third detection type inference modules (including module 1, module 2, module 5, and module 6) in the created visual task flow are displayed, along with the flow to which each module belongs (i.e., the visual task flow, including flow 1, flow 3, and flow 5). Users can select the modules (including module 2 and module 6) that need to be deployed in batches according to their needs; correspondingly, the module identifiers of module 2 and module 6 are displayed in the selected list area 3082. Additionally, users can also deselect the modules using the deselection component displayed in the selected list area 3082. Figure 12 (The "×" component) can be used to deselect the corresponding module.

[0217] As seen in the above embodiments, integrating existing model information display, annotation and training, training status display, and deployment functions into a single page (i.e., the global model management interface) satisfies users' full-process needs in model processing, improving model management efficiency and consequently, the overall efficiency of building the vision task workflow. Furthermore, regarding model information display, all models in the vision task workflow are automatically organized and summarized according to detection type and version, compatible with cloud / local external model configurations, and clearly displaying various model status information. In terms of training progress display, all training progress statuses are integrated with model management functions for unified management, collecting all training task progress information to form a unified information panel that clearly displays the required information.

[0218] See Figure 13 , Figure 13 The model management method shown includes the following steps:

[0219] S1301: The user opens the visual algorithm development platform.

[0220] S1302: When a trigger command is received for the global management component, the global model management interface is displayed.

[0221] S1303: In response to receiving a vision task flow creation / loading instruction, display the global model management interface.

[0222] S1304: Displays model information of the model represented by the model identifier selected by the user in the global model management interface.

[0223] S1305: In response to receiving a model creation command triggered in the global model management interface, create a model.

[0224] S1306: Train the model using sample images to obtain the trained model, and return to step S1304.

[0225] S1307: Displays the training status of the model in the global model management interface.

[0226] S1308: In response to receiving a model import command triggered in the global model management interface, obtain the model indicated by the model import command and return to the execution step S1304.

[0227] In steps S1306 and S1308 above, after obtaining the model, the process returns to step S1304, that is, a model identifier for displaying the model is added to the global model management interface to refresh the global model management interface.

[0228] S1309: Determine whether a model training instruction for the model displayed in the global model management interface has been received. If yes, execute step S1306; otherwise, execute steps S1310 and S1311.

[0229] S1310: In response to receiving a batch deployment instruction for a model displayed in the global model management interface, deploy the model to the inference module indicated by the batch deployment instruction.

[0230] S1311: Save the model displayed in the global model management interface to the local machine according to the user's instructions.

[0231] Based on the same inventive concept, this application also provides a model management device applied to a visual algorithm development platform, wherein the visual algorithm development platform includes a visual task module for building visual task schemes; see also Figure 14 The device includes:

[0232] Instruction receiving unit 1401 is used to receive global model management instructions;

[0233] Display unit 1402 is used to display a global model management interface; wherein, the global model management interface is used to display: model information and deployment status of the models carried in the vision task scheme.

[0234] In an optional embodiment, the apparatus further includes:

[0235] The model acquisition unit is used to acquire a second model in response to the add model operation;

[0236] The refresh module is used to add and display the model identifier of the second model and the model information of the second model in the global model management interface.

[0237] In an optional embodiment, the model acquisition unit is specifically configured to, in response to receiving a model creation instruction triggered in the global model management interface, create a first model; and train the first model using sample images to obtain a second model;

[0238] or,

[0239] In response to receiving a model import command triggered in the global model management interface, the second model indicated by the model import command is obtained.

[0240] In an optional embodiment, the model acquisition unit is specifically used to acquire a first model from the local source in response to receiving a local model import instruction triggered in the global model management interface;

[0241] or,

[0242] In response to receiving a cloud model import command triggered in the global model management interface, the first model is retrieved from the cloud.

[0243] In an optional embodiment, the model acquisition unit is specifically used to acquire the first model in the following manner:

[0244] Models are categorized and displayed according to their type.

[0245] In response to receiving the first selection instruction, the first model is retrieved from the displayed models.

[0246] In an optional embodiment, the model acquisition unit is specifically used to display the cloud login interface in response to receiving a cloud model import instruction triggered in the global model management interface;

[0247] After successful login, the model in the cloud will be displayed;

[0248] In response to receiving a second selection instruction, the first model is retrieved from the displayed models.

[0249] In an optional embodiment, the model acquisition unit is specifically used to train the first model using sample images to obtain a second model in response to receiving a model training instruction triggered in the global model management interface.

[0250] In an optional embodiment, the apparatus further includes: a sample image determination unit, configured to determine, from real-time acquired images, images for training the first model as sample images before training the first model using the sample images to obtain the second model.

[0251] In an optional embodiment, the sample image determination unit is specifically configured to determine the currently acquired single image as a sample image in response to receiving a single image addition instruction;

[0252] or,

[0253] In response to receiving a batch image addition instruction, the acquired multiple images are identified as sample images.

[0254] In an optional embodiment, the model acquisition unit is specifically used to display a training settings interface; wherein the training settings interface displays: sample images and a label setting area;

[0255] Obtain the labels that are used to annotate the sample image in the label setting area;

[0256] The first model is trained using sample images and acquired labels to obtain the second model.

[0257] In an optional embodiment, the apparatus further includes:

[0258] The training status display unit is used to display the training status of the model in the global model management interface.

[0259] In an optional embodiment, the training state includes at least one of the following:

[0260] Model training progress;

[0261] This indicates the training result, whether the training was successful or unsuccessful.

[0262] Reasons for training failure.

[0263] In an optional embodiment, the device further includes: a training details display unit, configured to display the training details of the model indicated by the details display instruction in response to a details display instruction for the training status of any model triggered in the global model management interface; wherein the training details include at least one of the following: basic training information, training parameters, and loss curve.

[0264] In an optional embodiment, the vision task module includes an inference module that uses a model for inference;

[0265] The device further includes a batch deployment unit for batch deploying the third model to at least one inference module included in the vision task scheme.

[0266] In an optional embodiment, the detection type of the at least one inference module is consistent with the detection type of the third model.

[0267] In an optional embodiment, the apparatus further includes: an inference module determination unit, configured to display, before deploying the third model in batches to at least one inference module included in the vision task scheme, an inference module in the vision task scheme that is consistent with the detection type of the third model;

[0268] In response to receiving a third selection instruction, determine from the displayed inference modules the inference modules for which the third model needs to be deployed in batches.

[0269] In an optional embodiment, the model information displayed in the global model management interface includes at least one of the following: the model's source, version number, detection type, testing status, and annotation status;

[0270] And / or,

[0271] In the global model management interface, models are displayed according to model type;

[0272] And / or,

[0273] The device further includes:

[0274] The saving unit is used to respond to a path modification instruction for the fourth model triggered in the global model management interface, and save the fourth model according to the saving path indicated by the path modification instruction.

[0275] In one alternative embodiment, the model carried in the vision task scheme includes a model with adjustable parameters.

[0276] This application also provides an electronic device, such as... Figure 15 As shown, it includes:

[0277] Memory 1501 is used to store computer programs;

[0278] When processor 1502 executes the program stored in memory 1501, it performs the following steps:

[0279] Receive global model management instructions;

[0280] The global model management interface is displayed; wherein, the global model management interface is used to display: the model information and deployment status of the models carried in the vision task solution.

[0281] Furthermore, the aforementioned electronic device may also include a communication bus and / or a communication interface, with the processor 1502, the communication interface, and the memory 1501 communicating with each other via the communication bus.

[0282] The communication bus mentioned in the above electronic devices can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This communication bus can be divided into address bus, data bus, control bus, etc. For ease of illustration, only one thick line is used to represent it in the diagram, but this does not mean that there is only one bus or one type of bus.

[0283] The communication interface is used for communication between the aforementioned electronic devices and other devices.

[0284] The memory may include random access memory (RAM) or non-volatile memory (NVM), such as at least one disk storage device. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor.

[0285] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0286] In another embodiment provided in this application, a computer-readable storage medium is also provided, which stores a computer program that, when executed by a processor, implements the steps of any of the above-described model management methods.

[0287] In another embodiment provided in this application, a computer program product containing instructions is also provided, which, when run on a computer, causes the computer to execute any of the model management methods described above.

[0288] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a solid-state drive (SSD), etc.

[0289] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0290] The various embodiments in this specification are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments of apparatus, devices, computer-readable storage media, and computer program products are basically similar to the method embodiments, and therefore the descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0291] The above description is merely a preferred embodiment of this application and is not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application are included within the scope of protection of this application.

Claims

1. A model management method, characterized in that, The method is applied to a visual algorithm development platform, which includes a visual task module for building visual task solutions; the method includes: Receive global model management instructions; Display the global model management interface; wherein, the global model management interface is used to display: model information and deployment status of the models carried in the vision task solution.

2. The method according to claim 1, characterized in that, The method further includes: In response to the add model operation, retrieve the second model; The global model management interface now displays the model identifier of the second model, as well as the model information of the second model.

3. The method according to claim 2, characterized in that, The step of obtaining the second model in response to the add model operation includes: In response to receiving a model creation command triggered in the global model management interface, a first model is created; the first model is trained using sample images to obtain a second model; or, In response to receiving a model import command triggered in the global model management interface, the second model indicated by the model import command is obtained.

4. The method according to claim 3, characterized in that, The step of responding to receiving a model import command triggered in the global model management interface and obtaining the second model indicated by the model import command includes: In response to receiving a local model import command triggered in the global model management interface, the first model is retrieved from the local machine; or, In response to receiving a cloud model import command triggered in the global model management interface, the first model is retrieved from the cloud.

5. The method according to claim 4, characterized in that, The first model is obtained through the following methods: Models are categorized and displayed according to their type. In response to receiving the first selection instruction, the first model is retrieved from the displayed models.

6. The method according to claim 4, characterized in that, The step of responding to receiving a cloud model import command triggered in the global model management interface and obtaining the first model from the cloud includes: In response to receiving a cloud model import command triggered in the global model management interface, the cloud login interface is displayed; After successful login, the model in the cloud will be displayed; In response to receiving a second selection instruction, the first model is retrieved from the displayed models.

7. The method according to claim 3, characterized in that, The step of training the first model using sample images to obtain the second model includes: In response to receiving a model training instruction triggered in the global model management interface, the first model is trained using sample images to obtain a second model.

8. The method according to claim 7, characterized in that, Before training the first model using sample images to obtain the second model, the method further includes: Images used to train the first model are determined from the real-time acquired images and used as sample images.

9. The method according to claim 8, characterized in that, The step of determining the images used to train the first model from the real-time acquired images, as sample images, includes: In response to receiving a single image addition instruction, the currently acquired single image is determined as a sample image; or, In response to receiving a batch image addition instruction, the acquired multiple images are identified as sample images.

10. The method according to claim 3, characterized in that, The step of training the first model using sample images to obtain the second model includes: The training settings interface is displayed; wherein, the training settings interface displays: sample images and label setting areas; Obtain the labels that are used to annotate the sample image in the label setting area; The first model is trained using sample images and acquired labels to obtain the second model.

11. The method according to claim 1, characterized in that, The method further includes: The training status of the model is displayed in the global model management interface.

12. The method according to claim 11, characterized in that, The training state includes at least one of the following: Model training progress; This indicates the training result, whether the training was successful or unsuccessful. Reasons for training failure.

13. The method according to claim 11, characterized in that, The method further includes: In response to a detailed display instruction for the training status of any model triggered in the global model management interface, the training details of the model indicated by the detailed display instruction are displayed; wherein, the training details include at least one of the following: basic training information, training parameters, and loss curve.

14. The method according to any one of claims 1 to 13, characterized in that, The vision task module includes a reasoning module that uses a model for reasoning; the method further includes: The third model is deployed in batches to at least one inference module included in the vision task scheme.

15. The method according to claim 14, characterized in that, The detection type of at least one inference module is consistent with the detection type of the third model.

16. The method according to claim 15, characterized in that, Before deploying the third model in batches to at least one inference module included in the vision task scheme, the method further includes: Display the inference module in the vision task scheme that has the same detection type as the third model; In response to receiving a third selection instruction, determine from the displayed inference modules the inference modules for which the third model needs to be deployed in batches.

17. The method according to claim 1, characterized in that, The model information displayed in the global model management interface includes at least one of the following: model source, version number, detection type, testing status, and annotation status; And / or, In the global model management interface, models are displayed according to model type; And / or, The method further includes: In response to receiving a path modification instruction for the fourth model triggered in the global model management interface, the fourth model is saved according to the save path indicated by the path modification instruction.

18. The method according to any one of claims 1 to 13, characterized in that, The vision task solution incorporates models with adjustable parameters.

19. A model management device, characterized in that, An application is made to a visual algorithm development platform, the visual algorithm development platform including a visual task module for building visual task solutions; the device includes: The instruction receiving unit is used to receive global model management instructions; The display unit is used to display the global model management interface; wherein, the global model management interface is used to display: the model information and deployment status of the models carried in the vision task solution.

20. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, when executing a program stored in memory, implements the method described in any one of claims 1-18.

21. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method described in any one of claims 1-18.

Images