Model training method and system, client, transfer terminal and server

Abstract

The embodiment of the invention discloses a model training method and system, a client, a transfer terminal and a server, which are used for saving human resources. The method comprises the steps of collecting a training image; the training image is sent to a transfer end, so that the transfer end marks the training image, and the marked training image is sent to a server; obtaining a first recognition model, wherein the first recognition model is an image recognition model obtained after the server uses the labeled training image to train an initial recognition model; obtaining a verification image, wherein the verification image is a marked image; and loading the first identification model through a first process, and verifying the first identification model by using the verification image.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of artificial intelligence technology, and in particular to a method, system, client, transfer terminal and server for training a model. Background Art

[0002] With the continuous development of science and technology, training models is a key technology in computer vision, image processing and artificial intelligence. The main purpose of training models is to train models that can accurately identify and classify objects through a large amount of sample data. These models usually need to be distributed and loaded through network communication means so that they can be used on multiple devices at the same time. In addition, the automatic collection of defect images is also an important technology, which can help us quickly collect and annotate a large number of defect samples, thereby improving the training efficiency and accuracy of the model.

[0003] In the prior art, a server is usually connected to multiple clients, and each client performs the tasks of image acquisition and image annotation. If the image annotation task is performed manually, each client needs to be equipped with staff, so a lot of human resources are needed for image annotation. Summary of the invention

[0004] The embodiments of the present application provide a method, system, client, transfer terminal and server for training a model, which can save human resources.

[0005] A first aspect of an embodiment of the present application provides a method for training a model, which is applied to a client, wherein the client is connected to a transfer terminal and a server via a local area network, and the method includes:

[0006] Collect training images;

[0007] Sending the training image to a transfer terminal so that the transfer terminal annotates the training image and sends the annotated training image to a server;

[0008] Acquire a first recognition model, where the first recognition model is an image recognition model obtained by the server training an initial recognition model using the labeled training image;

[0009] Acquire a verification image, where the verification image is a labeled image;

[0010] The first recognition model is loaded through a first process, and the first recognition model is verified using the verification image.

[0011] Optionally, after verifying the first recognition model using the verification image, the method further includes:

[0012] Acquire a second recognition model, where the second recognition model is an image recognition model obtained by the server training the first recognition model or the initial recognition model using the labeled training image;

[0013] Loading the second recognition model through a second process, and closing the first process after the second recognition model is loaded;

[0014] The second recognition model is verified using the verification image.

[0015] Optionally, obtaining the first recognition model includes:

[0016] Receiving a download notification sent by the server;

[0017] The first recognition model is downloaded from the server according to the download notification.

[0018] Optionally, obtaining a verification image includes:

[0019] Acquire the labeled training image from the transfer end; generate a verification image based on the labeled training image;

[0020] Or, obtain a verification image from a third party.

[0021] A second aspect of an embodiment of the present application provides a method for training a model, which is applied to a transfer terminal, wherein the transfer terminal is connected to a client and a server via a local area network, and the method includes:

[0022] Receiving a training image sent by the client;

[0023] Annotating the training image to obtain the annotated training image;

[0024] The labeled training image is sent to the server, so that the server trains the image recognition model by using the labeled training image.

[0025] Optionally, after labeling the training image to obtain the labeled training image, the method further includes:

[0026] The labeled training image is sent to the client, so that the client verifies the image recognition model using the labeled training image.

[0027] A third aspect of an embodiment of the present application provides a method for training a model, which is applied to a server, wherein the server is connected to a client and a transfer terminal via a local area network, and the method includes:

[0028] Receiving the annotated training image sent by the transfer end;

[0029] The initial recognition model is trained using the training image to obtain a first recognition model.

[0030] Optionally, after the initial recognition model is trained using the training image to obtain the first recognition model, the method further includes:

[0031] The first recognition model or the initial recognition model is trained using the labeled training image to obtain a second recognition model.

[0032] Optionally, after the initial recognition model is trained using the training image to obtain the first recognition model, the method further includes:

[0033] A download notification is sent to the client, so that the client downloads the first recognition model according to the download notification.

[0034] A fourth aspect of the embodiments of the present application provides a client, wherein the client is connected to a transfer terminal and a server via a local area network, and the client includes:

[0035] An acquisition unit, used for acquiring training images;

[0036] A first sending unit, configured to send the training image to a transfer end, so that the transfer end annotates the training image and sends the annotated training image to a server;

[0037] A first acquisition unit is used to acquire a first recognition model, where the first recognition model is an image recognition model obtained by the server training an initial recognition model using the labeled training image;

[0038] A second acquisition unit is used to acquire a verification image, where the verification image is a marked image;

[0039] The first verification unit is used to load the first recognition model through a first process and verify the first recognition model using the verification image.

[0040] Optionally, the client further includes:

[0041] A third acquisition unit is used to acquire a second recognition model, where the second recognition model is an image recognition model obtained by the server training the first recognition model or the initial recognition model using the annotated training image;

[0042] A loading unit, configured to load the second recognition model through a second process, and close the first process after the second recognition model is loaded;

[0043] The second verification unit is used to verify the second recognition model using the verification image.

[0044] Optionally, the first acquiring unit is specifically used for:

[0045] Receiving a download notification sent by the server;

[0046] The first recognition model is downloaded from the server according to the download notification.

[0047] Optionally, the second acquiring unit is specifically configured to:

[0048] Acquire the labeled training image from the transfer end; generate a verification image based on the labeled training image;

[0049] Or, obtain a verification image from a third party.

[0050] A fifth aspect of the embodiments of the present application provides a client, including:

[0051] Processor, memory, input-output unit, and bus;

[0052] The processor is connected to the memory, the input and output unit, and the bus;

[0053] A program is stored in the memory, and the processor calls the program to execute the method in the first aspect and any possible implementation manner of the first aspect.

[0054] A sixth aspect of the embodiment of the present application provides a transfer terminal, wherein the transfer terminal is connected to a client and a server via a local area network, and the transfer terminal includes:

[0055] A receiving unit, configured to receive a training image sent by the client;

[0056] A labeling unit, used for labeling the training image to obtain the labeled training image;

[0057] The first sending unit is used to send the labeled training image to the server, so that the server trains the image recognition model by using the labeled training image.

[0058] Optionally, the transfer end further includes:

[0059] The second sending unit is used to send the annotated training image to the client, so that the client verifies the image recognition model by using the annotated training image.

[0060] A seventh aspect of the embodiment of the present application provides a transfer terminal, including:

[0061] Processor, memory, input-output unit, and bus;

[0062] The processor is connected to the memory, the input and output unit, and the bus;

[0063] A program is stored in the memory, and the processor calls the program to execute the method in the second aspect and any possible implementation manner of the second aspect.

[0064] An eighth aspect of the present application provides a server, wherein the server is connected to a client and a transfer terminal via a local area network, and the server includes:

[0065] A receiving unit, used for receiving the annotated training image sent by the transfer end;

[0066] The first training unit is used to train the initial recognition model using the training image to obtain a first recognition model.

[0067] Optionally, the server further includes:

[0068] The second training unit is used to train the first recognition model or the initial recognition model by using the labeled training image to obtain a second recognition model.

[0069] Optionally, the server further includes:

[0070] A notification unit is used to send a download notification to the client, so that the client downloads the first recognition model according to the download notification.

[0071] A ninth aspect of an embodiment of the present application provides a server, including:

[0072] Processor, memory, input-output unit, and bus;

[0073] The processor is connected to the memory, the input and output unit, and the bus;

[0074] A program is stored in the memory, and the processor calls the program to execute the method in the third aspect and any possible implementation manner of the third aspect.

[0075] A tenth aspect of the embodiment of the present application provides a system for training a model, including at least one client, a transfer terminal and a server;

[0076] The client is connected to the transfer terminal and the server via a local area network;

[0077] The client executes the method in the first aspect and any possible implementation of the first aspect, the transfer end executes the method in the second aspect and any possible implementation of the second aspect, and the server executes the method in the third aspect and any possible implementation of the third aspect.

[0078] The eleventh aspect of the embodiments of the present application provides a computer-readable storage medium, on which a program is stored. When the program is executed on a computer, the computer executes the method in the first aspect and any possible implementation of the first aspect, the method in the second aspect and any possible implementation of the second aspect, and the method in the third aspect and any possible implementation of the third aspect.

[0079] It can be seen from the above technical solutions that the embodiments of the present application have the following advantages:

[0080] In the embodiment of the present application, the image annotation task is centralized at the transfer end, and the transfer end uniformly annotates all training images, and the client no longer needs to annotate images. If the image annotation task is performed by manual annotation, it is only necessary to arrange personnel at the transfer end to annotate images. Compared with the prior art of arranging personnel at each client to annotate images, the number of personnel handling image annotation work can be reduced, saving a lot of labor costs. At the same time, since the client, transfer end and server are connected via a local area network, the rate at which the server distributes image recognition models to each client can be increased, thereby improving the efficiency of training the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0081] Figure 1 A flowchart of an embodiment of a method for training a model applied to a client in an embodiment of the present application;

[0082] Figure 2 A flowchart of another embodiment of a method for training a model applied to a client in an embodiment of the present application;

[0083] Figure 3 A flowchart of an embodiment of a method for training a model applied to a transfer end in an embodiment of the present application;

[0084] Figure 4 A flowchart of another embodiment of a method for training a model applied to a transfer end in an embodiment of the present application;

[0085] Figure 5 A flowchart of an embodiment of a method for training a model applied to a server in an embodiment of the present application;

[0086] Figure 6A flowchart of another embodiment of a method for training a model applied to a server in an embodiment of the present application;

[0087] Figure 7 This is a schematic diagram of the structure of an embodiment of a client in the embodiments of the present application;

[0088] Figure 8 This is a schematic diagram of the structure of another embodiment of the client in the embodiment of the present application;

[0089] Fig. 9 This is a schematic diagram of the structure of an embodiment of a transfer end in the embodiments of the present application;

[0090] Fig.10 It is a structural schematic diagram of another embodiment of the transfer end in the embodiment of the present application;

[0091] Fig.11 This is a schematic diagram of the structure of an embodiment of a server in the embodiments of the present application;

[0092] Fig.12 This is a structural diagram of another embodiment of the server in the embodiment of the present application;

[0093] Fig.13 This is a structural diagram of an embodiment of a system for training a model in an embodiment of the present application. DETAILED DESCRIPTION

[0094] The embodiments of the present application provide a method, system, client, transfer terminal and server for training a model, which are used to save human resources.

[0095] The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0096] See also Figure 1 In one embodiment of the present application, a method for training a model applied to a client includes:

[0097] 101. Collect training images;

[0098] In actual applications, the client is connected to the transfer end and the server through a local area network. The client first obtains training images in real time through the connected image acquisition devices, such as cameras, sensors, etc. These images will be used as training data for the subsequent model training process. The acquired images may cover a variety of scenes, objects, and conditions to ensure that the model can cope with a variety of recognition tasks. The acquired training images are temporarily stored in the client's local storage device in preparation for subsequent operations.

[0099] 102. Sending the training image to the transfer end;

[0100] After the training images are collected, the client sends them to the transfer end through the local area network. The transfer end is responsible for annotating the images, which means it identifies key features or target objects and labels on the images. The annotation process can be automated or manual, depending on the complexity of the annotation task. The annotated images carry rich semantic information, allowing the data to be used for efficient model training. After the annotation is completed, the transfer end sends these training images with annotated information to the server for model training.

[0101] 103. Obtain a first recognition model;

[0102] After receiving the annotated training images, the server will use this data to train the initial recognition model. The training process includes multiple iterations, the purpose of which is to optimize the parameters of the model so that it can more accurately recognize and classify objects in the image. After the training is completed, the server generates a new image recognition model, namely the first recognition model. This model has been preliminarily verified on the server side to ensure that its performance meets certain standards. Subsequently, the client downloads this trained first recognition model from the server through the local area network and stores it in the local system.

[0103] 104. Obtain a verification image;

[0104] After downloading the first recognition model, the client also needs to obtain a set of verification images, which are labeled image data. These images will be used to evaluate the performance of the first recognition model and help verify the recognition accuracy and stability of the model.

[0105] Furthermore, there are multiple ways for these clients to obtain the verification image. Some possible ways of obtaining the image are described below:

[0106] 1. The client obtains the labeled training images from the transfer end through the local area network, and then selects images from the training images to form verification images.

[0107] 2. The client obtains annotated images from a third party, and then selects images from these annotated images to form verification images.

[0108] 3. The client obtains the labeled training images from the transfer end and the labeled images from the third party, and then selects images from these two sets of images as verification images.

[0109] The client can choose multiple ways to obtain verification images, which enriches the ways to obtain verification images.

[0110] 105. Load a first recognition model through a first process, and verify the first recognition model using a verification image.

[0111] The client loads the first recognition model just obtained through the first process, and verifies the model using the verification image obtained previously. During the verification process, the client compares the model's prediction results with the actual annotation information of the verification image to evaluate the model's recognition accuracy and performance. This step is critical because it determines whether the model can be put into use in a real environment or whether it needs further optimization. If the verification results show that the model performs as expected, it can be prepared to be deployed in a production environment.

[0112] In this embodiment, the image annotation task is centralized at the transfer end, which uniformly annotates all training images. The client no longer needs to annotate images. If the image annotation task is performed manually, only personnel need to be arranged at the transfer end to annotate images. Compared with the prior art of arranging personnel at each client to annotate images, the number of personnel handling image annotation can be reduced, saving a lot of labor costs. At the same time, since the client, transfer end and server are connected via a local area network, the rate at which the server distributes image recognition models to each client can be increased, thereby improving the efficiency of training the model.

[0113] See also Figure 2 Another embodiment of the method for training a model applied to a client in the embodiment of the present application includes:

[0114] 201. Collect training images;

[0115] 202. Sending a training image to a transfer terminal;

[0116] In this embodiment, steps 201 to 202 are similar to steps 101 to 102 in the aforementioned embodiment and are not described again here.

[0117] 203. Receive a download notification sent by the server;

[0118] The client receives a download notification from the server via the local area network. The notification usually contains information that the first recognition model is ready and instructs the client to download the model. The content of the notification may include the model version number, file path, and other relevant download instructions. This step ensures that the client can obtain the latest model in a timely manner and prepare for the subsequent download and loading process.

[0119] 204. Downloading a first recognition model from a server according to the download notification;

[0120] After receiving the download notification, the client downloads the first recognition model from the server through the local area network according to the download path and related information provided in the notification. During the download process, the client will store the model file in the local storage to prepare for subsequent loading and verification. After the download is complete, the client will have the latest recognition model and can proceed to the next loading and verification steps.

[0121] 205. Obtain a verification image;

[0122] 206. Loading a first recognition model through a first process, and verifying the first recognition model using a verification image;

[0123] In the embodiment of the present application, steps 205 to 206 are similar to steps 104 to 105 in the aforementioned embodiment and are not described again here.

[0124] 207. Obtain a second recognition model;

[0125] The client obtains the second recognition model from the server through the local area network. This model is generated by the server after further training the first recognition model or the initial recognition model using the previously annotated training images. Compared with the first recognition model, the second recognition model has undergone more training and optimization, so it is likely to be more outstanding in image recognition accuracy and performance. The client downloads and stores this newly generated second recognition model locally. It should be noted that the specific process of the client obtaining the second model is similar to the aforementioned steps 203 to 204, and will not be repeated here.

[0126] 208. Load the second recognition model through the second process, and when the second recognition model is loaded, close the first process;

[0127] The client starts a second process to load the second recognition model just obtained. The loading process may include reading the model file into memory and preparing it for use in the actual application. When the second recognition model is successfully loaded and ready, the client will close the previous first process to free up resources and avoid conflicts or resource occupation problems that may be caused by running multiple model processes at the same time.

[0128] 209. Verify the second recognition model using the verification image.

[0129] After loading the second recognition model, the client verifies the model using the previously acquired verification image. The verification process is similar to the previous one. The client compares the model's prediction results with the annotation information of the verification image to evaluate the recognition accuracy and performance of the second recognition model.

[0130] In this embodiment, after training the first recognition model, the server will broadcast a download notification to each client, so that the client can download the first recognition model in time, shortening the time to obtain the first recognition model, which is conducive to improving the efficiency of verifying the first recognition model. At the same time, when the client loads the first recognition model and the second recognition model, it loads the model using different processes respectively, so that the client can still verify the first recognition model during the process of downloading and loading the second recognition model, which is conducive to improving the efficiency of training the model. At the same time, through further training and optimization, the second recognition model has been improved in recognition accuracy and processing power. After loading the second recognition model and verifying its performance, the client ensures that the system runs under the condition of using the optimal model, thereby improving the accuracy and efficiency of the image recognition task. This process effectively improves the continuous improvement capability of the model, enables the system to adapt to changing application requirements, and provide more accurate recognition results.

[0131] See also Figure 3 In one embodiment of the present application, a method for training a model applied to a transfer terminal includes:

[0132] 301. Receive a training image sent by a client;

[0133] In actual applications, the client is connected to the transfer end and the server through a local area network. The transfer end receives training images sent from the client through the local area network. These images are obtained by the client through image acquisition equipment and will serve as the data source for subsequent model training. After receiving these images, the transfer end will perform further processing to prepare for subsequent labeling work.

[0134] 302. Labeling the training image to obtain a labeled training image;

[0135] After receiving the training images, the relay begins to annotate these images. The annotation process involves identifying and marking important features or target objects in the image and annotating the label of the image. Annotation may be done by an automated system or manually, with the goal of adding relevant semantic information to each image. This annotation information is key to training efficient and accurate image recognition models.

[0136] 303. Send the labeled training image to the server.

[0137] After the annotation is completed, the transfer end sends the training images with annotation information to the server through the local area network. The server will use these annotated image data to train the image recognition model to further optimize the recognition ability and accuracy of the model. By sending the annotated images, the transfer end provides high-quality data input for the server's model training, ensuring that the model can better understand and recognize the content in the image.

[0138] In this embodiment, the transfer terminal effectively undertakes the task of annotating the training images and sends high-quality annotated data to the server. If the task of image annotation is performed by manual annotation, it is only necessary to arrange personnel to annotate the images at the transfer terminal. Compared with the prior art of arranging personnel to annotate images at each client, the number of personnel handling image annotation can be reduced, saving a lot of labor costs. At the same time, since the client, transfer terminal and server are connected via a local area network, the rate at which the client transmits the training image to the transfer terminal and the rate at which the transfer terminal transmits the annotated training image to the server can be increased, thereby improving the efficiency of the training model.

[0139] See also Figure 4 Another embodiment of the method for training a model applied to a transfer terminal in the embodiment of the present application includes:

[0140] 401. Receive a training image sent by a client;

[0141] 402. Labeling the training image to obtain a labeled training image;

[0142] 403. Send the labeled training image to the server;

[0143] In this embodiment, steps 401 to 403 are similar to steps 301 to 303 in the aforementioned embodiment, and are not described again here.

[0144] 404. Send the labeled training image to the client.

[0145] In actual applications, the client can send a request to the transfer end to obtain the labeled training images in the transfer end, and use these labeled training images to generate verification images to verify the image recognition model. After receiving the request, the transfer end can send the labeled training images to the client.

[0146] In this embodiment, the client obtains the annotated training image from the transfer end, and generates a verification image using the annotated training image. Because the client and the transfer end are connected via a local area network, the efficiency of the client in obtaining the verification image can be improved.

[0147] See also Figure 5 In one embodiment of the present application, a method for training a model applied to a server includes:

[0148] 501. Receive the annotated training image sent by the transfer end;

[0149] In actual applications, the client is connected to the transfer end and the server through a local area network. The server receives the annotated training images sent by the transfer end through the local area network. These images have been annotated by the transfer end and contain important features in the image, identification information of the target object, and image labels. After receiving these images, the server stores them and prepares them for model training.

[0150] 502. Train the initial recognition model using the training image to obtain a first recognition model.

[0151] After receiving the annotated training images, the server will use them to train the initial recognition model. The training process involves using the annotated data to optimize the parameters of the model so that it can more accurately recognize and classify objects in the image. Through multiple training cycles, the recognition ability of the model is continuously improved, and finally a new recognition model is generated, namely the first recognition model. After training, this model will have better image recognition performance and can be used in practical applications.

[0152] In this embodiment, since the client, the transfer end and the server are connected via a local area network, the rate at which the transfer end transmits the annotated training images to the server can be increased, thereby improving the efficiency of the training model.

[0153] See also Figure 6 Another embodiment of the method for training a model applied to a server in the embodiment of the present application includes:

[0154] 601. Receive the annotated training image sent by the transfer end;

[0155] 602. Train the initial recognition model using the training image to obtain a first recognition model;

[0156] In this embodiment, steps 601 to 602 are similar to steps 501 to 502 in the aforementioned embodiment and are not described again here.

[0157] 603. Send a download notification to the client, so that the client downloads the first recognition model according to the download notification;

[0158] The server will send a download notification to the client, informing the client that the first recognition model is ready for download. This notification usually includes the model version information, download path, and related instructions. After receiving the notification, the client will download the first recognition model from the server according to the information in the notification for subsequent verification or application. Sending a download notification ensures that the client can obtain the latest recognition model in a timely manner and maintain the synchronization of the system.

[0159] 604. Train the first recognition model or the initial recognition model using the labeled training image to obtain a second recognition model.

[0160] After completing the training of the first recognition model, the server will continue to use the annotated training images for further training. At this time, the server can choose to retrain the first recognition model, or directly train the initial recognition model to generate a second recognition model. Through this further training, the performance of the model is further optimized, and the second recognition model may be superior to the first recognition model in recognition accuracy and robustness. This process is continuously iterated and optimized, so that the model performs better in practical applications. Furthermore, after obtaining the second recognition model, the server can also perform operations similar to step 603 to promptly notify the client to obtain the second recognition model.

[0161] In this embodiment, the server sends a download notification to the client, which helps to reduce the time for the client to obtain the first recognition model, thereby improving the efficiency of the client verifying the first recognition model. The server also further improves the recognition ability of the system by generating a second recognition model. This ensures that the image recognition system can be continuously optimized and provide higher accuracy and reliability, providing more powerful support for image processing tasks in actual application scenarios.

[0162] See also Figure 7 In the embodiment of the present application, one embodiment of the client includes:

[0163] A collection unit 701, used for collecting training images;

[0164] A first sending unit 702 is used to send the training image to the transfer end, so that the transfer end annotates the training image and sends the annotated training image to the server;

[0165] A first acquisition unit 703 is used to acquire a first recognition model, where the first recognition model is an image recognition model obtained by the server after training an initial recognition model using annotated training images;

[0166] A second acquisition unit 704 is used to acquire a verification image, where the verification image is a marked image;

[0167] The first verification unit 705 is used to load the first recognition model through the first process and verify the first recognition model using the verification image.

[0168] Optionally, the client also includes:

[0169] A third acquisition unit is used to acquire a second recognition model, where the second recognition model is an image recognition model obtained by the server training the first recognition model or the initial recognition model using the annotated training image;

[0170] A loading unit, used for loading the second recognition model through the second process, and closing the first process after the second recognition model is loaded;

[0171] The second verification unit is used to verify the second recognition model using the verification image.

[0172] Optionally, the first acquiring unit 703 is specifically configured to:

[0173] Receive download notifications sent by the server;

[0174] The first recognition model is downloaded from the server according to the download notification.

[0175] Optionally, the second acquiring unit 704 is specifically configured to:

[0176] Obtaining annotated training images from the transfer end; generating verification images based on the annotated training images;

[0177] Or, obtain a verification image from a third party.

[0178] In this implementation, the functions of each unit are the same as those mentioned above. Figure 1 to Figure 2 The steps in the illustrated embodiments correspond to each other and will not be described again here.

[0179] See also Figure 8 , another embodiment of the client in the embodiment of the present application includes:

[0180] Processor 801, memory 802, input and output unit 803 and bus 804;

[0181] The processor 801 is connected to the memory 802, the input and output unit 803 and the bus 804;

[0182] The memory 802 stores a program, and the processor 801 calls the program to execute Figure 1 to Figure 2 Steps in the illustrated embodiment.

[0183] In this embodiment, the function of the processor 801 is the same as that of the aforementioned Figure 1 to Figure 2 The steps in the illustrated embodiments correspond to each other and will not be described again here.

[0184] See also Fig. 9 , an embodiment of the transfer end in the embodiment of the present application includes:

[0185] The receiving unit 901 is used to receive the training image sent by the client;

[0186] A labeling unit 902 is used to label the training image to obtain a labeled training image;

[0187] The first sending unit 903 is used to send the annotated training image to the server, so that the server trains the image recognition model by using the annotated training image.

[0188] Optionally, the transfer end also includes:

[0189] The second sending unit is used to send the annotated training image to the client, so that the client can verify the image recognition model by using the annotated training image.

[0190] In this implementation, the functions of each unit are the same as those mentioned above. Figure 3 to Figure 4 The steps in the illustrated embodiments correspond to each other and will not be described again here.

[0191] See also Fig.10 Another embodiment of the transfer end in the embodiment of the present application includes:

[0192] Processor 1001, memory 1002, input and output unit 1003 and bus 1004;

[0193] The processor 1001 is connected to the memory 1002, the input and output unit 1003 and the bus 1004;

[0194] The memory 1002 stores a program, and the processor 1001 calls the program to execute Figure 3 to Figure 4 Steps in the illustrated embodiment.

[0195] In this embodiment, the function of the processor 1001 is the same as that of the aforementioned Figure 3 to Figure 4 The steps in the illustrated embodiments correspond to each other and will not be described again here.

[0196] See also Fig.11 In the embodiment of the present application, one embodiment of the server includes:

[0197] The receiving unit 1101 is used to receive the annotated training image sent by the transfer end;

[0198] The first training unit 1102 is used to train the initial recognition model using the training image to obtain a first recognition model.

[0199] Optionally, the server further includes:

[0200] The second training unit is used to train the first recognition model or the initial recognition model using the labeled training images to obtain a second recognition model.

[0201] Optionally, the server further includes:

[0202] The notification unit is used to send a download notification to the client, so that the client downloads the first recognition model according to the download notification.

[0203] In this implementation, the functions of each unit are the same as those mentioned above. Figures 5 and 6 The steps in the illustrated embodiments correspond to each other and will not be described again here.

[0204] See also Fig.12 , another embodiment of the server in the embodiment of the present application includes:

[0205] Processor 1201, memory 1202, input and output unit 1203 and bus 1204;

[0206] The processor 1201 is connected to the memory 1202, the input and output unit 1203 and the bus 1204;

[0207] The memory 1202 stores a program, and the processor 1201 calls the program to execute Figures 5 and 6 Steps in the illustrated embodiment.

[0208] In this embodiment, the function of the processor 1201 is the same as that of the aforementioned Figures 5 and 6 The steps in the illustrated embodiments correspond to each other and will not be described again here.

[0209] See also Fig.13 , an embodiment of a system for training a model in an embodiment of the present application includes:

[0210] At least one client 1301, a transfer terminal 1302 and a server 1303;

[0211] The client 1301 is connected to the transfer terminal 1302 and the server 1303 via a local area network;

[0212] Client 1301 executes Figure 1 to Figure 2 In any possible implementation method of the embodiment shown in the figure, the transfer end 1302 executes the following Figure 3 to Figure 4 In any possible implementation method of the embodiment shown in the figure, the server 1303 executes the following Figures 5 and 6 A method in any possible implementation of the illustrated embodiment.

[0213] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0214] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the 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.

[0215] The units described 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.

[0216] 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.

[0217] 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 is essentially 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 a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, read-only memory), random access memory (RAM, random access memory), disk or optical disk and other media that can store program code.

Claims

1. A method for training a model, applied to a client, characterized in that: The client is connected to the transfer terminal and the server via a local area network, and the method includes: Collect training images; Sending the training image to a transfer terminal so that the transfer terminal annotates the training image and sends the annotated training image to a server; Acquire a first recognition model, where the first recognition model is an image recognition model obtained by the server training an initial recognition model using the labeled training image; Acquire a verification image, where the verification image is a labeled image; The first recognition model is loaded through a first process, and the first recognition model is verified using the verification image.

2. The method according to claim 1, characterized in that: After verifying the first recognition model using the verification image, the method further includes: Acquire a second recognition model, where the second recognition model is an image recognition model obtained by the server training the first recognition model or the initial recognition model using the labeled training image; Loading the second recognition model through a second process, and closing the first process after the second recognition model is loaded; The second recognition model is verified using the verification image.

3. The method according to claim 1, characterized in that The obtaining of the first recognition model comprises: Receiving a download notification sent by the server; The first recognition model is downloaded from the server according to the download notification.

4. The method according to any one of claims 1 to 3, characterized in that The obtaining of the verification image comprises: Acquire the labeled training image from the transfer end; generate a verification image based on the labeled training image; Or, obtain a verification image from a third party.

5. A method for training a model, applied to a transfer end, characterized in that: The transfer end is connected to the client and the server via a local area network, and the method includes: Receiving a training image sent by the client; Annotating the training image to obtain the annotated training image; The labeled training image is sent to the server, so that the server trains the image recognition model by using the labeled training image.

6. The method according to claim 5, characterized in that After labeling the training image to obtain the labeled training image, the method further includes: The labeled training image is sent to the client, so that the client verifies the image recognition model using the labeled training image.

7. A method for training a model, applied to a server, characterized in that: The server is connected to the client and the transfer terminal via a local area network, and the method includes: Receiving the annotated training image sent by the transfer end; The initial recognition model is trained using the training image to obtain a first recognition model.

8. The method according to claim 7, characterized in that After the initial recognition model is trained using the training image to obtain the first recognition model, the method further includes: The first recognition model or the initial recognition model is trained using the labeled training image to obtain a second recognition model.

9. The method according to claim 7 or 8, characterized in that: After the initial recognition model is trained using the training image to obtain the first recognition model, the method further includes: A download notification is sent to the client, so that the client downloads the first recognition model according to the download notification.

10. A system for training a model, characterized in that: Includes at least one client, a transfer terminal and a server; The client is connected to the transfer terminal and the server via a local area network; The client executes the method according to any one of claims 1 to 4, the transfer end executes the method according to any one of claims 5 to 6, and the server executes the method according to any one of claims 7 to 9.

Images