Intelligent marking method and device, computer equipment and storage medium
Through the intelligent marking method, the convolutional neural network model is used to automatically identify the device feature tags, which solves the problem of low label refinement in device management, and achieves high-refinement and accurate device information management.
Patent Information
- Application Number
- CN202411516938.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-29
- Publication Date
- 2025-02-18
AI Technical Summary
Among the existing equipment management methods, the degree of refinement of equipment feature labels is low, resulting in insufficient comprehensive and accurate information. Especially in the management of charging stations, there is a lack of efficient automated marking methods.
The intelligent marking method is adopted to optimize the convolutional neural network model by obtaining the initial device image and using the stochastic gradient descent algorithm to automatically identify the device feature labels, thereby improving the degree of refinement and accuracy of the labels.
It realizes the high degree of refinement and accuracy of device feature labels, reduces labor costs, and improves the comprehensiveness and reliability of device information.
Smart Images
Figure CN119478625A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of computer technology, and more specifically, to an intelligent marking method, apparatus, computer equipment and storage medium. Background Art
[0002] In industrial production, equipment management is a vital link, especially in high-demand manufacturing industries (such as electronics, automobiles, aviation, etc.).
[0003] At present, the management of equipment is mainly carried out by operators who visit each city offline to confirm the equipment feature tags, such as whether there is a parking fee for the charging station, the location of the power station, whether parking is charged, the charging efficiency of the power station, whether the transportation is convenient, whether the surrounding facilities are complete, etc. The tag data is collected and then entered into the system through system import or batch labeling, so that charging users can search and use it on the client.
[0004] In addition, since some charging platforms are unable to provide too many operators to conduct offline visits to collect characteristic labels of charging stations, they can only manually determine the location of the charging station, parking fees, convenient transportation and other labels by combining data such as the power station name, power station address, and station instructions with tools such as Gaode Maps.
[0005] However, the above-mentioned device marking method results in a low degree of refinement of the device's feature labels. Summary of the invention
[0006] The main purpose of this application is to provide an intelligent labeling method, device, computer equipment and storage medium, which can improve the refinement of the feature labels of the equipment.
[0007] In order to achieve the above objectives, in a first aspect, the present application provides an intelligent marking method, comprising:
[0008] Acquire an initial device image, wherein no feature label of the device is identified in the initial device image;
[0009] Obtaining a target convolutional neural network model, wherein the target convolutional neural network model is obtained by optimizing parameters of the initial convolutional neural network model by using a stochastic gradient descent algorithm;
[0010] The initial device image is input into the target convolutional neural network model, and a target device image is output, wherein the target device image is marked with a feature label of the device.
[0011] In one embodiment, obtaining a target convolutional neural network model includes:
[0012] Acquire training data, wherein the training data includes a first image and a second image, the first image is used to represent an image of a device configured with a feature label of the device, and the second image is used to represent an image of a device not configured with a feature label of the device;
[0013] The initial convolutional neural network model is trained using the training data to obtain the target convolutional neural network model.
[0014] In one embodiment, the initial convolutional neural network model includes a basic module structure and a bottleneck structure;
[0015] The base structure consists of two convolutional layers and a shortcut connection;
[0016] The bottleneck structure consists of three convolutional layers and a shortcut connection.
[0017] In one embodiment, the first convolution layer of the two convolution layers in the base structure uses a 1×1 convolution kernel to perform a dimensionality reduction operation on the feature matrix;
[0018] The second of the two convolutional layers in the base structure uses a 3×3 convolution kernel to increase the dimension of the feature matrix.
[0019] In one embodiment, the first convolution layer of the three convolution layers in the bottleneck structure uses a 1×1 convolution kernel to perform a dimensionality reduction operation on the feature matrix;
[0020] The second convolutional layer of the three convolutional layers in the bottleneck structure uses a 1×1 convolution kernel to increase the dimension of the feature matrix;
[0021] The third convolutional layer of the three convolutional layers in the bottleneck structure uses a 3×3 convolution kernel to increase the dimension of the feature matrix.
[0022] In one embodiment, the method further comprises:
[0023] Acquire a device attribute library, wherein the device attributes in the device attribute library are obtained by processing the device data;
[0024] The device attribute library configures corresponding device attributes for the feature tags of the devices identified in the target device image.
[0025] In one embodiment, the method further comprises:
[0026] Receive a user's device query request;
[0027] Obtain the corresponding target device based on the device query request and push the target device to the user's client.
[0028] In a second aspect, an embodiment of the present application provides an intelligent marking device, including:
[0029] A first acquisition module is used to acquire an initial device image, wherein no feature tag of the device is marked in the initial device image;
[0030] A second acquisition module is used to acquire a target convolutional neural network model, wherein the target convolutional neural network model is acquired by optimizing the parameters of the initial convolutional neural network model by using a stochastic gradient descent algorithm;
[0031] The labeling module is used to input the initial device image into the target convolutional neural network model and output the target device image, wherein the target device image is marked with a feature label of the device.
[0032] In one embodiment, the second acquisition module is further used to acquire training data, wherein the training data includes a first image and a second image, the first image is used to represent an image of a device configured with a feature label of the device, and the second image is used to represent an image of a device not configured with a feature label of the device;
[0033] The initial convolutional neural network model is trained using the training data to obtain the target convolutional neural network model.
[0034] In one embodiment, the initial convolutional neural network model includes a basic module structure and a bottleneck structure;
[0035] The base structure consists of two convolutional layers and a shortcut connection;
[0036] The bottleneck structure consists of three convolutional layers and a shortcut connection.
[0037] In one embodiment, the first convolution layer of the two convolution layers in the base structure uses a 1×1 convolution kernel to perform a dimensionality reduction operation on the feature matrix;
[0038] The second of the two convolutional layers in the base structure uses a 3×3 convolution kernel to increase the dimension of the feature matrix.
[0039] In one embodiment, the first convolution layer of the three convolution layers in the bottleneck structure uses a 1×1 convolution kernel to perform a dimensionality reduction operation on the feature matrix;
[0040] The second convolutional layer of the three convolutional layers in the bottleneck structure uses a 1×1 convolution kernel to increase the dimension of the feature matrix;
[0041] The third convolutional layer of the three convolutional layers in the bottleneck structure uses a 3×3 convolution kernel to increase the dimension of the feature matrix.
[0042] In one embodiment, the apparatus further comprises: a configuration module, the configuration module being used to obtain a device attribute library, wherein the device attributes in the device attribute library are obtained by processing the device data;
[0043] The device attribute library configures corresponding device attributes for the feature tags of the devices identified in the target device image.
[0044] In one embodiment, the apparatus further comprises: a query module, the query module being configured to receive a device query request from a user;
[0045] Obtain the corresponding target device based on the device query request and push the target device to the user's client.
[0046] In a third aspect, an embodiment of the present application provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of any of the above methods when executing the computer program.
[0047] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the steps of any of the above methods are implemented.
[0048] The embodiments of the present application provide a smart labeling method, apparatus, computer equipment and storage medium, including: obtaining an initial device image, wherein the initial device image does not have a feature label of the device, then obtaining a target convolutional neural network model, wherein the target convolutional neural network model is obtained by optimizing the parameters of the initial convolutional neural network model by a stochastic gradient descent algorithm, inputting the initial device image into the target convolutional neural network model, and outputting a target device image, wherein the target device image has a feature label of the device. The present application implements device labeling through a target convolutional neural network model, which not only reduces labor costs, but also improves the refinement and accuracy of the device's feature label. BRIEF DESCRIPTION OF THE DRAWINGS
[0049] The drawings constituting a part of this application are used to provide a further understanding of this application, so that other features, purposes and advantages of this application become more obvious. The schematic embodiment drawings and their descriptions of this application are used to explain this application and do not constitute an improper limitation on this application. In the drawings:
[0050] Figure 1 This is an application scenario diagram of an intelligent marking method provided in an embodiment of the present application;
[0051] Figure 2 It is a flowchart of an intelligent marking method provided in an embodiment of the present application;
[0052] Figure 3 is a schematic diagram of a label of a charging station provided in an embodiment of the present application;
[0053] Figure 4 is a schematic diagram of a phrase of a charging station provided in an embodiment of the present application;
[0054] Figure 5 This is a schematic diagram of a main page of a device query APP provided in an embodiment of the present application;
[0055] Figure 6 It is a structural schematic diagram of an intelligent marking device provided in an embodiment of the present application;
[0056] Figure 7 It is a schematic diagram of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0057] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0058] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the application described herein can be implemented in an order other than those illustrated or described herein.
[0059] In the embodiments of the present application, the term "module" or "unit" refers to a computer program or a part of a computer program with a predetermined function, and works together with other related parts to achieve a predetermined goal, and can be implemented in whole or in part by using software, hardware (such as processing circuits or memories) or a combination thereof. Similarly, a processor (or multiple processors or memories) can be used to implement one or more modules or units. In addition, each module or unit can be part of an overall module or unit that includes the function of the module or unit.
[0060] It should be understood that in the various embodiments of the present application, the size of the serial number of each process does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.
[0061] It should be understood that in the present application, "include" and "have" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or apparatus comprising a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products or apparatuses.
[0062] It should be understood that in the present application, "plurality" refers to two or more than two. "And / or" is merely a description of the association relationship of associated objects, indicating that three relationships may exist. For example, and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "Contains A, B and C", "Contains A, B, C" means that A, B, and C are all included, "Contains A, B or C" means that one of A, B, and C is included, and "Contains A, B and / or C" means that any one, any two, or any three of A, B, and C are included.
[0063] It should be understood that in this application, "B corresponding to A", "B corresponding to A", "A corresponds to B" or "B corresponds to A" means that B is associated with A and B can be determined based on A. Determining B based on A does not mean determining B based only on A, but B can also be determined based on A and / or other information. A and B match when the similarity between A and B is greater than or equal to a preset threshold.
[0064] Depending on the context, "if" as used herein may be interpreted as "when" or "when" or "in response to determining" or "in response to detecting."
[0065] The data involved in this application may be data authorized by the tester or fully authorized by all parties. The collection, dissemination, and use of the data shall comply with the relevant laws, regulations, and standards of the relevant countries and regions. The implementation methods / examples of this application may be combined with each other.
[0066] The technical solution of the present application is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.
[0067] Next, the present application scheme will be described through specific embodiments in conjunction with the accompanying drawings.
[0068] See also Figure 1 , Figure 1 An intelligent marking method provided in the embodiment of the present application can be applied to Figure 1In the application environment shown, the user 102 clicks on the device query APP displayed on the screen of the client 104 and enters the main page of the device query APP. The client 104 responds to the device query request triggered by the user 102 and sends the device query request to the server 106. The server 106 obtains the corresponding target device based on the device query request and pushes the target device to the client 104 where the user is located.
[0069] Among them, the server 106 is also used to obtain an initial device image, wherein the feature label of the device is not identified in the initial device image, and then obtain a target convolutional neural network model, wherein the target convolutional neural network model is obtained by optimizing the parameters of the initial convolutional neural network model by a stochastic gradient descent algorithm, and the initial device image is input into the target convolutional neural network model, and the target device image is output, wherein the feature label of the device is identified in the target device image.
[0070] The client 104 and the server 106 can communicate with each other through any communication method, including but not limited to network communication, and the above network can include but not limited to: wired network, wireless network, wherein the wired network includes: local area network, metropolitan area network and wide area network, and the wireless network includes: Bluetooth, WIFI, and other networks that realize wireless communication. The client 104 includes but is not limited to at least one of the following: mobile phone (such as Android mobile phone, iOS mobile phone, etc.), laptop computer, tablet computer, PDA, MID (Mobile Internet Devices), PAD, desktop computer, smart TV, etc. The server 106 can be an on-site server or a remote server, wherein both the on-site server and the remote server can be implemented as independent servers or service clusters composed of multiple servers. The above is only an example, and no limitation is made to this in this embodiment.
[0071] See also Figure 2 , Figure 2 A flow chart of an intelligent marking method provided in an embodiment of the present application, which is applied to Figure 1 The server 106 shown includes the following steps:
[0072] Step S201: Acquire an initial device image.
[0073] The initial device image does not contain any feature labels of the device.
[0074] Equipment includes but is not limited to automotive equipment, positioning equipment, gate equipment, camera equipment, charging stations, etc.
[0075] Taking the device as a charging station as an example, its corresponding feature tags include the location of the power station, whether there is a parking fee, paid parking / free parking / limited time parking, applicable vehicle models, fast charging / slow charging, etc.
[0076] Step S202: Obtain a target convolutional neural network model.
[0077] Among them, the target convolutional neural network model is obtained by optimizing the parameters of the initial convolutional neural network model through the stochastic gradient descent algorithm.
[0078] The initial convolutional neural network models include but are not limited to ResNet (Residual Network), DenseNet (Densely Connected Convolutional Networks), etc.
[0079] Among them, the initial convolutional neural network model includes a basic module structure (BasicBlock) and a bottleneck structure (Bottleneck); the basic structure consists of two convolutional layers and a shortcut connection (shortcut); the bottleneck structure consists of three convolutional layers and a shortcut connection.
[0080] Among them, the first convolution layer of the two convolution layers in the basic structure uses a 1×1 convolution kernel to reduce the dimension of the feature matrix; the second convolution layer of the two convolution layers in the basic structure uses a 3×3 convolution kernel to increase the dimension of the feature matrix. This structure reduces the depth of the feature matrix, thereby reducing the number of parameters.
[0081] Among them, the first convolution layer of the three convolution layers in the bottleneck structure uses a 1×1 convolution kernel to reduce the dimension of the feature matrix; the second convolution layer of the three convolution layers in the bottleneck structure uses a 1×1 convolution kernel to increase the dimension of the feature matrix; the third convolution layer of the three convolution layers in the bottleneck structure uses a 3×3 convolution kernel to increase the dimension of the feature matrix. This structure further increases the depth of the feature matrix, thereby enhancing the learning ability of the network.
[0082] In one embodiment, obtaining a target convolutional neural network model includes: obtaining training data, wherein the training data includes a first image and a second image, the first image is used to represent an image of a device that has been configured with a feature label of the device, and the second image is used to represent an image of a device that has not been configured with a feature label of the device; and using the training data to train an initial convolutional neural network model to obtain a target convolutional neural network model.
[0083] The initial convolutional neural network model is ResNet. During the training process, ResNet uses the stochastic gradient descent (SGD) algorithm to optimize the network parameters. By normalizing the input data and the data in the middle layer, it can ensure that the network uses SGD in the back propagation, so that the network can converge. This training method effectively solves the problems of gradient disappearance and gradient explosion, so that deeper networks can also be trained to achieve good results.
[0084] Step S203: input the initial device image into the target convolutional neural network model and output the target device image.
[0085] The target device image is marked with a feature label of the device.
[0086] After obtaining the target convolutional neural network model, the intelligent labeling operation can be realized through the target convolutional neural network model. Specifically, the initial device image needs to be input into the target convolutional neural network model and the target device image is output. In other words, the output target device image has been labeled for the device.
[0087] like Figure 3 As shown, a device image is marked, and its configured feature tags include a primary tag name, a secondary tag name, a quantity, a data source, and an operation, etc. For example, the primary tag name is a service phone number, the secondary tag name is 400 phone number, the quantity is 65793, the data source is a service phone number, and the operation is an associated power station (i.e., a charging station).
[0088] For example, in the process of analyzing power station data, the analysis and processing of power station images and other data will be analyzed by different means. There are usually multiple images of power stations, and the locations and contents of the photos are different, so the analysis of power station images can only be labeled as the location of the power station (above ground, underground). The elements of the power station image include blue sky, white clouds, big trees, rivers, roads, light, etc., which can be judged as an above-ground power station. The elements with lights, dim light, underground garage railings, etc. can be judged as underground power stations. At the same time, the size of the parking space in the photo content can be used to preliminarily determine whether the power station can allow large logistics vehicles to enter. Similarly, the analysis and labeling results of the power station images are displayed on the page, and the operator manually confirms them. Through the number of confirmations, the system strengthens the analysis and learning ability of the pictures, and finally achieves the result without manual confirmation.
[0089] Through the target convolutional neural network model, intelligent labeling of equipment can be realized, which can determine the quality of equipment, whether the equipment is available, the equipment's surrounding facilities and other labels in real time, and ultimately assist the system to accurately maintain the equipment, save labor costs, and repair equipment in a timely manner.
[0090] In one embodiment, the method further includes: obtaining a device attribute library, wherein the device attributes in the device attribute library are obtained by processing device data; and configuring corresponding device attributes for feature tags of devices identified in the target device image through the device attribute library.
[0091] This application performs word segmentation processing on all power station data (such as power station name, power station address, parking fee instructions, etc.) through scheduled tasks every day, gives priority to the first-level filtering of phrases, filters out useless phrases such as provinces, cities and districts (which can be used for the management of useless phrases), establishes an equipment attribute library for the phrases segmented from the big data and displays it in the system, which can be viewed and maintained by operators, supports addition, modification and deletion, and records the number of times the phrases are used and the correlation and similarity between the phrases after the operators use these phrases for manual batch labeling.
[0092] like Figure 4 As shown in the figure, if the device is a charging station, the device attribute library includes phrases such as license plate number and free. It also records the data source and recording time corresponding to the phrase name. The phrases in the device attribute library can be added, modified, deleted, etc.
[0093] See also Figure 5 , Figure 5 A schematic diagram of the main page of a device query APP provided in an embodiment of the present application.
[0094] User 102 clicks on the device query APP displayed on the screen of client 104 and enters Figure 5 After the main page of the device query APP is displayed, the client 104 responds to the device query request triggered by the user 102, wherein the query request includes the charging pile power, charging pile voltage, etc. The user 102 can select through the options configured in the main page of the device query APP.
[0095] After the device query request is sent to the server 106 , the server 106 obtains the corresponding electric charging pile based on the device query request, and pushes the electric charging pile to the client 104 where the user is located.
[0096] The embodiment of the present application provides an intelligent labeling method, including: obtaining an initial device image, wherein the initial device image does not have a feature label of the device, then obtaining a target convolutional neural network model, wherein the target convolutional neural network model is obtained by optimizing the parameters of the initial convolutional neural network model by a stochastic gradient descent algorithm, inputting the initial device image into the target convolutional neural network model, and outputting a target device image, wherein the target device image has a feature label of the device. The present application implements device labeling through a target convolutional neural network model, which not only reduces labor costs, but also improves the refinement and accuracy of the device's feature label.
[0097] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.
[0098] The following is an embodiment of the device of the present application. For details not described in detail, please refer to the corresponding method embodiment described above.
[0099] Figure 6 A schematic diagram of the structure of an intelligent marking device provided in an embodiment of the present application is shown. For the sake of convenience, only the parts related to the embodiment of the present application are shown. An intelligent marking device includes a first acquisition module 601, a second acquisition module 602 and a marking module 603, which are as follows:
[0100] The first acquisition module 601 is used to acquire an initial device image, wherein the initial device image does not have a feature tag of the device;
[0101] A second acquisition module 602 is used to acquire a target convolutional neural network model, wherein the target convolutional neural network model is acquired by optimizing the parameters of the initial convolutional neural network model by using a stochastic gradient descent algorithm;
[0102] The labeling module 603 is used to input the initial device image into the target convolutional neural network model and output the target device image, wherein the target device image is marked with a feature label of the device.
[0103] In one embodiment, the second acquisition module 602 is further used to acquire training data, wherein the training data includes a first image and a second image, the first image is used to represent a device image configured with a feature label of the device, and the second image is used to represent a device image not configured with a feature label of the device;
[0104] The initial convolutional neural network model is trained using the training data to obtain the target convolutional neural network model.
[0105] In one embodiment, the initial convolutional neural network model includes a basic module structure and a bottleneck structure;
[0106] The base structure consists of two convolutional layers and a shortcut connection;
[0107] The bottleneck structure consists of three convolutional layers and a shortcut connection.
[0108] In one embodiment, the first convolution layer of the two convolution layers in the base structure uses a 1×1 convolution kernel to perform a dimensionality reduction operation on the feature matrix;
[0109] The second of the two convolutional layers in the base structure uses a 3×3 convolution kernel to increase the dimension of the feature matrix.
[0110] In one embodiment, the first convolution layer of the three convolution layers in the bottleneck structure uses a 1×1 convolution kernel to perform a dimensionality reduction operation on the feature matrix;
[0111] The second convolutional layer of the three convolutional layers in the bottleneck structure uses a 1×1 convolution kernel to increase the dimension of the feature matrix;
[0112] The third convolutional layer of the three convolutional layers in the bottleneck structure uses a 3×3 convolution kernel to increase the dimension of the feature matrix.
[0113] In one embodiment, the apparatus further comprises: a configuration module, the configuration module being used to obtain a device attribute library, wherein the device attributes in the device attribute library are obtained by processing the device data;
[0114] The device attribute library configures corresponding device attributes for the feature tags of the devices identified in the target device image.
[0115] In one embodiment, the apparatus further comprises: a query module, the query module being configured to receive a device query request from a user;
[0116] Obtain the corresponding target device based on the device query request and push the target device to the user's client.
[0117] The embodiment of the present application provides an intelligent marking device, which is specifically used to: obtain an initial device image, wherein the initial device image does not have a feature label of the device, then obtain a target convolutional neural network model, wherein the target convolutional neural network model is obtained by optimizing the parameters of the initial convolutional neural network model by a stochastic gradient descent algorithm, input the initial device image into the target convolutional neural network model, and output a target device image, wherein the target device image has a feature label of the device. The present application implements device marking through a target convolutional neural network model, which not only reduces labor costs, but also improves the refinement and accuracy of the device's feature label.
[0118] This application Figure 7 A schematic diagram of a computer device is provided. Figure 7 As shown, the computer device 7 of this embodiment includes: a processor 701, a memory 702, and a computer program 703 stored in the memory 702 and executable on the processor 701. When the processor 701 executes the computer program 703, the steps in the above-mentioned various intelligent marking method embodiments are implemented, for example Figure 2Alternatively, when the processor 701 executes the computer program 703, the functions of each module / unit in the above-mentioned embodiments of the intelligent marking device are realized, for example Figure 6 Functionality of modules / units 601 to 603 shown.
[0119] The present application also provides a readable storage medium, in which a computer program is stored. When the computer program is executed by a processor, it is used to implement the intelligent marking method provided by the various embodiments described above.
[0120] Among them, the readable storage medium can be a computer storage medium or a communication medium. The communication medium includes any medium that facilitates the transmission of a computer program from one place to another. The computer storage medium can be any available medium that can be accessed by a general or special-purpose computer. For example, a readable storage medium is coupled to a processor so that the processor can read information from the readable storage medium and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can be located in an application-specific integrated circuit (ASIC). In addition, the ASIC can be located in a user device. Of course, the processor and the readable storage medium can also exist in a communication device as discrete components. The readable storage medium can be a read-only memory (ROM), a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, and an optical data storage device, etc.
[0121] The present application also provides a program product, which includes an execution instruction, which is stored in a readable storage medium. At least one processor of the device can read the execution instruction from the readable storage medium, and at least one processor executes the execution instruction so that the device implements the intelligent marking method provided by the various embodiments described above.
[0122] In the embodiments of the above-mentioned devices, it should be understood that the processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), etc. The general-purpose processor may be a microprocessor or the processor may be any conventional processor, etc. The steps of the method disclosed in the present application may be directly embodied as being executed by a hardware processor, or may be executed by a combination of hardware and software modules in the processor.
[0123] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.
Claims
1. An intelligent marking method, characterized in that: include: Acquire an initial device image, wherein no feature tag of the device is identified in the initial device image; Obtaining a target convolutional neural network model, wherein the target convolutional neural network model is obtained by optimizing parameters of an initial convolutional neural network model using a stochastic gradient descent algorithm; The initial device image is input into the target convolutional neural network model, and a target device image is output, wherein the target device image is marked with a feature label of the device.
2. The intelligent marking method according to claim 1, characterized in that: The obtaining of the target convolutional neural network model comprises: Acquire training data, wherein the training data includes a first image and a second image, the first image is used to represent a device image configured with a feature label of the device, and the second image is used to represent a device image not configured with a feature label of the device; The initial convolutional neural network model is trained using the training data to obtain the target convolutional neural network model.
3. The intelligent marking method according to claim 2, characterized in that: The initial convolutional neural network model includes a basic module structure and a bottleneck structure; The base structure consists of two convolutional layers and a shortcut connection; The bottleneck structure consists of three convolutional layers and a shortcut connection.
4. The intelligent marking method according to claim 3, characterized in that: The first convolution layer of the two convolution layers in the basic structure uses a 1×1 convolution kernel to perform a dimensionality reduction operation on the feature matrix; The second convolution layer of the two convolution layers in the basic structure uses a 3×3 convolution kernel to perform a dimensionality increase operation on the feature matrix.
5. The intelligent marking method according to claim 3, characterized in that: The first convolution layer of the three convolution layers in the bottleneck structure uses a 1×1 convolution kernel to perform a dimensionality reduction operation on the feature matrix; The second convolution layer of the three convolution layers in the bottleneck structure uses a 1×1 convolution kernel to perform a dimensionality increase operation on the feature matrix; The third convolution layer of the three convolution layers in the bottleneck structure uses a 3×3 convolution kernel to perform a dimensionality increase operation on the feature matrix.
6. The intelligent marking method according to claim 1, characterized in that: The method further comprises: Acquire a device attribute library, wherein the device attributes in the device attribute library are obtained by processing device data; The device attribute library configures corresponding device attributes for the feature tag of the device identified in the target device image.
7. The intelligent marking method according to claim 1, characterized in that: The method further comprises: Receive a user's device query request; The corresponding target device is obtained based on the device query request, and the target device is pushed to the client where the user is located.
8. An intelligent marking device, characterized in that: include: A first acquisition module is used to acquire an initial device image, wherein the initial device image does not contain a feature tag of the device; A second acquisition module is used to acquire a target convolutional neural network model, wherein the target convolutional neural network model is acquired by optimizing the parameters of the initial convolutional neural network model by using a stochastic gradient descent algorithm; The labeling module is used to input the initial device image into the target convolutional neural network model and output a target device image, wherein the target device image is marked with a feature label of the device.
9. A computer device, characterized in that: comprising a memory, and one or more processors communicatively connected to the memory; The memory stores instructions that can be executed by the one or more processors, and the instructions are executed by the one or more processors to enable the one or more processors to implement the intelligent marking method as described in any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that: The method comprises a program or an instruction, and when the program or the instruction is executed on a computer, the intelligent marking method according to any one of claims 1 to 7 is implemented.