A brand identification method, apparatus, electronic device, and storage medium

Abstract

This invention provides a brand recognition method, apparatus, electronic device, and storage medium. On one hand, it addresses the difficulties in data collection and the issue of whether the collected logo data matches the distribution of real-world application scenarios when building large-scale brand logo datasets. On the other hand, it enables the rapid construction of brand logo datasets, accelerates model training iterations, improves the generalization ability of the brand logo recognition model by adding numerous adversarial examples from real-world application scenarios, and solves the problem of class imbalance by expanding the sample size for brands with insufficient data, thereby improving the performance of the brand logo model and ultimately enabling more effective brand logo recognition.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and more specifically, to a brand identification method, apparatus, electronic device, and storage medium. Background Technology

[0002] With the continuous development of the self-media industry, the viewership of short videos on various platforms is also constantly increasing, with the vast majority of content disseminated in the form of images and videos. In image or video content with little or no text information, brand logo recognition technology can help marketing teams identify identifiable brand logos in a specific area of ​​social media content images or videos, quickly and accurately locating specific brands and helping brands track brand content published on social media in real time. Existing brand logo recognition methods typically employ computer technology based on object detection, using computer vision and image processing. This mainly involves classifying identified brands in an image and identifying the logo's location within the image. Currently, mainstream brand logo recognition solutions on the market typically use end-to-end deep learning methods based on convolutional neural networks. However, this method suffers from difficulties in data collection and the collected logo data not conforming to the distribution of real-world application scenarios. Brand logo recognition technology based on deep convolutional neural networks is data-driven; therefore, training the brand logo recognition model using imbalanced class samples, insufficient brand training sets, or brand logo datasets that do not cover all real-world application scenarios will lead to a decline in brand logo recognition performance.

[0003] Therefore, how to establish a more efficient brand logo recognition model to achieve higher-performance brand logo recognition is an urgent problem to be solved. Summary of the Invention

[0004] To address the aforementioned issues, this invention provides a brand identification method, apparatus, electronic device, and storage medium.

[0005] A first aspect of the present invention provides a brand identification method, the method comprising:

[0006] Collect logo data from multiple brands according to a preset number for the first time, and build a brand logo dataset;

[0007] The recognition model is trained using a deep convolutional neural network-based training method to obtain the brand logo recognition model with the best performance based on the brand logo dataset.

[0008] The brand logo recognition model is used to recognize the brand logo data information to be recognized, and the recognition results and corresponding application scenario data are saved.

[0009] Receive manually labeled information from external input and process the recognition results and application scenario data accordingly;

[0010] The processed recognition results and application scenario data are updated in the brand logo dataset for the next recognition model training.

[0011] Optionally, the manually labeled information specifically includes:

[0012] This indicates that the brand logo data for this application scenario has been correctly identified.

[0013] This indicates the labeling information for incorrect identification and the correct identification result, used to indicate that the brand logo data for this application scenario was identified incorrectly.

[0014] Blank annotation information is used to indicate that no brand logo data was identified for this application scenario.

[0015] Optionally, the step of processing the recognition results and application scenario data accordingly specifically includes:

[0016] For those correctly identified, the identified brand logo data, application scenario data, and annotation information will be saved;

[0017] For those labeled as having recognition errors, the recognition results and application scenario data are cleaned and corrected to the correct recognition results.

[0018] Optionally, the step of updating the processed recognition results and application scenario data to the brand logo dataset specifically includes:

[0019] Add the saved identified brand logo data, application scenario data, and annotation information to the brand logo dataset;

[0020] Add the saved, cleaned brand logo data, application scenario data, and annotation information to the brand logo dataset;

[0021] Application scenario data that did not identify brand logo data were added to the brand logo dataset as background scenario data and corresponding blank annotation information.

[0022] A second aspect of the present invention provides a brand identification device, the device comprising:

[0023] The dataset creation unit is used to acquire logo data from multiple brands collected for the first time according to a preset number, and to create a brand logo dataset;

[0024] The recognition model training unit is used to train the recognition model using a deep convolutional neural network-based training method to obtain the brand logo recognition model with the best performance based on the brand logo dataset.

[0025] The brand data recognition unit is used to recognize the brand logo data information to be recognized using the brand logo recognition model, and to save the recognition results and corresponding application scenario data.

[0026] The identification data processing unit is used to receive manually labeled information from external input and to process the identification results and application scenario data accordingly.

[0027] The dataset management unit is used to update the processed recognition results and application scenario data into the brand logo dataset for the next recognition model training.

[0028] Optionally, the manually labeled information specifically includes:

[0029] This indicates that the brand logo data for this application scenario has been correctly identified.

[0030] This indicates the labeling information for incorrect identification and the correct identification result, used to indicate that the brand logo data for this application scenario was identified incorrectly.

[0031] Blank annotation information is used to indicate that no brand logo data was identified for this application scenario.

[0032] Optionally, the identification data processing unit is specifically used for:

[0033] For those correctly identified, the identified brand logo data, application scenario data, and annotation information will be saved;

[0034] For those labeled as having recognition errors, the recognition results and application scenario data are cleaned and corrected to the correct recognition results.

[0035] Optionally, the dataset management unit is specifically used for:

[0036] Add the saved identified brand logo data, application scenario data, and annotation information to the brand logo dataset;

[0037] Add the saved, cleaned brand logo data, application scenario data, and annotation information to the brand logo dataset;

[0038] Application scenario data that did not identify brand logo data were added to the brand logo dataset as background scenario data and corresponding blank annotation information.

[0039] A third aspect of the present invention provides an electronic device, characterized in that it comprises:

[0040] One or more processors; memory; one or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the one or more processors, the one or more applications being configured to perform the method as described in the first aspect.

[0041] A fourth aspect of the present invention provides a computer-readable storage medium, characterized in that the computer-readable storage medium stores program code that can be invoked by a processor to execute the method described in the first aspect.

[0042] In summary, this invention provides a brand recognition method, apparatus, electronic device, and storage medium. On one hand, it addresses the difficulties in data collection and the issue of whether the collected logo data conforms to real-world application scenarios when building large-scale brand logo datasets. On the other hand, it enables rapid construction of brand logo datasets, accelerates model training iterations, improves the generalization ability of the brand logo recognition model by adding numerous adversarial examples from real-world application scenarios, and resolves class imbalance by expanding the sample size for brands with insufficient data, thereby improving the brand logo model's recognition performance and ultimately enabling more effective brand logo recognition. Attached Figure Description

[0043] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0044] Figure 1 This is a schematic diagram illustrating an application scenario of the brand recognition method and apparatus according to an embodiment of the present invention;

[0045] Figure 2 This is a flowchart of the brand recognition method according to an embodiment of the present invention;

[0046] Figure 3 This is a flowchart of a brand recognition method according to another embodiment of the present invention;

[0047] Figure 4This is a functional module block diagram of the scene image text data generation device according to an embodiment of the present invention;

[0048] Figure 5 This is a structural block diagram of an electronic device for performing a brand recognition method according to an embodiment of the present invention.

[0049] Figure 6 This is a structural block diagram of a computer-readable storage medium for storing or carrying program code implementing a brand identification method according to an embodiment of the present application.

[0050] icon:

[0051] Cloud server 100; user client 200; dataset creation unit 110; recognition model training unit 120; brand data recognition unit 130; recognition data processing unit 140; dataset management unit 150; electronic device 300; processor 310; memory 320; computer-readable storage medium 400; program code 410. Detailed Implementation

[0052] With the continuous development of the self-media industry, the viewership of short videos on various platforms is also constantly increasing, with the vast majority of content disseminated in the form of images and videos. In image or video content with little or no text information, brand logo recognition technology can help marketing teams identify identifiable brand logos in a specific area of ​​social media content images or videos, quickly and accurately locating specific brands and helping brands track brand content published on social media in real time. Existing brand logo recognition methods typically employ computer technology based on object detection, using computer vision and image processing. This primarily involves classifying identified brands within an image and identifying the logo's location within the image. Currently, mainstream brand logo recognition solutions on the market typically use end-to-end deep learning methods based on convolutional neural networks.

[0053] In traditional logo recognition methods based on deep convolutional neural networks, the first step is to collect logo data containing all the required brands, which is necessary for training the logo recognition model. All data is then manually labeled to construct a brand logo dataset. The deep convolutional neural network is then trained using the brand logo dataset to obtain a logo recognition model with the best performance. Finally, the logo recognition model recognition service is deployed to perform brand logo recognition.

[0054] However, in practical applications, there may be tens of thousands of brand logos that need to be recognized, and there may also be situations where one brand corresponds to multiple different logo areas. Therefore, collecting a sufficient amount of brand logo data requires a significant investment of manpower and time. However, brand logo recognition technology based on deep convolutional neural networks is data-driven. Therefore, training the brand logo recognition model using brand logo datasets with imbalanced class samples, insufficient brand training sets, or that do not cover all real-world application scenarios will lead to a decrease in brand logo recognition performance.

[0055] During training, when there are significant differences in the amount of data between classes in the dataset, the model tends to classify samples as those with larger datasets to obtain lower losses, thus lacking the motivation to further optimize parameters. This leads to undertraining and weakened model performance. Consequently, some brand logo samples with insufficient data may be classified as background samples, easily resulting in missed recognition. If brand logo data from all real-world application scenarios cannot be collected, the model will fail to accurately detect and recognize brand logos in unfamiliar contexts, leading to misidentification and reduced generalization ability.

[0056] In view of this, the inventors of this invention have designed a brand recognition method, device, electronic device, and storage medium. On the one hand, this solves the problems of data collection difficulties and whether the collected logo data conforms to the distribution of real-world application scenarios when building large-scale brand logo datasets. On the other hand, it enables the rapid construction of brand logo datasets, accelerates model training iteration, improves the generalization ability of the brand logo recognition model by adding a large number of adversarial examples from real-world application scenarios, and solves the problem of class imbalance by expanding the sample size for brands with insufficient data, thereby improving the brand logo model's recognition performance and ultimately enabling more effective brand logo recognition.

[0057] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0058] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0059] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0060] In the description of this invention, it should be noted that the terms "top," "bottom," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship commonly used when the product of the invention is in use. They are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. In addition, the terms "first," "second," etc., are only used to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0061] In the description of this invention, it should also be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," and "link" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0062] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0063] Example

[0064] Please see Figure 1 This embodiment provides an application scenario diagram of a brand recognition method and device.

[0065] like Figure 1 As shown, the brand recognition method and apparatus provided by this invention can be applied to a cloud server 100, which is connected to a user client 200 via the Internet or other means. When performing brand logo recognition, the cloud server 100 obtains manually labeled information via the Internet or the user client 200, and sends the saved brand logo data to the user client 200 or other locations for storage. The dataset used to store the brand logo data can be located on the cloud server 100, or on other devices or terminals connected to the cloud server 100.

[0066] It should be noted that the text content classification method and apparatus provided by the present invention can also be applied to local terminals other than the cloud server 100, such as PCs, smartphones, tablets, or other devices with data processing and data interaction functions.

[0067] Based on the above, such as Figure 2 As shown, a brand identification method is provided according to an embodiment of the present invention. The method includes:

[0068] Step S101: Collect logo data from multiple brands for the first time according to a preset number, and establish a brand logo dataset.

[0069] Before performing brand logo recognition, a recognition model needs to be built, which in turn requires a certain amount of data. Compared to traditional deep convolutional neural network-based training methods, this method requires a much smaller initial amount of logo data, primarily relying on subsequent steps of introducing manually labeled information to train the model. As one implementation method in this example, collecting logo data from a total of 3000 brands is sufficient when initially building the brand logo dataset.

[0070] Step S102: Train the recognition model using a deep convolutional neural network-based training method to obtain the brand logo recognition model with the best performance based on the brand logo dataset.

[0071] In this step, the training method based on deep convolutional neural networks can be any currently available method, as long as it can produce a brand logo recognition model with optimal performance based on the current dataset.

[0072] Step S103: Use the brand logo recognition model to recognize the brand logo data information to be recognized, and save the recognition results and corresponding application scenario data.

[0073] Once the initial brand logo recognition model is obtained, it can be used to recognize brand logos. Specifically, the brand logo recognition model is used to deploy an automatic brand logo annotation algorithm service. This algorithm service is then used on various application platforms to recognize brand logos and automatically annotate brand logo data, obtaining and saving the recognition results. Simultaneously, the application scenario data used during recognition is also saved.

[0074] Step S104: Receive manually labeled information from external input and process the recognition results and application scenario data accordingly.

[0075] Due to the insufficient number of samples in the initial dataset, even if a brand logo recognition model with optimal performance is obtained through deep convolutional neural network training, numerous problems will inevitably arise when deployed for recognition in real-world applications. Therefore, at this stage, it is necessary to further introduce external human annotation. Based on the content of the human-annotated information, the recognition results obtained through the algorithm model are reprocessed to improve the accuracy of recognition.

[0076] Step S105: Update the processed recognition results and application scenario data to the brand logo dataset for the next recognition model training.

[0077] After executing step S105, step S102 is executed next. Using the updated dataset, the algorithm model is trained again. Based on the updated data content, the retrained recognition model will have better performance. This process continues. As the dataset iterates a certain number of times, the recognition model is trained a certain number of times. Furthermore, each update to the dataset involves adjustments based on manually standardized information, ensuring a better match to real-world application scenarios even with a smaller dataset.

[0078] The processed data is then expanded into the existing dataset, making the dataset more suitable for the application scenario. By using this dataset, the model can be iterated to achieve better recognition performance and generalization ability in the application scenario.

[0079] The brand recognition method provided in this embodiment addresses two main issues: firstly, the difficulty of data collection and the question of whether the collected logo data matches the distribution of real-world application scenarios when building large-scale brand logo datasets; secondly, it enables rapid construction of brand logo datasets, accelerates model training iterations, improves the generalization ability of the brand logo recognition model by adding numerous adversarial examples from real-world application scenarios, and resolves the problem of class imbalance by expanding the sample size for brands with insufficient data, thereby improving the performance of the brand logo model and ultimately enabling more effective brand logo recognition.

[0080] like Figure 3 As shown, another embodiment of the present invention provides a brand identification method, which includes:

[0081] Step S201: Collect logo data from multiple brands for the first time according to a preset number, and establish a brand logo dataset.

[0082] Step S202: Train the recognition model using a deep convolutional neural network-based training method to obtain the brand logo recognition model with the best performance based on the brand logo dataset.

[0083] Step S203: Use the brand logo recognition model to recognize the brand logo data information to be recognized, and save the recognition results and corresponding application scenario data.

[0084] Step S204: Receive manually labeled information from external input. For those labeled as correctly identified, save the identified brand logo data, application scenario data, and labeling information. For those labeled as incorrectly identified, clean the identification results and application scenario data to correct the identification results.

[0085] In this embodiment, the manually labeled information includes three types of information, namely:

[0086] The first label indicates that the brand logo data for this application scenario was correctly identified; the second label indicates that the brand logo data for this application scenario was incorrectly identified, along with the correct identification result; the third label indicates that no brand logo data was identified for this application scenario.

[0087] In other words, when external manual annotation is performed, there are three scenarios: the algorithm model identifies the logo correctly, the identification is incorrect, and the brand logo is not present in the application scenario, resulting in no identification. The annotation information differs for each scenario. For correctly identified logos, the annotation is "correctly identified." For incorrectly identified logos, in addition to annotating the incorrect identification, the corrected identification result also needs to be annotated. For logos not identified, no annotation is provided; that is, blank annotation information is provided.

[0088] Based on the saved recognition results and corresponding application scenario data, perform corresponding processing operations according to the content of the manually labeled information.

[0089] Step S205: Add the saved identified brand logo data, application scenario data, and annotation information to the brand logo dataset; add the saved cleaned brand logo data, application scenario data, and annotation information to the brand logo dataset; add the application scenario data for which no brand logo data was identified as background scenario data and corresponding blank annotation information to the brand logo dataset.

[0090] After processing the recognition results and the corresponding application scenario data, the processed information needs to be updated in the brand logo dataset. For the three different situations mentioned above, the corresponding data update methods are used to update the brand logo dataset so that the brand logo dataset can use the new data for the next recognition model training.

[0091] Therefore, after executing step S205, step S202 is executed again, and the loop is repeated continuously. The updated dataset is used to train the algorithm model again. The recognition model obtained by retraining based on the updated data content will have better performance.

[0092] In summary, the brand recognition method provided in this embodiment addresses two key issues. First, it solves the difficulties in data collection and the question of whether the collected logo data accurately reflects real-world application scenarios when building large-scale brand logo datasets. Second, it enables rapid construction of brand logo datasets, accelerates model training iterations, improves the generalization ability of the brand logo recognition model by adding numerous adversarial examples from real-world application scenarios, and addresses class imbalance by expanding the sample size for brands with insufficient data. These improvements enhance the brand logo recognition model's performance, leading to more effective brand logo recognition.

[0093] like Figure 4 As shown, the brand identification device provided by this invention includes:

[0094] Data set creation unit 110 is used to acquire logo data from multiple brands collected for the first time according to a preset quantity, and to create a brand logo dataset;

[0095] The recognition model training unit 120 is used to train the recognition model using a deep convolutional neural network-based training method to obtain a brand logo recognition model with the best performance based on the brand logo dataset.

[0096] The brand data recognition unit 130 is used to recognize the brand logo data information to be recognized using the brand logo recognition model, and save the recognition results and corresponding application scenario data.

[0097] The identification data processing unit 140 is used to receive manually labeled information input from the outside and to process the identification results and application scenario data accordingly.

[0098] The dataset management unit 150 is used to update the processed recognition results and application scenario data into the brand logo dataset for the next recognition model training.

[0099] In a preferred embodiment of this invention, the manually labeled information specifically includes:

[0100] This indicates that the brand logo data for this application scenario has been correctly identified.

[0101] This indicates the labeling information for incorrect identification and the correct identification result, used to indicate that the brand logo data for this application scenario was identified incorrectly.

[0102] Blank annotation information is used to indicate that no brand logo data was identified for this application scenario.

[0103] In a preferred embodiment of this invention, the identification data processing unit 140 is specifically used for:

[0104] For those correctly identified, the identified brand logo data, application scenario data, and annotation information will be saved;

[0105] For those labeled as having recognition errors, the recognition results and application scenario data are cleaned and corrected to the correct recognition results.

[0106] In a preferred embodiment of this invention, the dataset management unit 150 is specifically used for:

[0107] Add the saved identified brand logo data, application scenario data, and annotation information to the brand logo dataset;

[0108] Add the saved, cleaned brand logo data, application scenario data, and annotation information to the brand logo dataset;

[0109] Application scenario data that did not identify brand logo data were added to the brand logo dataset as background scenario data and corresponding blank annotation information.

[0110] The brand recognition device provided in this embodiment of the invention is used to implement the above-described brand recognition method. Therefore, the specific implementation method is the same as the above-described method and will not be repeated here.

[0111] like Figure 5 The diagram shows a structural block diagram of an electronic device 300 provided in an embodiment of the present invention. This electronic device 300 can be a smartphone, tablet computer, e-reader, or other electronic device capable of running applications. The electronic device 300 in this application may include one or more of the following components: a processor 310, a memory 320, and one or more applications, wherein the one or more applications can be stored in the memory 320 and configured to be executed by one or more processors 310, and the one or more applications are configured to perform the methods described in the foregoing method embodiments.

[0112] Processor 310 may include one or more processing cores. Processor 310 connects to various parts within the electronic device 300 using various interfaces and lines, and performs various functions and processes data of the electronic device 300 by running or executing instructions, programs, code sets, or instruction sets stored in memory 320, and by calling data stored in memory 320. Optionally, processor 310 may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). Processor 310 may integrate one or a combination of several of the following: Central Processing Unit (CPU), Graphics Processing Unit (GPU), and modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the displayed content; and the modem handles wireless communication. It is understood that the modem may also not be integrated into processor 310 and may be implemented separately using a communication chip.

[0113] The memory 320 may include random access memory (RAM) or read-only memory (ROM). The memory 320 can be used to store instructions, programs, code, code sets, or instruction sets. The memory 320 may include a program storage area and a data storage area. The program storage area may store instructions for implementing an operating system, instructions for implementing at least one function (such as touch functionality, sound playback functionality, image playback functionality, etc.), and instructions for implementing the various method embodiments described below. The data storage area may also store data created by the terminal during use (such as phonebook data, audio and video data, chat log data, etc.).

[0114] like Figure 6 The diagram shows a structural block diagram of a computer-readable storage medium 400 provided in an embodiment of the present invention. The computer-readable medium stores program code 410, which can be called by a processor to execute the methods described in the above method embodiments.

[0115] The computer-readable storage medium 400 may be an electronic memory such as flash memory, EEPROM (Electrically Erasable Programmable Read-Only Memory), EPROM, hard disk, or ROM. Optionally, the computer-readable storage medium 400 includes a non-transitory computer-readable storage medium. The computer-readable storage medium 400 has storage space for program code 410 that performs any of the method steps described above. This program code 410 can be read from or written to one or more computer program products. The program code 410 may, for example, be compressed in a suitable form.

[0116] In summary, this invention provides a brand recognition method, device, electronic device, and storage medium that rapidly generates a large number of scene image-text datasets using corpus information, background image libraries, and font libraries. This not only reduces the difficulty of data acquisition and annotation but also allows for the rapid acquisition of images with diverse backgrounds and fonts, facilitating model training and testing for researchers without being limited by insufficient training data or long annotation cycles. This solution is an important source of training datasets for OCR text detection and recognition tasks, featuring diverse backgrounds and fonts. It not only compensates for insufficient image-text training datasets, thus reducing manual annotation, but also increases the diversity of image-text datasets, thereby improving the generalization ability of OCR text recognition.

[0117] In the embodiments disclosed in this application, it should be understood that the disclosed apparatus and methods can also be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0118] In addition, the functional modules in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0119] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A brand identification method, characterized in that, The method includes: Collect logo data from multiple brands according to a preset number for the first time, and build a brand logo dataset; The recognition model is trained using a deep convolutional neural network-based training method to obtain the brand logo recognition model with the best performance based on the brand logo dataset. The brand logo recognition model is used to recognize the brand logo data information to be recognized, and the recognition results and corresponding application scenario data are saved. Receive manually labeled information from external input and process the recognition results and application scenario data accordingly; The processed recognition results and application scenario data are updated to the brand logo dataset for the next recognition model training. This indicates that the brand logo data for this application scenario has been correctly identified. This indicates the labeling information for incorrect identification and the correct identification result, used to indicate that the brand logo data for this application scenario was identified incorrectly. Blank annotation information is used to indicate that no brand logo data was identified for this application scenario; The steps for processing the identification results and application scenario data specifically include: For those correctly identified, the identified brand logo data, application scenario data, and annotation information will be saved; For those labeled as incorrect, the recognition results and application scenario data are cleaned and corrected to the correct recognition results. The step of updating the processed recognition results and application scenario data to the brand logo dataset specifically includes: Add the saved identified brand logo data, application scenario data, and annotation information to the brand logo dataset; Add the saved, cleaned brand logo data, application scenario data, and annotation information to the brand logo dataset; Application scenario data that did not identify brand logo data were added to the brand logo dataset as background scenario data and corresponding blank annotation information.

2. A brand identification device, characterized in that, The device includes: The dataset creation unit is used to acquire logo data from multiple brands collected for the first time according to a preset number, and to create a brand logo dataset; The recognition model training unit is used to train the recognition model using a deep convolutional neural network-based training method to obtain the brand logo recognition model with the best performance based on the brand logo dataset. The brand data recognition unit is used to recognize the brand logo data information to be recognized using the brand logo recognition model, and to save the recognition results and corresponding application scenario data. The identification data processing unit is used to receive manually labeled information from external input and to process the identification results and application scenario data accordingly. The dataset management unit is used to update the processed recognition results and application scenario data into the brand logo dataset for the next recognition model training. The manually labeled information specifically includes: This indicates that the brand logo data for this application scenario has been correctly identified. This indicates the labeling information for incorrect identification and the correct identification result, used to indicate that the brand logo data for this application scenario was identified incorrectly. Blank annotation information is used to indicate that no brand logo data was identified for this application scenario; The identification data processing unit is specifically used for: For those correctly identified, the identified brand logo data, application scenario data, and annotation information will be saved; For those labeled as incorrect, the recognition results and application scenario data are cleaned and corrected to the correct recognition results. The dataset management unit is specifically used for: Add the saved identified brand logo data, application scenario data, and annotation information to the brand logo dataset; Add the saved, cleaned brand logo data, application scenario data, and annotation information to the brand logo dataset; Application scenario data that did not identify brand logo data were added to the brand logo dataset as background scenario data and corresponding blank annotation information.

3. An electronic device, characterized in that, include: One or more processors; Memory; One or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the one or more processors, the one or more applications being configured to perform the method as described in claim 1.

4. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores program code that can be invoked by a processor to execute the method as described in claim 1.

Images