A data annotation system, method, terminal and storage medium

Abstract

The present application relates to a data annotation system, method, terminal and storage medium. The system includes a data visualization module, a parameter setting module and an annotation interaction module; the parameter setting module is used to import the file to be annotated and set the annotation parameters of the file to be annotated; the data visualization module is used to display the file to be annotated through a window according to the set annotation parameters; the annotation interaction module is used to display the label elements according to the set annotation parameters, and receive the label elements selected by the user for the file to be annotated, and generate the annotated file data. The embodiment of the present application provides a unified visual data annotation platform for data that needs to be manually annotated, so that the annotator can quickly and easily generate data labels based on visualization, which reduces the professional ability requirements for the annotator. For annotators with no programming experience, data annotation can also be performed simply, quickly and accurately, and the process of data preprocessing for algorithm engineers can also be simplified.

Description

Technical Field

[0001] The present application belongs to the field of machine learning technology, and in particular relates to a data annotation system, method, terminal and storage medium. Background Art

[0002] Whether it is traditional machine learning or deep learning algorithms, the classification algorithms in them are all supervised training methods. For supervised learning that requires a large number of samples, the data labeling process is the most complicated and labor-intensive work on the basis of the relatively complete machine learning framework, especially the labeling process of time series data. The reason why supervised learning is called supervised is that there are real labels as a reference for the predicted values ​​of the machine learning algorithm, so that indicators such as loss can be calculated based on the distance between the predicted values ​​and the real data, thereby continuously iterating and optimizing the algorithm. The real labels come from manual labeling, which often requires a lot of time and energy from professionals.

[0003] At present, in the data labeling process represented by digital signals, the labelers often need to first obtain the waveform data of all cycles or a certain time length, and then the programmers draw the waveform of a certain period of data or the waveform drawings provided by other tools to determine which category of labels the data in this time period belongs to, and then label the waveform in this time period accordingly. Another method is to use a trained model to label new data. This method is essentially a prediction and often requires manual review. In addition, for many fields where the model has not yet been trained, the initial data labeling method still requires manual operation.

[0004] In summary, the existing data labeling lacks a unified data labeling platform and requires extremely high manual programming capabilities. For staff with no programming experience, it is impossible to label data quickly and accurately. Summary of the invention

[0005] The present application provides a data annotation system, method, terminal and storage medium, which aim to solve at least one of the above-mentioned technical problems in the prior art to a certain extent.

[0006] In order to solve the above problems, this application provides the following technical solutions:

[0007] A data annotation system comprises a data visualization module, a parameter setting module and an annotation interaction module; the parameter setting module is used to import a file to be annotated and set annotation parameters of the file to be annotated; the data visualization module is used to display the file to be annotated through a window according to the set annotation parameters; the annotation interaction module is used to display label elements according to the set annotation parameters, receive label elements selected by a user for the file to be annotated, and generate annotated file data.

[0008] The technical solution adopted by the embodiment of the present application also includes: the data visualization module includes a previous window data visualization unit, a current window data visualization unit and a next window data visualization unit;

[0009] The previous window data visualization unit, the current window data visualization unit and the next window data visualization unit are respectively used to display the waveform data of the previous window, the waveform data of the current window and the waveform data of the next window according to the label parameters set by the parameter setting module.

[0010] The technical solution adopted by the embodiment of the present application also includes: the parameter setting module includes a file selection unit, a label export unit, a window size setting unit, a step size setting unit, a label element setting unit and a label management unit;

[0011] The file selection unit is used to select an import path for the file to be annotated and import the file;

[0012] The label exporting unit is used to export labels;

[0013] The window size setting unit is used to set the window size win_size of the waveform data, where win_size represents the number of data in a window;

[0014] The step size setting unit is used to set the step size stride_size of the waveform data; the starting position index of the data in a window is s=i*stride_size, and the ending position index is e=s+win_size; where i represents the current window, starting from 0;

[0015] The tag element setting unit is used to set the tag element of the file to be marked;

[0016] The tag management unit is used to store or delete the set tag elements.

[0017] The technical solution adopted in the embodiment of the present application also includes: assuming that the length of the file to be annotated is T, the constraint relationship between the length T, the window size win_size and the step size stride_size is: T>=(stride_size+win_size).

[0018] The technical solution adopted by the embodiment of the present application also includes: the parameter setting module also includes a check unit, the check unit is used to determine whether the file to be marked is high-dimensional data, and if it is high-dimensional data, set the main channel main_channel parameter of the high-dimensional data;

[0019] The high-dimensional data indicates that the number of channels of the file to be labeled is greater than 1.

[0020] The technical solution adopted by the embodiment of the present application also includes: the annotation interaction module includes a window selection unit, a label checking unit and a label selection unit;

[0021] The window selection unit is used to select the previous window or the next window;

[0022] The label checking unit is used to display the label elements set by the parameter setting module and select the label elements of the current window from the label elements;

[0023] The label selection unit is used to determine whether to mark the data waveform of the current window with the selected label element.

[0024] Another technical solution adopted by the embodiment of the present application is: a data labeling method, comprising:

[0025] Import the file to be annotated and set the annotation parameters of the file to be annotated; the annotation parameters include window size, step size and label elements;

[0026] Display the files to be annotated through a window according to the set annotation parameters;

[0027] Display label elements according to the set labeling parameters, receive label elements selected by the user for the document to be labeled, and generate labeled file data.

[0028] The technical solution adopted by the embodiment of the present application also includes: the setting of annotation parameters of the file to be annotated also includes:

[0029] Determine whether the file to be annotated is high-dimensional data. If it is high-dimensional data, set the main channel main_channel parameter of the high-dimensional data;

[0030] The high-dimensional data indicates that the number of channels of the file to be labeled is greater than 1.

[0031] Another technical solution adopted by the embodiment of the present application is: a terminal, the terminal includes a processor and a memory coupled to the processor, wherein:

[0032] The memory stores program instructions for implementing the data labeling method;

[0033] The processor is used to execute the program instructions stored in the memory to control data labeling.

[0034] Another technical solution adopted by the embodiment of the present application is: a storage medium storing program instructions executable by a processor, wherein the program instructions are used to execute the data labeling method.

[0035] Compared with the prior art, the beneficial effects of the embodiments of the present application are as follows: the data labeling system, method, terminal and storage medium of the embodiments of the present application provide a unified visual data labeling platform for data that needs to be manually labeled, so that labelers can quickly and easily generate data labels based on visualization, reducing the professional ability requirements for labelers, and labelers with no programming experience can also perform data labeling work simply, quickly and accurately, and at the same time can also simplify the process of data preprocessing for algorithm engineers. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 is a structural diagram of a data annotation system according to an embodiment of the present application;

[0037] Figure 2 A prototype diagram of a data annotation system according to an embodiment of the present application;

[0038] Figure 3 A flowchart of a data labeling method according to an embodiment of the present application;

[0039] Figure 4 This is a schematic diagram of the terminal structure of an embodiment of the present application;

[0040] Figure 5 A schematic diagram of the structure of a storage medium according to an embodiment of the present application. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0042] See also Figure 1 , is a structural diagram of the data annotation system of the embodiment of the present application. The data annotation system of the embodiment of the present application includes a data visualization module 10, a parameter setting module 20 and an annotation interaction module 30. Among them, the parameter setting module 20 is used to import the file to be annotated, and set the annotation parameters of the file to be annotated, and store the set annotation parameters in the memory; the data visualization module 10 is used to display the file to be annotated through a window according to the set annotation parameters; the annotation interaction module 30 is used to display the label elements according to the set annotation parameters, and receive the label elements selected by the user for the file to be annotated, and generate the annotated file data.

[0043] Specifically, Figure 2As shown, it is a prototype diagram of the data annotation system of the embodiment of the present application. Among them, the data visualization module 10 includes a previous window data visualization unit 1, a current window data visualization unit 2 and a next window data visualization unit 3. The previous window data visualization unit 1, the current window data visualization unit 2 and the next window data visualization unit 3 are respectively used to display the waveform data of the previous window, the waveform data of the current window and the waveform data of the next window according to the label parameters set by the parameter setting module 20. The waveform data of the three windows represent the historical, current and future annotation data respectively.

[0044] The parameter setting module 20 includes a file selection unit 4, a label export unit 5, a window size setting unit 6, a step size setting unit 7, a label element setting unit 8, a label management unit 9 and a check unit 11; wherein, the file selection unit 4 is used to select the import path of the file to be annotated; the label export unit 5 is used to export labels; the window size setting unit 6 is used to set the window size win_size of the waveform data; the step size setting unit 7 is used to set the step size stride_size of the waveform data; taking the file to be annotated as time series data as an example, the time series data is usually a data point of length T obtained by sampling at a certain frequency fs, and the number of data in a window is win_size; the starting position index of the data in the window is s=i*stride_size (i represents the current window, starting from 0), and the end position index is e=s+win_size. There is the following constraint relationship between the length T, the window size win_size and the step size stride_size: T>=(stride_size+win_size). The label element setting unit 8 is used to set the element tags of the label, that is, to set the label to be annotated for the data. After a tag element is set by the tag element setting unit 8, the tag element is stored in the memory by the tag management unit 9 and displayed in the form of a drop-down list; the tag element that has been added can also be deleted by the tag management unit 9.

[0045] The check unit 11 is used to determine whether the file to be labeled is high-dimensional data, and displays a high-dimensional check box is_G1D. If it is high-dimensional data, the is_G1D option is checked; otherwise, the high-dimensional check box is_G1D is empty. High-dimensional data means that the imported file to be labeled includes multiple channels. For high-dimensional data, the main channel main_channel parameter needs to be set to determine which channel's data is used as the main channel for labeling.

[0046] After all parameters are set, you can jump to the annotation interaction module 30 to start annotation by clicking the start button 12. After clicking the start button 12, according to the window size win_size and the starting position index s=i*stride_size (i represents the current window, starting from 0), the data waveform of the current (first) window and the data waveform of the next window are displayed respectively through the current window data visualization unit 2 and the next window data visualization unit 3.

[0047] The annotation interaction module 30 includes a window selection unit 14, a label check unit 16 and a label selection unit 17. The window selection unit 14 includes a previous window button and a next window button, which are used to select the previous window or the next window. The label check unit 16 is used to display label elements, and select label elements of the data waveform of the current window from the label elements or delete the selected label elements; the label selection unit 17 is used to determine whether to mark the selected label elements for the data waveform of the current window. Specifically, when the start button 12 is clicked, the label check unit 16 will display the corresponding label elements in the form of a drop-down list. The corresponding label elements can be selected from the drop-down list according to the data waveform diagram of the current window, and the label selection unit 17 determines whether to mark the selected label elements for the data waveform of the current window. After completing the data annotation of the current window, the next window can be selected by clicking the window selection unit 14, and the above annotation process is repeated until the data annotation of the file to be annotated is completed. During the annotation process, if a certain annotated window needs to be modified, it can also be returned to the window to be modified by clicking the window selection unit 14 and reselect the label. After completing the labeling of all data, you can click the Finish Labeling button 13 to confirm the completion, and export the labeled file data through the label export unit 5. The exported file data can be directly used for model training of machine learning or deep learning.

[0048] Based on the above, the annotation process of the data annotation system of the embodiment of the present application specifically includes: first, the import path of the file to be annotated is selected by the file selection unit 4 and the file is imported, and the window size win_size, step size stride_size, tag element tags and high-dimensional check box is_G1D and other parameters are set for the imported file to be annotated by the window size setting unit 6, the step size setting unit 7, the tag element setting unit 8 and the check unit 11, and the tag management unit 9 adds or deletes the set tag elements to complete the parameter setting of the imported file. Then, by clicking the start button 12, jump to the annotation interaction module 30 to start annotation. After clicking the start button 12, the data waveform of the current window and the data waveform of the next window are displayed respectively by the current window data visualization unit 2 and the next window data visualization unit 3, and the set tag elements are displayed in the form of a drop-down list by the tag check unit 16. The corresponding tag element can be selected from the drop-down list according to the data waveform of the current window, and the tag selection unit 17 determines whether the data waveform of the current window is marked with the selected tag element. After the data annotation of the current window is completed, the next window is selected through the window selection unit 14. At this time, the data waveforms of the previous window, the current window and the next window are respectively displayed through the previous window data visualization unit 1, the current window data visualization unit 2 and the next window data visualization unit 3, and the above annotation process is repeated to annotate the data waveform of the current window, and the cycle is repeated until the data annotation of the entire file to be annotated is completed. Finally, after all data annotations are completed, the completion is confirmed by clicking the completion annotation button 13, and the annotated file is exported through the label export unit 5 to complete this data annotation.

[0049] The data labeling system of the embodiment of the present application provides a unified visual data labeling platform for data that needs to be manually labeled, so that labelers can quickly and easily generate data labels based on visualization, reducing the professional ability requirements for labelers. Labelers with no programming experience can also perform data labeling work simply, quickly and accurately, and can also simplify the process of data preprocessing for algorithm engineers.

[0050] See also Figure 3 , is a flow chart of a data labeling method according to an embodiment of the present application. The data labeling method according to an embodiment of the present application comprises the following steps:

[0051] S1: Obtain the file to be annotated imported by the user;

[0052] S2: Setting the annotation parameters such as window size, step size, label elements, etc. of the file to be annotated through the parameter setting module;

[0053] S3: Determine whether the file to be labeled is high-dimensional data. If it is high-dimensional data, execute S4; otherwise, execute S5;

[0054] Among them, high-dimensional data refers to data with a channel number greater than 1. For high-dimensional data, it is necessary to set the main channel main_channel parameter to determine which channel's data is used as the main channel for labeling.

[0055] S4: Set the main channel main_channel parameters used for labeling;

[0056] S5: Based on the set annotation parameters, the data waveform of the current window is displayed through the visualization module, and the set label elements are displayed in the form of a drop-down list through the annotation interaction module. The label element of the waveform of the current window is selected through the drop-down list to complete the data annotation of the current window;

[0057] S6: Select the next window and re-execute S5 until the data labeling of the file to be labeled is completed and the labeled file data is obtained;

[0058] S7: Export the marked file data.

[0059] Based on the above, the data labeling method of the embodiment of the present application provides a unified visual data labeling platform for data that needs to be manually labeled, so that labelers can quickly and easily generate data labels based on visualization, reducing the professional ability requirements for labelers. Labelers with no programming experience can also perform data labeling work simply, quickly and accurately, and at the same time can also simplify the process of data preprocessing for algorithm engineers.

[0060] See also Figure 4 , is a schematic diagram of a terminal structure of an embodiment of the present application. The terminal 50 includes a processor 51 and a memory 52 coupled to the processor 51.

[0061] The memory 52 stores program instructions for implementing the above data labeling method.

[0062] The processor 51 is used to execute program instructions stored in the memory 52 to control data labeling.

[0063] The processor 51 may also be referred to as a CPU (Central Processing Unit). The processor 51 may be an integrated circuit chip having the ability to process signals. The processor 51 may also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0064] See also Figure 5 , which is a schematic diagram of the structure of the storage medium of the embodiment of the present application. The storage medium of the embodiment of the present application stores a program file 61 that can implement all the above methods, wherein the program file 61 can be stored in the above storage medium in the form of a software product, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor (processor) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or terminal devices such as a computer, a server, a mobile phone, and a tablet.

[0065] The above description of the disclosed embodiments enables professionals and technicians in the field to implement or use the present application. Various modifications to these embodiments will be apparent to professionals and technicians in the field, and the general principles defined in this application can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown in the present application, but will conform to the widest range consistent with the principles and novel features disclosed in the present application.

Claims

1. A data annotation system, characterized in that: It includes a data visualization module, a parameter setting module and a marking interaction module; the parameter setting module is used to import the file to be marked and set the marking parameters of the file to be marked; the data visualization module is used to display the file to be marked through a window according to the set marking parameters; The annotation interaction module is used to display the label elements according to the set annotation parameters, receive the label elements selected by the user for the document to be annotated, and generate the annotated document data; The data visualization module includes a previous window data visualization unit, a current window data visualization unit and a next window data visualization unit; The previous window data visualization unit, the current window data visualization unit and the next window data visualization unit are respectively used to display the waveform data of the previous window, the waveform data of the current window and the waveform data of the next window according to the label parameters set by the parameter setting module.

2. The data annotation system according to claim 1, characterized in that: The parameter setting module includes a file selection unit, a label export unit, a window size setting unit, a step size setting unit, a label element setting unit and a label management unit; The file selection unit is used to select an import path for the file to be annotated and import the file; The label exporting unit is used to export labels; The window size setting unit is used to set the window size win_size of the waveform data, where win_size represents the number of data in a window; The step size setting unit is used to set the step size stride_size of the waveform data; the starting position index of the data in a window is s=i*stride_size, and the ending position index is e=s+win_size; where i represents the current window, starting from 0; The tag element setting unit is used to set the tag element of the file to be marked; The tag management unit is used to store or delete the set tag elements.

3. The data annotation system according to claim 2, characterized in that: Assuming that the length of the file to be annotated is T, the constraint relationship among the length T, the window size win_size and the step size stride_size is: T>=(stride_size+win_size).

4. The data annotation system according to claim 3, characterized in that: The parameter setting module also includes a check unit, which is used to determine whether the file to be labeled is high-dimensional data, and if it is high-dimensional data, set the main channel main_channel parameter of the high-dimensional data; The high-dimensional data indicates that the number of channels of the file to be labeled is greater than 1.

5. The data annotation system according to any one of claims 1 to 4, characterized in that: The annotation interaction module includes a window selection unit, a label checking unit and a label selection unit; The window selection unit is used to select the previous window or the next window; The label checking unit is used to display the label elements set by the parameter setting module and select the label elements of the current window from the label elements; The label selection unit is used to determine whether to mark the data waveform of the current window with the selected label element.

6. A data annotation method using the data annotation system of claim 1, characterized in that: include: Import the file to be annotated and set the annotation parameters of the file to be annotated; The annotation parameters include window size, step size and label elements; Display the files to be annotated through a window according to the set annotation parameters; Display label elements according to the set labeling parameters, receive label elements selected by the user for the document to be labeled, and generate labeled file data.

7. The data labeling method according to claim 6, characterized in that: The step of setting the annotation parameters of the file to be annotated also includes: Determine whether the file to be annotated is high-dimensional data. If it is high-dimensional data, set the main channel main_channel parameter of the high-dimensional data; The high-dimensional data indicates that the number of channels of the file to be labeled is greater than 1.

8. A terminal, characterized in that: The terminal includes a processor and a memory coupled to the processor, wherein: The memory stores program instructions for implementing the data labeling method according to claim 6 or 7; The processor is used to execute the program instructions stored in the memory to control data labeling.

9. A storage medium, characterized in that: Program instructions executable by a processor are stored, and the program instructions are used to execute the data labeling method described in claim 6 or 7.

Images