CN116361279A - Cross-platform artificial intelligent data labeling method - Google Patents
Cross-platform artificial intelligent data labeling method Download PDFInfo
- Publication number
- CN116361279A CN116361279A CN202310645802.7A CN202310645802A CN116361279A CN 116361279 A CN116361279 A CN 116361279A CN 202310645802 A CN202310645802 A CN 202310645802A CN 116361279 A CN116361279 A CN 116361279A
- Authority
- CN
- China
- Prior art keywords
- data
- image
- annotation
- labeling
- data set
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000002372 labelling Methods 0.000 title claims abstract description 65
- 238000011161 development Methods 0.000 claims abstract description 13
- 238000013473 artificial intelligence Methods 0.000 claims abstract description 9
- 238000004140 cleaning Methods 0.000 claims description 41
- 238000000034 method Methods 0.000 claims description 15
- 238000010276 construction Methods 0.000 claims description 13
- 230000018109 developmental process Effects 0.000 claims description 11
- 238000012545 processing Methods 0.000 claims description 9
- 230000000694 effects Effects 0.000 claims description 6
- 238000012549 training Methods 0.000 claims description 6
- 238000000605 extraction Methods 0.000 claims description 5
- 238000012795 verification Methods 0.000 claims description 5
- 238000009877 rendering Methods 0.000 claims description 4
- 239000003086 colorant Substances 0.000 claims description 3
- 238000001514 detection method Methods 0.000 claims description 3
- 238000003709 image segmentation Methods 0.000 claims description 3
- 238000012544 monitoring process Methods 0.000 description 8
- 230000008569 process Effects 0.000 description 6
- VYZAMTAEIAYCRO-UHFFFAOYSA-N Chromium Chemical compound [Cr] VYZAMTAEIAYCRO-UHFFFAOYSA-N 0.000 description 1
- 230000002457 bidirectional effect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000008676 import Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012827 research and development Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/21—Design, administration or maintenance of databases
- G06F16/215—Improving data quality; Data cleansing, e.g. de-duplication, removing invalid entries or correcting typographical errors
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/28—Databases characterised by their database models, e.g. relational or object models
- G06F16/284—Relational databases
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/51—Indexing; Data structures therefor; Storage structures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/583—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
- G06F16/5846—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content using extracted text
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/5866—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using information manually generated, e.g. tags, keywords, comments, manually generated location and time information
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Databases & Information Systems (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Library & Information Science (AREA)
- Quality & Reliability (AREA)
- Software Systems (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a cross-platform artificial intelligence data labeling method, which establishes an artificial intelligence data labeling mode based on an integrated development environment, can realize rapid cross-platform deployment of data labeling, has the advantages of rich page styles of the integrated development environment and the like, and provides a more efficient and convenient data labeling realization way for development and users.
Description
Technical Field
The invention belongs to the technical field of computer software development, and particularly relates to a cross-platform artificial intelligent data labeling method.
Background
The diversity and the large-scale property of the data set are important to the research and development of related algorithms in the artificial intelligence field, the primary condition for improving the accuracy of the algorithm is to acquire massive and high-quality application scene data, and the labeling of the massive data is an important basis for data application, so that the rapid development and the wide application in the artificial intelligence field bring great challenges to the data labeling technology. The existing data labeling mode mainly comprises data labeling of a B/S architecture and data labeling of a C/S architecture, wherein the data labeling mode of the B/S architecture has the problem of high deployment difficulty during cross-platform development because of deployment aiming at different platforms, and the data labeling of the C/S architecture has the problem of poor display effect of a labeling page. Therefore, the existing data labeling method cannot meet the use requirement of a user across platforms.
Disclosure of Invention
In view of the above, the invention provides a cross-platform artificial intelligence data labeling method, which realizes cross-platform data labeling based on an integrated development environment.
The invention provides a cross-platform artificial intelligence data labeling method, which comprises the following steps:
step 1, creating a data annotation window by adopting an integrated development environment, and providing functions of data set creation, data set introduction, image data annotation and text data annotation;
step 2, selecting a data set to be created in a data labeling window, and creating a storage structure of the data set to be labeled and corresponding construction configuration information, wherein the construction configuration information comprises names, data amounts, labeling types and description information of the data set to be labeled, the storage structure is a three-layer directory structure consisting of a type directory, a version directory and data files, and the labeling types comprise image types and text types;
step 3, selecting a data set to be imported in the data labeling window, copying the data to be labeled, which needs to be imported, into the data set to be labeled according to the construction configuration information, and refreshing the data labeling window; if the type of the data set to be marked is the image type, executing the step 5; if the type of the data set to be marked is a text type, executing the step 4;
step 4, acquiring text cleaning parameters, including objects, output paths and text cleaning modes of text cleaning, calling a text cleaning algorithm to finish cleaning a data set to be marked, and executing step 8;
step 5, acquiring a first storage path of the data set to be marked, and executing step 6 if the data set to be marked has quality problems; if the data quantity in the data set to be marked is less, executing the step 7; otherwise, executing the step 8;
step 6, taking the first storage path as an object of image cleaning, acquiring image cleaning parameters, executing image cleaning according to the image cleaning parameters, and storing a cleaning result into the second storage path; comparing the image data in the first storage path and the second storage path one by one, and updating the image data in the first storage path after removing the data with poor quality; if the data quantity in the data set to be marked is less, executing the step 7, otherwise executing the step 9;
step 7, taking the first storage path as an object of image enhancement, executing image enhancement according to the image enhancement parameters, and updating the image data in the first storage path according to the enhancement result;
step 8, calling a labeling tool to label part of the images in the first storage path, and updating the data set to be labeled after labeling is completed and labeling information is generated;
and 9, training the intelligent annotation model in the remote annotation server by adopting the annotated data, executing annotation of the unlabeled data in the to-be-annotated data set by adopting the remote annotation server according to the intelligent annotation parameters, and ending the flow.
Further, in the step 2, the image class includes object detection, image segmentation and image classification, and the text class includes entity identification and relation extraction.
Further, after the step 9 is completed, verification is performed on the image class data, which specifically includes:
creating a plurality of thread splitting data by using a workbench_threads () in node. Js to distinguish images with marked information and images without marked information; when the image with the marking information is previewed, a check API is called to map the marking information into the image to form new image data, then the image is previewed, the image which does not meet the requirement is singly marked, an artificial intelligent data marking window is rendered and refreshed after marking is finished, and a current check data set to be marked is updated.
Further, the separate labeling is to label a single image with a vscore.
Further, after the step 9 is completed, the text data is processed, which specifically includes:
the artificial intelligent data annotation window adopts a page paging mode to respectively display the data with annotation information and the data without annotation information in the pages with annotation information and the pages without annotation information, and the text category is rendered by carrying out form splitting and category adding columns in the pages with annotation information; determining a display mode of a verification effect according to the annotation type, adding a highlight class to entity elements in the marked page when the annotation type is entity identification, rendering different types of entities into different colors, and carrying out connection processing to relation elements in the marked page when the annotation type is relation extraction; and finally, sending the processing result to an artificial intelligent data labeling window for updating.
Further, the method also comprises the step of reading text data by adopting a database mode.
Further, the database is an SQLite database.
Advantageous effects
The invention establishes an artificial intelligent data annotation mode based on an integrated development environment, can realize rapid cross-platform deployment of data annotation, has the advantages of rich page styles of the integrated development environment and the like, and provides a more efficient and convenient data annotation realization way for development and use personnel.
Detailed Description
The present invention will be described in detail with reference to the following examples.
The invention provides a cross-platform artificial intelligent data labeling method, which specifically comprises the following steps:
and step 1, creating an artificial intelligent data annotation window by adopting an instruction createWebViewPanel in an integrated development environment, wherein the artificial intelligent data annotation window is used for providing operation functions of data set creation, data set introduction, image data annotation and text data annotation.
And 2, selecting a data set to be created in an artificial intelligent data annotation window, and creating a storage structure of the data set and corresponding construction configuration information, wherein the construction configuration information comprises names, data amounts, annotation types, description information and the like of the data set, and the storage structure is a three-layer directory structure consisting of a type directory, a version directory and data files. The annotation type comprises an image class and a text class, wherein the image class comprises object detection, image segmentation and image classification, and the text class comprises entity identification, relation extraction and the like.
The construction configuration information of the data set can be stored by adopting a Json file, and in particular, the construction configuration information of the data set can be written into the Json file by adopting a fs.writeFile () function.
The process of creating the dataset is: firstly, creating a type directory, calling command to acquire an input type name, and creating the type directory by adopting a fs.mkdir () function in node.js; then creating a version directory in the type directory, calling command to acquire the input version name, data quantity, labeling type, description information and the like, creating the version directory by adopting a fs.mkdir () function in node.js, and then writing the acquired construction configuration information into a Json file by using a fs.writeFile () function; finally, the fs.lstatsync () function is used to obtain the unique index value of the Json file to distinguish the versions of the data set, for example, to generate the V1.0 version.
For example, the format of the Json configuration information is as follows:
{
"ino" (document index): {
"name": "data set name",
"tagType": "label type",
"size": "data amount",
"descriptions": descriptive information ",
"index" serial number "
}
...
}。
And 3, selecting data set import in the artificial intelligent data annotation window, using a fs.readfile () function of node.js to read construction configuration information of the data set, opening a catalog according to the construction configuration information by adopting a vscore.window.showopendialog () function and a fs.readfile () function to select data to be annotated to be imported, copying the data to be annotated to the created data set to be annotated in a data stream mode by adopting fs.createReadstream () and fs.createWritestream (), for example, using a postMessage API to send the information successfully imported to the artificial intelligent data annotation window, and refreshing the artificial intelligent data annotation window after receiving the information.
Wherein node. Js is JavaScript operating environment based on Chrome V8 engine.
Step 4, if the type of the data set to be marked is an image type, executing a step 6; and if the type of the data set to be marked is a text type, executing the step 5.
Step 5, defining a text cleaning parameter comprising a text cleaning object, an output path and a text cleaning mode; calling a vscode, window, showopendialog () to acquire an output path of text cleaning, acquiring other parameters of the text cleaning through a UI event, calling a text cleaning algorithm by using child_process, exec () to execute text cleaning operation, and comparing the original text content with a cleaning result after cleaning is finished to delete the cleaning result so as to finish cleaning the data set to be marked; step 9 is performed.
Step 6, calling command to acquire a first storage path of the data set to be marked, checking the data set to be marked, and executing step 7 if the data set to be marked has quality problems; if the data quantity in the data set to be marked is less, executing the step 8; otherwise, step 9 is performed.
Step 7, defining image cleaning parameters including objects of image cleaning, de-approximation threshold values, de-blurring threshold values, whether de-blurring, whether de-approximation and the like; taking the first storage path as an object of image cleaning through a postMessage API, acquiring the rest image cleaning parameters through a UI event, and defining a second storage path for storing cleaned image class data; calling a cleaning command through child_process.exec (), executing image cleaning according to the image cleaning parameters, and storing a cleaning result into a second storage path after the cleaning is finished; comparing the image class data in the first storage path and the second storage path one by one, and updating the image class data in the first storage path by removing the data with poor quality through mv to finish the data cleaning of the data set to be marked; if the data amount in the data set to be marked is less, executing the step 8, otherwise executing the step 9.
Step 8, defining an image enhancement parameter comprising an image enhancement object and a third storage path serving as an output path, calling a postMessage API to take the first storage path as the image enhancement object, and acquiring the third storage path by using a vscore. And calling command to execute image enhancement operation according to the image enhancement parameters, and updating the image class data in the first storage path according to the enhancement result after the execution is finished, so as to complete the image enhancement of the data set to be annotated.
And 9, acquiring the annotation type of the data set to be annotated according to the unique index of the Json file corresponding to the data set to be annotated, acquiring a first storage path through an API of the node. Js, then calling an annotation tool to annotate part of the images in the first storage path after the first storage path is sent to the annotation tool by adopting the postMessage API, generating annotation information under the peer path of the annotation file after the annotation is completed, updating the data set to be annotated, and completing the annotation of part of the images in the data set to be annotated.
Step 10, training the intelligent labeling model in the remote labeling server by adopting the labeled data obtained in the step 9; defining intelligent labeling parameters comprising a data set to be labeled, a data set output version, a remote labeling Json file path, a remote labeling data path and a remote non-labeling data path, executing labeling of non-labeling data in the data set to be labeled by adopting a remote labeling server which completes training according to the intelligent labeling parameters, and ending the process.
The ssh2 module in node. Js can be used for achieving interaction between the local data set and the remote annotation server.
The step 10 is a process of executing marking of unlabeled data in a set of data to be marked by adopting a remote marking server for finishing training according to intelligent marking parameters, and specifically comprises the following steps:
and 10.1, configuring a server by adopting a package json configuration server field in an integrated development environment, wherein the package json configuration server field comprises a server IP address, a server user name and a server password and is used for locally connecting a server.
The format of the Json configuration information is as follows:
{
“mainconfig.severAddress”: {
"type": "object",
"default": {
"SeverIP": "default IP",
"username": default user name ",
"password" means "default password"
},
"properties": {
"severIP": {
"type": "string",
"description" server IP Address "
},
"username": {
"type": "string",
"description" Server user name "
},
"password": {
"type": "string",
"description" server password "
}
}
}
}
And step 10.2, acquiring a server configuration object by using a vscore.workspace.getconfiguration (). Get () API, calling a connect connection server in the ssh2 module, and calling a connect method to fill the server configuration information into connect.
And 10.3, calling command to acquire intelligent annotation parameters, uploading a data set to be annotated, a data set output version, a remote annotation Json file path, a remote annotation data path and a remote non-annotation data path in the intelligent annotation parameters to different addresses of a server by using an sftp.fastput () uploading file in a ssh2 module, asynchronously uploading by using a Promise mode, defining more than 3 Promise functions, transferring the remote annotation Json file path, the remote annotation data path and the remote non-annotation data path to the Promise functions, calling by using Promise.all (), monitoring whether uploading is successful or not by the thene function and the catch function, describing that uploading is successful after the thene function is executed, marking can be carried out, and outputting error information log functions after the catch function is executed.
10.4, calling exec () in ssh2 to execute command for marking, wherein the marking time length is related to the number of uploaded files and the configuration of a server, continuously monitoring marking information through an on monitoring event in the execution process, calling command to start a visual interface to display marking process in real time, displaying a stop marking button when monitoring that marking starts, clicking a stop marking button to stop marking work at the time, and executing command to delete a process occupied by server marking by calling exec () in ssh 2; after the labeling work is finished, after the monitoring event monitoring labeling is finished through on, sftp.fastget () in ssh2 is called to carry out file downloading after labeling and model file downloading after training, a model path is stored in an intelligent model object model, and a data set to be labeled is updated at the same time. The downloaded model file may be used for secondary labeling.
Step 10.5, acquiring intelligent model parameters, selecting data to be marked by using vscore.window.showopendialog () and then sending the selected result to an artificial intelligent data marking window for rendering through a postMessage API, and filling the data in the output version acquired page elements according to the bidirectional binding principle; uploading intelligent model parameters and data to be marked by using sftp.fastput () in the ssh2 module, if the uploaded intelligent model parameters and the data to be marked are at different positions, respectively encapsulating the uploaded functions by new Promise (), calling by Promise.all (), monitoring whether uploading is successful or not by the thene function and the catch function, executing the thene function to explain that uploading is successful, marking, and executing the catch function to output error information log; after the completion of the labeling is monitored through the on monitoring event, the sftp.fastget () in ssh2 is called to download the labeled file, and the data set to be labeled is updated.
In order to further improve the reliability of data labeling and reduce missed labeling data, the invention increases the verification of image data after the intelligent labeling of the data set to be labeled is completed in the execution step 10, and specifically comprises the following steps:
creating multithreaded new workbench () API split data to distinguish images with marked information and images without marked information by using workbench_threads () in node. Js; when the image with the marking information is previewed, a check API is called to map the marking information into the image to form new image data and then preview the new image data, and the image which does not meet the requirement is singly marked, namely, a single image is marked by using a vscore.
In addition, in order to improve the data display effect, the following processing of text data is added after the intelligent labeling of the data set to be labeled is completed in the execution step 10:
the artificial intelligent data annotation window displays the data with annotation information and the data without annotation information respectively in a page form of the page with annotation information and the page without annotation information, the text class is rendered by carrying out form splitting and class adding columns in the page with annotation information, and the form is dynamically rendered by a render function of JSX; determining a display mode of a verification effect according to the annotation type, adding a highlight class to entity elements in the annotated page when the annotation type is entity identification, rendering different types of entities into different colors, and carrying out connection processing on the relation elements in the annotated page by using jsPlumb. And finally, sending the processing result to an artificial intelligent data labeling window for updating through a postMessage API.
For text data, when the text file is too large, the text data can be directly read with lower efficiency, so the invention adopts a database mode to realize the reading of the text data, namely, the text data is firstly stored in the database, and then the text data is read through the operation of the database. Specifically, because the SQLite belongs to a lightweight database and does not depend on any third party software, the SQLite database can be selected to realize the reading of text class data, an API for creating the database is called by child_process.exec () to create the text class database by taking the text class data as a parameter, after the creation is successful, a creation completion message is sent through a postMessage API, the reading of the database is executed after the message is received, the text class database is read by adopting an all function of the SQLite to be transferred into an sql statement, and after the reading is monitored through a callback function of a callback, the text class database is sent to an artificial intelligent data annotation window through the postMessage API to be updated.
In summary, the above embodiments are only preferred embodiments of the present invention, and are not intended to limit the scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (7)
1. The cross-platform artificial intelligent data labeling method is characterized by comprising the following steps of: image processing apparatus
Step 1, creating a data annotation window by adopting an integrated development environment, and providing functions of data set creation, data set introduction, image data annotation and text data annotation;
step 2, selecting a data set to be created in a data labeling window, and creating a storage structure of the data set to be labeled and corresponding construction configuration information, wherein the construction configuration information comprises names, data amounts, labeling types and description information of the data set to be labeled, the storage structure is a three-layer directory structure consisting of a type directory, a version directory and data files, and the labeling types comprise image types and text types;
step 3, selecting a data set to be imported in the data labeling window, copying the data to be labeled, which needs to be imported, into the data set to be labeled according to the construction configuration information, and refreshing the data labeling window; if the type of the data set to be marked is the image type, executing the step 5; if the type of the data set to be marked is a text type, executing the step 4;
step 4, acquiring text cleaning parameters, including objects, output paths and text cleaning modes of text cleaning, calling a text cleaning algorithm to finish cleaning a data set to be marked, and executing step 8;
step 5, acquiring a first storage path of the data set to be marked, and executing step 6 if the data set to be marked has quality problems; if the data quantity in the data set to be marked is less, executing the step 7; otherwise, executing the step 8;
step 6, taking the first storage path as an object of image cleaning, acquiring image cleaning parameters, executing image cleaning according to the image cleaning parameters, and storing a cleaning result into the second storage path; comparing the image class data in the first storage path and the second storage path one by one, and updating the image class data in the first storage path after removing the data with poor quality; if the data quantity in the data set to be marked is less, executing the step 7, otherwise executing the step 9;
step 7, taking the first storage path as an object of image enhancement, executing image enhancement according to the image enhancement parameters, and updating the image class data in the first storage path according to the enhancement result;
step 8, calling a labeling tool to label part of the images in the first storage path, and updating the data set to be labeled after labeling is completed and labeling information is generated;
and 9, training the intelligent annotation model in the remote annotation server by adopting the annotated data, executing annotation of the unlabeled data in the to-be-annotated data set by adopting the remote annotation server according to the intelligent annotation parameters, and ending the flow.
2. The method according to claim 1, wherein the image class in step 2 includes object detection, image segmentation and image classification, and the text class includes entity recognition and relationship extraction.
3. The artificial intelligence data labeling method according to claim 1, wherein the verifying the image class data after completing the step 9 specifically comprises:
creating a plurality of thread splitting data by using a workbench_threads () in node. Js to distinguish images with marked information and images without marked information; when the image with the marking information is previewed, a check API is called to map the marking information into the image to form new image data, then the image is previewed, the image which does not meet the requirement is singly marked, an artificial intelligent data marking window is rendered and refreshed after marking is finished, and a current check data set to be marked is updated.
4. The method of claim 3, wherein the single labeling is labeling a single image using a vscore.
5. The artificial intelligence data labeling method according to claim 1, wherein the processing of the text data after the completion of the step 9 specifically comprises:
the artificial intelligent data annotation window adopts a page paging mode to respectively display the data with annotation information and the data without annotation information in the pages with annotation information and the pages without annotation information, and the text category is rendered by carrying out form splitting and category adding columns in the pages with annotation information; determining a display mode of a verification effect according to the annotation type, adding a highlight class to entity elements in the marked page when the annotation type is entity identification, rendering different types of entities into different colors, and carrying out connection processing to relation elements in the marked page when the annotation type is relation extraction; and finally, sending the processing result to an artificial intelligent data labeling window for updating.
6. The method of claim 5, further comprising implementing the reading of the text-like data by means of a database.
7. The artificial intelligence data tagging method of claim 6, wherein the database is a SQLite database.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310645802.7A CN116361279A (en) | 2023-06-02 | 2023-06-02 | Cross-platform artificial intelligent data labeling method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310645802.7A CN116361279A (en) | 2023-06-02 | 2023-06-02 | Cross-platform artificial intelligent data labeling method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116361279A true CN116361279A (en) | 2023-06-30 |
Family
ID=86928595
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310645802.7A Pending CN116361279A (en) | 2023-06-02 | 2023-06-02 | Cross-platform artificial intelligent data labeling method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116361279A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020010724A1 (en) * | 2018-07-13 | 2020-01-16 | 平安科技（深圳）有限公司 | Front-end static resource management method, apparatus, computer device and storage medium |
| CN111367445A (en) * | 2020-03-31 | 2020-07-03 | 中国建设银行股份有限公司 | Image annotation method and device |
| CN112099773A (en) * | 2020-09-14 | 2020-12-18 | 济南浪潮高新科技投资发展有限公司 | Method for realizing desktop application development |
| CN112579803A (en) * | 2020-11-16 | 2021-03-30 | 北京迈格威科技有限公司 | Image data cleaning method and device, electronic equipment and storage medium |
| WO2021208380A1 (en) * | 2020-04-15 | 2021-10-21 | 长沙中联重科环境产业有限公司 | Disinfection and cleaning operation effect testing method and device for epidemic prevention disinfection and cleaning robot |
-
2023
- 2023-06-02 CN CN202310645802.7A patent/CN116361279A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020010724A1 (en) * | 2018-07-13 | 2020-01-16 | 平安科技（深圳）有限公司 | Front-end static resource management method, apparatus, computer device and storage medium |
| CN111367445A (en) * | 2020-03-31 | 2020-07-03 | 中国建设银行股份有限公司 | Image annotation method and device |
| WO2021208380A1 (en) * | 2020-04-15 | 2021-10-21 | 长沙中联重科环境产业有限公司 | Disinfection and cleaning operation effect testing method and device for epidemic prevention disinfection and cleaning robot |
| CN112099773A (en) * | 2020-09-14 | 2020-12-18 | 济南浪潮高新科技投资发展有限公司 | Method for realizing desktop application development |
| CN112579803A (en) * | 2020-11-16 | 2021-03-30 | 北京迈格威科技有限公司 | Image data cleaning method and device, electronic equipment and storage medium |
Non-Patent Citations (2)
| Title |
|---|
| "基于Node.js开发跨平台窗口程序", Retrieved from the Internet <URL:http://www.manongjc.com/detail/50-pwjrqcbxlumirbo.html> * |
| WEB_BOOM: "01-Node.js安装及fs、path模块", Retrieved from the Internet <URL:https://blog.csdn.net/Android_boom/article/details/128431851> * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11023666B2 (en) | Narrative-based media organizing system for transforming and merging graphical representations of digital media within a work area | |
| Villán | Mastering OpenCV 4 with Python: a practical guide covering topics from image processing, augmented reality to deep learning with OpenCV 4 and Python 3.7 | |
| EP3433732B1 (en) | Converting visual diagrams into code | |
| US10838697B2 (en) | Storing logical units of program code generated using a dynamic programming notebook user interface | |
| US20180314882A1 (en) | Sorting and displaying digital notes on a digital whiteboard | |
| US8347206B2 (en) | Interactive image tagging | |
| US20190050378A1 (en) | Serializable and serialized interaction representations | |
| WO2020140932A1 (en) | Method and device for generating interface codes | |
| US20150128017A1 (en) | Enabling interactive screenshots within collaborative applications | |
| US11443144B2 (en) | Storage and automated metadata extraction using machine teaching | |
| US11443239B2 (en) | Interface for machine teaching modeling | |
| EP3665629A1 (en) | Creation, management, and transfer of interaction representation sets | |
| CN112714908A (en) | Dependency graph controlled object and compute pipeline migration | |
| CN109299452A (en) | A kind of document combination method, device and equipment | |
| JP2022500802A (en) | Image management with region-based metadata indexing | |
| US20120192046A1 (en) | Generation of a source complex document to facilitate content access in complex document creation | |
| CN116361279A (en) | Cross-platform artificial intelligent data labeling method | |
| CN116400914A (en) | Method for quickly constructing web application based on data model | |
| Rathnavibushana et al. | Cross-platform annotation development for real-time collaborative learning | |
| CN113010400B (en) | Computer processing technology document intelligent generation and multiple disk system and method | |
| TWI757733B (en) | Network data collection method | |
| Thakurdesai et al. | Smart Gallery using Google Vision | |
| US20130054607A1 (en) | Method and System for Preparing Digital Information for Long-Term Preservation | |
| CN115048339A (en) | Method and device for efficiently browsing pdf document | |
| WITTENBERG et al. | Assess, annotate, export |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20230630 |