CN102542014A - Image searching feedback method based on contents - Google Patents
Image searching feedback method based on contents Download PDFInfo
- Publication number
- CN102542014A CN102542014A CN2011104239785A CN201110423978A CN102542014A CN 102542014 A CN102542014 A CN 102542014A CN 2011104239785 A CN2011104239785 A CN 2011104239785A CN 201110423978 A CN201110423978 A CN 201110423978A CN 102542014 A CN102542014 A CN 102542014A
- Authority
- CN
- China
- Prior art keywords
- image
- class label
- feedback
- user
- retrieval
- 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.)
- Granted
Links
Images
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to an image searching feedback method based on contents, which comprises selecting a training sample from an image data base, training the training sample by utilizing a support vector machine to obtain a feature classification model, and classifying an image into a visual category; determining the visual category of the image submitted by a user according to the feature classification model, searching an image similar to the image submitted by the user, and returning a searching result; selecting feedback images in the searching result, marking images to be positive and negative feedback images respectively according to the relevance between the feedback images and the image submitted by the user, and feeding back the mark result to a searching system; judging the accurate category of the image submitted by the user according to text keywords of the feedback images, category labels and a mapping table, searching the image similar to the image submitted by the user in the accurate category, and returning a secondary searching result. The image searching feedback method based on contents can fast and accurately locate the category the image submitted by the user belongs to and improve secondary searching accuracy.
Description
Technical field
The invention belongs to image retrieval and identification field, more particularly, the present invention relates to a kind of CBIR feedback method.
Background technology
The retrieval precision of traditional CBIR technology is not very desirable usually, and the relevant feedback technology can remedy such shortcoming to a certain extent.People have done the research work of many relevant feedback aspect, have also obtained many achievements.The query point moving method that more typically has Rui etc. to propose, this method copy the proper vector of the image that the Rachio formula in the text retrieval submits to the user to make amendment, and make its direction towards user expectation move.Amended proper vector is the weighted sum of proper vector of proper vector and negative feedback image of former proper vector, the positive feedback image of the image submitted to of user; Make the proper vector of its deflection positive feedback image; The proper vector that departs from the negative feedback image, thus the direction as a result towards user expectation moves when quadratic search.In recent years; Machine learning method becomes the main flow of related feedback method gradually; Its thought is relevant feedback to be regarded as the classification problem of a band supervision: with the positive and negative training sample of positive and negative feedback samples as machine learning; Train a sorter, and calculate ordering back output result for retrieval as carrying out similarity between the image of new similarity measurement function to all images in the image library and user's submission with this.
Yet there is following problem in existing image retrieval feedback method: because the existence of semantic wide gap, the method for the characteristics of image that traditional change user submits to is little to the effect of precision improvement as a result of quadratic search; Just to working as time retrieval, precision is still very low when submitting to identical image to retrieve next time in the effect of feedback; The feedback method of machine learning has been introduced machine learning when retrieval, real-time is difficult to guarantee that because sample size is less, training effect is not obvious, and is little to the retrieval precision castering action simultaneously.
Summary of the invention
The object of the present invention is to provide a kind of CBIR feedback method; This method has been introduced SVMs (SVM) image has been classified in searching system, obtain class label, and according to class label, text keyword and the text keyword of feedback image and the mapping relations of feedback image class label; Carry out the relevant feedback of image retrieval; Remedied first result for retrieval precision shortcoming on the low side, promoted the precision of quadratic search, simultaneously the precision of elevator system retrieval in repeatedly feeding back; Automatically revise training sample, the reduction human cost.
The present invention realizes through following technical scheme:
A kind of CBIR feedback method may further comprise the steps:
(1) obtain all images in the image library, the visual category of definition image, and the quantity of definite visual category, each visual category is represented by a class label;
(2) extract the text keyword of image, and set up the mapping table from the text keyword to the class label;
(3) from image library, select training sample, and utilize SVMs that training sample is trained, to obtain the tagsort model;
(4) according to the tagsort model with image division in visual category;
(5) confirm the visual category of the image that the user submits to according to the tagsort model, the image of the image similarity that retrieval and user submit in visual category, and return result for retrieval;
(6) in result for retrieval, select feedback image, the correlativity of the image of submitting to according to feedback image and user is labeled as the positive-negative feedback image respectively with it, and annotation results is fed back to searching system;
(7) searching system is according to the accurate classification of the image of text keyword, class label and the submission of mapping table judges of feedback image;
(8) image of submitting to according to feedback image wrong in the class label of classification error in the accurate classification correction feedback image, the training sample and user;
(9) accurately retrieve the image of the image similarity of submitting to the user in the classification, and returning the quadratic search result;
(10) judge whether the quadratic search result satisfies the retrieval requirement;
(11) if the quadratic search result does not satisfy the retrieval requirement, then return step (6),, then get into step (12) if the quadratic search result satisfies the retrieval requirement;
(12) whether the correction quantity of training of judgement sample reaches 10% of total number of images in the training sample, if reach, then gets into step (13), and else process finishes;
(13) by the training sample of revising training characteristics disaggregated model again,, and upgrade its tag along sort to the image classification in the image library.
Step (2) comprises following substep: extract the web page text of the image in the image library, the analyzing web page text is rejected wherein contained HTML label; And extract its body text; Utilize the Chinese lexical analytic system of the Computer Department of the Chinese Academy of Science that body text is carried out participle, and reject irrelevant word, obtain the text keyword of image; Text keyword is divided in the visual category according to its semanteme, sets up the mapping table of text keyword to the visual category label.
Step (7) comprises following substep: the text keyword corresponding class label that obtains the positive feedback image according to mapping table; Different classes of number of tags in the statistics class label; And the maximum class label of acquisition class label quantity; If the class label kind that quantity is maximum is unique, then judge the accurate classification of this class label as the image of user's submission.
Step (7) also comprises following substep: the class label kind as if quantity is maximum is not unique; Then obtain the text keyword corresponding class label of negative feedback image according to mapping table; Different classes of number of tags in the statistics class label; And according to quantity from more to less with class label ordering, and be stored in the negative feedback list of categories, from the maximum class label of quantity, reject the class label that in the negative feedback list of categories, occurs in order; Class label kind up to quantity is maximum is unique, and judges the accurate classification of this class label as the image of user's submission.
Step (8) comprises following substep: if the accurate class label of the image that the class label of positive feedback image and user submit to is inconsistent; Then the classification of positive feedback image is modified to the accurate classification of the image of user's submission; And the positive feedback image added in the training sample of accurate classification; If the class label of negative feedback image is consistent with the accurate class label of the image that the user submits to; And accurately comprise the negative feedback image in the training sample of classification; Then from the training sample of accurate classification, delete the negative feedback image, if the accurate classification of the image that the visual category of the image that the user submits to and user submit to is inconsistent, the image of then user being submitted to adds in the training sample of accurate classification.
The present invention has following advantage and beneficial effect:
(1) owing to introduced the mapping relations of text keyword and image category label, utilize text keyword to feed back, the precision of feedback searching will be more a lot of than the feedback method precision improvement of the level image characteristic of simple modification user submission;
(2) in feedback procedure, the classification of the image in the image library has one from the process of revising, and feedback has not only promoted the precision when time result for retrieval each time, and has revised the image category of the classification error in the image library, has promoted the precision of next retrieval.Along with the increase of user's access times, retrieval precision will be increasingly high;
(3) mainly be to utilize the mapping relations of text keyword and image category label simply to calculate in the feedback procedure, judge the classification of the image that the user submits to, calculated amount is little, and real-time is high, and is more a lot of soon than the feedback method speed based on machine learning;
(4) training sample to image library carries out having promoted the precision of next training from revising in the feedback procedure, has reduced the human cost of selecting sample simultaneously.
Description of drawings
Fig. 1 is the process flow diagram that the present invention is based on the image retrieval feedback method of content.
Fig. 2 is the refinement process flow diagram of step in the inventive method (2).
Fig. 3 is the refinement process flow diagram of step in the inventive method (7).
Embodiment
At first the technical term among the present invention is made an explanation and explains:
Visual category: the sets definition that visually has the image of similarity on certain meaning is a visual category.
Class label: each visual category representes that with a unique number designation this number designation is defined as the class label of this visual category.Class label is an another name of visual category, is mainly used in the expression of simplifying visual category.
Text keyword: the source of the image in the native system image library is a network, and the image on the network all has the explanation of certain web page text, and text keyword is defined as in the web page text can those semantic words of token image.
Training sample: the sorting algorithm SVMs in the machine Learning Theory needs the process of a prior learning, and this learning process needs the handmarking to cross the sample of the some of visual category, and this sample is defined as training sample.
The tagsort model: the disaggregated model that obtains after utilizing SVMs according to the image low-level image feature training sample to be trained, this model is used for all images is classified.
Feedback image: the image that feeds back to searching system after the user marks parts of images in the result for retrieval is defined as feedback image.
As shown in Figure 1, the image retrieval feedback method that the present invention is based on content may further comprise the steps:
(1) obtain all images in the image library, the visual category of definition image, and the quantity of definite visual category, each visual category is represented by a class label;
(2) extract the text keyword of image, and set up the mapping table from the text keyword to the class label, specifically comprise following substep (see figure 2):
(21) web page text of the image in the extraction image library;
(22) analyzing web page text is rejected wherein contained HTML label, and is extracted its body text;
(23) utilize the Chinese lexical analytic system of the Computer Department of the Chinese Academy of Science that body text is carried out participle, and reject irrelevant word, obtain the text keyword of image;
(24) text keyword is divided in the visual category according to its semanteme;
(25) set up the mapping table of text keyword to the visual category label.
(3) from image library, select training sample, and utilize SVMs that training sample is trained, to obtain the tagsort model;
(4) according to the tagsort model with image division in visual category;
(5) confirm the visual category of the image that the user submits to according to the tagsort model, the image of the image similarity that retrieval and user submit in visual category, and return result for retrieval;
(6) in result for retrieval, select feedback image, the correlativity of the image of submitting to according to feedback image and user is labeled as the positive-negative feedback image respectively with it, and annotation results is fed back to searching system;
(7) searching system specifically comprises following substep (see figure 3) according to the accurate classification of the image of text keyword, class label and the submission of mapping table judges of feedback image:
(71) obtain the text keyword corresponding class label of positive feedback image according to mapping table;
(72) different classes of number of tags in the statistics class label, and obtain the maximum class label of class label quantity;
(73) maximum as if quantity class label kinds is unique, gets into step (77), otherwise gets into step (74);
(74) maximum as if quantity class label kinds is not unique, then obtains the text keyword corresponding class label of negative feedback image according to mapping table;
(75) add up different classes of number of tags in the class label, and from more to less class label is sorted, and be stored in the negative feedback list of categories according to quantity;
(76) from the maximum class label of quantity, reject the class label that in the negative feedback list of categories, occurs in order, unique up to the class label kind that quantity is maximum;
(77) judge the accurate classification of the maximum class label of quantity as the image of user's submission.
(8) image of submitting to according to feedback image wrong in the class label of classification error in the accurate classification correction feedback image, the training sample and user.Specifically comprise following substep:
(81), then the classification of positive feedback image is modified to the accurate classification of the image of user's submission, and the positive feedback image is added in the training sample of accurate classification if the accurate class label of the image that the class label of positive feedback image and user submit to is inconsistent;
(82) if the class label of negative feedback image is consistent with the accurate class label of the image of user's submission, and comprise the negative feedback image in the training sample of accurate classification, then deletion negative feedback image from the training sample of accurate classification;
(83) if the accurate classification of the image that the visual category of the image that the user submits to and user submit to is inconsistent, the image of then user being submitted to adds in the training sample of accurate classification.
(9) accurately retrieve the image of the image similarity of submitting to the user in the classification, and returning the quadratic search result;
(10) judge whether the quadratic search result satisfies the retrieval requirement;
(11) if the quadratic search result does not satisfy the retrieval requirement, then return step (6),, then get into step (12) if the quadratic search result satisfies the retrieval requirement;
(12) whether the correction quantity of training of judgement sample reaches 10% of total number of images in the training sample, if reach, then gets into step (13), and else process finishes;
(13) by the training sample of revising training characteristics disaggregated model again,, and upgrade its tag along sort to the image classification in the image library.
What be worth explanation is, this feedback scheme is based upon on the basis that the user is fully trusted, and promptly each feedback image mark of user is all accurate.The user may identify mistake to feedback image owing to reasons such as negligences in the reality; In order to prevent the bug patch of this situation to the image information in the image library; Can postpone revising the class label and the training sample of feedback image, add a statistical number to them, repeatedly feed back (such as 3 times) same information the user after; It is accurate to confirm to be somebody's turn to do feedback, revises class label and training sample label to feedback image then.
Claims (5)
1. a CBIR feedback method is characterized in that, may further comprise the steps:
(1) obtain all images in the image library, define the visual category of said image, and confirm the quantity of said visual category, each visual category is represented by a class label;
(2) extract the text keyword of said image, and set up mapping table from said text keyword to said class label;
(3) from said image library, select training sample, and utilize SVMs that said training sample is trained, to obtain the tagsort model;
(4) according to said tagsort model with said image division in said visual category;
(5) confirm the visual category of the image that the user submits to according to said tagsort model, the image of the image similarity that retrieval and said user submit in said visual category, and return result for retrieval;
(6) in said result for retrieval, select feedback image, the correlativity of the image of submitting to according to said feedback image and said user is labeled as the positive-negative feedback image respectively with it, and annotation results is fed back to said searching system;
(7) said searching system is judged the accurate classification of the image that said user submits to according to text keyword, class label and the said mapping table of said feedback image;
(8) image of submitting to according to feedback image wrong in the class label of classification error in the said feedback image of said accurate classification correction, the said training sample and said user;
The image of the image similarity that (9) retrieval and said user submit in said accurate classification, and return the quadratic search result;
(10) judge whether said quadratic search result satisfies the retrieval requirement;
(11) if said quadratic search result does not satisfy the retrieval requirement, then return step (6),, then get into step (12) if said quadratic search result satisfies the retrieval requirement;
Whether the correction quantity of (12) judging said training sample reaches 10% of total number of images in the said training sample, if reach, then gets into step (13), and else process finishes;
(13) train said tagsort model again by the said training sample of revising,, and upgrade its tag along sort the image classification in the said image library.
2. image retrieval feedback method according to claim 1 is characterized in that, said step (2) comprises following substep:
Extract the web page text of the image in the said image library;
Analyze said web page text, reject wherein contained HTML label, and extract its body text;
Utilize the Chinese lexical analytic system of the Computer Department of the Chinese Academy of Science that said body text is carried out participle, and reject irrelevant word, obtain the text keyword of said image;
Said text keyword is divided in the said visual category according to its semanteme;
Set up the mapping table of text keyword to the visual category label.
3. image retrieval feedback method according to claim 1 is characterized in that, said step (7) comprises following substep:
Obtain the text keyword corresponding class label of said positive feedback image according to said mapping table;
Add up different classes of number of tags in the said class label, and obtain the maximum class label of said class label quantity;
If the class label kind that said quantity is maximum is unique, then judge the accurate classification of this class label as the image of said user's submission.
4. image retrieval feedback method according to claim 1 is characterized in that, said step (7) also comprises following substep:
If the class label kind that said quantity is maximum is not unique, then obtain the text keyword corresponding class label of said negative feedback image according to said mapping table;
Add up different classes of number of tags in the said class label, and from more to less said class label is sorted, and be stored in the negative feedback list of categories according to quantity;
From the maximum class label of said quantity, reject the class label that in the negative feedback list of categories, occurs in order, unique up to the class label kind that said quantity is maximum, and judge the accurate classification of this class label as the image of said user's submission.
5. image retrieval feedback method according to claim 1 is characterized in that, said step (8) comprises following substep:
If the accurate class label of the image that the class label of said positive feedback image and said user submit to is inconsistent; Then the classification of said positive feedback image is modified to the accurate classification of the image of said user's submission, and said positive feedback image is added in the training sample of said accurate classification;
If the class label of said negative feedback image is consistent with the accurate class label of the image that said user submits to, and comprises said negative feedback image in the training sample of said accurate classification, then from the training sample of said accurate classification, delete said negative feedback image;
If the accurate classification of the image that the visual category of the image that said user submits to and said user submit to is inconsistent, the image of then said user being submitted to adds in the training sample of said accurate classification.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 201110423978 CN102542014B (en) | 2011-12-16 | 2011-12-16 | Image searching feedback method based on contents |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 201110423978 CN102542014B (en) | 2011-12-16 | 2011-12-16 | Image searching feedback method based on contents |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102542014A true CN102542014A (en) | 2012-07-04 |
| CN102542014B CN102542014B (en) | 2013-09-18 |
Family
ID=46348901
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 201110423978 Expired - Fee Related CN102542014B (en) | 2011-12-16 | 2011-12-16 | Image searching feedback method based on contents |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102542014B (en) |
Cited By (21)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103064900A (en) * | 2012-12-17 | 2013-04-24 | 北京奇虎科技有限公司 | Device and method for image configuration |
| CN103605992A (en) * | 2013-11-28 | 2014-02-26 | 国家电网公司 | Sensitive image recognizing method in interaction of inner and outer power networks |
| CN104102699A (en) * | 2014-07-01 | 2014-10-15 | 华中科技大学 | Subgraph searching method in cluster graph set |
| CN105022752A (en) * | 2014-04-29 | 2015-11-04 | 中国电信股份有限公司 | Image retrieval method and apparatus |
| CN105653723A (en) * | 2016-01-19 | 2016-06-08 | 中国科学技术大学 | Inquired image characteristic cutting method for image retrieval |
| CN106250916A (en) * | 2016-07-22 | 2016-12-21 | 西安酷派软件科技有限公司 | A kind of screen the method for picture, device and terminal unit |
| CN107577682A (en) * | 2016-07-05 | 2018-01-12 | 上海交通大学 | Users' Interests Mining and user based on social picture recommend method and system |
| CN107818160A (en) * | 2017-10-31 | 2018-03-20 | 上海掌门科技有限公司 | Expression label updates and realized method, equipment and the system that expression obtains |
| CN107918624A (en) * | 2016-10-11 | 2018-04-17 | 富士通株式会社 | Image retrieving apparatus and method, electronic equipment |
| CN108563683A (en) * | 2018-03-08 | 2018-09-21 | 北京达佳互联信息技术有限公司 | Label addition method, device and terminal |
| CN108563792A (en) * | 2018-05-02 | 2018-09-21 | 百度在线网络技术（北京）有限公司 | Image retrieval processing method, server, client and storage medium |
| WO2018188410A1 (en) * | 2017-04-14 | 2018-10-18 | 优酷信息技术（北京）有限公司 | Feedback response method and apparatus |
| CN109447937A (en) * | 2017-08-29 | 2019-03-08 | 中国移动通信有限公司研究院 | A kind of determination method and related device of image processing model |
| CN110020078A (en) * | 2017-12-01 | 2019-07-16 | 北京搜狗科技发展有限公司 | A kind of method and relevant apparatus for generating relevance map dictionary and its verifying correlation |
| CN110163296A (en) * | 2019-05-29 | 2019-08-23 | 北京达佳互联信息技术有限公司 | Method, apparatus, equipment and the storage medium of image recognition |
| CN110210535A (en) * | 2019-05-21 | 2019-09-06 | 北京市商汤科技开发有限公司 | Neural network training method and device and image processing method and device |
| WO2019206251A1 (en) * | 2018-04-27 | 2019-10-31 | Shanghai Truthvision Information Technology Co., Ltd. | Systems and methods for image archiving |
| CN110909776A (en) * | 2019-11-11 | 2020-03-24 | 维沃移动通信有限公司 | Image identification method and electronic equipment |
| CN111444372A (en) * | 2019-04-04 | 2020-07-24 | 北京京东尚科信息技术有限公司 | System and method for image processing |
| CN111783903A (en) * | 2020-08-05 | 2020-10-16 | 腾讯科技（深圳）有限公司 | Text processing method, text model processing method and device and computer equipment |
| CN113989610A (en) * | 2021-12-27 | 2022-01-28 | 广州思德医疗科技有限公司 | Intelligent image labeling method, device and system |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101833565A (en) * | 2010-03-31 | 2010-09-15 | 南京大学 | Method for actively selecting related feedbacks of representative image |
-
2011
- 2011-12-16 CN CN 201110423978 patent/CN102542014B/en not_active Expired - Fee Related
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101833565A (en) * | 2010-03-31 | 2010-09-15 | 南京大学 | Method for actively selecting related feedbacks of representative image |
Non-Patent Citations (2)
| Title |
|---|
| EN CHENG ET AL: "USING IMPLICIT RELEVANE FEEDBACK TO ADVANCE WEB IMAGE SEARCH", 《2006IEEE》 * |
| 姜文超 等: "Web图像搜索系统中用户交互及反馈系统模型", 《华中科技大学学报（自然科学版）》 * |
Cited By (34)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103064900A (en) * | 2012-12-17 | 2013-04-24 | 北京奇虎科技有限公司 | Device and method for image configuration |
| CN103605992B (en) * | 2013-11-28 | 2018-01-19 | 国家电网公司 | A kind of nude picture detection method in electric power intranet and extranet interaction |
| CN103605992A (en) * | 2013-11-28 | 2014-02-26 | 国家电网公司 | Sensitive image recognizing method in interaction of inner and outer power networks |
| CN105022752B (en) * | 2014-04-29 | 2019-04-05 | 中国电信股份有限公司 | Image search method and device |
| CN105022752A (en) * | 2014-04-29 | 2015-11-04 | 中国电信股份有限公司 | Image retrieval method and apparatus |
| CN104102699B (en) * | 2014-07-01 | 2018-01-26 | 华中科技大学 | A kind of subgraph search method in the set of graphs that clusters |
| CN104102699A (en) * | 2014-07-01 | 2014-10-15 | 华中科技大学 | Subgraph searching method in cluster graph set |
| CN105653723A (en) * | 2016-01-19 | 2016-06-08 | 中国科学技术大学 | Inquired image characteristic cutting method for image retrieval |
| CN105653723B (en) * | 2016-01-19 | 2019-03-01 | 中国科学技术大学 | A kind of query image feature method of cutting out for image retrieval |
| CN107577682B (en) * | 2016-07-05 | 2021-06-29 | 上海交通大学 | Social picture-based user interest mining and user recommending method and system |
| CN107577682A (en) * | 2016-07-05 | 2018-01-12 | 上海交通大学 | Users' Interests Mining and user based on social picture recommend method and system |
| CN106250916A (en) * | 2016-07-22 | 2016-12-21 | 西安酷派软件科技有限公司 | A kind of screen the method for picture, device and terminal unit |
| CN106250916B (en) * | 2016-07-22 | 2020-02-21 | 西安酷派软件科技有限公司 | Method and device for screening pictures and terminal equipment |
| CN107918624B (en) * | 2016-10-11 | 2022-03-15 | 富士通株式会社 | Image retrieval device and method and electronic equipment |
| CN107918624A (en) * | 2016-10-11 | 2018-04-17 | 富士通株式会社 | Image retrieving apparatus and method, electronic equipment |
| WO2018188410A1 (en) * | 2017-04-14 | 2018-10-18 | 优酷信息技术（北京）有限公司 | Feedback response method and apparatus |
| CN109447937A (en) * | 2017-08-29 | 2019-03-08 | 中国移动通信有限公司研究院 | A kind of determination method and related device of image processing model |
| CN107818160A (en) * | 2017-10-31 | 2018-03-20 | 上海掌门科技有限公司 | Expression label updates and realized method, equipment and the system that expression obtains |
| CN110020078A (en) * | 2017-12-01 | 2019-07-16 | 北京搜狗科技发展有限公司 | A kind of method and relevant apparatus for generating relevance map dictionary and its verifying correlation |
| CN110020078B (en) * | 2017-12-01 | 2021-08-20 | 北京搜狗科技发展有限公司 | Method and related device for generating relevance mapping dictionary and verifying relevance |
| CN108563683A (en) * | 2018-03-08 | 2018-09-21 | 北京达佳互联信息技术有限公司 | Label addition method, device and terminal |
| WO2019206251A1 (en) * | 2018-04-27 | 2019-10-31 | Shanghai Truthvision Information Technology Co., Ltd. | Systems and methods for image archiving |
| CN112074827A (en) * | 2018-04-27 | 2020-12-11 | 上海趋视信息科技有限公司 | System and method for image archiving |
| CN108563792A (en) * | 2018-05-02 | 2018-09-21 | 百度在线网络技术（北京）有限公司 | Image retrieval processing method, server, client and storage medium |
| CN111444372A (en) * | 2019-04-04 | 2020-07-24 | 北京京东尚科信息技术有限公司 | System and method for image processing |
| CN111444372B (en) * | 2019-04-04 | 2024-03-08 | 北京京东尚科信息技术有限公司 | System and method for image processing |
| CN110210535A (en) * | 2019-05-21 | 2019-09-06 | 北京市商汤科技开发有限公司 | Neural network training method and device and image processing method and device |
| CN110210535B (en) * | 2019-05-21 | 2021-09-10 | 北京市商汤科技开发有限公司 | Neural network training method and device and image processing method and device |
| CN110163296B (en) * | 2019-05-29 | 2020-12-18 | 北京达佳互联信息技术有限公司 | Image recognition method, device, equipment and storage medium |
| CN110163296A (en) * | 2019-05-29 | 2019-08-23 | 北京达佳互联信息技术有限公司 | Method, apparatus, equipment and the storage medium of image recognition |
| CN110909776A (en) * | 2019-11-11 | 2020-03-24 | 维沃移动通信有限公司 | Image identification method and electronic equipment |
| CN111783903A (en) * | 2020-08-05 | 2020-10-16 | 腾讯科技（深圳）有限公司 | Text processing method, text model processing method and device and computer equipment |
| CN111783903B (en) * | 2020-08-05 | 2023-11-28 | 腾讯科技（深圳）有限公司 | Text processing method, text model processing method and device and computer equipment |
| CN113989610A (en) * | 2021-12-27 | 2022-01-28 | 广州思德医疗科技有限公司 | Intelligent image labeling method, device and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102542014B (en) | 2013-09-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102542014B (en) | Image searching feedback method based on contents | |
| CN108121829B (en) | Software defect-oriented domain knowledge graph automatic construction method | |
| Sebastiani | Text categorization | |
| US10366093B2 (en) | Query result bottom retrieval method and apparatus | |
| Burger et al. | Discriminating gender on Twitter | |
| WO2016179938A1 (en) | Method and device for question recommendation | |
| CN111026671A (en) | Test case set construction method and test method based on test case set | |
| CN109766438A (en) | Biographic information extracting method, device, computer equipment and storage medium | |
| CN112035653A (en) | Policy key information extraction method and device, storage medium and electronic equipment | |
| CN106202380B (en) | Method and system for constructing classified corpus and server with system | |
| CN107392143A (en) | A kind of resume accurate Analysis method based on SVM text classifications | |
| CN102542067A (en) | Automatic image semantic annotation method based on scale learning and correlated label dissemination | |
| CN112182148B (en) | Standard aided writing method based on full text retrieval | |
| Malik et al. | Accurate information extraction for quantitative financial events | |
| CN110134799B (en) | BM25 algorithm-based text corpus construction and optimization method | |
| CN112035675A (en) | Medical text labeling method, device, equipment and storage medium | |
| CN103577462A (en) | Document classification method and document classification device | |
| Feng et al. | Practical duplicate bug reports detection in a large web-based development community | |
| CN111930933A (en) | Detection case processing method and device based on artificial intelligence | |
| CN108038099A (en) | Low frequency keyword recognition method based on term clustering | |
| CN106339459B (en) | The method that Chinese web page is presorted is carried out based on Keywords matching | |
| CN106294786A (en) | A kind of code search method and system | |
| CN104881446A (en) | Searching method and searching device | |
| CN109948015B (en) | Meta search list result extraction method and system | |
| CN112215002A (en) | Electric power system text data classification method based on improved naive Bayes |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20130918 Termination date: 20151216 |
|
| EXPY | Termination of patent right or utility model |