CN104679902B - A kind of informative abstract extracting method of combination across Media Convergence - Google Patents
A kind of informative abstract extracting method of combination across Media Convergence Download PDFInfo
- Publication number
- CN104679902B CN104679902B CN201510123093.1A CN201510123093A CN104679902B CN 104679902 B CN104679902 B CN 104679902B CN 201510123093 A CN201510123093 A CN 201510123093A CN 104679902 B CN104679902 B CN 104679902B
- Authority
- CN
- China
- Prior art keywords
- image
- data
- dimensional data
- text label
- dimension
- 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.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 43
- 230000004927 fusion Effects 0.000 claims abstract description 11
- 238000007689 inspection Methods 0.000 claims abstract description 4
- 238000000605 extraction Methods 0.000 claims description 17
- 238000005516 engineering process Methods 0.000 claims description 5
- 239000000284 extract Substances 0.000 claims description 5
- 238000005065 mining Methods 0.000 claims description 4
- 238000003672 processing method Methods 0.000 claims description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000002068 genetic effect Effects 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 230000003321 amplification Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000000151 deposition Methods 0.000 description 1
- 230000003447 ipsilateral effect Effects 0.000 description 1
- 230000007935 neutral effect Effects 0.000 description 1
- 238000003199 nucleic acid amplification method Methods 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Image Analysis (AREA)
Abstract
The present invention proposes a kind of informative abstract extracting method of combination across Media Convergence.The multi-medium data (word, image, audio, video etc.) of input is classified by data type first;Isomery multi-medium data with dimensionization and is established into the text label of data again, obtained with dimension image and text label;Then same dimensional data image is clustered and carries out the relevance inspection of text label;Several same dimension images of sub-category fusion are a sub-picture again;Ultimately produce across media information summary.User can check the fused images of every category information by informative abstract, and can quickly access corresponding multi-medium data.
Description
Technical field
The present invention relates to a kind of combination across the informative abstract extracting method of Media Convergence, belong to information extraction field.
Background technology
We live in an information age, and magnanimity information amplification, internet is daily in newly-increased substantial amounts of information, and information
Storage mode it is increasingly diversified, text, image, audio, video are the basic existence forms of multimedia resource.Nowadays multiple types
Type media data is mixed and deposited, and media data organization is complicated, but never ipsilateral expression is same for different types of media data
One is semantic, needs, according to existing various contacts between media, another media to be crossed from a kind of media in information extraction.Cause
How this, how across the boundary between media, extract the potential relevance between media, turn into current information extraction institute
Facing challenges.
The big data for mixing and depositing for media form, existing method are mainly distinguished by the feature of same media
Know to realize, it is difficult to across the semantic gap between multimedia, such as the visual signature of image and the aural signature of audio it
Between intrinsic dimensionality it is different and can not directly measure the similitude between them, therefore, existing information extracting method can not be fine
Provide the user visual thumbnail(Or informative abstract), how by a large amount of multimedia data classifications of mixing with extraction, turn into letter
One of key technology difficulty of breath extraction urgent need to resolve, and the heat subject studied at present.
Existing ripe Text Mining Technology, image characteristics extraction algorithm, audio scene identification, speech recognition, video field
The methods of scape segmentation, key-frame extraction, can extract the semantic information of single medium, how be combined these algorithms, will not
With the feature information extraction of dimension, formed and handle multimedia information extracting system, this is middle-dimentional by image for we
Media solve this problem.
The content of the invention
In view of the above-mentioned problems, the present invention proposes a kind of informative abstract extracting method of combination across Media Convergence.By using
Different dimension data is tieed up together to the method for turning to image, solves the problems, such as to be difficult to cross over semantic information of multimedia wide gap.Pass through image clustering
Method, so as to indirectly by multimedia data classification and extraction, generate across media information summary.
The present invention proposes a kind of informative abstract extracting method of combination across Media Convergence.First by the multimedia number of input
Classified according to (word, image, audio, video etc.) by data type;Different dimension multi-medium data with dimensionization and is established into number again
According to text label, obtain with dimension image and text label;Then same dimensional data image is clustered and carries out the pass of text label
Connection property is examined;Several same dimension images of sub-category fusion are a sub-picture again;Ultimately produce across media information summary.User passes through
Informative abstract can check the fused images of every category information, and can quickly access corresponding multi-medium data.
The present invention proposes that a kind of combination across the informative abstract extracting method of Media Convergence, comprises the following steps:
Step 1:It is original that (word, image, audio, video) in the multi-medium data of input is pressed into data type classifications
Text data, raw image data, original audio data,
Original video data;
Step 2:View data dimension is set(Image pixel)Standard value, establish the same dimension image sample with text label
This storehouse, carry out different dimension multi-medium data and handled with dimensionization, corresponding processing method is used according to the difference of data type;
Step 3:To processed same dimensional data image, according to the degree of accuracy threshold value required for cluster, gather according to image
Class algorithm is clustered, and text label relevance inspection is carried out according to the text label of every class data, by the data for the condition that is unsatisfactory for
Cluster again, until the data bulk for being unsatisfactory for condition is less than threshold value, can obtainSimilar dimensional data image
Address, that is, index;
Step 4:To the same dimensional data image clustered, according to a kind of fusion rule, merged, it is each so as to obtain
The fused images of similar dimensional data image;
Step 5:According to the fused images and index of each similar dimensional data image, informative abstract is generated.
Compared with the conventional method, advantage of the invention is that:
1st, the multi-medium data semanteme of different dimension is expressed with same dimensional data image, spans the boundary between media,
And with the related algorithm processing multi-medium data of image procossing;
2nd, image clustering method with text label relevance examine be combined, ensure that classification accuracy and data it
Between High relevancy.
Brief description of the drawings
Fig. 1 is the flow chart of the present invention;
Fig. 2 is the same dimensionization method flow diagram of different dimension data in the present invention;
Fig. 3 is to examine schematic diagram with text label relevance with dimensional data image cluster in the present invention.
Specific implementation method
The present invention is described in further detail with reference to the accompanying drawings and detailed description:
Step 1:It is original that (word, image, audio, video) in the multi-medium data of input is pressed into data type classifications
Text data, raw image data, original audio data,
Original video data。
Step 2:Referring to Fig. 2, view data dimension is set(Image pixel)Standard value, establish same with text label
Wei Tuxiangyangbenku, carry out different dimension multi-medium data and handled with dimensionization, according to the different using corresponding place of data type
Reason method;
Existing classification results areGroup urtext data,Group raw image data,Group original audio data,Group is original
Video data, willGroup urtext dataHandle as same dimensional data image, will
Group raw image dataHandle as same dimensional data image, willGroup original audio
DataHandle as same dimensional data image, willGroup original video dataHandle as same dimensional data image, detailed step is as follows;
1)By urtext dataHandle as same dimensional data imageProcess
And associative operation;
a)Pretreatment, utilizes certain Text Mining Technology(Such as text mining based on semantic understanding), by urtext number
According toIn the keyword extraction of every group of text message paragraph be label;
b)WillGroup text data corresponds to same dimensional data image according to label keyword and Sample Storehouse
, wherein, one group of text can correspond to multiple labels and same dimensional data image,Corresponding sample image is represented by。
2)By raw image dataHandle as same dimensional data imageMistake
Journey and associative operation;
a)Pre-process raw image data, strengthen key feature using related algorithm and (carried on the back as rejected
Scene area), the image after being handled;
b)For image, utilize certain image scaling techniques(As bicubic interpolation is inverse with small echo
To interpolation)It is scaled same dimensional data image(With Sample Storehouse with dimension);
c)By same dimensional data imageUsing certain recognition methods (such as figure of view-based access control model information
As feature extraction algorithm) compared with Sample Storehouse, the text label of image is obtained, is as a result deposited in。
3)By original audio dataHandle as same dimensional data image's
Process and associative operation;
a)Pre-process original audio data, audio scene is extracted using related algorithm(Such as based on general
The audio scene recognition method of rate latent semantic analysis), language semantic(Such as speech recognition based on neutral net)Deng crucial special
Sign, the text label extracted;
b)For the text label of extraction, text label is corresponding with Sample Storehouse, obtains same tie up
View data, wherein, multiple labels and same dimensional data image can be corresponded to audio is organized,It is right
The multiple sample images answered are represented by。
4)By original video dataHandle as same dimensional data imageMistake
Journey and associative operation;
a)Pre-process original video data, utilize a certain Algorithm of Scene(As based on semanteme
Video scene partitioning algorithm), for each video, after obtaining split senceIndividual video segment;
b)For each video segment, using a certain extraction method of key frame(Such as multiple features fusion based on clustering algorithm
Key-frame extraction), obtain key frame images, the set of the key frame images of each video is designated as;
c)For key frame images, strengthen key feature using related algorithm(Such as reject the back of the body
Scene area);
d)To certain image scaling techniques of processed imagery exploitation(Such as bicubic interpolation and small echo reversed interpolation)Scaling
For same dimensional data image(With Sample Storehouse with dimension);
e)By same dimensional data image, compared, schemed with Sample Storehouse using certain recognition methods
The text label of picture, is as a result deposited in。
Step 3:Referring to Fig. 3, to processed same dimensional data image, according to the degree of accuracy threshold value required for cluster
, clustered according to certain image clustering algorithm(Such as image clustering based on genetic algorithm), according to the text mark of every class data
Label carry out text label relevance inspection, the data for the condition that is unsatisfactory for are clustered again, the data bulk until being unsatisfactory for condition
Less than threshold value, must can index, it isSimilar dimensional data imageAddress, in detail
Thin step is as follows:
1)To processed same dimensional data image, according to the degree of accuracy threshold value required for cluster
,Smaller, classification quantity is more, and classification is more accurate, conversely, classification quantity is fewer;
2)Clustered according to certain image clustering algorithm(Such as image clustering based on genetic algorithm), what storage had clustered
With dimension image address, the same class for having clustered extracts its corresponding text label with image is tieed up, carries out text label and figure
As the text label relevance of cluster result is examined;
3)For the same dimensional data image clustered, if the quantity for being unsatisfactory for text label relevance test condition is more than threshold
Value, then the data for the condition that is unsatisfactory for are rejected into this class, turn into the same dimensional data image that does not cluster again, and according to identical or not
Same clustering method clusters again, until the data bulk for being unsatisfactory for condition is less than threshold value;
4)Classification results are stored in the form of address, indexed, it isSimilar dimension picture number
According toAddress.
Step 4:To the same dimensional data image clustered, according to a certain fusion rule(Such as choose the more width figure of target
Picture), merged, so as to obtain the fused images of each similar dimensional data image;
Successively by indexTake outSimilar dimensional data image, according to a certain
Fusion rule, merged, so as to obtain the fused images of each similar dimensional data image。
Step 5:According to the fused images and index of each similar dimensional data image, informative abstract is generated;
By the fused images of acquisitionAnd indexGenerate informative abstract, user can check fused images, multi-medium data corresponding to access.
Claims (4)
1. a kind of combination is across the informative abstract extracting method of Media Convergence, it is characterised in that first by the multi-medium data of input,
Including word, image, audio, video, classified by data type;Different dimension multi-medium data with dimensionization and is established into data again
Text label, obtain with dimension image and text label;Then same dimensional data image is clustered and carries out the association of text label
Property examine;Several same dimension images of sub-category fusion are a sub-picture again;Ultimately produce across media information summary;Methods described is extremely
Comprise the following steps less:
Step 1:By in the multi-medium data of input, including word, image, audio, video, it is original by data type classifications
Text data T { T1, T2, T3..., Tt, raw image data P { P1, P2, P3..., Pp, original audio data A { A1, A2,
A3..., Aa, original video data V { V1, V2, V3..., Vv};
Step 2:View data dimension (image pixel) standard value is set, establishes the same Wei Tuxiangyangbenku with text label,
Carry out different dimension multi-medium data to handle with dimensionization, corresponding processing method is used according to the difference of data type;
1) by urtext data T { T1, T2, T3..., TtHandle as with dimensional data image Ft { Ft1,Ft2,Ft3,...,Ftt,
Step includes, pretreatment, using Text Mining Technology, by urtext data T { T1, T2, T3..., TtIn every group of text envelope
The keyword extraction for ceasing paragraph is label Lt{Lt1, Lt2, Lt3..., Ltt};Then by T groups text data according to label keyword
Same dimensional data image F is corresponded to Sample Storehouset{Ft1,Ft2,Ft3,...,Ftt, wherein, one group of text can correspond to multiple labels with
And same dimensional data image;
2) by raw image data P { P1,P2,P3,...,PpHandle as with dimensional data image Fp{Fp1, Fp2, Fp3..., Fpp, step
Suddenly include:Pre-process raw image data P { P1, P2, P3..., Pp, strengthen key feature using related algorithm, handled
Image P ' { P ' afterwards1, P '2, P '3..., P 'p};For image P ' { P '1, P '2, P '3..., P 'p, utilize image scaling techniques
It is scaled same dimensional data image Fp{Fp1, Fp2, Fp3..., Fpp(with Sample Storehouse with dimension);Will be with dimensional data image Fp{Fp1, Fp2,
Fp3..., FppCompared using image-recognizing method with Sample Storehouse, the text label of image is obtained, as a result deposits in Lp{Lp1,
Lp2, Lp3..., Lpp};
3) by original audio data A { A1, A2, A3..., AaHandle as with dimensional data image Fa{Fa1, Fa2, Fa3..., Faa, step
Suddenly include:Pre-process original audio data A { A1, A2, A3..., Aa, extract audio scene, language semantic using related algorithm
Feature, the text label L extracteda{La1, La2, La3..., Laa};
For the text label L of extractiona{La1, La2, La3..., Laa, text label is corresponding with Sample Storehouse, obtains with dimension image
Data Fa{Fa1, Fa2, Fa3..., Faa, wherein, it can correspond to multiple labels and same dimensional data image with audio is organized;
4) by original video data V { V1, V2, V3..., VvHandle as with dimensional data image Fv{Fv1, Fv2, Fv3..., Fvv, step
Suddenly include:Pre-process original video data Vi{V1, V2, V3..., Vv, using Algorithm of Scene, for each video Vi, obtain
J video segment V after to split senceij{Vi1, Vi2, Vi3..., Vij};
For each video segment Vij, using extraction method of key frame, obtain key frame images SVij, the key of each video
The set of two field picture is designated as SV { SV1, SV2, SV3..., SVv};
For key frame images SV { SV1, SV2, SV3..., SVv, strengthen key feature using related algorithm;
Same dimensional data image F is scaled to processed imagery exploitation image scaling techniquesv{Fv1, Fv2, Fv3..., Fvv(with sample
This storehouse is with dimension);
Will be with dimensional data image Fv{Fv1, Fv2, Fv3..., Fvv, compared using image-recognizing method and Sample Storehouse, obtain image
Text label, as a result deposit in Lv{Lv1, Lv2, Lv3..., Lvv};
Step 3:To processed same dimensional data image, according to the degree of accuracy threshold value N required for cluster, gather according to image
Class algorithm is clustered, and text label relevance inspection is carried out according to the text label of every class data, by the number for the condition that is unsatisfactory for
According to clustering again, until the data bulk for being unsatisfactory for condition is less than threshold value N, the similar dimensional data image C { C of m can be obtained1, C2,
C3..., CmAddress, that is, index K { K1, K2, K3..., Km};
Step 4:To the same dimensional data image clustered, according to a kind of fusion rule, merged, it is each similar so as to obtain
The fused images F of dimensional data imageC{FC1, FC2, FC3..., FCm};
Step 5:According to the fused images and index of each similar dimensional data image, informative abstract is generated.
2. a kind of combination according to claim 1 is across the informative abstract extracting method of Media Convergence, it is characterised in that step
It is at least further comprising the steps of to the process with dimension image clustering and foundation index in three:
1) to processed same dimensional data image, according to the degree of accuracy threshold value N (N > 0, N ∈ Z) required for cluster, N is got over
Small, classification quantity is more, and classification is more accurate, conversely, classification quantity is fewer;
2) clustered according to image clustering algorithm, store the same dimension image address that has clustered, it is same for the same class that has clustered
Image is tieed up, extracts its corresponding text label, text label is carried out and the relevance of image clustering result is examined;
3) for the same dimensional data image clustered, if the quantity for being unsatisfactory for test condition is more than threshold value N, condition will be unsatisfactory for
Data reject this class, turn into the same dimensional data image that does not cluster again, and cluster again according to identical or different method, directly
Data bulk to the condition that is unsatisfactory for is less than threshold value N;
4) classification results are stored in the form of address, obtains indexing K { K1,K2,K3,...,Km, it is the similar dimensional data image C of m
{C1,C2,C3,...,CmAddress.
3. a kind of combination according to claim 1 is across the informative abstract extracting method of Media Convergence, it is characterised in that step
Sub-category fusion is at least further comprising the steps of with the process that dimensional data image is a width in four:
1) successively by index K { K1,K2,K3,...,KmTake out the similar dimensional data image C { C of m1, C2, C3..., Cm, according to one kind
Fusion rule, merged, so as to obtain the fused images F of each similar dimensional data imageC{FC1, FC2, FC3..., FCm}。
4. a kind of combination according to claim 1 is across the informative abstract extracting method of Media Convergence, it is characterised in that step
The process of informative abstract is generated, at least also includes following step with dimension image and index according to the fusion of each category information in five
Suddenly:
1) by the fused images F of acquisitionC{FC1, FC2, FC3..., FCmAnd index K { K1, K2, K3..., KmGeneration information pluck
Want I { I1{FC1, K1, I2{FC2, K2....
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510123093.1A CN104679902B (en) | 2015-03-20 | 2015-03-20 | A kind of informative abstract extracting method of combination across Media Convergence |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510123093.1A CN104679902B (en) | 2015-03-20 | 2015-03-20 | A kind of informative abstract extracting method of combination across Media Convergence |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104679902A CN104679902A (en) | 2015-06-03 |
CN104679902B true CN104679902B (en) | 2017-11-28 |
Family
ID=53314944
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510123093.1A Expired - Fee Related CN104679902B (en) | 2015-03-20 | 2015-03-20 | A kind of informative abstract extracting method of combination across Media Convergence |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104679902B (en) |
Families Citing this family (22)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105142096B (en) * | 2015-08-14 | 2018-10-19 | 湘潭大学 | Across media data fusion method based on neural network in Internet of Things |
CN105245370B (en) * | 2015-10-13 | 2019-03-19 | 湘潭大学 | Across the media data fusion method of adaptive layered in a kind of Internet of Things |
CN106815253B (en) * | 2015-12-01 | 2020-04-10 | 慧科讯业有限公司 | Mining method based on mixed data type data |
WO2017128438A1 (en) * | 2016-01-31 | 2017-08-03 | 深圳市博信诺达经贸咨询有限公司 | Method and system for application of big data |
CN106686403B (en) * | 2016-12-07 | 2019-03-08 | 腾讯科技(深圳)有限公司 | A kind of video preview drawing generating method, device, server and system |
CN106997387B (en) * | 2017-03-28 | 2019-08-09 | 中国科学院自动化研究所 | Based on the multi-modal automaticabstracting of text-images match |
CN107437100A (en) * | 2017-08-08 | 2017-12-05 | 重庆邮电大学 | A kind of picture position Forecasting Methodology based on the association study of cross-module state |
CN107885845B (en) * | 2017-11-10 | 2020-11-17 | 广州酷狗计算机科技有限公司 | Audio classification method and device, computer equipment and storage medium |
CN108388942A (en) * | 2018-02-27 | 2018-08-10 | 四川云淞源科技有限公司 | Information intelligent processing method based on big data |
CN110472075A (en) * | 2018-05-09 | 2019-11-19 | 中国互联网络信息中心 | A kind of isomeric data classification storage method and system based on machine learning |
CN109299315B (en) | 2018-09-03 | 2023-03-28 | 腾讯科技(深圳)有限公司 | Multimedia resource classification method and device, computer equipment and storage medium |
CN110489475B (en) * | 2019-08-14 | 2021-01-26 | 广东电网有限责任公司 | Multi-source heterogeneous data processing method, system and related device |
CN110532426A (en) * | 2019-08-27 | 2019-12-03 | 新华智云科技有限公司 | It is a kind of to extract the method and system that Multi-media Material generates video based on template |
CN110837560B (en) * | 2019-11-15 | 2022-03-15 | 北京字节跳动网络技术有限公司 | Label mining method, device, equipment and storage medium |
CN111291204B (en) * | 2019-12-10 | 2023-08-29 | 河北金融学院 | Multimedia data fusion method and device |
CN111488490A (en) * | 2020-03-31 | 2020-08-04 | 北京奇艺世纪科技有限公司 | Video clustering method, device, server and storage medium |
CN111767395B (en) * | 2020-06-30 | 2023-12-26 | 平安国际智慧城市科技股份有限公司 | Abstract generation method and system based on pictures |
CN112925902B (en) * | 2021-02-22 | 2024-01-30 | 新智认知数据服务有限公司 | Method, system and electronic equipment for intelligently extracting text abstract from case text |
CN112860905A (en) * | 2021-04-08 | 2021-05-28 | 深圳壹账通智能科技有限公司 | Text information extraction method, device and equipment and readable storage medium |
CN113505201A (en) * | 2021-07-29 | 2021-10-15 | 宁波薄言信息技术有限公司 | Contract extraction method based on SegaBert pre-training model |
CN117371533B (en) * | 2023-11-01 | 2024-05-24 | 深圳市马博士网络科技有限公司 | Method and device for generating data tag rule |
CN117573870B (en) * | 2023-11-20 | 2024-05-07 | 中国人民解放军国防科技大学 | Text label extraction method, device, equipment and medium for multi-mode data |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101021849A (en) * | 2006-09-14 | 2007-08-22 | 浙江大学 | Transmedia searching method based on content correlation |
CN102693321A (en) * | 2012-06-04 | 2012-09-26 | 常州南京大学高新技术研究院 | Cross-media information analysis and retrieval method |
CN103646094A (en) * | 2013-12-18 | 2014-03-19 | 上海紫竹数字创意港有限公司 | System and method for automatic extraction and generation of audiovisual product content abstract |
CN104166982A (en) * | 2014-06-30 | 2014-11-26 | 复旦大学 | Image optimization clustering method based on typical correlation analysis |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20030004922A1 (en) * | 2001-06-27 | 2003-01-02 | Ontrack Data International, Inc. | System and method for data management |
-
2015
- 2015-03-20 CN CN201510123093.1A patent/CN104679902B/en not_active Expired - Fee Related
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101021849A (en) * | 2006-09-14 | 2007-08-22 | 浙江大学 | Transmedia searching method based on content correlation |
CN102693321A (en) * | 2012-06-04 | 2012-09-26 | 常州南京大学高新技术研究院 | Cross-media information analysis and retrieval method |
CN103646094A (en) * | 2013-12-18 | 2014-03-19 | 上海紫竹数字创意港有限公司 | System and method for automatic extraction and generation of audiovisual product content abstract |
CN104166982A (en) * | 2014-06-30 | 2014-11-26 | 复旦大学 | Image optimization clustering method based on typical correlation analysis |
Also Published As
Publication number | Publication date |
---|---|
CN104679902A (en) | 2015-06-03 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104679902B (en) | A kind of informative abstract extracting method of combination across Media Convergence | |
Wang et al. | Event driven web video summarization by tag localization and key-shot identification | |
Zhu et al. | Video synopsis by heterogeneous multi-source correlation | |
CN107292642B (en) | Commodity recommendation method and system based on images | |
CN106709032A (en) | Method and device for extracting structured information from spreadsheet document | |
CN109886330A (en) | Method for text detection, device, computer readable storage medium and computer equipment | |
Papadopoulos et al. | Image clustering through community detection on hybrid image similarity graphs | |
Elkasrawi et al. | What you see is what you get? Automatic Image Verification for Online News Content | |
CN109272440B (en) | Thumbnail generation method and system combining text and image content | |
Zhu et al. | Learning from multiple sources for video summarisation | |
CN113762138A (en) | Method and device for identifying forged face picture, computer equipment and storage medium | |
CN108268875A (en) | A kind of image meaning automatic marking method and device based on data smoothing | |
US20220292809A1 (en) | Methods and systems for grouping of media based on similarities between features of the media | |
Wang et al. | MindCamera: Interactive sketch-based image retrieval and synthesis | |
US20150356353A1 (en) | Method for identifying objects in an audiovisual document and corresponding device | |
Wang et al. | Task-dependent and query-dependent subspace learning for cross-modal retrieval | |
Wan et al. | Efficient virtual data search for annotation‐free vehicle reidentification | |
Guo et al. | Bag of surrogate parts feature for visual recognition | |
Luo et al. | An optimization framework of video advertising: using deep learning algorithm based on global image information | |
Guo et al. | Saliency detection on sampled images for tag ranking | |
Saravanan | Segment based indexing technique for video data file | |
CN116955707A (en) | Content tag determination method, device, equipment, medium and program product | |
KR102444172B1 (en) | Method and System for Intelligent Mining of Digital Image Big-Data | |
CN114998809A (en) | False news detection method and system based on ALBERT and multi-mode cycle fusion | |
CN114463552A (en) | Transfer learning and pedestrian re-identification method and related equipment |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20171128 |
|
CF01 | Termination of patent right due to non-payment of annual fee |