CN109710792A - A kind of fast face searching system application based on index - Google Patents
A kind of fast face searching system application based on index Download PDFInfo
- Publication number
- CN109710792A CN109710792A CN201811580645.1A CN201811580645A CN109710792A CN 109710792 A CN109710792 A CN 109710792A CN 201811580645 A CN201811580645 A CN 201811580645A CN 109710792 A CN109710792 A CN 109710792A
- Authority
- CN
- China
- Prior art keywords
- face
- characteristic data
- cluster
- face picture
- picture
- 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
Landscapes
- Processing Or Creating Images (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The fast face searching system application based on index that the present invention relates to a kind of, including face picture storage method and face picture search method are had big data bearing capacity, can be supported 10,000,000,000 grades of data scale using completely new logical design；And characteristic dimension is high, can preferably carry out characterizing semantics；Moreover, in practical applications, the response time is fast, has faster retrieval rate for entire design.
Description
Technical field
The fast face searching system application based on index that the present invention relates to a kind of, belongs to magnanimity face technical field.
Background technique
With the prevalence of social network sites, the unstructured datas such as image, video are daily all with surprising speed in internet
Degree increases.For the mass picture comprising enriching visual information, how in these immense image libraries easily and fast, accurately
Ground is inquired and is retrieved needed for user or interested image, and the hot spot of information retrieval field research is become.Existing face
Image retrieval technologies mainly have text based image retrieval technologies and two kinds of content-based image retrieval technology.
Text based image retrieval technologies characterize facial image feature by the way of manually marking.Big
In the facial image retrieving of scale, the keyword message of iamge description is extracted first, utilizes Inverted Index Technique later
Keyword message is established and is indexed.User needs the key message to picture to be checked to retouch when carrying out face picture retrieval
It states, then matches the description information of extraction with the keyword message in inverted index table, approximate figure is inquired with this
Sheet data.
Content-based image retrieval technology is analyzed and is mentioned to facial image feature using computer vision technique
It takes, and the characteristic of extraction is put in storage.When user carries out inquiry operation, using identical feature extracting method to figure
As extracting feature, then using feature difference is calculated, finally it is ranked up according to the size of feature difference, and according to preset threshold
Value exports the picture met the requirements.
Text based image retrieval technologies mark when need manpower work handle so that this mode may be only available for it is small
The image data retrieval of scale, and mode this for the image data of magnanimity seems awkward.And carrying out image mark
There is very strong subjectivity when note, used by the human-subject test of labeler, speech and subjective judgement etc. is influenced so that marking
The accuracy and integrality of note hardly result in guarantee.
Content-based image retrieval technology has higher requirement for the quality for extracting feature, if feature can not be very
It characterizes well or is effectively different from other images, be then likely to result in retrieval failure.And the image extracted at this stage is special
Sign is mostly high dimensional feature, needs to occupy a large amount of time and resource carrying out aspect ratio clock synchronization, performance is lower.
Summary of the invention
The fast face searching system application based on index that technical problem to be solved by the invention is to provide a kind of, passes through
Quantization and coding to high dimensional feature, reduce the memory space of high dimensional feature；Construction feature index accelerates the comparison of feature simultaneously
And retrieving, to meet the demand that data volume is big, characteristic dimension is high, retrieval time is fast.
In order to solve the above-mentioned technical problem the present invention uses following technical scheme: the present invention devises a kind of based on index
Fast face searching system application, including face picture characteristic storage method, face picture storage method include the following steps:
Step A1. extraction is obtained wait store face characteristic data corresponding to face picture, subsequently into step A2；
Step A2. is based on presetting each cluster centre, and it is poly- to carry out level for face picture face characteristic data to be stored
Class obtains cluster corresponding to face picture face characteristic data to be stored, and distributes uniquely for face picture to be stored
Identification index ID, subsequently into step A3；
Step A3. be directed to wait store face characteristic data corresponding to face picture carry out Hash it is encoded translated, obtain to
Feature hash corresponding to face picture face characteristic data is stored, subsequently into step A4；
The unique identification of face picture to be stored is indexed ID by step A4., with face picture face characteristic data to be stored
Corresponding feature hash is associated, and is stored in memory database, face picture face characteristic data to be stored
The storage region of corresponding cluster, subsequently into step A5；
Face picture to be stored is stored in disk database by step A5., and will face picture unique identification rope be stored
Draw ID, wait store corresponding to face characteristic data corresponding to face picture, face picture face characteristic data to be stored cluster, with
And face picture store path to be stored is associated, and is stored in disk database.
As a preferred technical solution of the present invention, further includes face picture search method, include the following steps:
Step B1., which is extracted, obtains face characteristic data corresponding to face picture to be retrieved, subsequently into step B2；
Step B2. is based on presetting each cluster centre, and it is poly- to carry out level for face picture face characteristic data to be retrieved
Class obtains cluster corresponding to face picture face characteristic data to be retrieved, and meets default rule of similarity with the cluster
Other each clusters, as each cluster to be processed, subsequently into step B3；
Step B3. scans memory database, obtains all unique marks in storage region corresponding to each cluster to be processed
Index ID and corresponding feature hash are known, as each group data to be matched, subsequently into step B4；
It is encoded translated that step B4. for face characteristic data corresponding to face picture to be retrieved carries out Hash, obtain to
Feature hash corresponding to face picture face characteristic data is retrieved, as feature hash to be matched, subsequently into
Step B5；
Step B5. calculates in each group data to be matched feature hash respectively between feature hash to be matched
Error obtains each group data to be matched that error is lower than default error threshold, and extracts unique in each group data to be matched
Identification index ID, subsequently into step B6；
Step B6. obtains each unique identification according to step B5 and indexes ID, and in disk database, it is each unique to extract this
The corresponding face characteristic data of identification index ID difference and corresponding face picture store path, as each group similar candidate
Data, subsequently into step B7；
Step B7. is according to face picture store path in each group similar candidate data, by extracting each in disk database
Face picture, the search result as face picture to be retrieved.
As a preferred technical solution of the present invention: further including step B6-7 as follows, after executing the step B6, enter
Step B6-7 after executing the step B6-7, enters step B7；
It is special with face picture face to be retrieved respectively that step B6-7. calculates face characteristic data in each group similar candidate data
The Euclidean distance between data is levied, each group similar candidate data that Euclidean distance is higher than pre-determined distance threshold value are deleted, for residue
Each group similar candidate data, enter step B7.
As a preferred technical solution of the present invention: face characteristic data in the step A1 are extracted and described
Face characteristic data in step B1 are extracted, and are realized by following operation；
First against face picture, by least two convolutional neural networks, by convolutional neural networks output accuracy by low
To high sequence, successively the position of face in face picture is positioned；Then, using residual error network from the face of acquisition
Extract characteristic.
As a preferred technical solution of the present invention: special in each group data to be matched by calculating in the step B5
The hash Hamming code distance between feature hash to be matched respectively is levied, is breathed out as feature in each group data to be matched
The uncommon data error between feature hash to be matched respectively.
As a preferred technical solution of the present invention: obtaining face picture face characteristic number to be stored in the step A2
According to cluster corresponding to face picture face characteristic data to be retrieved is obtained in corresponding cluster and the step B2,
It is realized by following operation:
It is primarily based on and presets each cluster centre, carry out hierarchical clustering for the face characteristic data of face picture, obtain
Cluster corresponding to the face characteristic data of the face picture；
Then, based on preset each sub- cluster centre in the corresponding cluster, then the face characteristic number to face picture
According to hierarchical clustering is carried out, the cluster of son corresponding to the face characteristic data of the face picture is further obtained；
By above-mentioned secondary cluster, cluster corresponding to the face characteristic data by the face picture and its son cluster are made
Cluster corresponding to face characteristic data for the face picture.
As a preferred technical solution of the present invention: special obtaining face picture face to be retrieved in the step B2
After levying cluster corresponding to data, Euclidean distance of the cluster centre respectively between remaining each cluster centre is calculated, and select
Each cluster that Euclidean distance is lower than pre-determined distance threshold value is selected, is gathered in conjunction with corresponding to face picture face characteristic data to be retrieved
Class, as each cluster to be processed.
A kind of fast face searching system based on index of the present invention is applied using above technical scheme and existing skill
Art is compared, and is had following technical effect that
A kind of fast face searching system application based on index designed by the present invention has big data bearing capacity, energy
Enough support 10,000,000,000 grades of data scale；And characteristic dimension is high, can preferably carry out characterizing semantics；Moreover, it entirely sets
In practical applications, the response time is fast, has faster retrieval rate for meter.
Detailed description of the invention
Fig. 1 is the flow diagram of face picture storage method designed by the present invention；
Fig. 2 is the flow diagram of face picture search method designed by the present invention.
Specific embodiment
Specific embodiments of the present invention will be described in further detail with reference to the accompanying drawings of the specification.
The present invention devises a kind of fast face searching system application based on index, in practical application, including face
Picture feature storage method and face picture search method, wherein as shown in Figure 1, face picture storage method includes following step
Rapid A1 to step A5.
Step A1. extraction is obtained wait store face characteristic data corresponding to face picture, subsequently into step A2.
In practical application, the face characteristic data in above-mentioned steps A1 are extracted, and are realized by following operation.
First against face picture, by least two convolutional neural networks, by convolutional neural networks output accuracy by low
To high sequence, successively the position of face in face picture is positioned；Then, using residual error network from the face of acquisition
Extract characteristic.
Step A2. is based on presetting each cluster centre, and it is poly- to carry out level for face picture face characteristic data to be stored
Class obtains cluster corresponding to face picture face characteristic data to be stored, and distributes uniquely for face picture to be stored
Identification index ID, subsequently into step A3.
In practical application, above-mentioned steps A2 is obtained corresponding to face picture face characteristic data to be stored by following operation
Cluster.
It is primarily based on and presets each cluster centre, carry out hierarchical clustering for the face characteristic data of face picture, obtain
Cluster corresponding to the face characteristic data of the face picture.
Then, based on preset each sub- cluster centre in the corresponding cluster, then the face characteristic number to face picture
According to hierarchical clustering is carried out, the cluster of son corresponding to the face characteristic data of the face picture is further obtained.
By above-mentioned secondary cluster, cluster corresponding to the face characteristic data by the face picture and its son cluster are made
Cluster corresponding to face characteristic data for the face picture.
In this way, cluster operation twice is passed sequentially through, so as to effectively improve recall precision.
Due to the real-coded GA that the feature of original face is higher-dimension, calculating process is complicated and amount of storage is big, and inconvenient
Among storage to memory.Therefore, primitive character is encoded by the way of Hash coding, is translated into convenient for processing
Binary coded form, data volume can be reduced, improve memory service efficiency, therefore continue to execute following steps A3.
Step A3. be directed to wait store face characteristic data corresponding to face picture carry out Hash it is encoded translated, obtain to
Feature hash corresponding to face picture face characteristic data is stored, subsequently into step A4.
The unique identification of face picture to be stored is indexed ID by step A4., with face picture face characteristic data to be stored
Corresponding feature hash is associated, and is stored in memory database, face picture face characteristic data to be stored
The storage region of corresponding cluster, subsequently into step A5.
Face picture to be stored is stored in disk database by step A5., and will face picture unique identification rope be stored
Draw ID, wait store corresponding to face characteristic data corresponding to face picture, face picture face characteristic data to be stored cluster, with
And face picture store path to be stored is associated, and is stored in disk database.
In practical application, face picture search method, as shown in Fig. 2, including the following steps B1 to step B7.
Step B1., which is extracted, obtains face characteristic data corresponding to face picture to be retrieved, subsequently into step B2.
In practical application, the face characteristic data in above-mentioned steps B1 are extracted, and are realized by following operation.
First against face picture, by least two convolutional neural networks, by convolutional neural networks output accuracy by low
To high sequence, successively the position of face in face picture is positioned；Then, using residual error network from the face of acquisition
Extract characteristic.
Step B2. is based on presetting each cluster centre, and it is poly- to carry out level for face picture face characteristic data to be retrieved
Class obtains cluster corresponding to face picture face characteristic data to be retrieved, and meets default rule of similarity with the cluster
Other each clusters, as each cluster to be processed, subsequently into step B3.
In practical application, above-mentioned steps B2 is obtained corresponding to face picture face characteristic data to be retrieved by following operation
Cluster.
It is primarily based on and presets each cluster centre, carry out hierarchical clustering for the face characteristic data of face picture, obtain
Cluster corresponding to the face characteristic data of the face picture.
Then, based on preset each sub- cluster centre in the corresponding cluster, then the face characteristic number to face picture
According to hierarchical clustering is carried out, the cluster of son corresponding to the face characteristic data of the face picture is further obtained.
By above-mentioned secondary cluster, cluster corresponding to the face characteristic data by the face picture and its son cluster are made
Cluster corresponding to face characteristic data for the face picture.
Here same, cluster operation twice is passed sequentially through, so as to effectively improve recall precision.
Step B2 in practical applications, after obtaining cluster corresponding to face picture face characteristic data to be retrieved, is counted
Euclidean distance of the cluster centre respectively between remaining each cluster centre is calculated, and selects Euclidean distance lower than pre-determined distance threshold
Each cluster of value, the cluster in conjunction with corresponding to face picture face characteristic data to be retrieved, as each cluster to be processed.
Step B3. scans memory database, obtains all unique marks in storage region corresponding to each cluster to be processed
Index ID and corresponding feature hash are known, as each group data to be matched, subsequently into step B4.
It is encoded translated that step B4. for face characteristic data corresponding to face picture to be retrieved carries out Hash, obtain to
Feature hash corresponding to face picture face characteristic data is retrieved, as feature hash to be matched, subsequently into
Step B5.
Step B5. calculates in each group data to be matched feature hash respectively between feature hash to be matched
Hamming code distance, as the mistake between feature hash to be matched respectively of feature hash in each group data to be matched
Difference obtains each group data to be matched that error is lower than default error threshold, and extracts unique mark in each group data to be matched
Index ID is known, subsequently into step B6.
Step B6. obtains each unique identification according to step B5 and indexes ID, and in disk database, it is each unique to extract this
The corresponding face characteristic data of identification index ID difference and corresponding face picture store path, as each group similar candidate
Data, subsequently into step B6-7.
It is special with face picture face to be retrieved respectively that step B6-7. calculates face characteristic data in each group similar candidate data
The Euclidean distance between data is levied, each group similar candidate data that Euclidean distance is higher than pre-determined distance threshold value are deleted, for residue
Each group similar candidate data, enter step B7.
Step B7. is according to face picture store path in each group similar candidate data, by extracting each in disk database
Face picture, the search result as face picture to be retrieved.
There is big data to carry energy for a kind of fast face searching system application based on index designed by above-mentioned technical proposal
Power can support 10,000,000,000 grades of data scale；And characteristic dimension is high, can preferably carry out characterizing semantics；Moreover, whole
In practical applications, the response time is fast, has faster retrieval rate for a design.
Embodiments of the present invention are explained in detail above in conjunction with attached drawing, but the present invention is not limited to above-mentioned implementations
Mode within the knowledge of a person skilled in the art can also be without departing from the purpose of the present invention
It makes a variety of changes.
Claims (7)
1. a kind of fast face searching system application based on index, including face picture characteristic storage method, which is characterized in that
Face picture storage method includes the following steps:
Step A1. extraction is obtained wait store face characteristic data corresponding to face picture, subsequently into step A2；
Step A2. is based on presetting each cluster centre, carries out hierarchical clustering for face picture face characteristic data to be stored, obtains
Cluster corresponding to face picture face characteristic data to be stored is obtained, and distributes unique identification rope for face picture to be stored
Draw ID, subsequently into step A3；
Step A3. be directed to wait store face characteristic data corresponding to face picture carry out Hash it is encoded translated, obtain wait store
Feature hash corresponding to face picture face characteristic data, subsequently into step A4；
The unique identification of face picture to be stored is indexed ID by step A4., right with face picture face characteristic data institute to be stored
The feature hash answered is associated, and be stored in memory database, face picture face characteristic data to be stored institute it is right
The storage region that should be clustered, subsequently into step A5；
Step A5. by face picture to be stored deposit disk database, and will face picture unique identification index ID be stored,
Wait store corresponding to face characteristic data corresponding to face picture, face picture face characteristic data to be stored cluster and to
Storage face picture store path is associated, and is stored in disk database.
2. a kind of fast face searching system application based on index according to claim 1, which is characterized in that further include people
Face picture retrieval method, includes the following steps:
Step B1., which is extracted, obtains face characteristic data corresponding to face picture to be retrieved, subsequently into step B2；
Step B2. is based on presetting each cluster centre, carries out hierarchical clustering for face picture face characteristic data to be retrieved, obtains
Cluster corresponding to face picture face characteristic data to be retrieved is obtained, and meets the other each of default rule of similarity with the cluster
A cluster, as each cluster to be processed, subsequently into step B3；
Step B3. scans memory database, obtains all unique identification ropes in storage region corresponding to each cluster to be processed
Draw ID and corresponding feature hash, as each group data to be matched, subsequently into step B4；
Step B4. is encoded translated for the progress Hash of face characteristic data corresponding to face picture to be retrieved, obtains to be retrieved
Feature hash corresponding to face picture face characteristic data, as feature hash to be matched, subsequently into step
B5；
Step B5. calculates the feature hash mistake between feature hash to be matched respectively in each group data to be matched
Difference obtains each group data to be matched that error is lower than default error threshold, and extracts unique mark in each group data to be matched
Index ID is known, subsequently into step B6；
Step B6. obtains each unique identification index ID according to step B5 and extracts each unique identification in disk database
ID corresponding face characteristic data and corresponding face picture store path respectively are indexed, as each group similar candidate data,
Subsequently into step B7；
Step B7. is according to face picture store path in each group similar candidate data, by extracting each face in disk database
Picture, the search result as face picture to be retrieved.
3. a kind of fast face searching system application based on index according to claim 2, it is characterised in that: further include step
Rapid B6-7 is as follows, after executing the step B6, enters step B6-7, after executing the step B6-7, enters step B7；
Step B6-7. calculate each group similar candidate data in face characteristic data respectively with face picture face characteristic number to be retrieved
Euclidean distance between deletes each group similar candidate data that Euclidean distance is higher than pre-determined distance threshold value, for remaining each group
Similar candidate data, enter step B7.
4. a kind of fast face searching system application based on index according to claim 2, it is characterised in that: the step
Face characteristic data extraction in A1 and the face characteristic data in the step B1 are extracted, by following operation realization；
First against face picture, by least two convolutional neural networks, from low to high by convolutional neural networks output accuracy
Sequence, successively the position of face in face picture is positioned；Then, it is extracted from the face of acquisition using residual error network
Characteristic out.
5. a kind of fast face searching system application based on index according to claim 2, it is characterised in that: the step
In B5, by calculating the feature hash Hamming code between feature hash to be matched respectively in each group data to be matched
Distance, as the error between feature hash to be matched respectively of feature hash in each group data to be matched.
6. a kind of fast face searching system application based on index according to claim 2, it is characterised in that: the step
It is obtained in A2 in cluster corresponding to face picture face characteristic data to be stored and the step B2 and obtains face to be retrieved
Cluster corresponding to picture face characteristic data is realized by following operation:
It is primarily based on and presets each cluster centre, carry out hierarchical clustering for the face characteristic data of face picture, obtain the people
Cluster corresponding to the face characteristic data of face picture；
Then, based on preset each sub- cluster centre in the corresponding cluster, then to the face characteristic data of face picture into
Row hierarchical clustering further obtains the cluster of son corresponding to the face characteristic data of the face picture；
By above-mentioned secondary cluster, cluster corresponding to the face characteristic data by the face picture and its son cluster, as this
Cluster corresponding to the face characteristic data of face picture.
7. a kind of fast face searching system application based on index according to claim 2, it is characterised in that: the step
In B2, after obtaining cluster corresponding to face picture face characteristic data to be retrieved, calculate the cluster centre respectively with remaining
Euclidean distance between each cluster centre, and each cluster of the Euclidean distance lower than pre-determined distance threshold value is selected, in conjunction with to be checked
Cluster corresponding to rope face picture face characteristic data, as each cluster to be processed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811580645.1A CN109710792B (en) | 2018-12-24 | 2018-12-24 | Index-based rapid face retrieval system application |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811580645.1A CN109710792B (en) | 2018-12-24 | 2018-12-24 | Index-based rapid face retrieval system application |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109710792A true CN109710792A (en) | 2019-05-03 |
| CN109710792B CN109710792B (en) | 2023-02-03 |
Family
ID=66257402
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811580645.1A Active CN109710792B (en) | 2018-12-24 | 2018-12-24 | Index-based rapid face retrieval system application |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109710792B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021027555A1 (en) * | 2019-08-15 | 2021-02-18 | 华为技术有限公司 | Face retrieval method and apparatus |
| CN112445892A (en) * | 2019-09-02 | 2021-03-05 | 百度在线网络技术（北京）有限公司 | Method and device for determining brand mentioning rate, electronic equipment and storage medium |
| WO2021047305A1 (en) * | 2019-09-10 | 2021-03-18 | 中兴通讯股份有限公司 | Face search method and device, apparatus, and computer-readable storage medium |
| CN112633051A (en) * | 2020-09-11 | 2021-04-09 | 博云视觉（北京）科技有限公司 | Online face clustering method based on image search |
| CN113806582A (en) * | 2021-11-17 | 2021-12-17 | 腾讯科技（深圳）有限公司 | Image retrieval method, image retrieval device, electronic equipment and storage medium |
| CN114036324A (en) * | 2021-10-27 | 2022-02-11 | 青岛海纳云科技控股有限公司 | Storage method and device of face picture file |
| CN114048344A (en) * | 2021-11-25 | 2022-02-15 | 天翼数字生活科技有限公司 | Similar face searching method, device, equipment and readable storage medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090238464A1 (en) * | 2008-03-21 | 2009-09-24 | Masakazu Ohira | Image processing method, image processing apparatus, image forming apparatus and storage medium |
| US20150227557A1 (en) * | 2014-02-10 | 2015-08-13 | Geenee Ug | Systems and methods for image-feature-based recognition |
| CN105260412A (en) * | 2015-09-24 | 2016-01-20 | 东方网力科技股份有限公司 | Image storage method and device, and image retrieval method and device |
| CN105589929A (en) * | 2015-12-09 | 2016-05-18 | 东方网力科技股份有限公司 | Image retrieval method and device |
| CN105930834A (en) * | 2016-07-01 | 2016-09-07 | 北京邮电大学 | Face identification method and apparatus based on spherical hashing binary coding |
| US9922052B1 (en) * | 2013-04-26 | 2018-03-20 | A9.Com, Inc. | Custom image data store |
-
2018
- 2018-12-24 CN CN201811580645.1A patent/CN109710792B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090238464A1 (en) * | 2008-03-21 | 2009-09-24 | Masakazu Ohira | Image processing method, image processing apparatus, image forming apparatus and storage medium |
| US9922052B1 (en) * | 2013-04-26 | 2018-03-20 | A9.Com, Inc. | Custom image data store |
| US20150227557A1 (en) * | 2014-02-10 | 2015-08-13 | Geenee Ug | Systems and methods for image-feature-based recognition |
| CN105260412A (en) * | 2015-09-24 | 2016-01-20 | 东方网力科技股份有限公司 | Image storage method and device, and image retrieval method and device |
| CN105589929A (en) * | 2015-12-09 | 2016-05-18 | 东方网力科技股份有限公司 | Image retrieval method and device |
| CN105930834A (en) * | 2016-07-01 | 2016-09-07 | 北京邮电大学 | Face identification method and apparatus based on spherical hashing binary coding |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021027555A1 (en) * | 2019-08-15 | 2021-02-18 | 华为技术有限公司 | Face retrieval method and apparatus |
| CN112395448A (en) * | 2019-08-15 | 2021-02-23 | 华为技术有限公司 | Face retrieval method and device |
| CN112445892A (en) * | 2019-09-02 | 2021-03-05 | 百度在线网络技术（北京）有限公司 | Method and device for determining brand mentioning rate, electronic equipment and storage medium |
| CN112445892B (en) * | 2019-09-02 | 2023-09-29 | 百度在线网络技术（北京）有限公司 | Method, device, electronic equipment and storage medium for determining brand mention rate |
| WO2021047305A1 (en) * | 2019-09-10 | 2021-03-18 | 中兴通讯股份有限公司 | Face search method and device, apparatus, and computer-readable storage medium |
| CN112633051A (en) * | 2020-09-11 | 2021-04-09 | 博云视觉（北京）科技有限公司 | Online face clustering method based on image search |
| CN114036324A (en) * | 2021-10-27 | 2022-02-11 | 青岛海纳云科技控股有限公司 | Storage method and device of face picture file |
| CN113806582A (en) * | 2021-11-17 | 2021-12-17 | 腾讯科技（深圳）有限公司 | Image retrieval method, image retrieval device, electronic equipment and storage medium |
| CN113806582B (en) * | 2021-11-17 | 2022-02-25 | 腾讯科技（深圳）有限公司 | Image retrieval method, image retrieval device, electronic equipment and storage medium |
| CN114048344A (en) * | 2021-11-25 | 2022-02-15 | 天翼数字生活科技有限公司 | Similar face searching method, device, equipment and readable storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109710792B (en) | 2023-02-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109710792A (en) | A kind of fast face searching system application based on index | |
| CN111177446B (en) | Method for searching footprint image | |
| Wang et al. | Annosearch: Image auto-annotation by search | |
| CN102254015B (en) | Image retrieval method based on visual phrases | |
| CN109165294A (en) | Short text classification method based on Bayesian classification | |
| CN107169106B (en) | Video retrieval method, device, storage medium and processor | |
| CN105843850B (en) | Search optimization method and device | |
| CN109165273A (en) | General Chinese address matching method facing big data environment | |
| CN104346438A (en) | Data management service system based on large data | |
| CN114461839B (en) | Multi-mode pre-training-based similar picture retrieval method and device and electronic equipment | |
| CN107423309A (en) | Magnanimity internet similar pictures detecting system and method based on fuzzy hash algorithm | |
| CN105678244B (en) | A kind of near video search method based on improved edit-distance | |
| CN110765266B (en) | Method and system for merging similar dispute focuses of referee documents | |
| CN109657063A (en) | A kind of processing method and storage medium of magnanimity environment-protection artificial reported event data | |
| CN108304502A (en) | Quick hot spot detecting method and system based on magnanimity news data | |
| CN113282834A (en) | Web search intelligent ordering method, system and computer storage medium based on mobile internet data deep mining | |
| CN109446399A (en) | A kind of video display entity search method | |
| CN105183746A (en) | Method for realizing image retrieval by mining distinguishing features from multiple relevant pictures | |
| KR20120087214A (en) | Friend recommendation method for SNS user, recording medium for the same, and SNS and server using the same | |
| CN104615734A (en) | Community management service big data processing system and processing method thereof | |
| CN113449111A (en) | Social governance hot topic automatic identification method based on time-space semantic knowledge migration | |
| CN108121806A (en) | One kind is based on the matched image search method of local feature and system | |
| CN102663053A (en) | Distributed server system based on image content search | |
| CN114943285B (en) | Intelligent auditing system for internet news content data | |
| CN107423294A (en) | A kind of community image search method and system |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |