CN109710792B - Index-based rapid face retrieval system application - Google Patents
Index-based rapid face retrieval system application Download PDFInfo
- Publication number
- CN109710792B CN109710792B CN201811580645.1A CN201811580645A CN109710792B CN 109710792 B CN109710792 B CN 109710792B CN 201811580645 A CN201811580645 A CN 201811580645A CN 109710792 B CN109710792 B CN 109710792B
- Authority
- CN
- China
- Prior art keywords
- face
- data
- face picture
- picture
- cluster
- 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.)
- Active
Links
Images
Landscapes
- Processing Or Creating Images (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention relates to an index-based rapid face retrieval system application, which comprises a face picture storage method and a face picture retrieval method, adopts brand-new logic design, has large data bearing capacity and can support billion-level data scale; the feature dimension is high, and semantic representation can be better performed; moreover, the whole design has fast response time and faster retrieval speed in practical application.
Description
Technical Field
The invention relates to an application of an index-based rapid face retrieval system, belonging to the technical field of massive faces.
Background
With the popularity of social networking sites, unstructured data such as images, videos, etc. in the internet is growing at an alarming rate each day. For massive pictures containing rich visual information, how to conveniently, quickly and accurately query and retrieve images required by or interested by users in these vast image libraries becomes a hotspot of research in the field of information retrieval. The existing face image retrieval technology mainly comprises a text-based image retrieval technology and a content-based image retrieval technology.
The image retrieval technology based on the text adopts a manual labeling mode to represent the characteristics of the human face image. In the large-scale face image retrieval process, firstly, keyword information described by an image is extracted, and then an index is established for the keyword information by utilizing an inverted index technology. When a user searches a face picture, the key information of the picture to be searched needs to be described, and then the extracted description information is matched with the key word information in the inverted index table, so that approximate picture data can be searched.
The image retrieval technology based on the content utilizes the computer vision technology to analyze and extract the human face image characteristics, and stores the extracted characteristic data in a warehouse. When a user performs query operation, the same feature extraction method is adopted to extract features from the image, then the feature difference is calculated, and finally, sorting is performed according to the size of the feature difference, and pictures meeting requirements are output according to a preset threshold value.
The image retrieval technology based on the text needs manual processing during marking, so that the method is only suitable for small-scale image data retrieval, and the method looks very poor for massive image data. And the method has strong subjectivity when image annotation is carried out, and the accuracy and the integrity of the annotation are difficult to ensure due to the influences of the cognitive level, the speech use, the subjective judgment and the like of an annotator.
The content-based image retrieval technology has high requirements on the quality of extracted features, and if the features cannot be well characterized or are effectively distinguished from other images, the retrieval is likely to fail. Moreover, most of the image features extracted at the present stage are high-dimensional features, and a large amount of time and resources are required to be occupied when feature comparison is performed, so that the performance is low.
Disclosure of Invention
The invention aims to solve the technical problem of providing an index-based rapid face retrieval system application, which reduces the storage space of high-dimensional features by quantizing and coding the high-dimensional features; meanwhile, the comparison and retrieval process of the characteristic index acceleration characteristic is constructed, so that the requirements of large data volume, high characteristic dimension and quick retrieval time are met.
The invention adopts the following technical scheme for solving the technical problems: the invention designs an index-based rapid face retrieval system application, which comprises a face picture feature storage method, wherein the face picture storage method comprises the following steps:
a1, extracting face feature data corresponding to a face picture to be stored, and then entering A2;
step A2, based on each preset cluster center, carrying out hierarchical clustering on the face feature data of the face picture to be stored to obtain clusters corresponding to the face feature data of the face picture to be stored, allocating a unique identification index ID to the face picture to be stored, and then entering the step A3;
a3, carrying out Hash coding conversion on the face feature data corresponding to the face picture to be stored to obtain feature Hash data corresponding to the face feature data of the face picture to be stored, and then entering the step A4;
step A4, associating the unique identification index ID of the face picture to be stored with the feature hash data corresponding to the face feature data of the face picture to be stored, storing the unique identification index ID in a memory database, storing the unique identification index ID in a cluster storage area corresponding to the face feature data of the face picture to be stored, and entering the step A5;
and step A5, storing the face picture to be stored into a disk database, associating the unique identification index ID of the face picture to be stored, the face feature data corresponding to the face picture to be stored, the cluster corresponding to the face feature data of the face picture to be stored and the storage path of the face picture to be stored, and storing the associated data into the disk database.
As a preferred technical solution of the present invention, the present invention further includes a face image retrieval method, including the steps of:
b1, extracting and obtaining face feature data corresponding to the face picture to be retrieved, and then entering a step B2;
b2, based on each preset cluster center, carrying out hierarchical clustering on the face feature data of the face picture to be retrieved to obtain a cluster corresponding to the face feature data of the face picture to be retrieved and other clusters meeting preset similar rules with the cluster as each cluster to be processed, and then entering the step B3;
b3, scanning a memory database, acquiring all unique identification index IDs in storage areas corresponding to the clusters to be processed and corresponding characteristic hash data as each group of data to be matched, and then entering a step B4;
b4, carrying out Hash coding conversion on the face characteristic data corresponding to the face picture to be retrieved to obtain characteristic Hash data corresponding to the face characteristic data of the face picture to be retrieved, taking the characteristic Hash data as characteristic Hash data to be matched, and then entering the step B5;
b5, calculating errors between the characteristic hash data in each group of data to be matched and the characteristic hash data to be matched respectively, obtaining each group of data to be matched with the errors lower than a preset error threshold value, extracting unique identification index ID in each group of data to be matched, and then entering the step B6;
b6, extracting the face feature data respectively corresponding to each unique identification index ID and the corresponding face picture storage path in the disk database according to each unique identification index ID obtained in the step B5 to serve as each group of similar alternative data, and then entering a step B7;
and B7, extracting each human face picture from the disk database according to the storage path of the human face pictures in each group of similar alternative data to be used as the retrieval result of the human face picture to be retrieved.
As a preferred technical scheme of the invention: the method also comprises the following steps B6-7, after the step B6 is executed, the step B6-7 is executed, and after the step B6-7 is executed, the step B7 is executed;
and B6-7, calculating Euclidean distances between the face feature data in each group of similar alternative data and the face feature data of the face picture to be retrieved respectively, deleting each group of similar alternative data with the Euclidean distance being higher than a preset distance threshold value, and entering the step B7 aiming at the rest groups of similar alternative data.
As a preferred technical scheme of the invention: the extraction of the face feature data in the step A1 and the extraction of the face feature data in the step B1 are both realized according to the following operations;
firstly, sequentially positioning the positions of the human faces in the human face picture by at least two convolutional neural networks according to the sequence of the output accuracy of the convolutional neural networks from low to high; and then, extracting feature data from the obtained human face by adopting a residual error network.
As a preferred technical scheme of the invention: in the step B5, the hamming code distance between the characteristic hash data in each group of data to be matched and the characteristic hash data to be matched is calculated to be used as the error between the characteristic hash data in each group of data to be matched and the characteristic hash data to be matched.
As a preferred technical scheme of the invention: the cluster corresponding to the face feature data of the face picture to be stored obtained in the step A2 and the cluster corresponding to the face feature data of the face picture to be retrieved obtained in the step B2 are all realized according to the following operations:
firstly, carrying out hierarchical clustering on face feature data of a face picture based on preset clustering centers to obtain clusters corresponding to the face feature data of the face picture;
then, based on each pre-set sub-cluster center in the corresponding cluster, hierarchical clustering is carried out on the face feature data of the face picture, and sub-clusters corresponding to the face feature data of the face picture are further obtained;
and by the secondary clustering, the cluster corresponding to the face characteristic data of the face picture and the sub-cluster thereof are used as the cluster corresponding to the face characteristic data of the face picture.
As a preferred technical scheme of the invention: in the step B2, after the cluster corresponding to the face feature data of the face picture to be retrieved is obtained, the Euclidean distance between the cluster center and each of the rest cluster centers is calculated, and selecting each cluster with the Euclidean distance lower than a preset distance threshold value, and combining the cluster corresponding to the facial feature data of the facial picture to be retrieved to serve as each cluster to be processed.
Compared with the prior art, the index-based rapid face retrieval system provided by the invention has the following technical effects by adopting the technical scheme:
the index-based rapid face retrieval system has large data carrying capacity and can support billion-level data scale; the feature dimension is high, and semantic representation can be better performed; moreover, in practical application, the whole design has fast response time and faster retrieval rate.
Drawings
FIG. 1 is a schematic flow chart of a face image storage method according to the present invention;
fig. 2 is a schematic flow diagram of the face image retrieval method according to the present invention.
Detailed Description
The following description will explain embodiments of the present invention in further detail with reference to the accompanying drawings.
The invention designs an index-based rapid face retrieval system application, which comprises a face picture feature storage method and a face picture retrieval method in practical application, wherein the face picture storage method comprises the following steps A1 to A5 as shown in figure 1.
And A1, extracting the face feature data corresponding to the face picture to be stored, and then entering the step A2.
In practical application, the face feature data extraction in the step A1 is implemented as follows.
Firstly, sequentially positioning the positions of the human faces in the human face picture by at least two convolutional neural networks according to the sequence of the output accuracy of the convolutional neural networks from low to high; and then, extracting feature data from the acquired human face by adopting a residual error network.
Step A2, based on each preset cluster center, carrying out hierarchical clustering on the face feature data of the face picture to be stored to obtain clusters corresponding to the face feature data of the face picture to be stored, allocating a unique identification index ID to the face picture to be stored, and then entering the step A3.
In practical application, in the step A2, the cluster corresponding to the face feature data of the face picture to be stored is obtained as follows.
Firstly, hierarchical clustering is carried out on the face feature data of the face picture based on preset clustering centers, and a cluster corresponding to the face feature data of the face picture is obtained.
Then, based on each pre-set sub-cluster center in the corresponding cluster, hierarchical clustering is performed on the face feature data of the face picture, and sub-clusters corresponding to the face feature data of the face picture are further obtained.
And by the secondary clustering, the cluster corresponding to the face characteristic data of the face picture and the sub-cluster thereof are used as the cluster corresponding to the face characteristic data of the face picture.
Therefore, the searching efficiency can be effectively improved by sequentially carrying out twice clustering operations.
Because the original human face features are high-dimensional floating point type data, the calculation process is complex and the storage capacity is large, and the human face is not convenient to store in the memory. Therefore, the original features are encoded by using a hash encoding method, and converted into a binary encoding form convenient for processing, so that the data size can be reduced, and the memory use efficiency can be improved, and therefore the following step A3 is continuously executed.
And A3, carrying out Hash coding conversion on the face feature data corresponding to the face picture to be stored to obtain feature Hash data corresponding to the face feature data of the face picture to be stored, and then entering the step A4.
And A4, associating the unique identification index ID of the face picture to be stored with the feature hash data corresponding to the face feature data of the face picture to be stored, storing the unique identification index ID in the memory database and the storage area of the cluster corresponding to the face feature data of the face picture to be stored, and then entering the step A5.
And step A5, storing the face picture to be stored into a disk database, associating the unique identification index ID of the face picture to be stored, the face feature data corresponding to the face picture to be stored, the cluster corresponding to the face feature data of the face picture to be stored and the storage path of the face picture to be stored, and storing the associated data into the disk database.
In practical application, the face image retrieval method, as shown in fig. 2, includes the following steps B1 to B7.
And B1, extracting the face feature data corresponding to the face picture to be retrieved, and then entering the step B2.
In practical application, the face feature data extraction in the step B1 is implemented as follows.
Firstly, sequentially positioning the positions of the human faces in the human face picture by at least two convolutional neural networks according to the sequence of the output accuracy of the convolutional neural networks from low to high; and then, extracting feature data from the obtained human face by adopting a residual error network.
And B2, based on the preset cluster centers, carrying out hierarchical clustering on the face feature data of the face picture to be retrieved to obtain clusters corresponding to the face feature data of the face picture to be retrieved and other clusters meeting preset similarity rules with the clusters as clusters to be processed, and then entering the step B3.
In practical application, in the step B2, the cluster corresponding to the face feature data of the face picture to be retrieved is obtained as follows.
Firstly, hierarchical clustering is carried out on the face feature data of the face picture based on preset clustering centers, and a cluster corresponding to the face feature data of the face picture is obtained.
Then, based on each pre-set sub-cluster center in the corresponding cluster, hierarchical clustering is performed on the face feature data of the face picture, and sub-clusters corresponding to the face feature data of the face picture are further obtained.
And by the secondary clustering, the cluster corresponding to the face characteristic data of the face picture and the sub-cluster thereof are used as the cluster corresponding to the face characteristic data of the face picture.
Here, similarly, the search efficiency can be effectively improved by sequentially performing clustering operations twice.
And step B2, in practical application, after the clusters corresponding to the facial feature data of the facial picture to be retrieved are obtained, calculating Euclidean distances between the cluster center and the rest cluster centers respectively, selecting each cluster with the Euclidean distance lower than a preset distance threshold value, and taking the cluster corresponding to the facial feature data of the facial picture to be retrieved as each cluster to be processed.
And B3, scanning the memory database, acquiring all unique identification index IDs in the storage areas corresponding to the clusters to be processed and corresponding characteristic hash data as each group of data to be matched, and then entering the step B4.
And B4, carrying out Hash coding conversion on the face characteristic data corresponding to the face picture to be retrieved to obtain characteristic Hash data corresponding to the face characteristic data of the face picture to be retrieved, taking the characteristic Hash data as characteristic Hash data to be matched, and then entering the step B5.
And B5, calculating Hamming code distances between the characteristic hash data in each group of data to be matched and the characteristic hash data to be matched respectively, taking the Hamming code distances as errors between the characteristic hash data in each group of data to be matched and the characteristic hash data to be matched, obtaining each group of data to be matched with the errors lower than a preset error threshold, extracting unique identification index ID in each group of data to be matched, and entering the step B6.
And B6, extracting the face feature data respectively corresponding to each unique identification index ID and the corresponding face image storage path in the disk database according to each unique identification index ID obtained in the step B5 to serve as each group of similar alternative data, and then entering the step B6-7.
And B6-7, calculating Euclidean distances between the face feature data in each group of similar alternative data and the face feature data of the face picture to be retrieved respectively, deleting each group of similar alternative data with the Euclidean distance being higher than a preset distance threshold value, and entering the step B7 aiming at the rest groups of similar alternative data.
And B7, extracting each human face picture from the disk database according to the storage path of the human face picture in each group of similar alternative data to be used as a retrieval result of the human face picture to be retrieved.
The index-based rapid face retrieval system designed by the technical scheme has large data carrying capacity and can support billion-level data scale; the feature dimension is high, and semantic representation can be better performed; moreover, in practical application, the whole design has fast response time and faster retrieval rate.
The embodiments of the present invention have been described in detail with reference to the drawings, but the present invention is not limited to the above embodiments, and various changes can be made within the knowledge of those skilled in the art without departing from the gist of the present invention.
Claims (6)
1. An index-based rapid face retrieval system application comprises a face image feature storage method, and is characterized in that the face image storage method comprises the following steps:
a1, extracting face feature data corresponding to a face picture to be stored, and then entering A2;
step A2, based on preset cluster centers, carrying out hierarchical clustering on the face feature data of the face picture to be stored to obtain clusters corresponding to the face feature data of the face picture to be stored, allocating unique identification index IDs to the face picture to be stored, and then entering step A3;
step A3, carrying out Hash coding conversion on the face characteristic data corresponding to the face picture to be stored to obtain characteristic Hash data corresponding to the face characteristic data of the face picture to be stored, and then entering step A4;
step A4, associating the unique identification index ID of the face picture to be stored with the characteristic hash data corresponding to the face characteristic data of the face picture to be stored, storing the unique identification index ID in a memory database and the storage area of the cluster corresponding to the face characteristic data of the face picture to be stored, and then entering the step A5;
step A5, storing the face picture to be stored into a disk database, associating the unique identification index ID of the face picture to be stored, the face feature data corresponding to the face picture to be stored, the cluster corresponding to the face feature data of the face picture to be stored and the storage path of the face picture to be stored, and storing the cluster and the storage path in the disk database;
the method also comprises a face picture retrieval method, which comprises the following steps:
b1, extracting and obtaining face feature data corresponding to the face picture to be retrieved, and then entering a step B2;
b2, based on each preset cluster center, carrying out hierarchical clustering on the face feature data of the face picture to be retrieved to obtain a cluster corresponding to the face feature data of the face picture to be retrieved and other clusters meeting preset similar rules with the cluster as each cluster to be processed, and then entering the step B3;
b3, scanning the memory database, acquiring all unique identification index IDs in the storage areas corresponding to the clusters to be processed and corresponding characteristic hash data as each group of data to be matched, and then entering the step B4;
b4, carrying out Hash coding conversion on the face characteristic data corresponding to the face picture to be retrieved to obtain characteristic Hash data corresponding to the face characteristic data of the face picture to be retrieved, taking the characteristic Hash data as characteristic Hash data to be matched, and then entering the step B5;
b5, calculating errors between the characteristic hash data in each group of data to be matched and the characteristic hash data to be matched respectively, obtaining each group of data to be matched with the errors lower than a preset error threshold value, extracting a unique identification index ID in each group of data to be matched, and entering the step B6;
step B6, extracting the face feature data respectively corresponding to each unique identification index ID and the corresponding face picture storage path in the disk database according to each unique identification index ID obtained in the step B5, using the face feature data and the corresponding face picture storage path as each group of similar alternative data, and then entering the step B7;
and B7, extracting each human face picture from the disk database according to the storage path of the human face picture in each group of similar alternative data to serve as a retrieval result of the human face picture to be retrieved.
2. The index-based rapid face retrieval system application of claim 1, wherein: the method also comprises the following steps B6-7, after the step B6 is executed, the step B6-7 is executed, and after the step B6-7 is executed, the step B7 is executed;
and B6-7, calculating Euclidean distances between the face feature data in each group of similar alternative data and the face feature data of the face picture to be retrieved respectively, deleting each group of similar alternative data with the Euclidean distance being higher than a preset distance threshold value, and entering the step B7 aiming at the rest groups of similar alternative data.
3. The application of the index-based rapid face retrieval system of claim 1, wherein: the extraction of the face feature data in the step A1 and the extraction of the face feature data in the step B1 are both realized according to the following operations;
firstly, sequentially positioning the positions of the human faces in the human face picture by at least two convolutional neural networks according to the sequence of the output accuracy of the convolutional neural networks from low to high; and then, extracting feature data from the obtained human face by adopting a residual error network.
4. The index-based rapid face retrieval system application of claim 1, wherein: in the step B5, the hamming code distance between the characteristic hash data in each group of data to be matched and the characteristic hash data to be matched is calculated to be used as the error between the characteristic hash data in each group of data to be matched and the characteristic hash data to be matched.
5. The application of the index-based rapid face retrieval system of claim 1, wherein: the cluster corresponding to the face feature data of the face picture to be stored obtained in the step A2 and the cluster corresponding to the face feature data of the face picture to be retrieved obtained in the step B2 are all realized according to the following operations:
firstly, carrying out hierarchical clustering on face feature data of a face picture based on preset clustering centers to obtain clusters corresponding to the face feature data of the face picture;
then, based on each pre-set sub-cluster center in the corresponding cluster, hierarchical clustering is carried out on the face feature data of the face picture, and sub-clusters corresponding to the face feature data of the face picture are further obtained;
and taking the cluster corresponding to the face characteristic data of the face picture and the sub-cluster thereof as the cluster corresponding to the face characteristic data of the face picture through secondary clustering.
6. The index-based rapid face retrieval system application of claim 1, wherein: in the step B2, after the clusters corresponding to the face feature data of the face picture to be retrieved are obtained, euclidean distances between the cluster center and the rest of cluster centers are calculated respectively, each cluster with the Euclidean distance lower than a preset distance threshold is selected, and the cluster corresponding to the face feature data of the face picture to be retrieved is combined to serve 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 CN109710792A (en) | 2019-05-03 |
| CN109710792B true 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) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112395448A (en) * | 2019-08-15 | 2021-02-23 | 华为技术有限公司 | Face retrieval method and device |
| 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 |
| CN112559789A (en) * | 2019-09-10 | 2021-03-26 | 中兴通讯股份有限公司 | Face searching method, device, equipment 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 |
| 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 |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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 |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP4604100B2 (en) * | 2008-03-21 | 2010-12-22 | シャープ株式会社 | Image processing method, image processing apparatus, image forming apparatus, program, and storage medium |
| CA2939161C (en) * | 2014-02-10 | 2017-05-09 | Geenee Gmbh | Systems and methods for image-feature-based recognition |
-
2018
- 2018-12-24 CN CN201811580645.1A patent/CN109710792B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9922052B1 (en) * | 2013-04-26 | 2018-03-20 | A9.Com, Inc. | Custom image data store |
| 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 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109710792A (en) | 2019-05-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109710792B (en) | Index-based rapid face retrieval system application | |
| US11048966B2 (en) | Method and device for comparing similarities of high dimensional features of images | |
| Wang et al. | Annosearch: Image auto-annotation by search | |
| CN106033426B (en) | Image retrieval method based on latent semantic minimum hash | |
| Zhou et al. | BSIFT: Toward data-independent codebook for large scale image search | |
| CN109166615B (en) | Medical CT image storage and retrieval method based on random forest hash | |
| CN103345496B (en) | multimedia information retrieval method and system | |
| Xie et al. | Fast and accurate near-duplicate image search with affinity propagation on the ImageWeb | |
| CN102254015A (en) | Image retrieval method based on visual phrases | |
| CN111177432B (en) | Large-scale image retrieval method based on hierarchical depth hash | |
| CN114461839B (en) | Multi-mode pre-training-based similar picture retrieval method and device and electronic equipment | |
| Adamu et al. | A survey on big data indexing strategies | |
| CN105760875B (en) | The similar implementation method of differentiation binary picture feature based on random forests algorithm | |
| Zhang et al. | Large‐scale video retrieval via deep local convolutional features | |
| CN112836008B (en) | Index establishing method based on decentralized storage data | |
| CN109918529A (en) | A kind of image search method based on the quantization of tree-like cluster vectors | |
| Liang et al. | An efficient hierarchical near-duplicate video detection algorithm based on deep semantic features | |
| Wan et al. | Cd-tree: A clustering-based dynamic indexing and retrieval approach | |
| JP2013246739A (en) | Search device and program for precisely searching content represented by set of high dimensional feature vectors | |
| CN107423294A (en) | A kind of community image search method and system | |
| Zhou et al. | Adaptive subspace symbolization for content-based video detection | |
| Deniziak et al. | The scalable distributed two-layer content based image retrieval data store | |
| CN114610960A (en) | Real-time recommendation method based on item2vec and vector clustering | |
| CN114943285A (en) | Intelligent auditing system for internet news content data | |
| Zhu et al. | Reverse Spatial Visual Top-$ k $ Query |
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 |