CN114139000A - Image retrieval system based on image global and local feature reordering - Google Patents
Image retrieval system based on image global and local feature reordering Download PDFInfo
- Publication number
- CN114139000A CN114139000A CN202111431412.7A CN202111431412A CN114139000A CN 114139000 A CN114139000 A CN 114139000A CN 202111431412 A CN202111431412 A CN 202111431412A CN 114139000 A CN114139000 A CN 114139000A
- Authority
- CN
- China
- Prior art keywords
- module
- unit
- global
- orb
- image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/53—Querying
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/51—Indexing; Data structures therefor; Storage structures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/55—Clustering; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/583—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Library & Information Science (AREA)
- Software Systems (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Processing Or Creating Images (AREA)
Abstract
The application discloses image retrieval system based on image global and local feature is reordered, including data storage unit, internet unit, off-line training unit, online inquiry unit and display element, off-line training unit and online inquiry unit all are connected with data storage unit, off-line training unit and online inquiry unit all are connected with the display element, off-line training unit and online inquiry unit connect, data storage unit, off-line training unit, online inquiry unit and display element all are connected with the internet unit. According to the method, the global features extracted by deep learning and the local features extracted by a traditional method are combined and applied to image retrieval, and then the Faiss vector is used for fast query, so that the same and similar images in the database can be retrieved and sequenced quickly and well.
Description
Technical Field
The application relates to the field of image retrieval, in particular to an image retrieval system based on image global and local feature reordering.
Background
At present, which is prosperous in the internet industry, a great amount of picture data is generated every day, many of the data are original and many of the data are repeated pictures, the repeated pictures may not only involve the copyright of original pictures, but also occupy a great amount of resources of internet companies, and the repeated pictures need to be retrieved and avoided and deleted.
The existing image retrieval methods are divided into two types, one type is that the features are extracted by using a traditional manual local feature extraction method and then feature similarity matching is carried out, the other type is that the features are extracted by using a deep learning model extraction method and then feature similarity matching is carried out, and the method is based on the traditional manual local feature extraction method: on one hand, the manually extracted features are difficult to adapt to different scenes; on the other hand, the manually extracted features only concern local parts, the current scenes are different, the backgrounds are the same or similar, and non-repeated pictures are often mistakenly detected as repetition. The method for extracting the image features based on the deep learning model usually uses the deep learning classification model to extract the features, the initial purpose is to classify the images according to categories, the images of different categories can be well classified, but the similarity of the features among the same categories is large, and the method is not convenient to be directly used for retrieval. Therefore, an image retrieval system based on image global and local feature reordering is proposed to solve the above problems.
Disclosure of Invention
The image retrieval system based on image global and local feature reordering is provided in the embodiment to solve the problem that the existing image retrieval method in the prior art has large similarity of features between the same types and is not convenient to be directly used as a retrieval problem.
According to one aspect of the application, an image retrieval system based on image global and local feature reordering is provided, and comprises a data storage unit, an internet unit, an offline training unit, an online query unit and a display unit, wherein the offline training unit and the online query unit are connected with the data storage unit, the offline training unit and the online query unit are connected with the display unit, the offline training unit and the online query unit are connected, and the data storage unit, the offline training unit, the online query unit and the display unit are connected with the internet unit;
the off-line training unit comprises an acquisition training picture set module, an off-line feature extraction module, a coarse screening model component module and a construction mapping dictionary table module, the off-line feature extraction module comprises a first orb feature extraction module and a first solar global feature extraction module, the on-line query unit comprises a picture acquisition module, an on-line feature extraction module, a loading coarse screening model module, a coarse arrangement retrieval module, a loading mapping dictionary table module, a fine arrangement model construction module and a fine arrangement retrieval module, and the on-line feature extraction module comprises a second orb feature extraction module and a second solar global feature extraction module.
Further, the orb features in the first orb and second orb feature extraction modules are all referred to as orientided Fast and Rotated Brief, and 160 orb feature vectors are selected per picture, the dimension of each orb feature vector being 32 dimensions.
Furthermore, the SOLAR in the first and Second SOLAR global feature extraction modules refers to a network model for Image Retrieval, which is proposed in SOLAR, Second-Order Loss and Attention for Image Retrieval, and the global feature refers to a global feature.
Further, the coarse screening model in the coarse screening model component module is a faiss vector coarse screening model, and the dictionary table in the mapping dictionary table building module is a key value pair, wherein a "key" is an index entry of the faiss in the faiss vector coarse screening model, and the "value" stores orb features and the address of the training set picture.
Further, the pictures in the picture acquisition module include a local picture and a network picture.
Further, the coarse screening model loaded by the coarse screening model loading module is a faiss vector coarse screening model constructed in the coarse screening model component module.
Further, the coarse ranking retrieval module sends the solar global features extracted from the second solar global feature extraction module into a faiss vector coarse screening model in the loading coarse screening model module for retrieval, and outputs a top100 sequence which is the most matched with the solar global features, wherein the top100 sequence is the sequence number of the images which are under the inquiry image co-vehicle department and the most similar images.
Further, the dictionary table loaded by the loading mapping dictionary table module is the dictionary table constructed in the construction mapping dictionary table module.
Further, the refined model building module firstly inputs a top100 sequence in the rough-arrangement retrieval module into a mapping dictionary table for query, and obtains orb features of a top100 matching picture and an address of the top100 matching picture, wherein the top100 matching picture orb features are used for building a faiss vector refined model, the address of the top100 matching picture is used for subsequent result output, and the faiss vector refined model is dynamically generated by orb 160.
Further, the fine-ranking search module sends orb features in the second orb feature extraction module into a faiss vector fine-ranking model in the fine-ranking model construction module for fine-ranking search, so as to obtain a fine-ranking top100 sequence, wherein the size of orb features in the query picture is 160 times 32, a total of 160 vectors are obtained, and each vector has 32 dimensions.
By adopting the embodiment of the application, the problems that the existing image retrieval method has larger similarity of characteristics among the same types and is inconvenient to directly use as retrieval are solved by adopting the off-line training unit, the on-line query unit and the like, the same and similar images in the database can be quickly and well retrieved and sequenced, and the retrieval of the images is facilitated.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without inventive exercise.
FIG. 1 is a general schematic view of an embodiment of the present application;
FIG. 2 is a schematic diagram of an offline training unit according to an embodiment of the present application;
fig. 3 is a schematic diagram of an online query unit according to an embodiment of the present application.
In the figure: 1. a data storage unit; 2. an Internet unit; 3. an offline training unit; 4. an online query unit; 5. a display unit; 6. acquiring a training picture set module; 7. an offline feature extraction module; 8. a coarse screen model component module; 9. constructing a mapping dictionary table module; 10. a first orb feature extraction module; 11. a first solar global feature extraction module; 12. a picture acquisition module; 13. an online feature extraction module; 14. loading a coarse screening model module; 15. a coarse sorting retrieval module; 16. loading a mapping dictionary table module; 17. a fine model building module; 18. a fine row search module; 19. a second orb feature extraction module; 20. and the second solar global feature extraction module.
Detailed Description
In order to make the technical solutions better understood by those skilled in the art, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only partial embodiments of the present application, but not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
It should be noted that the terms "first," "second," and the like in the description and claims of this application and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It should be understood that the data so used may be interchanged under appropriate circumstances such that embodiments of the application described herein may be used. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
In this application, the terms "upper", "lower", "left", "right", "front", "rear", "top", "bottom", "inner", "outer", "middle", "vertical", "horizontal", "lateral", "longitudinal", and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the drawings. These terms are used primarily to better describe the present application and its embodiments, and are not used to limit the indicated devices, elements or components to a particular orientation or to be constructed and operated in a particular orientation.
Moreover, some of the above terms may be used to indicate other meanings besides the orientation or positional relationship, for example, the term "on" may also be used to indicate some kind of attachment or connection relationship in some cases. The specific meaning of these terms in this application will be understood by those of ordinary skill in the art as appropriate.
Furthermore, the terms "mounted," "disposed," "provided," "connected," and "sleeved" are to be construed broadly. For example, it may be a fixed connection, a removable connection, or a unitary construction; can be a mechanical connection, or an electrical connection; may be directly connected, or indirectly connected through intervening media, or may be in internal communication between two devices, elements or components. The specific meaning of the above terms in the present application can be understood by those of ordinary skill in the art as appropriate.
It should be noted that the embodiments and features of the embodiments in the present application may be combined with each other without conflict. The present application will be described in detail below with reference to the embodiments with reference to the attached drawings.
Referring to fig. 1-3, the image retrieval system based on image global and local feature reordering comprises a data storage unit 1, an internet unit 2, an offline training unit 3, an online query unit 4 and a display unit 5, wherein the offline training unit 3 and the online query unit 4 are both connected with the data storage unit 1, the offline training unit 3 and the online query unit 4 are both connected with the display unit 5, the offline training unit 3 and the online query unit 4 are connected, the data storage unit 1, the offline training unit 3, the online query unit 4 and the display unit 5 are all connected with the internet unit 2, and image retrieval results and the like are displayed by the display unit 5;
the off-line training unit 3 comprises an acquisition training picture set module 6, an off-line feature extraction module 7, a coarse screening model component module 8 and a construction mapping dictionary table module 9, the off-line feature extraction module 7 comprises a first orb feature extraction module 10 and a first solar global feature extraction module 11, the on-line query unit 4 comprises a picture acquisition module 12, an on-line feature extraction module 13, a loading coarse screening model module 14, a coarse ranking search module 15, a loading mapping dictionary table module 16, a fine ranking model construction module 17 and a fine ranking search module 18, and the on-line feature extraction module 13 comprises a second orb feature extraction module 19 and a second solar global feature extraction module 20.
Orb features in the first orb and second orb feature extraction modules 10 and 19 are all called orientided Fast and Rotated Brief, and 160 orb feature vectors are selected for each picture, and the dimension of each orb feature vector is 32 dimensions.
The SOLAR in the first and Second SOLAR global feature extraction modules 11 and 20 refers to a network model for Image Retrieval, which is proposed in the text of SOLAR, Second-Order Loss and Attention for Image Retrieval, and the global feature refers to a global feature.
The coarse screening model in the coarse screening model component module 8 is a faiss vector coarse screening model, the dictionary table in the mapping dictionary table building module 9 is a key value pair, wherein a key is an index entry of the faiss in the faiss vector coarse screening model, and the value stores orb features and the address of a training set picture.
The pictures in the picture acquiring module 12 include a local picture and a network picture.
The coarse screening model loaded by the loading coarse screening model module 14 is a faiss vector coarse screening model constructed in the coarse screening model component module 8.
The coarse ranking search module 15 sends the solar global features extracted by the second solar global feature extraction module 20 to the faiss vector coarse screening model loaded in the coarse screening model module 14 for searching, and outputs the most matched top100 sequence, where the top100 sequence is the serial number of the searched pictures under the same train and the most similar pictures.
The dictionary table loaded by the loading mapping dictionary table module 16 is the dictionary table constructed in the mapping dictionary table constructing module 9.
The refined model building module 17 inputs the top100 sequence in the rough-ranking retrieval module 15 into a mapping dictionary table for query, and obtains orb features of a top100 matching picture and an address of the top100 matching picture, wherein the top100 matching picture orb feature is used for building a faiss vector refined model, the address of the top100 matching picture is used for subsequent result output, and the built faiss vector refined model is dynamically generated by orb 160.
The fine ranking search module 18 sends orb features in the second orb feature extraction module 19 to a faiss vector fine ranking model in the fine ranking model construction module 17 for fine ranking search, so as to obtain a fine ranking top100 sequence, where the size of orb features in the query picture is 160 times 32, and there are 160 vectors in total, and each vector has 32 dimensions.
When the method is used, the global depth feature of the image and the manual local feature of the image are combined to perform image retrieval; in the block of extracting the global features of the Image, a network model of Image Retrieval, which is proposed in SOLAR, namely, Second-ordered-Loss and Attention for Image Retrieval, is used, compared with the method of extracting the features by using a deep learning classification network, the features extracted by the network are more concerned about the difference in classes; in the manual extraction of local features, orb operator named Oriented Fast and Rotated Brief is adopted to extract local features, the operator can quickly extract key points in images and obtain feature vectors corresponding to each key point, the extracted feature vectors can be used for an open-source frame vector fine-ranking model for quickly calculating vector similarity developed by Faissfacebook company, the main technology is divided into two parts, one part is an offline training part, the other part is a new retrieval part, the offline training part comprises the specific steps of acquiring a training picture set by using a display unit 5, and at present prosperous in the Internet industry, massive picture data can be generated every day, the data are many original and repeated pictures, the repeated pictures not only can possibly occupy the copyright of the original picture, but also occupy a great amount of resources of the Internet company, the repeated pictures need to be retrieved and evaded and deleted; the method has the main scene that repeated vehicle pictures in the internet forum are retrieved; the vehicle pictures are respectively stored according to different vehicle series; when a user uploads a vehicle picture newly, the vehicle picture is uploaded to a picture library under a designated vehicle system, and at the moment, the picture library needs to be matched with all pictures under the vehicle system, and if the picture library is repeated, operation needs to be informed for confirmation and deletion; in the step, vehicle pictures in the database need to be sorted and named uniformly according to vehicle systems, so that the subsequent output is facilitated, and then orb features are extracted from the pictures in the training picture set by using a first orb feature extraction module 10; orb are collectively referred to as Oriented Fast and Rotated Brief; orb, the key points in the image can be extracted quickly, and in order to obtain the corresponding feature vector of each key point, the extracted feature vectors can be used for training a Faiss vector refinement model later; faiss is an open source framework developed by Facebook corporation for fast computing vector similarity; the method is used for constructing a coarse screening model and a fine discharging model; according to the method, OpenCV is utilized to extract orb features of all images in a training set according to a vehicle system, 160 orb feature vectors are selected for each picture, the dimensionality of each orb feature vector is 32-dimensional, and then a first solar global feature extraction module 11 is utilized to extract solar global features from the pictures in the training picture set; SOLAR refers to a network model for Image Retrieval proposed in SOLAR, Second-Order Loss and Attention for Image Retrieval; global features refer to global features; compared with the features extracted by using a deep learning classification model, the features extracted by the solar model have more distinctiveness in class, not only pay attention to the features of vehicles, but also pay attention to the differences among different vehicles, and then the solar global features extracted from the S1 offline training unit 3 are trained by using the coarse screening model component module 8 to train a faiss vector coarse screening model for later retrieval; the method comprises the following steps of training a coarse screening model based on faiss vectors, grasping the global characteristics of pictures by using solar global characteristics, and then constructing a mapping dictionary table by using a mapping dictionary table constructing module 9; the step of constructing a mapping dictionary table is important; the step is to connect the bridge with the coarse screen and the fine row; the dictionary table is a key-value pair, wherein a 'key' is an index entry of the faiss in the faiss vector coarse screening model, and the 'value' stores orb features and the address of the picture of the training set, and the address is used for outputting the address of the subsequent repeated picture.
An online retrieval part comprises the specific steps of utilizing a picture acquisition module 12 to acquire a query picture; the inquired pictures can be local pictures or pictures pointed by a network, orb characteristics of the inquired pictures are extracted by using a second orb characteristic extraction module 19, 160 orb characteristic vectors of the inquired pictures are extracted, solar global characteristics are extracted from the inquired pictures by using a second solar global characteristic extraction module 20, then a loaded coarse screen model module 14 is used for loading a constructed faiss vector coarse screen model in a coarse screen model component module 8, then a coarse sorting module 15 is used for sending solar global characteristics of the inquired pictures in the second solar global characteristic extraction module 20 into an S2 online inquiry unit 4 to carry out coarse screening on the faiss vector model for retrieval, the best matching top100 sequences are output, the top100 sequences are serial numbers of the inquired pictures in the same vehicle department and the most similar pictures, and then a mapping dictionary table constructed by a loading mapping dictionary table module 16 is loaded into an internal memory, the method is convenient for quick query, then a top100 sequence in a rough-ranking retrieval module 15 is firstly input into a mapping dictionary table for query by using a refined model building module 17, orb characteristics of a top100 matching picture and an address of the top100 matching picture are obtained, wherein orb characteristics of the top100 matching picture are used for building a faiss vector refined model, and the address of the top100 matching picture is used for outputting a subsequent result; the faiss vector fine ranking model is dynamically generated by orb160, local features are paid more attention to, and pertinence is achieved, and then orb features of the second orb feature extraction module 19 are sent to the faiss vector fine ranking model constructed in the fine ranking model construction module 17 by the fine ranking search module 18 to be subjected to fine ranking search, and a fine ranking top100 sequence is obtained; it is noted that the size of the features of the query picture orb is 160 by 32, and there are 160 vectors, each with 32 dimensions; the 160 vectors need to be sequentially sent to a faiss vector fine ranking model for retrieval, and finally, sequencing is performed according to the repetition times to obtain a sequence of top 100.
The application has the advantages that:
according to the method, the global features extracted by deep learning and the local features extracted by a traditional method are combined and applied to image retrieval, and then the Faiss vector is used for fast query, so that the same and similar images in the database can be retrieved and sequenced quickly and well.
It is well within the skill of those in the art to implement, without undue experimentation, the present application is not directed to software and process improvements, as they relate to circuits and electronic components and modules.
The above description is only a preferred embodiment of the present application and is not intended to limit the present application, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, improvement and the like made within the spirit and principle of the present application shall be included in the protection scope of the present application.
Claims (10)
1. The image retrieval system based on image global and local feature reordering is characterized in that: the online training system comprises a data storage unit (1), an internet unit (2), an offline training unit (3), an online query unit (4) and a display unit (5), wherein the offline training unit (3) and the online query unit (4) are connected with the data storage unit (1), the offline training unit (3) and the online query unit (4) are connected with the display unit (5), the offline training unit (3) is connected with the online query unit (4), and the data storage unit (1), the offline training unit (3), the online query unit (4) and the display unit (5) are connected with the internet unit (2);
offline training unit (3) is including obtaining training picture set module (6), offline feature extraction module (7), coarse screening model component module (8) and constructing mapping dictionary table module (9), offline feature extraction module (7) includes first orb feature extraction module (10) and first solar global feature extraction module (11), online query unit (4) includes picture acquisition module (12), online feature extraction module (13), load coarse screening model module (14), coarse row search module (15), load mapping dictionary table module (16), fine row model construction module (17) and fine row search module (18), online feature extraction module (13) includes second orb feature extraction module (19) and second solar global feature extraction module (20).
2. The image retrieval system based on image global and local feature reordering of claim 1, wherein: the orb features in the first orb feature extraction module (10) and the second orb feature extraction module (19) are all called orintered Fast and Rotated Brief, and 160 orb feature vectors are selected for each picture, and the dimension of each orb feature vector is 32 dimensions.
3. The image retrieval system based on image global and local feature reordering of claim 1, wherein: the SOLAR in the first SOLAR global feature extraction module (11) and the Second SOLAR global feature extraction module (20) refers to a network model for image Retrieval, which is proposed in the text of SOLAR: Second-Order Loss and Attention for image Retrieval, and the global features refer to global features.
4. The image retrieval system based on image global and local feature reordering of claim 1, wherein: the coarse screening model in the coarse screening model component module (8) is a faiss vector coarse screening model, the dictionary table in the mapping dictionary table building module (9) is a key value pair, wherein a key is an index item of the faiss in the faiss vector coarse screening model, and orb features and the address of a training set picture are stored in the key value.
5. The image retrieval system based on image global and local feature reordering of claim 1, wherein: the pictures in the picture acquisition module (12) comprise local pictures and network pictures.
6. The image retrieval system based on image global and local feature reordering of claim 1, wherein: the coarse screening model loaded by the coarse screening model loading module (14) is a faiss vector coarse screening model constructed in the coarse screening model component module (8).
7. The image retrieval system based on image global and local feature reordering of claim 1, wherein: the coarse sorting retrieval module (15) sends the solar global features extracted from the second solar global feature extraction module (20) into a faiss vector coarse screening model loaded in the coarse screening model module (14) for retrieval, and outputs a top100 sequence which is the most matched, wherein the top100 sequence is the serial number of the images which are inquired under the same train and are the most similar to the images.
8. The image retrieval system based on image global and local feature reordering of claim 1, wherein: the dictionary table loaded by the loading mapping dictionary table module (16) is a dictionary table constructed in the construction mapping dictionary table module (9).
9. The image retrieval system based on image global and local feature reordering of claim 1, wherein: the refined model building module (17) firstly inputs the top100 sequence in the rough arrangement retrieval module (15) into a mapping dictionary table for query, and obtains orb characteristics of a top100 matching picture and an address of the top100 matching picture, wherein the top100 matching picture orb characteristics are used for building a faiss vector refined model, the address of the top100 matching picture is used for outputting a subsequent result, and the faiss vector refined model is dynamically generated by orb 160.
10. The image retrieval system based on image global and local feature reordering of claim 1, wherein: the fine ranking search module (18) sends orb features in the second orb feature extraction module (19) to a faiss vector fine ranking model in the fine ranking model construction module (17) for fine ranking search, so as to obtain a fine ranking top100 sequence, wherein the size of orb features in the query picture is 160 times 32, the total number of 160 vectors is 160, and each vector has 32 dimensions.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111431412.7A CN114139000A (en) | 2021-11-29 | 2021-11-29 | Image retrieval system based on image global and local feature reordering |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111431412.7A CN114139000A (en) | 2021-11-29 | 2021-11-29 | Image retrieval system based on image global and local feature reordering |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114139000A true CN114139000A (en) | 2022-03-04 |
Family
ID=80389137
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111431412.7A Pending CN114139000A (en) | 2021-11-29 | 2021-11-29 | Image retrieval system based on image global and local feature reordering |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114139000A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116910296A (en) * | 2023-09-08 | 2023-10-20 | 上海任意门科技有限公司 | Method, system, electronic device and medium for identifying transport content |
-
2021
- 2021-11-29 CN CN202111431412.7A patent/CN114139000A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116910296A (en) * | 2023-09-08 | 2023-10-20 | 上海任意门科技有限公司 | Method, system, electronic device and medium for identifying transport content |
| CN116910296B (en) * | 2023-09-08 | 2023-12-08 | 上海任意门科技有限公司 | Method, system, electronic device and medium for identifying transport content |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP7058669B2 (en) | Vehicle appearance feature identification and vehicle search methods, devices, storage media, electronic devices | |
| CN110210542B (en) | Picture character recognition model training method and device and character recognition system | |
| US9710695B2 (en) | Characterizing pathology images with statistical analysis of local neural network responses | |
| CN111125435B (en) | Video tag determination method and device and computer equipment | |
| CN109325148A (en) | The method and apparatus for generating information | |
| CN111046275B (en) | User label determining method and device based on artificial intelligence and storage medium | |
| US20170124618A1 (en) | Methods and Systems for Image-Based Searching of Product Inventory | |
| CN106708940A (en) | Method and device used for processing pictures | |
| CN110598019B (en) | Repeated image identification method and device | |
| CN109377494B (en) | Semantic segmentation method and device for image | |
| CN111242119B (en) | Vehicle image processing method and device, electronic equipment and computer readable medium | |
| CN112487242A (en) | Method and device for identifying video, electronic equipment and readable storage medium | |
| CN107133854A (en) | Information recommendation method and device | |
| CN109145139A (en) | A kind of image search method and device, equipment and storage medium | |
| CN108734159A (en) | The detection method and system of sensitive information in a kind of image | |
| CN112183078A (en) | Text abstract determining method and device | |
| CN114139000A (en) | Image retrieval system based on image global and local feature reordering | |
| CN114741581A (en) | Image classification method and device, computer equipment and medium | |
| CN109492120A (en) | Model training method, search method, device, electronic equipment and storage medium | |
| CN107291774A (en) | Error sample recognition methods and device | |
| CN109086737B (en) | Convolutional neural network-based shipping cargo monitoring video identification method and system | |
| CN113989476A (en) | Object identification method and electronic equipment | |
| CN110188277B (en) | Resource recommendation method and device | |
| CN112348107A (en) | Image data cleaning method and apparatus, electronic device, and medium | |
| CN115687676B (en) | Information retrieval method, terminal and computer-readable storage medium |
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 |