CN112528066B - Trademark retrieval method, system, computer device and storage medium based on attention mechanism - Google Patents
Trademark retrieval method, system, computer device and storage medium based on attention mechanism Download PDFInfo
- Publication number
- CN112528066B CN112528066B CN202011511790.1A CN202011511790A CN112528066B CN 112528066 B CN112528066 B CN 112528066B CN 202011511790 A CN202011511790 A CN 202011511790A CN 112528066 B CN112528066 B CN 112528066B
- Authority
- CN
- China
- Prior art keywords
- trademark
- module
- image
- characteristic
- representing
- 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
- 238000000034 method Methods 0.000 title claims abstract description 35
- 238000012549 training Methods 0.000 claims abstract description 29
- 238000000605 extraction Methods 0.000 claims abstract description 19
- 238000007781 pre-processing Methods 0.000 claims description 18
- 230000006870 function Effects 0.000 claims description 12
- 238000004590 computer program Methods 0.000 claims description 10
- 238000005259 measurement Methods 0.000 claims description 9
- 238000010606 normalization Methods 0.000 claims description 6
- 238000012163 sequencing technique Methods 0.000 claims description 5
- 238000012545 processing Methods 0.000 claims description 3
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
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
- G06F16/58—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
-
- 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
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Abstract
The invention provides a trademark retrieval method, a system, computer equipment and a storage medium based on an attention mechanism, wherein the method is characterized in that trademark images in a trademark data set are preprocessed, more reasonable weight distribution is carried out on the characteristics of the initial trademark images through a channel attention network module, the difference of the characteristics of the initial trademark images is learned through an example distinguishing module, loss of the characteristics of the initial trademark images is optimized, an obtained training data set is utilized for training an unsupervised network, and a characteristic extraction network is formed, wherein the characteristic extraction network can accurately extract key characteristics of the trademark images.
Description
Technical Field
The present invention relates to the field of image retrieval, and more particularly, to a brand retrieval method, system, computer device, and storage medium based on an attention mechanism.
Background
Trademarks are becoming more and more important for various industries as important marks of goods and factories. The core technology in trademark retrieval is extraction and measurement of trademark characteristics. Whether the trademark characteristics are extracted accurately or not directly influences the subsequent retrieval result. In the past, trademark management departments have performed search work in the form of "classification numbers" and manually judge the similarity between images, and the method has the problems of large workload and low efficiency. In order to solve the problem of searching work, researchers at home and abroad begin to capture more accurate trademark characteristic information through a content-based image searching mode, and deviation caused by text description is avoided.
In the conventional image trademark retrieval method, people prefer to perform feature extraction through shallow visual features of an image, such as a shape context method, extraction of SIFT features, a trademark retrieval algorithm based on zernike moment and edge gradient symbiotic moment proposed from the viewpoint of a shape descriptor of a trademark, and the like. The content of the trademark image does not have the same importance, and information elements which are more important than surrounding features exist, so that the conventional method cannot judge the importance of the information differently, and the capability of extracting key features is insufficient.
Chinese patent CN111797260a published on 10 months and 20 days in 2020 provides a trademark retrieval system based on image recognition, which comprises a database, the database signal is connected with a classification unit, the classification unit signal is connected with a plurality of data sets, the data set signal is connected with an algorithm module, the algorithm module signal is connected with a training module, the training module signal is connected with a storage module, the storage module signal is connected with a plurality of initial centroid sets, the initial centroid set signal is connected with a comparison unit and a search module, the comparison unit signal is connected with an output module, the search module signal is connected with a feature recognition unit, the feature recognition unit signal is connected with a temporary storage module, and the temporary storage module signal is connected with an input module. The invention has the characteristics of independently comparing the characteristics of the trademark and being favorable for the retrieval effect, but has the condition of inaccurate characteristics of the extracted trademark.
Disclosure of Invention
The invention provides a trademark retrieval method, a system, computer equipment and a storage medium based on an attention mechanism, which are used for overcoming the defect that key features of trademark images extracted in the prior art are inaccurate.
The technical scheme of the invention is as follows:
the invention provides a trademark retrieval method based on an attention mechanism, which comprises the following steps:
s1: acquiring a trademark data set;
s2: constructing a channel attention network module;
s3: the preprocessing module preprocesses the trademark image in the trademark data set to obtain a preprocessed trademark image x i I=1, 2, …, n, n is the number of trademark images after preprocessing, and inputs the trademark images to the channel attention network module;
s4: channel attention networkThe module weights omega according to the characteristic channel domain i For pretreated trademark image x i Training and distributing to obtain preprocessed trademark image x i Corresponding initial trademark characteristics v i I=1, 2, … n, will be the original trademark characteristic v i An initial trademark characteristic set V is formed, and the initial trademark characteristic set V is input into an instance distinguishing module;
s5: the instance distinguishing module is used for distinguishing the initial trademark characteristics V in the initial trademark characteristic set V i Calculating loss and optimizing the loss, and continuously and iteratively updating a trademark characteristic set V to obtain trademark characteristic training data to form a training data set, and training an unsupervised network by using the training data set to obtain a characteristic extraction network;
s6: inputting all trademark images in a trademark data set into a feature extraction network to obtain a trademark image feature database, and storing the trademark image feature database into a retrieval module;
s7: inputting the trademark image to be searched into a feature extraction network to obtain the trademark image feature to be searched, and conveying the trademark image feature to be searched to a searching module;
s8: and the retrieval module carries out similarity measurement on the trademark image features to be retrieved and the trademark features in the trademark image feature database, and outputs a trademark image sequencing result similar to the trademark images to be retrieved.
Preferably, in the step S2, the method for building the channel attention network module is as follows: resNet50 is selected as a base network, and a channel attention module is inserted between all convolution layers of the ResNet50 to form a channel attention network module.
Preferably, in the step S3, the specific method for preprocessing the trademark image in the dataset by the preprocessing module is: converting the trademark image into a gray level image after cutting, converting the gray level image into tensors, and carrying out normalization processing on the tensors.
Preferably, in S3, the pretreated trademark image x i The input channel is input in time-division batch by the attention network module.
Preferably, in the step S4, the characteristic channel domain weight ω i Calculated from the following formula:
where σ () is a Sigmoid function, k is the convolution kernel size of the channel attention module,is a tilt parameter.
Preferably, in the step S5, a trademark characteristic v is calculated i The loss J of (2) is calculated from the following loss function:
wherein ,representing a trademark image x i Is (are) desirable to be (are)>Representing the desire of noise data, P n Representing noise data, v i Representing a trademark image x i Corresponding trademark characteristics in trademark characteristic set V, V j Representing a trademark characteristic set V different from V i M represents that the noise data is m times the trademark image, < >>Representing feature channel domain weights ω i Is the transpose of τ to the temperature parameter, Z i To normalize constants, exp () is expressed in eAn exponential function of the base.
Preferably, in the step S8, the specific method for measuring the similarity is as follows: the searching module calculates Euclidean distance between the trademark image feature to be searched and the trademark feature in the trademark image feature database, and the smaller the Euclidean distance is, the more similar the corresponding trademark image is to the trademark image to be searched.
The invention also provides a trademark retrieval system based on the attention mechanism, which comprises a preprocessing module, a channel attention network module, an instance distinguishing module and a retrieval module;
the preprocessing module preprocesses trademark images in the trademark data set to obtain preprocessed trademark images;
the channel attention network module trains and distributes the preprocessed trademark images to obtain an initial trademark characteristic set;
the example distinguishing module calculates loss of the initial trademark characteristics in the initial trademark characteristic set and optimizes the loss of the initial trademark characteristics to obtain trademark characteristic training data to form a training data set, and the training data set is utilized to train the unsupervised network to obtain a characteristic extraction network;
and the retrieval module carries out similarity measurement on the trademark image features to be retrieved and the trademark features in the trademark image feature database, and outputs a trademark image sequencing result similar to the trademark images to be retrieved.
The invention also provides a computer device, which comprises a memory and a processor, wherein the memory stores a computer program, and the processor realizes the steps of the trademark retrieval method based on the attention mechanism when executing the computer program.
The present invention also proposes a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the above-described attention-mechanism-based brand search method.
Compared with the prior art, the technical scheme of the invention has the beneficial effects that:
according to the invention, more reasonable weight distribution is carried out on trademark image features through the channel attention network module, the difference of the trademark image features is learned through the example distinguishing module, loss of the trademark image features is optimized, and the obtained training data set is utilized to train the unsupervised network, so that a feature extraction network is formed, and key features of the trademark images can be accurately extracted by the feature extraction network.
Drawings
FIG. 1 is a flow chart of a trademark retrieval method based on attention mechanism according to embodiment 1;
FIG. 2 is a schematic diagram showing the effect of a trademark retrieval method based on attention mechanism according to embodiment 1;
fig. 3 is a schematic block diagram of a trademark retrieval system based on an attention mechanism in embodiment 2.
Detailed Description
The drawings are for illustrative purposes only and are not to be construed as limiting the present patent;
for the purpose of better illustrating the embodiments, certain elements of the drawings may be omitted, enlarged or reduced and do not represent the actual product dimensions;
it will be appreciated by those skilled in the art that certain well-known structures in the drawings and descriptions thereof may be omitted.
The technical scheme of the invention is further described below with reference to the accompanying drawings and examples.
Example 1
The embodiment provides a trademark retrieval method based on an attention mechanism, as shown in fig. 1, the method comprises the following steps:
s1: acquiring a trademark data set; the trademark dataset is a METU trademark dataset.
S2: constructing a channel attention network module;
s3: the preprocessing module preprocesses the trademark image in the trademark data set to obtain a preprocessed trademark image x i I=1, 2, …, n, n is the number of trademark images after preprocessing, and inputs the trademark images to the channel attention network module;
s4: the channel attention network module weights omega according to the characteristic channel domain i For pretreated trademark image x i Training and distributing to obtain pretreatmentPost trademark image x i Corresponding initial trademark characteristics v i I=1, 2, … n, will be the original trademark characteristic v i An initial trademark characteristic set V is formed, and the initial trademark characteristic set V is input into an instance distinguishing module;
s5: the instance distinguishing module is used for distinguishing the initial trademark characteristics V in the initial trademark characteristic set V i Calculating loss and optimizing the loss, and continuously and iteratively updating a trademark characteristic set V to obtain trademark characteristic training data to form a training data set, and training an unsupervised network by using the training data set to obtain a characteristic extraction network; the unsupervised network is an ECA_ResNet50 network;
s6: inputting all trademark images in a trademark data set into a feature extraction network to obtain a trademark image feature database, and storing the trademark image feature database into a retrieval module;
s7: inputting the trademark image to be searched into a feature extraction network to obtain the trademark image feature to be searched, and conveying the trademark image feature to be searched to a searching module;
s8: the retrieval module performs similarity measurement on the trademark image features to be retrieved and the trademark features in the trademark image feature database, and outputs a trademark image ordering result similar to the trademark images to be retrieved as shown in fig. 2.
In the step S2, the method for constructing the channel attention network module is as follows: resNet50 is selected as a base network, and a channel attention module is inserted between all convolution layers of the ResNet50 to form a channel attention network module.
In the step S3, the specific method for preprocessing the trademark image in the dataset by the preprocessing module is as follows: converting the trademark image into a gray level image after cutting, converting the gray level image into tensors, and carrying out normalization processing on the tensors.
In S3, the pretreated trademark image x i The input channel is input in time-division batch by the attention network module.
In the S4, the characteristic channel domain weight omega i Calculated from the following formula:
where σ () is a Sigmoid function, k is the convolution kernel size of the channel attention module,is a tilt parameter.
In the S5, trademark characteristic v is calculated i The loss J of (2) is calculated from the following loss function:
wherein ,representing a trademark image x i Is (are) desirable to be (are)>Representing the desire of noise data, P n Representing noise data, v i Representing a trademark image x i Corresponding trademark characteristics in trademark characteristic set V, V j Representing a trademark characteristic set V different from V i M represents that the noise data is m times the trademark image, < >>Representing feature channel domain weights ω i Is the transpose of τ to the temperature parameter, Z i For the normalization constant exp () represents an exponential function based on e.
In the step S8, the specific method for similarity measurement is as follows: the searching module calculates Euclidean distance between the trademark image feature to be searched and the trademark feature in the trademark image feature database, and the smaller the Euclidean distance is, the more similar the corresponding trademark image is to the trademark image to be searched.
The embodiment also provides a computer device, which comprises a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to realize the steps of the trademark retrieval method based on the attention mechanism.
The present embodiment also proposes a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the above-described attention-mechanism-based trademark retrieval method.
Example 2
The embodiment provides a trademark retrieval system based on an attention mechanism, as shown in fig. 3, wherein the system comprises a preprocessing module, a channel attention network module, an instance distinguishing module and a retrieval module;
the preprocessing module preprocesses trademark images in the trademark data set to obtain preprocessed trademark images;
the channel attention network module trains and distributes the preprocessed trademark images to obtain an initial trademark characteristic set;
the example distinguishing module calculates loss of the initial trademark characteristics in the initial trademark characteristic set and optimizes the loss of the initial trademark characteristics to obtain trademark characteristic training data to form a training data set, and the training data set is utilized to train the unsupervised network to obtain a characteristic extraction network;
and the retrieval module carries out similarity measurement on the trademark image features to be retrieved and the trademark features in the trademark image feature database, and outputs a trademark image sequencing result similar to the trademark images to be retrieved.
It is to be understood that the above examples of the present invention are provided by way of illustration only and not by way of limitation of the embodiments of the present invention. Other variations or modifications of the above teachings will be apparent to those of ordinary skill in the art. It is not necessary here nor is it exhaustive of all embodiments. Any modification, equivalent replacement, improvement, etc. which come within the spirit and principles of the invention are desired to be protected by the following claims.
Claims (8)
1. A brand search method based on an attention mechanism, the method comprising the steps of:
s1: acquiring a trademark data set;
s2: constructing a channel attention network module;
s3: the preprocessing module preprocesses the trademark image in the trademark data set to obtain a preprocessed trademark image x i I=1, 2, …, n, n is the number of trademark images after preprocessing, and inputs the trademark images to the channel attention network module;
s4: the channel attention network module weights omega according to the characteristic channel domain i For pretreated trademark image x i Training and distributing to obtain preprocessed trademark image x i Corresponding initial trademark characteristics v i I=1, 2, … n, will be the original trademark characteristic v i An initial trademark characteristic set V is formed, and the initial trademark characteristic set V is input into an instance distinguishing module;
feature channel domain weight omega i Calculated from the following formula:
where σ () is a Sigmoid function, k is the convolution kernel size of the channel attention module,is a tilt parameter;
s5: the instance distinguishing module is used for distinguishing the initial trademark characteristics V in the initial trademark characteristic set V i Calculating loss and optimizing the loss to obtain trademark characteristic training data to form a training data set, and training an unsupervised network by using the training data set to obtain a characteristic extraction network;
calculating trademark characteristics v i The loss J of (2) is calculated from the following loss function:
wherein ,representing a trademark image x i Is (are) desirable to be (are)>Representing the desire of noise data, P n Representing noise data, v i Representing a trademark image x i Corresponding trademark characteristics in trademark characteristic set V, V j Representing a trademark characteristic set V different from V i M represents that the noise data is m times the trademark image, < >>Representing feature channel domain weights ω i Is the transpose of τ to the temperature parameter, Z i For the normalization constant, exp () represents an exponential function based on e;
s6: inputting all trademark images in a trademark data set into a feature extraction network to obtain a trademark image feature database, and storing the trademark image feature database into a retrieval module;
s7: inputting the trademark image to be searched into a feature extraction network to obtain the trademark image feature to be searched, and conveying the trademark image feature to be searched to a searching module;
s8: and the retrieval module carries out similarity measurement on the trademark image features to be retrieved and the trademark features in the trademark image feature database, and outputs a trademark image sequencing result similar to the trademark images to be retrieved.
2. The trademark retrieval method based on attention mechanism of claim 1, wherein in S2, the method for constructing the channel attention network module is as follows: resNet50 is selected as a base network, and a channel attention module is inserted between all convolution layers of the ResNet50 to form a channel attention network module.
3. The trademark retrieval method based on the attention mechanism of claim 2, wherein in S3, the specific method of preprocessing the trademark image in the dataset by the preprocessing module is as follows: converting the trademark image into a gray level image after cutting, converting the gray level image into tensors, and carrying out normalization processing on the tensors.
4. A trademark retrieving method based on attention mechanism as set forth in claim 3, wherein in S3, the preprocessed trademark image x i The input channel is input in time-division batch by the attention network module.
5. The trademark retrieving method based on attention mechanism of claim 4, wherein in S8, the specific method of similarity measurement is as follows: the searching module calculates Euclidean distance between the trademark image feature to be searched and the trademark feature in the trademark image feature database, and the smaller the Euclidean distance is, the more similar the corresponding trademark image is to the trademark image to be searched.
6. A trademark retrieval system based on an attention mechanism, which is characterized by comprising a preprocessing module, a channel attention network module, an instance distinguishing module and a retrieval module;
the preprocessing module preprocesses trademark images in the trademark data set to obtain preprocessed trademark images;
the channel attention network module trains and distributes the preprocessed trademark images according to the characteristic channel domain weight to obtain an initial trademark characteristic set; feature channel domain weight omega i Calculated from the following formula:
where σ () is a Sigmoid function, k is the convolution kernel size of the channel attention module,is a tilt parameter;
the example distinguishing module calculates loss of the initial trademark characteristics in the initial trademark characteristic set and optimizes the loss to obtain trademark characteristic training data, a training data set is formed, an unsupervised network is trained by using the training data set, and a characteristic extraction network is obtained; calculating trademark characteristics v i The loss J of (2) is calculated from the following loss function:
wherein ,representing a trademark image x i Is (are) desirable to be (are)>Representing the desire of noise data, P n Representing noise data, v i Representing a trademark image x i Corresponding trademark characteristics in trademark characteristic set V, V j Representing a trademark characteristic set V different from V i M represents that the noise data is m times the trademark image, < >>Representing feature channel domain weights ω i Is the transpose of τ to the temperature parameter, Z i For the normalization constant, exp () represents an exponential function based on e;
and the retrieval module carries out similarity measurement on the trademark image features to be retrieved and the trademark features in the trademark image feature database, and outputs a trademark image sequencing result similar to the trademark images to be retrieved.
7. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the steps of the method of any one of claims 1 to 5 when the computer program is executed.
8. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 5.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011511790.1A CN112528066B (en) | 2020-12-18 | 2020-12-18 | Trademark retrieval method, system, computer device and storage medium based on attention mechanism |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011511790.1A CN112528066B (en) | 2020-12-18 | 2020-12-18 | Trademark retrieval method, system, computer device and storage medium based on attention mechanism |
Publications (2)
Publication Number | Publication Date |
---|---|
CN112528066A CN112528066A (en) | 2021-03-19 |
CN112528066B true CN112528066B (en) | 2023-08-04 |
Family
ID=75001771
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011511790.1A Active CN112528066B (en) | 2020-12-18 | 2020-12-18 | Trademark retrieval method, system, computer device and storage medium based on attention mechanism |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112528066B (en) |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108681555A (en) * | 2018-04-08 | 2018-10-19 | 天津大学 | A kind of sketch image search method returned based on shape |
CN108920720A (en) * | 2018-07-30 | 2018-11-30 | 电子科技大学 | The large-scale image search method accelerated based on depth Hash and GPU |
CN109086437A (en) * | 2018-08-15 | 2018-12-25 | 重庆大学 | A kind of image search method merging Faster-RCNN and Wasserstein self-encoding encoder |
CN110059206A (en) * | 2019-03-29 | 2019-07-26 | 银江股份有限公司 | A kind of extensive hashing image search method based on depth representative learning |
-
2020
- 2020-12-18 CN CN202011511790.1A patent/CN112528066B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108681555A (en) * | 2018-04-08 | 2018-10-19 | 天津大学 | A kind of sketch image search method returned based on shape |
CN108920720A (en) * | 2018-07-30 | 2018-11-30 | 电子科技大学 | The large-scale image search method accelerated based on depth Hash and GPU |
CN109086437A (en) * | 2018-08-15 | 2018-12-25 | 重庆大学 | A kind of image search method merging Faster-RCNN and Wasserstein self-encoding encoder |
CN110059206A (en) * | 2019-03-29 | 2019-07-26 | 银江股份有限公司 | A kind of extensive hashing image search method based on depth representative learning |
Also Published As
Publication number | Publication date |
---|---|
CN112528066A (en) | 2021-03-19 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110162593B (en) | Search result processing and similarity model training method and device | |
EP2565804B1 (en) | Text-based searching of image data | |
US9218364B1 (en) | Monitoring an any-image labeling engine | |
US9037600B1 (en) | Any-image labeling engine | |
CN111104466A (en) | Method for rapidly classifying massive database tables | |
CN110688474B (en) | Embedded representation obtaining and citation recommending method based on deep learning and link prediction | |
CN107679070B (en) | Intelligent reading recommendation method and device and electronic equipment | |
CN112559684A (en) | Keyword extraction and information retrieval method | |
CN112949740B (en) | Small sample image classification method based on multilevel measurement | |
CN109213886B (en) | Image retrieval method and system based on image segmentation and fuzzy pattern recognition | |
CN114780746A (en) | Knowledge graph-based document retrieval method and related equipment thereof | |
Sadique et al. | Content-based image retrieval using color layout descriptor, gray-level co-occurrence matrix and k-nearest neighbors | |
CN112528066B (en) | Trademark retrieval method, system, computer device and storage medium based on attention mechanism | |
CN116128846B (en) | Visual transducer hash method for lung X-ray image retrieval | |
CN114077682B (en) | Intelligent recognition matching processing method and system for image retrieval and storage medium | |
CN116363460A (en) | High-resolution remote sensing sample labeling method based on topic model | |
CN115797795A (en) | Remote sensing image question-answering type retrieval system and method based on reinforcement learning | |
CN107423294A (en) | A kind of community image search method and system | |
CN116258938A (en) | Image retrieval and identification method based on autonomous evolution loss | |
CN111460206B (en) | Image processing method, apparatus, electronic device, and computer-readable storage medium | |
CN113806472B (en) | Method and equipment for realizing full-text retrieval of text picture and image type scanning piece | |
CN113887422A (en) | Table picture content extraction method, device and equipment based on artificial intelligence | |
CN114511027A (en) | Method for extracting English remote data through big data network | |
Dey et al. | Evaluation of word spotting under improper segmentation scenario | |
CN117788850B (en) | Trademark similarity evaluation method and device |
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 |