CN117934891A - Image contrast clustering method and system based on graph structure - Google Patents
Image contrast clustering method and system based on graph structure Download PDFInfo
- Publication number
- CN117934891A CN117934891A CN202410343857.7A CN202410343857A CN117934891A CN 117934891 A CN117934891 A CN 117934891A CN 202410343857 A CN202410343857 A CN 202410343857A CN 117934891 A CN117934891 A CN 117934891A
- Authority
- CN
- China
- Prior art keywords
- contrast
- original image
- network
- graph
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 45
- 238000013528 artificial neural network Methods 0.000 claims abstract description 28
- 230000009466 transformation Effects 0.000 claims abstract description 26
- 238000000605 extraction Methods 0.000 claims abstract description 21
- 239000011159 matrix material Substances 0.000 claims abstract description 18
- 238000000547 structure data Methods 0.000 claims abstract description 16
- 230000006870 function Effects 0.000 claims description 39
- 238000005065 mining Methods 0.000 claims description 9
- 238000010276 construction Methods 0.000 claims description 7
- 230000007246 mechanism Effects 0.000 claims description 7
- 238000003860 storage Methods 0.000 claims description 5
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 3
- 230000000052 comparative effect Effects 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 238000005096 rolling process Methods 0.000 claims 1
- 238000000844 transformation Methods 0.000 abstract description 4
- 238000009826 distribution Methods 0.000 abstract description 2
- 238000010801 machine learning Methods 0.000 abstract description 2
- 238000004220 aggregation Methods 0.000 abstract 1
- 230000002776 aggregation Effects 0.000 abstract 1
- 238000004590 computer program Methods 0.000 description 7
- 238000010586 diagram Methods 0.000 description 6
- 230000008569 process Effects 0.000 description 4
- 230000009471 action Effects 0.000 description 3
- 238000004422 calculation algorithm Methods 0.000 description 3
- 230000004931 aggregating effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000002059 diagnostic imaging Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000010191 image analysis Methods 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000001131 transforming effect Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/762—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using clustering, e.g. of similar faces in social 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
- G06N3/042—Knowledge-based neural networks; Logical representations of neural 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
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Biophysics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Databases & Information Systems (AREA)
- Medical Informatics (AREA)
- Image Analysis (AREA)
Abstract
The invention provides an image contrast clustering method and system based on a graph structure, and relates to the field of machine learning and image processing. The method comprises the steps of obtaining an original image dataset, carrying out two different transformations on the original image dataset, obtaining characteristics and contrast characteristics of the transformed dataset, calculating instance-level contrast loss, obtaining and updating neighbors of each sample, obtaining an adjacent matrix, adding different random masks, constructing graph structure data, inputting graph neural network aggregation characteristics and neighbor characteristics thereof, obtaining cluster distribution, calculating cluster-level contrast loss, optimizing a characteristic extraction network, a contrast network and a graph convolution neural network through the instance-level contrast loss and the cluster-level contrast loss, repeating the steps until the iteration times are met, and outputting a cluster grouping result; the method of the invention not only can acquire the semantic information of the image, but also can effectively mine the structural information among the image data.
Description
Technical Field
The invention relates to the technical field of machine learning and image processing, in particular to an image contrast clustering method and system based on a graph structure.
Background
Image contrast clustering is a technology combining contrast learning and clustering algorithms, and by means of the contrast learning method, similar and different modes are identified in unlabeled image data sets without relying on manual labeling. And clustering the images into different clusters by using a clustering algorithm, so that the images in the same cluster are similar to a certain extent, and the images in different clusters have a certain difference. The method is widely applied to a plurality of fields such as medical imaging, satellite image analysis, social media content classification and the like.
The image data, such as attribute graph, contains the structure information of edges between adjacent nodes besides the characteristic information of the nodes, and the image data set usually only contains the characteristic information of images, so that most image clustering algorithms only consider the characteristic information in the images and cannot effectively fuse the structure information between the images. Furthermore, the contrast learning method treats different transformations of the same image as positive pairs of samples, while the remaining images as negative pairs of samples, even samples of the same class, resulting in a bias in clustering.
Disclosure of Invention
(One) solving the technical problems
Aiming at the defects of the prior art, the invention provides an image contrast clustering method and system based on a graph structure, and aims to solve the problems in the background art.
(II) technical scheme
In order to achieve the above purpose, the invention is realized by the following technical scheme:
In a first aspect, an image contrast clustering method based on a graph structure is provided, which is characterized by comprising the following steps:
Acquiring an original image dataset And two kinds of transformation are carried out on the original image data set to respectively obtain two transformed data sets/>; Wherein the two transforms include, weak transform/>And strong transformation/>，/>，/>Is a set of transforms;
To collect original image data And transformed dataset/>Input to a feature extraction network to obtain features/>, of the original image dataset and the two transformed datasets, respectively；
Characterizing an original image dataset with features of two transformed datasetsInput to a comparison network to obtain the comparison features/>, of the original image dataset and the two transformed datasets respectively；
Will contrast characteristicsBy loss function/>Calculating an instance contrast loss;
In contrast to features Upper mining of neighboring images, acquiring and updating neighbor/>, per sample；
Utilizing neighborsConstruction of adjacency matrix/>And adding different random masks to obtain two adjacent matrixes;
By means of featuresAnd adjacency matrix A,/>Constructing graph structure data；
Map structure dataInputting the cluster characteristics and the neighbor characteristics of the cluster characteristics into a graph neural network, and obtaining cluster allocation/>；
Using cluster allocationBy loss function/>Calculating cluster level contrast loss;
Using loss functions ,/>Calculating a final loss function L, and jointly optimizing a feature extraction network, a comparison network and a graph convolution neural network through a back propagation mechanism of the deep neural network;
repeating the steps until the iteration times or network convergence is met, and obtaining and outputting a final clustering grouping result 。
Preferably, the original image is used for neighbor mining and obtaining the final clustering result.
Preferably, the weak transformationFor random clipping and scaling, random horizontal flipping, random brightness, contrast, saturation, tone transformation, and random gray-scale transformation.
Preferably, the strong transformationIn addition to including random horizontal flipping, random clipping, and randomly selecting four components from normalized contrast, balanced histogram, rotation, inversion, enhanced color balance, adjusted contrast, adjusted brightness, adjusted sharpness, affine transformation, reduced number of bits per color channel.
Preferably, the adjacency matrixTo add the bipartite graph of the self-loop:
wherein, And/>Index representing contrast features of different samples, respectively,/>Representing acquisition neighbors, the random mask randomly deletes/>, with 10% probabilityEdges between adjacent nodes.
Preferably, the graph neural network is composed of a graph convolution network and a full-connection network, the output dimension is K, and K represents the category number contained in the data.
Preferably, the loss function L is defined as:
wherein, Representing example contrast loss,/>Representing cluster level contrast loss,/>Representing the balance coefficient.
In a second aspect, an image contrast clustering system based on a graph structure is provided, which is characterized by comprising:
An acquisition module for acquiring an original image dataset And two kinds of transformation are carried out on the original image data set to respectively obtain two transformed data sets/>; Wherein the two transforms include weak transforms/>And strong transformation/>，/>，/>Is a set of transforms;
A feature extraction module for extracting original image data set And two transformed datasets/>Input to a feature extraction network to obtain features/>, of the original image dataset and the two transformed datasets, respectively；
A contrast feature extraction module for extracting features of the original image dataset and the two transformed datasetsInputting into a comparison network to obtain an original image data set and two comparison features/>；
Example contrast Module for comparing featuresBy loss function/>Calculation of examples comparative loss/>
An updating module for comparing the characteristicsUpper mining of neighboring images, acquiring and updating neighbor/>, per sample；
A processing module for utilizing neighborsConstruction of adjacency matrix/>And adding different random masks to obtain two adjacent matrixes/>;
A building module for utilizing the characteristicsAnd adjacency matrix/>Constructing graph structure data；
Clustering module for clustering graph structure dataInputting the cluster characteristics and the neighbor characteristics of the cluster characteristics into a graph neural network, and obtaining cluster allocation/>；
Cluster level comparison module for assigning using clustersBy loss function/>Calculate cluster level contrast loss/>
An optimization module for utilizing the loss function,/>And calculating a final loss function L, and jointly optimizing a feature extraction network, a comparison network and a graph neural network through a back propagation mechanism of the deep neural network.
In a third aspect, there is provided a computer readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by a computing device, cause the computing device to perform the method of the first aspect.
In a fourth aspect, there is provided a computing device comprising:
one or more processors, memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs comprising instructions for performing the method of the first aspect.
(III) beneficial effects
(1) The image contrast clustering method and system based on the graph structure solve the problem that an image dataset generally only contains intrinsic characteristic information of images and cannot effectively fuse structural information among the images.
(2) The invention discloses an image contrast clustering method and system based on a graph structure, which solve the problem that the contrast learning method treats different transformations of the same image as positive sample pairs, and the rest images as negative sample pairs, even samples of the same type, so as to cause clustering deviation.
Drawings
FIG. 1 is a schematic flow chart of an image contrast clustering method based on a graph structure;
fig. 2 is a flowchart of an embodiment of the present invention when the image contrast clustering method is implemented.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are only some, but not all embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Examples
As shown in fig. 1, an embodiment of the present invention provides an image contrast clustering method based on a graph structure, including:
Acquiring an original image dataset And two kinds of transformation are carried out on the original image data set to respectively obtain two transformed data sets/>; Wherein the two transforms include, weak transform/>And strong transformation/>，/>，/>Is a set of transforms;
To collect original image data And transformed dataset/>Input to a feature extraction network to obtain features/>, of the original image dataset and the two transformed datasets, respectively；
Characterizing an original image dataset with features of two transformed datasetsInput to a comparison network to obtain the comparison features/>, of the original image dataset and the two transformed datasets respectively；
Will contrast characteristicsBy loss function/>Calculating an instance contrast loss;
In contrast to features Upper mining of neighboring images, acquiring and updating neighbor/>, per sample；
Utilizing neighborsConstruction of adjacency matrix/>And adding different random masks to obtain two adjacent matrixes;
By means of featuresAnd adjacency matrix A,/>Constructing graph structure data；
Map structure dataInputting the cluster characteristics and the neighbor characteristics of the cluster characteristics into a graph neural network, and obtaining cluster allocation/>；
Using cluster allocationBy loss function/>Calculating cluster level contrast loss;
Using loss functions ,/>Calculating a final loss function L, and jointly optimizing a feature extraction network, a comparison network and a graph convolution neural network through a back propagation mechanism of the deep neural network;
repeating the steps until the iteration times or network convergence is met, and obtaining and outputting a final clustering grouping result 。
As shown in fig. 2, in the implementation, the steps are as follows:
step S1, for an original image dataset Subjecting it to different transformations/>, respectivelyObtaining transformed dataset/>，/>For transforming a set
In one embodiment of the present invention, the weak transform is performed using two transform modesFor random clipping and scaling, random horizontal flipping, random brightness, contrast, saturation, tone transformation, and random gray-scale transformation. Strong transform/>Including random horizontal flipping, random clipping, and selecting four components from normalized contrast, balanced histogram, rotation, inversion, enhanced color balance, adjusted contrast, adjusted brightness, adjusted sharpness, affine transformation, reduced number of bits per color channel. The weak and strong transforms are used to train the deep neural network.
Step S2, acquiring a feature Z through a feature extraction network by using the original image data set and the transformed data set obtained in the step S1,；
In one embodiment of the invention, the features are acquired using a deep neural network ResNet with an output dimension of 512.
Step S3, utilizing the characteristics obtained in the step S2Obtaining contrast characteristics through a contrast network；
In one embodiment of the present invention, the two-layer fully connected network is utilized to obtain the contrast characteristics, and the output dimension is 128.
Step S4, utilizing the contrast characteristic obtained in the step S3Calculating contrast loss/>, by example level contrast loss function；
In one embodiment of the invention, the example level contrast loss function is implemented by the NT-Xent loss function, and the different transformation characteristics of the same image are @) The remaining samples are considered as negative pairs of samples. The loss function is widely applied to contrast learning and will not be described in detail.
Step S5, obtaining contrast characteristics by utilizing the step S3Mining neighboring images on the feature, acquiring and updating neighbors/>, of each sample；
In an embodiment of the invention, the nearest neighbors are mined by using the inner product similarity among the features, and the nearest 10 neighbors of each image are obtained, so that the network iterates for a plurality of times to update the neighbors in order to ensure the stability of clustering。
Step S6, utilizing the neighbors obtained in the step S5Construction of adjacency matrix/>And adding different random masks to obtain two adjacent matrixes/>；
In one embodiment of the invention, the adjacency matrixTo add the bipartite graph of the self-loop:
wherein, And/>Index representing contrast features of different samples, respectively,/>Representing acquisition neighbors, the random mask randomly deletes/>, with 10% probabilityEdges between adjacent nodes.
Step S7, constructing graph structure data by using the characteristics obtained in the step S2 and the adjacency matrix obtained in the step S6；
In an embodiment of the present invention, similar to the graph contrast learning method, two different graph structure data are respectively formed by different transformed image features and two randomly transformed adjacent graphsAnd/>，/>For final verification.
Step S8, utilizing the graph structure data obtained in the step S7, aggregating the characteristics and the neighbor characteristics thereof through a graph neural network, and obtaining cluster distribution。
In an embodiment of the present invention, the graph neural network is a layer of graph convolution, and the output dimension is k, where k represents the number of categories contained in the data.
Step S9, utilizing the cluster allocation obtained in the step S8Calculating contrast loss/>, by cluster-level contrast function；
In one embodiment of the invention, the cluster-level contrast function is implemented by the NT-Xent loss function, with the same class of different transform characteristicsThe remaining classes are considered as positive pairs of samples and negative pairs of samples.
Step S10, utilizing the loss functions obtained in the steps S4 and S9 to jointly optimize a feature extraction network through a back propagation mechanism of a deep neural network, and comparing the network with a graph convolution neural network;
in one embodiment of the invention, the final loss function is defined as:
wherein, Representing example contrast loss,/>Representing cluster level contrast loss,/>Representing the balance coefficient.
Step S11, returning to the step S1 for iteration until the iteration times are met, and obtaining and outputting a final clustering grouping result。
According to the method, neighbors are mined through the contrast characteristics, the image data set is constructed into a graph data structure, so that semantic information of images can be obtained, and structural information among the image data can be effectively mined; and the neighbor information is aggregated through graph convolution, the features with consistent semantic information are fused, and the deviation of contrast learning is reduced, so that the clustering performance is improved.
Yet another embodiment of the present invention provides an image contrast clustering system based on a graph structure, which is characterized by comprising:
An acquisition module for acquiring an original image dataset And two kinds of transformation are carried out on the original image data set to respectively obtain two transformed data sets/>; Wherein the two transforms include weak transforms/>And strong transformation/>，/>，Is a set of transforms;
A feature extraction module for extracting original image data set And two transformed datasets/>Input to a feature extraction network to obtain features/>, of the original image dataset and the two transformed datasets, respectively；
A contrast feature extraction module for extracting features of the original image dataset and the two transformed datasetsInputting into a comparison network to obtain an original image data set and two comparison features/>；
Example contrast Module for comparing featuresBy loss function/>Calculation of examples comparative loss/>
An updating module for comparing the characteristicsUpper mining of neighboring images, acquiring and updating neighbor/>, per sample；
A processing module for utilizing neighborsConstruction of adjacency matrix/>And adding different random masks to obtain two adjacent matrixes/>;
A building module for utilizing the characteristicsAnd adjacency matrix/>Constructing graph structure data；
Clustering module for clustering graph structure dataInputting the cluster characteristics and the neighbor characteristics of the cluster characteristics into a graph neural network, and obtaining cluster allocation/>；
Cluster level comparison module for assigning using clustersBy loss function/>Calculate cluster level contrast loss/>
An optimization module for utilizing the loss function,/>Calculating a final loss function L, and jointly optimizing a feature extraction network, a comparison network and a graph neural network through a back propagation mechanism of a deep neural network
Embodiments of the present application may be provided as a method or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein. The aspects of embodiments of the present application may be implemented in a variety of computer languages, such as object oriented python, etc.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
It is noted that relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
Claims (10)
1. The image contrast clustering method based on the graph structure is characterized by comprising the following steps of:
Acquiring an original image dataset And two kinds of transformation are carried out on the original image data set to respectively obtain two transformed data sets/>; Wherein the two transforms include, weak transform/>And strong transformation/>，/>，/>Is a set of transforms;
To collect original image data And transformed dataset/>Input to a feature extraction network to obtain features/>, of the original image dataset and the two transformed datasets, respectively；
Characterizing an original image dataset with features of two transformed datasetsInput to a comparison network to obtain the comparison features/>, of the original image dataset and the two transformed datasets respectively；
Will contrast characteristicsBy loss function/>Calculating an instance contrast loss;
In contrast to features Upper mining of neighboring images, acquiring and updating neighbor/>, per sample；
Utilizing neighborsConstruction of adjacency matrix/>And adding different random masks to obtain two adjacent matrixes;
By means of featuresAnd adjacency matrix A,/>Constructing graph structure data；
Map structure dataInputting the cluster characteristics and the neighbor characteristics of the cluster characteristics into a graph neural network, and obtaining cluster allocation/>；
Using cluster allocationBy loss function/>Calculating cluster level contrast loss;
Using loss functions , />Calculating a final loss function L, and jointly optimizing a feature extraction network, a comparison network and a graph convolution neural network through a back propagation mechanism of the deep neural network;
repeating the steps until the iteration times or network convergence is met, and obtaining and outputting a final clustering grouping result 。
2. The image contrast clustering method based on the graph structure of claim 1, wherein the image contrast clustering method is characterized by comprising the following steps of: the original image is used for neighbor mining and obtaining a final clustering result.
3. The image contrast clustering method based on the graph structure according to claim 2, wherein the image contrast clustering method is characterized in that: the weak transformationFor random clipping and scaling, random horizontal flipping, random brightness, contrast, saturation, tone transformation, and random gray-scale transformation.
4. The image contrast clustering method based on graph structure of claim 3, wherein: the strong transformationIn addition to including random horizontal flipping, random clipping, and randomly selecting four components from normalized contrast, balanced histogram, rotation, inversion, enhanced color balance, adjusted contrast, adjusted brightness, adjusted sharpness, affine transformation, reduced number of bits per color channel.
5. The image contrast clustering method based on the graph structure of claim 1, wherein the image contrast clustering method is characterized by comprising the following steps of: the adjacency matrixTo add the bipartite graph of the self-loop:
wherein, And/>Index representing contrast features of different samples, respectively,/>Representing acquisition neighbors, the random mask randomly deletes/>, with 10% probabilityEdges between adjacent nodes.
6. The image contrast clustering method based on graph structure of claim 5, wherein the image contrast clustering method is characterized by comprising the following steps: the graph neural network is composed of a graph rolling network and a full-connection network, the output dimension is K, and K represents the category number contained in data.
7. The image contrast clustering method based on the graph structure of claim 6, wherein the image contrast clustering method is characterized by comprising the following steps of: the loss function L is defined as:
wherein, Representing example contrast loss,/>Representing cluster level contrast loss,/>Representing the balance coefficient.
8. An image contrast clustering system based on a graph structure, which is characterized by comprising:
An acquisition module for acquiring an original image dataset And two kinds of transformation are carried out on the original image data set to respectively obtain two transformed data sets/>; Wherein the two transforms include weak transforms/>And strong transformation/>，/>，/>Is a set of transforms;
A feature extraction module for extracting original image data set And two transformed datasets/>Input to a feature extraction network to obtain features/>, of the original image dataset and the two transformed datasets, respectively；
A contrast feature extraction module for extracting features of the original image dataset and the two transformed datasetsInputting into a comparison network to obtain an original image data set and two comparison features/>；
Example contrast Module for comparing featuresBy loss function/>Calculation of examples comparative loss/>
An updating module for comparing the characteristicsUpper mining of neighboring images, acquiring and updating neighbor/>, per sample；
A processing module for utilizing neighborsConstruction of adjacency matrix/>And adding different random masks to obtain two adjacent matrixes/>;
A building module for utilizing the characteristicsAnd adjacency matrix/>Constructing graph structure data；
Clustering module for clustering graph structure dataInputting the cluster characteristics and the neighbor characteristics of the cluster characteristics into a graph neural network, and obtaining cluster allocation/>；
Cluster level comparison module for assigning using clustersBy loss function/>Calculate cluster level contrast loss/>
An optimization module for utilizing the loss function, />And calculating a final loss function L, and jointly optimizing a feature extraction network, a comparison network and a graph neural network through a back propagation mechanism of the deep neural network.
9. A computer readable storage medium storing one or more programs, wherein the one or more programs comprise instructions, which when executed by a computing device, cause the computing device to perform any of the methods of claims 1-7.
10. A computing device for image contrast clustering method based on graph structure, comprising:
one or more processors, memory, and one or more programs, wherein one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs comprising instructions for performing any of the methods of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410343857.7A CN117934891B (en) | 2024-03-25 | 2024-03-25 | Image contrast clustering method and system based on graph structure |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410343857.7A CN117934891B (en) | 2024-03-25 | 2024-03-25 | Image contrast clustering method and system based on graph structure |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117934891A true CN117934891A (en) | 2024-04-26 |
| CN117934891B CN117934891B (en) | 2024-06-07 |
Family
ID=90761448
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410343857.7A Active CN117934891B (en) | 2024-03-25 | 2024-03-25 | Image contrast clustering method and system based on graph structure |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117934891B (en) |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112435239A (en) * | 2020-11-25 | 2021-03-02 | 南京农业大学 | Scindapsus aureus leaf shape parameter estimation method based on MRE-PointNet and self-encoder model |
| CN112668627A (en) * | 2020-12-24 | 2021-04-16 | 四川大学 | Large-scale image online clustering system and method based on contrast learning |
| CN112925977A (en) * | 2021-02-26 | 2021-06-08 | 中国科学技术大学 | Recommendation method based on self-supervision graph representation learning |
| CN114332853A (en) * | 2021-12-08 | 2022-04-12 | 腾讯科技（深圳）有限公司 | Training method, device and equipment of image interpretation model and storage medium |
| CN114792113A (en) * | 2022-05-10 | 2022-07-26 | 中国人民解放军国防科技大学 | Graph clustering method and device based on multi-order neighbor information transfer fusion clustering network |
| CN115168632A (en) * | 2022-07-12 | 2022-10-11 | 合肥中聚源智能科技有限公司 | Image-text matching model training method with robustness |
| CN115457395A (en) * | 2022-09-22 | 2022-12-09 | 南京信息工程大学 | Lightweight remote sensing target detection method based on channel attention and multi-scale feature fusion |
| CN115456093A (en) * | 2022-09-20 | 2022-12-09 | 中山大学 | High-performance graph clustering method based on attention-graph neural network |
| CN115730143A (en) * | 2022-11-22 | 2023-03-03 | 上海交通大学 | Recommendation system, method, terminal and medium based on task alignment meta learning and augmentation graph |
| CN116562274A (en) * | 2023-05-18 | 2023-08-08 | 中国工商银行股份有限公司 | Target theme determining method and device |
| WO2023165506A1 (en) * | 2022-03-01 | 2023-09-07 | Insilico Medicine Ip Limited | Structure-based deep generative model for binding site descriptors extraction and de novo molecular generation |
| CN117036760A (en) * | 2023-07-31 | 2023-11-10 | 南京邮电大学 | Multi-view clustering model implementation method based on graph comparison learning |
| CN117219291A (en) * | 2023-09-06 | 2023-12-12 | 大连海事大学 | Drug repositioning prediction method and device based on contrast learning and graph neural network |
-
2024
- 2024-03-25 CN CN202410343857.7A patent/CN117934891B/en active Active
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112435239A (en) * | 2020-11-25 | 2021-03-02 | 南京农业大学 | Scindapsus aureus leaf shape parameter estimation method based on MRE-PointNet and self-encoder model |
| CN112668627A (en) * | 2020-12-24 | 2021-04-16 | 四川大学 | Large-scale image online clustering system and method based on contrast learning |
| CN112925977A (en) * | 2021-02-26 | 2021-06-08 | 中国科学技术大学 | Recommendation method based on self-supervision graph representation learning |
| CN114332853A (en) * | 2021-12-08 | 2022-04-12 | 腾讯科技（深圳）有限公司 | Training method, device and equipment of image interpretation model and storage medium |
| WO2023165506A1 (en) * | 2022-03-01 | 2023-09-07 | Insilico Medicine Ip Limited | Structure-based deep generative model for binding site descriptors extraction and de novo molecular generation |
| CN114792113A (en) * | 2022-05-10 | 2022-07-26 | 中国人民解放军国防科技大学 | Graph clustering method and device based on multi-order neighbor information transfer fusion clustering network |
| CN115168632A (en) * | 2022-07-12 | 2022-10-11 | 合肥中聚源智能科技有限公司 | Image-text matching model training method with robustness |
| CN115456093A (en) * | 2022-09-20 | 2022-12-09 | 中山大学 | High-performance graph clustering method based on attention-graph neural network |
| CN115457395A (en) * | 2022-09-22 | 2022-12-09 | 南京信息工程大学 | Lightweight remote sensing target detection method based on channel attention and multi-scale feature fusion |
| CN115730143A (en) * | 2022-11-22 | 2023-03-03 | 上海交通大学 | Recommendation system, method, terminal and medium based on task alignment meta learning and augmentation graph |
| CN116562274A (en) * | 2023-05-18 | 2023-08-08 | 中国工商银行股份有限公司 | Target theme determining method and device |
| CN117036760A (en) * | 2023-07-31 | 2023-11-10 | 南京邮电大学 | Multi-view clustering model implementation method based on graph comparison learning |
| CN117219291A (en) * | 2023-09-06 | 2023-12-12 | 大连海事大学 | Drug repositioning prediction method and device based on contrast learning and graph neural network |
Non-Patent Citations (3)
| Title |
|---|
| 朱云华等: "图对比学习引导的多视图聚类网络", 《计算机应用》, 8 January 2024 (2024-01-08), pages 1 - 10 * |
| 杨国亮;谢乃俊;王艳芳;梁礼明;: "基于低秩稀疏评分的非监督特征选择", 计算机工程与科学, no. 04, 15 April 2015 (2015-04-15), pages 33 - 40 * |
| 罗海银;郑钰辉: "图像修复方法研究综述", 《计算机科学与探索》, vol. 16, no. 10, 17 July 2022 (2022-07-17), pages 2193 - 2218 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117934891B (en) | 2024-06-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108132968B (en) | Weak supervision learning method for associated semantic elements in web texts and images | |
| CN110929029A (en) | Text classification method and system based on graph convolution neural network | |
| CN108614997B (en) | Remote sensing image identification method based on improved AlexNet | |
| CN109934826A (en) | A kind of characteristics of image dividing method based on figure convolutional network | |
| CN106780639B (en) | Hash coding method based on significance characteristic sparse embedding and extreme learning machine | |
| CN112711652B (en) | Term standardization method and device | |
| CN111178196B (en) | Cell classification method, device and equipment | |
| CN111784699B (en) | Method and device for carrying out target segmentation on three-dimensional point cloud data and terminal equipment | |
| CN112836820B (en) | Deep convolution network training method, device and system for image classification task | |
| CN114581451A (en) | Scattering map neural network-based brain magnetic resonance image segmentation method | |
| CN117934890B (en) | Prototype comparison image clustering method and system based on local and global neighbor alignment | |
| CN112508966B (en) | Interactive image segmentation method and system | |
| CN111245820A (en) | Phishing website detection method based on deep learning | |
| CN113850811B (en) | Three-dimensional point cloud instance segmentation method based on multi-scale clustering and mask scoring | |
| CN117934891B (en) | Image contrast clustering method and system based on graph structure | |
| CN113408418A (en) | Calligraphy font and character content synchronous identification method and system | |
| CN110781936B (en) | Construction method of threshold learnable local binary network based on texture description and deep learning and remote sensing image classification method | |
| CN111260659A (en) | Image interactive segmentation method based on initial annotation point guidance | |
| CN113836244B (en) | Sample acquisition method, model training method, relation prediction method and device | |
| CN115775340A (en) | Feature modulation-based self-adaptive small sample image classification method and device | |
| CN107644251B (en) | Object classification method, device and system | |
| US11165646B1 (en) | Network node clustering | |
| CN114780649A (en) | Method and device for identifying structured data entity type | |
| CN109800384B (en) | Basic probability assignment calculation method based on rough set information decision table | |
| CN112836763A (en) | Graph structure data classification 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 |