CN108564102A - Image clustering evaluation of result method and apparatus - Google Patents
Image clustering evaluation of result method and apparatus Download PDFInfo
- Publication number
- CN108564102A CN108564102A CN201810007796.1A CN201810007796A CN108564102A CN 108564102 A CN108564102 A CN 108564102A CN 201810007796 A CN201810007796 A CN 201810007796A CN 108564102 A CN108564102 A CN 108564102A
- Authority
- CN
- China
- Prior art keywords
- cluster
- image
- result
- result evaluation
- target 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Probability & Statistics with Applications (AREA)
- Image Analysis (AREA)
Abstract
The embodiment of the present application discloses image clustering evaluation of result method and apparatus.One specific implementation mode of this method includes：Each image in the target image cluster obtained to cluster carries out feature extraction；The corresponding Eigen Covariance matrix of target image cluster is generated based on the feature extracted；Covariance matrix is inputted in cluster result evaluation model, the cluster result evaluation information of target image cluster is obtained, wherein sample image cluster training of the cluster result evaluation model based on marked cluster result evaluation information obtains.The embodiment realizes the assessment of reliable image clustering effect.
Description
Technical field
The invention relates to field of computer technology, and in particular to technical field of image processing more particularly to image
Cluster result evaluation method and device.
Background technology
Cluster is a kind of method of data mining.In the image processing arts, image clustering be based on characteristics of image will be more
Width image is divided into the process for the multiple classes being made of similar image.Image clustering has in fields such as image segmentation, target followings
There is important role.
In large-scale image clustering scene, it is possible that error, will originally be not belonging to of a sort in cluster process
Image is classified as the same image cluster, then the center of the image cluster can shift so that subsequent cluster result is more and more inaccurate
Really, it is therefore desirable to the order of accuarcy for evaluating cluster result, the image clustering result corrected mistake in time.
Invention content
The embodiment of the present application proposes image clustering evaluation of result method and apparatus.
In a first aspect, the embodiment of the present application provides a kind of image clustering evaluation of result method, including：Cluster is obtained
Each image in target image cluster carries out feature extraction；The corresponding feature association of target image cluster is generated based on the feature extracted
Variance matrix；Covariance matrix is inputted in cluster result evaluation model, the cluster result evaluation information of target image cluster is obtained；
Wherein, sample image cluster training of the cluster result evaluation model based on marked cluster result evaluation information obtains.
In some embodiments, above-mentioned cluster result evaluation model is trained as follows obtains：It obtains multiple
The label result of the cluster result evaluation information of sample image cluster and each sample image cluster；Feature is carried out to each sample image cluster to carry
It takes；The sample characteristics covariance matrix of each sample image cluster is generated based on the feature extracted；Using deep learning method, by sample
Input of the eigen covariance matrix as the corresponding neural network of cluster result evaluation model, the cluster based on sample image cluster
The label result of evaluation of result information and preset loss function train to obtain cluster result evaluation model, wherein preset damage
Lose the label result nerve corresponding with cluster result evaluation model of cluster result evaluation information of the function based on sample image cluster
Difference structure between the prediction result of the evaluation information of the cluster result of the sample image cluster of network output.
In some embodiments, above-mentioned to use deep learning method, using sample characteristics covariance matrix as cluster result
The input of the corresponding neural network of evaluation model, the label result of the cluster result evaluation information based on sample image cluster and default
Loss function train to obtain cluster result evaluation model, including：Determine the corresponding neural network of cluster result evaluation model
The initial value of parameter executes and compares step；Comparing step includes：Sample characteristics covariance matrix input cluster result is evaluated into mould
The corresponding neural network of type obtains the prediction result of the cluster result evaluation information of corresponding sample image cluster, judges to lose letter
Whether several values meets the preset condition of convergence；If the judging result for comparing step is no, it is based on loss function, using under gradient
Drop method updates the corresponding parameter of the corresponding neural network of cluster result evaluation model, and executes comparison step；If comparing step
Judging result is yes, the parameter of the corresponding neural network of output cluster result evaluation model.
In some embodiments, each image in the above-mentioned target image cluster obtained to cluster carries out feature extraction, packet
It includes：Feature extraction is carried out to each image in target image cluster using convolutional neural networks, obtains every width in target image cluster
The feature set of graphs of image；It is above-mentioned that the corresponding Eigen Covariance matrix of target image cluster is generated based on the feature extracted, including：
The eigenmatrix of feature set of graphs structure target image cluster based on each image in target image cluster；Calculate target image cluster
Eigenmatrix covariance matrix.
In some embodiments, the above method further includes：According to the cluster result evaluation information of target image cluster retain or/
Delete target image cluster.
Second aspect, the embodiment of the present application provide a kind of image clustering evaluation of result device, including：Extraction unit is used
Each image in the target image cluster obtained to cluster carries out feature extraction；Generation unit, for based on the spy extracted
Sign generates the corresponding Eigen Covariance matrix of target image cluster；Evaluation unit is tied for clustering Eigen Covariance Input matrix
In fruit evaluation model, the cluster result evaluation information of target image cluster is obtained；Wherein, cluster result evaluation model is based on marked
The sample image cluster training of cluster result evaluation information obtains.
In some embodiments, above-mentioned cluster result evaluation model is trained as follows obtains：It obtains multiple
The label result of the cluster result evaluation information of sample image cluster and each sample image cluster；Feature is carried out to each sample image cluster to carry
It takes；The sample characteristics covariance matrix of each sample image cluster is generated based on the feature extracted；Using deep learning method, by sample
Input of the eigen covariance matrix as the corresponding neural network of cluster result evaluation model, the cluster based on sample image cluster
The label result of evaluation of result information and preset loss function train to obtain cluster result evaluation model, wherein preset damage
Lose the label result nerve corresponding with cluster result evaluation model of cluster result evaluation information of the function based on sample image cluster
Difference structure between the prediction result of the evaluation information of the cluster result of the sample image cluster of network output.
In some embodiments, above-mentioned cluster result evaluation model is to use deep learning method, by sample characteristics association side
Input of the poor matrix as the corresponding neural network of cluster result evaluation model, the cluster result evaluation letter based on sample image cluster
The label result of breath and preset loss function, further according to being trained such as under type：Determine that cluster result evaluates mould
The initial value of the parameter of the corresponding neural network of type executes and compares step；Comparing step includes：By sample characteristics covariance matrix
The corresponding neural network of cluster result evaluation model is inputted, the pre- of the cluster result evaluation information of corresponding sample image cluster is obtained
It surveys as a result, judging whether the value of loss function meets the preset condition of convergence；If the judging result for comparing step is no, based on damage
Function is lost, using the corresponding parameter of the corresponding neural network of gradient descent method update cluster result evaluation model, and executes comparison
Step；If the judging result for comparing step is yes, the parameter of the corresponding neural network of output cluster result evaluation model.
In some embodiments, said extracted unit is further used for the target image obtained as follows to cluster
Each image in cluster carries out feature extraction：Feature is carried out using convolutional neural networks to each image in target image cluster to carry
It takes, obtains the feature set of graphs of each image in target image cluster；Above-mentioned generation unit is further used for giving birth to as follows
At the corresponding Eigen Covariance matrix of target image cluster：Feature set of graphs based on each image in target image cluster builds mesh
The eigenmatrix of logo image cluster；Calculate the covariance matrix of the eigenmatrix of target image cluster.
In some embodiments, above-mentioned apparatus further includes：Processing unit, for being commented according to the cluster result of target image cluster
Valence information retains or/delete target image cluster.
The third aspect, the embodiment of the present application provide a kind of electronic equipment, including：One or more processors；Storage dress
It sets, for storing one or more programs, when one or more programs are executed by one or more processors so that one or more
A processor realizes the image clustering evaluation of result method provided such as first aspect.
Fourth aspect, the embodiment of the present application provide a kind of computer readable storage medium, are stored thereon with computer journey
Sequence, wherein the image clustering evaluation of result method that first aspect provides is realized when program is executed by processor.
The image clustering evaluation of result method and apparatus of the above embodiments of the present application pass through the target image obtained to cluster
Each image in cluster carries out feature extraction, is then based on the feature extracted and generates the corresponding Eigen Covariance of target image cluster
Matrix；Covariance matrix is inputted in cluster result evaluation model later, obtains the cluster result evaluation information of target image cluster,
Wherein, sample image cluster training of the cluster result evaluation model based on marked cluster result evaluation information show that realizing can
The assessment for the image clustering effect leaned on.
Description of the drawings
By reading a detailed description of non-restrictive embodiments in the light of the attached drawings below, the application's is other
Feature, objects and advantages will become more apparent upon：
Fig. 1 is that this application can be applied to exemplary system architecture figures therein；
Fig. 2 is the flow chart according to one embodiment of the image clustering evaluation of result method of the application；
Fig. 3 is the schematic diagram according to an application scenarios of the image clustering evaluation of result method of the application；
Fig. 4 is a structural schematic diagram according to the image clustering evaluation of result device of the embodiment of the present application；
Fig. 5 is adapted for the structural schematic diagram of the computer system of the electronic equipment for realizing the embodiment of the present application.
Specific implementation mode
The application is described in further detail with reference to the accompanying drawings and examples.It is understood that this place is retouched
The specific embodiment stated is used only for explaining related invention, rather than the restriction to the invention.It also should be noted that in order to
Convenient for description, is illustrated only in attached drawing and invent relevant part with related.
It should be noted that in the absence of conflict, the features in the embodiments and the embodiments of the present application can phase
Mutually combination.The application is described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.
Fig. 1 shows the image clustering evaluation of result method or image clustering evaluation of result device that can apply the application
The exemplary system architecture 100 of embodiment.
As shown in Figure 1, system architecture 100 may include terminal device 101,102,103, network 104 and server 105.
Network 104 between terminal device 101,102,103 and server 105 provide communication link medium.Network 104 can be with
Including various connection types, such as wired, wireless communication link or fiber optic cables etc..
Terminal device 101,102,103 can be interacted by network 104 with server 105, with reception or transmission data.Eventually
End equipment 101,102,103 can have the equipment of image collecting function, such as monitoring camera, smart mobile phone, tablet are electric
Brain, PC etc..Terminal device 101,102,103 can have network interface, can receive the image that server 105 is sent out
Obtain request, can also in response to user 110 request by network interface by the image collected be uploaded to server 105 into
Row processing.
Server 105 can be to provide the server of various services, such as the figure to the upload of terminal device 101,102,103
Server as carrying out image clustering.After terminal device 101,102 uploads image, server 105 can be to the figure of upload
As carrying out the processing such as feature extraction, cluster, target identification, and handling result is returned into terminal device 101,102,103.
It should be noted that the image clustering evaluation of result method that the embodiment of the present application is provided can be by terminal device
101,102,103 or server 105 execute, correspondingly, image clustering evaluation of result device can be set to terminal device 101,
102,103 or server 105 in.
It should be understood that the number of the terminal device, network and server in Fig. 1 is only schematical.According to realization need
It wants, can have any number of terminal device, network and server.Such as server can be the server of concentrating type, packet
Include the multiple servers for deploying different processes.
With continued reference to Fig. 2, it illustrates the streams according to one embodiment of the image clustering evaluation of result method of the application
Journey 200.The image clustering evaluation of result method, includes the following steps：
Step 201, each image in the target image cluster obtained to cluster carries out feature extraction.
In the present embodiment, the electronic equipment of above-mentioned image clustering evaluation of result method operation thereon is (such as shown in Fig. 1
Terminal device or server) the image cluster obtained to image clustering can be obtained, each image cluster may include multiple image,
Such as may include several facial images.Herein, target image cluster can be image cluster to be evaluated, can be that cluster obtains
Multiple images cluster in any one.
It, can be successively using each image cluster as target figure after obtaining image cluster to image clustering in actual scene
As cluster, its cluster result is evaluated.The instruction that image cluster to be evaluated can also be selected in response to receiving, determines selected figure
As cluster is target image cluster.
Various features extracting method may be used, feature extraction is carried out to each width image in target image cluster.Such as it can be with
Extract the statistical nature, including area, gray value, histogram statistical features etc. of image；Edge detection, angle point inspection may be used
Survey, Scale invariant features transform (Scale Invariant Feature Transform, SIFT), principal component analysis scheduling algorithm
Extract the feature of image.The expression of the forms such as array, vector or matrix may be used in the feature extracted.
Optionally, each width image in target image cluster carries out feature extraction using identical feature extracting method, in this way,
The dimension of the feature of each width image extracted is identical.
Step 202, the corresponding Eigen Covariance matrix of target image cluster is generated based on the feature extracted.
It, can be based on the feature construction target of all images in target image cluster after extracting the feature of each image
The corresponding Eigen Covariance matrix of image cluster.The covariance of all features of target image cluster between any two can be specifically calculated,
Each covariance is as an element in covariance matrix.
In some optional realization methods of the present embodiment, every width figure in the above-mentioned target image cluster obtained to cluster
As the step 201 of progress feature extraction may include：The each image in target image cluster is carried out using convolutional neural networks
Feature extraction obtains the feature set of graphs of each image in target image cluster.At this moment, under type such as may be used and execute step
202：The eigenmatrix of feature set of graphs structure target image cluster based on each image in target image cluster；Calculate target figure
As the covariance matrix of the eigenmatrix of cluster.That is, convolutional neural networks may be used as feature extracting method, will carry
Several characteristic patterns of each image of taking-up form a feature set of graphs.It then can be by every width in the same target image cluster
The characteristic pattern collective combinations of image together, form the feature set of graphs of target image cluster, and based on the feature of target image cluster
Set of graphs construction feature matrix.This feature matrix can be by the pixel grey scale of each characteristic pattern in the feature set of graphs of target image cluster
Matrix combines to be formed successively along matrix line direction or column direction or preset combination.Feature square can be calculated later
The covariance matrix of battle array.
Specifically, it is assumed that target image cluster includes n width images X1、X2、…、Xn, each image XiIt is characterized as fi, special
Levy fiCan be the feature of D dimensions, fi∈RD, then the corresponding eigenmatrix S of target image cluster is：
Wherein, fijIt is characterized fiJth tie up component, j=1,2 ..., D.
Then the covariance matrix of the corresponding eigenmatrix S of target image cluster, the i.e. corresponding feature of image cluster can be calculated
Covariance matrix E, such as formula (2)：
Wherein,
In this way, the Eigen Covariance matrix E that a size is D × D can be obtained.
If similarity is higher between each width image in image cluster, clustering precision is higher, then the Eigen Covariance obtained
The distribution of regularity is presented in matrix E；If instead the similarity between each width image in image cluster is low, clustering precision is low, then
The regularity of the distribution of obtained Eigen Covariance matrix E is poor.
Step 203, by Eigen Covariance Input matrix cluster result evaluation model, the cluster knot of target image cluster is obtained
Fruit evaluation information.
In the present embodiment, features described above covariance matrix can be inputted to the cluster result evaluation model trained to carry out
Processing.Herein, cluster result evaluation model is the model evaluated for the cluster result to image cluster, it can be determined that poly-
Whether class result is accurate.
In the present embodiment, cluster result evaluation model can be built based on convolutional neural networks, the convolutional Neural
Network may include at least one convolutional layer, can also include at least one down-sampled layer.Each convolutional layer includes convolution kernel, can be with
Convolution algorithm is carried out to the image of input using convolution kernel, the image information of redundancy is removed, extracts characteristics of image.It extracts
Characteristics of image can be characteristic pattern, can characteristic pattern be converted to the feature vector including at least one-dimensional characteristic to indicate.
Above-mentioned cluster result evaluation model can be obtained using the machine learning method training for having supervision.It specifically can be with
Sample image cluster training based on marked cluster result evaluation information obtains.The sample graph of marked cluster result evaluation information
As cluster may include multiple sample image clusters, and the cluster result evaluation information of each sample image cluster label result.
When building sample image cluster, some sample images can be chosen, randomly sorts out and forms negative sample, according to similar between image
Degree is sorted out to form positive sample, marks the cluster result evaluation information of negative sample and positive sample respectively.It in the training process, can be with
The cluster result evaluation information of sample image cluster is predicted using cluster result evaluation model, prediction result and label are tied
Fruit compares, and constantly adjusts the parameter of cluster result evaluation model so that prediction result is approached with label result.In this way, cluster knot
Fruit evaluation model may learn the logic of handmarking's cluster result evaluation information.
In actual scene, can according in application to the requirement of clustering precision come the cluster result of marker samples image cluster
Evaluation information.Usually, the cluster result of sample image cluster can be labeled as " accurate " and " inaccuracy ", respectively with numerical value or
Alphanumeric tag marks.In some scenes, when label, can be by the evaluation information of the cluster result of sample image cluster according to accurate
Degree is divided into multiple levels, for example, for example, sample image cluster a, b, c of structure separately include 100 width images, sample
Similarity degree in this image cluster a between 100 width images is higher, 5 width images and other 95 width image phases in sample image cluster b
Relatively low like degree, the similarity degree of 20 width images and other 80 width images is relatively low in sample image cluster c；It then can be by sample image
The cluster result of cluster a, b, c are respectively labeled as " accurate ", " good ", " inaccuracy ".In this way, the image clustering result that training obtains
The level that evaluation model uses when can be according to label evaluates the target image cluster of input.
The image clustering evaluation of result method that the above embodiments of the present application provide passes through the target image cluster obtained to cluster
In each image carry out feature extraction, be then based on the feature that extracts and generate the corresponding Eigen Covariance square of target image cluster
Battle array；Covariance matrix is inputted in cluster result evaluation model later, obtains the cluster result evaluation information of target image cluster,
In, sample image cluster training of the cluster result evaluation model based on marked cluster result evaluation information obtains, realizes image
The assessment of Clustering Effect can promote the reliability of Clustering Effect assessment.
In some optional realization methods of the present embodiment, above-mentioned cluster result evaluation model can be according to such as lower section
Formula training obtains：
First, the label result of the cluster result evaluation information of multiple sample image clusters and each sample image cluster is obtained.It can
Sample image cluster is built to select some pictures in the image data bases such as existing network image library, monitoring image library.It needs
It is noted that including at least the accurate image cluster of cluster result in the sample image cluster of structure, that is, need to build at least one
Sample image cluster, the similarity degree in the sample image cluster between each image are all higher.For example, field is clustered in facial image
Jing Zhong needs the sample image cluster for building at least one several different facial images comprising same user.
The cluster result evaluation information of above-mentioned sample image cluster can be that handmarking is good, and number or symbol may be used
Label indicates, such as clusters the label result of the cluster result evaluation information of accurate sample image cluster and can use label " 1 "
It indicates, the label result for clustering the evaluation information of the cluster result of inaccurate sample image cluster can with label " 0 " come table
Show.In this way, after training is completed, cluster result evaluation model can export corresponding label " 1 " or " 0 " to indicate cluster knot
Fruit.
Then, feature extraction is carried out to each sample image cluster.
A variety of image characteristic extracting methods may be used, feature extraction is carried out to the image in sample image cluster.Such as it can be with
The feature of each image is extracted using the methods of principal component analysis, Scale invariant features transform, or neural network may be used
Carry out feature extraction.
It is alternatively possible to using same feature extracting method to sample image cluster and to every width figure in target image cluster
As carrying out feature extraction, the type or attribute of the feature extracted to avoid different feature extracting methods (such as extract
Characteristic type can be textural characteristics, edge feature, gray distribution features etc.) between difference to image clustering evaluation of result knot
The order of accuarcy of fruit impacts.
Later, the sample characteristics covariance matrix of each sample image cluster is generated based on the feature extracted.
It is then possible to be counted using the identical method of computational methods of Eigen Covariance matrix corresponding with target image cluster
Calculate the covariance matrix for the matrix being made of the feature of sample image cluster, you can special with the sample that sample image cluster is calculated
Levy covariance matrix.Specific computational methods are with reference to above formula (1), (2), (3), and details are not described herein again.
Finally, using deep learning method, sample characteristics covariance matrix is corresponding as cluster result evaluation model
The input of neural network, the label result of the cluster result evaluation information based on sample image cluster and the training of preset loss function
Obtain cluster result evaluation model.Wherein, the mark of cluster result evaluation information of the preset loss function based on sample image cluster
Remember the evaluation information of the cluster result of the sample image cluster of result neural network output corresponding with cluster result evaluation model
Difference structure between prediction result.
The neural network of cluster result evaluation model can be built, which can be convolutional neural networks, including
Multiple convolutional layers and down-sampled layer.Then sample characteristics covariance matrix is inputted into convolutional neural networks, obtains sample image cluster
Cluster result evaluation information prediction result.The difference being then based between the prediction result of output and label result is built
Loss function, then by repeating the parameter of adjustment cluster result evaluation model, inputting sample characteristics covariance matrix
New prediction result is obtained with cluster result evaluation model to minimize the value of loss function, by the value minimum when institute of loss function
The parameter for the cluster result evaluation model that the parameter used is completed as training.
It is above-mentioned to use deep learning method in further realization method, using sample characteristics covariance matrix as poly-
The input of the corresponding neural network of class evaluation of result model, the label result of the cluster result evaluation information based on sample image cluster
The step of training to obtain cluster result evaluation model with preset loss function can execute as follows：
First, it determines the initial value of the parameter of the corresponding neural network of cluster result evaluation model, executes and compare step.Than
Include to step：By the corresponding neural network of sample characteristics covariance matrix input cluster result evaluation model, obtain corresponding
The prediction result of the cluster result evaluation information of sample image cluster, judges whether the value of above-mentioned loss function meets preset convergence
Condition.
Herein, loss function can indicate that the prediction result of the cluster result evaluation information of sample image cluster is tied with label
Difference between fruit.The prediction result of above-mentioned cluster result evaluation information can have probability distribution, then may be used and be based on
The cross entropies of softmax functions builds loss function.
In the neural network using the cluster result evaluation model that parameter value is determined to the cluster result of sample image cluster
Evaluation information predicted after obtaining prediction result, the value of loss function can be calculated.Then it may determine that loss
Whether the value of function meets the preset condition of convergence, wherein the preset condition of convergence can be that the value of loss function is less than first
Threshold value can also be the difference between the value for comparing the loss function that step obtains several times recently less than second threshold.
It is then possible to be determined whether to carry out parameter update according to the judging result for comparing step.Specifically, if comparing step
Judging result be no, loss function is based on, using the corresponding neural network of gradient descent method update cluster result evaluation model
Corresponding parameter, and execute comparison step；If the judging result for comparing step is yes, output cluster result evaluation model is corresponding
The parameter of neural network.
Specifically, it if the value that the judging result for comparing step is loss function is unsatisfactory for the preset condition of convergence, counts
Loss function is calculated to the gradient of the parameter in the corresponding neural network of above-mentioned cluster result evaluation model, using gradient descent method come
Update each parameter in the corresponding neural network of above-mentioned cluster result evaluation model.It is then based on the cluster result after undated parameter
Evaluation model executes above-mentioned comparison step.
If the value that the judging result for comparing step is loss function meets the preset condition of convergence, can determine current
Cluster result evaluation model realizes the study of the logic of the cluster result evaluation to sample image cluster, at this moment can export current
Parameter as training complete cluster result evaluation model parameter.
The cluster result evaluation model that training obtains through the above way, takes full advantage of the image that similar image is formed
Otherness between the feature for the image cluster that consistency between the feature of cluster, dissimilar image are formed is assisted in character pair
Embodiment in variance, the parameter of continuous iteration update neural network are capable of the prediction of effectively training for promotion efficiency, lift scheme
Effect.
In some embodiments, above-mentioned image clustering evaluation of result method can also include：According to the poly- of target image cluster
Class evaluation of result information retains or/delete target image cluster.If the cluster result evaluation information of target image cluster indicates target
The Clustering Effect of image cluster is poor, then can be clustered again to the image in target image cluster with delete target image cluster；Such as
The Clustering Effect of the cluster result evaluation information instruction target image cluster of fruit target image cluster is good, then can retain target image
Cluster.Since inaccurate cluster result can cause the offset of cluster centre, can be deviated to avoid cluster centre by the above method
The caused further cluster subsequently based on the cluster result generates the error of bigger.
Referring to FIG. 3, it illustrates showing according to application scenarios of the image clustering evaluation of result method of the application
It is intended to.
As shown in figure 3, image collection can be sent to server B by terminal device A, and request server B is to image set
Conjunction is clustered.Server B obtains intermediate cluster result, including cluster 1, cluster during executing the cluster to image collection
2, cluster 3 ..., cluster m.Then server can carry out cluster result evaluation to each cluster in preliminary cluster result, specifically may be used
To carry out feature extraction to the image in each cluster, the Eigen Covariance matrix of each cluster is generated based on the feature extracted, it
In the cluster result evaluation model that Eigen Covariance Input matrix has been trained afterwards, the cluster result evaluation letter of each cluster is obtained
Breath, wherein the cluster result evaluation information of cluster 2 is " unclean ", indicates that the cluster result of cluster 2 is inaccurate；Cluster 1, cluster 3 ..., cluster
The cluster result evaluation information of m is " clean ", indicates that cluster 1, the cluster result of cluster 3 ..., cluster m are accurate.At this moment, cluster can be deleted
2, the image in cluster 2 is clustered again.From figure 3, it can be seen that the image clustering evaluation of result method of the present embodiment can be with
Any time when all image clusterings in image collection being applied to complete and before the completion of all image clusterings executes,
The reliability of cluster result can be promoted.
With further reference to Fig. 4, as the realization to method shown in above-mentioned each figure, this application provides a kind of image clustering knots
One embodiment of fruit evaluating apparatus, the device embodiment is corresponding with embodiment of the method shown in Fig. 2, which specifically can be with
Applied in various electronic equipments.
As shown in figure 4, the image clustering evaluation of result device 400 of the present embodiment includes：Extraction unit 401, generation unit
402 and evaluation unit 403.Each image in the target image cluster that extraction unit 401 can be used for obtaining cluster carries out special
Sign extraction；Generation unit 402 can be used for generating the corresponding Eigen Covariance matrix of target image cluster based on the feature extracted；
Evaluation unit 403 can be used for that the poly- of target image cluster in Eigen Covariance Input matrix cluster result evaluation model, will be obtained
Class evaluation of result information；Wherein, sample image cluster instruction of the cluster result evaluation model based on marked cluster result evaluation information
It gets out.
In the present embodiment, extraction unit 401 can obtain the image cluster obtained to image clustering, and use characteristics of image
Feature in the specified target image cluster of extracting method extraction per piece image.The target image cluster can be cluster obtain it is more
Any one in a image cluster.
After extraction unit 401 extracts the feature of each image, generation unit 402 can be based in target image cluster
The corresponding eigenmatrix of feature construction target image cluster of all images, specifically can be by the spy of each width image in target image cluster
Sign is indicated with vector or matrix, and the feature of each width image is combined as to the eigenmatrix of the target image, then can be calculated
The covariance matrix of eigenmatrix is to get to Eigen Covariance matrix.
The cluster that can have been trained the Eigen Covariance Input matrix that generation unit 402 generates with Utilization assessment unit 403
Evaluation of result model is evaluated, and the evaluation information of the cluster result of target image cluster is obtained.
In some embodiments, above-mentioned cluster result evaluation model trained as follows can obtain：It obtains
The label result of the cluster result evaluation information of multiple sample image clusters and each sample image cluster；Each sample image cluster is carried out special
Sign extraction；The sample characteristics covariance matrix of each sample image cluster is generated based on the feature extracted；Using deep learning method,
Using sample characteristics covariance matrix as the input of the corresponding neural network of cluster result evaluation model, based on sample image cluster
The label result of cluster result evaluation information and preset loss function train to obtain cluster result evaluation model, wherein default
Cluster result evaluation information of the loss function based on sample image cluster label result it is corresponding with cluster result evaluation model
Difference structure between the prediction result of the evaluation information of the cluster result of the sample image cluster of neural network output.
In a further embodiment, above-mentioned cluster result evaluation model can use deep learning method, by sample
Input of the Eigen Covariance matrix as the corresponding neural network of cluster result evaluation model, the cluster knot based on sample image cluster
The label result of fruit evaluation information and preset loss function, further according to being trained such as under type：Determine cluster knot
The initial value of the parameter of the corresponding neural network of fruit evaluation model executes and compares step；Comparing step includes：Sample characteristics are assisted
Variance matrix inputs the corresponding neural network of cluster result evaluation model, obtains the cluster result evaluation of corresponding sample image cluster
The prediction result of information, judges whether the value of loss function meets the preset condition of convergence；If the judging result for comparing step is
It is no, it is based on loss function, using the corresponding parameter of the corresponding neural network of gradient descent method update cluster result evaluation model, and
Execute above-mentioned comparison step；If the judging result for comparing step is yes, the corresponding neural network of output cluster result evaluation model
Parameter.
In some embodiments, said extracted unit can be further used for the target obtained as follows to cluster
Each image in image cluster carries out feature extraction：The each image in target image cluster is carried out using convolutional neural networks special
Sign extraction, obtains the feature set of graphs of each image in target image cluster；Above-mentioned generation unit can be further according to such as lower section
Formula generates the corresponding Eigen Covariance matrix of target image cluster：Feature set of graphs structure based on each image in target image cluster
Build the eigenmatrix of target image cluster；Calculate the covariance matrix of the eigenmatrix of target image cluster.
In some embodiments, above-mentioned apparatus 400 can also include：Processing unit, for according to the poly- of target image cluster
Class evaluation of result information retains or/delete target image cluster.
It should be appreciated that all units described in device 400 are corresponding with each step in the method described with reference to figure 2.
It is equally applicable to device 400 and unit wherein included above with respect to the operation and feature of method description as a result, it is no longer superfluous herein
It states.
The image clustering evaluation of result device 400 of the above embodiments of the present application, the mesh that cluster is obtained by extraction unit
Each image in logo image cluster carries out feature extraction, and then generation unit generates target image cluster pair based on the feature extracted
The Eigen Covariance matrix answered；Evaluation unit inputs covariance matrix in cluster result evaluation model later, obtains target figure
As the cluster result evaluation information of cluster, wherein sample of the cluster result evaluation model based on marked cluster result evaluation information
The training of image cluster obtains, realizes the assessment of reliable image clustering effect.
Below with reference to Fig. 5, it illustrates the computer systems 500 suitable for the electronic equipment for realizing the embodiment of the present application
Structural schematic diagram.Electronic equipment shown in Fig. 5 is only an example, to the function of the embodiment of the present application and should not use model
Shroud carrys out any restrictions.
As shown in figure 5, computer system 500 includes central processing unit (CPU) 501, it can be read-only according to being stored in
Program in memory (ROM) 502 or be loaded into the program in random access storage device (RAM) 503 from storage section 508 and
Execute various actions appropriate and processing.In RAM 503, also it is stored with system 500 and operates required various programs and data.
CPU 501, ROM 502 and RAM 503 are connected with each other by bus 504.Input/output (I/O) interface 505 is also connected to always
Line 504.
It is connected to I/O interfaces 505 with lower component：Importation 506 including keyboard, mouse etc.；It is penetrated including such as cathode
The output par, c 507 of spool (CRT), liquid crystal display (LCD) etc. and loud speaker etc.；Storage section 508 including hard disk etc.；
And the communications portion 509 of the network interface card including LAN card, modem etc..Communications portion 509 via such as because
The network of spy's net executes communication process.Driver 510 is also according to needing to be connected to I/O interfaces 505.Detachable media 511, such as
Disk, CD, magneto-optic disk, semiconductor memory etc. are mounted on driver 510, as needed in order to be read from thereon
Computer program be mounted into storage section 508 as needed.
Particularly, in accordance with an embodiment of the present disclosure, it may be implemented as computer above with reference to the process of flow chart description
Software program.For example, embodiment of the disclosure includes a kind of computer program product comprising be carried on computer-readable medium
On computer program, which includes the program code for method shown in execution flow chart.In such reality
It applies in example, which can be downloaded and installed by communications portion 509 from network, and/or from detachable media
511 are mounted.When the computer program is executed by central processing unit (CPU) 501, limited in execution the present processes
Above-mentioned function.It should be noted that the computer-readable medium of the application can be computer-readable signal media or calculating
Machine readable storage medium storing program for executing either the two arbitrarily combines.Computer readable storage medium for example can be --- but it is unlimited
In --- electricity, system, device or the device of magnetic, optical, electromagnetic, infrared ray or semiconductor, or the arbitrary above combination.It calculates
The more specific example of machine readable storage medium storing program for executing can include but is not limited to：Being electrically connected, be portable with one or more conducting wires
Formula computer disk, hard disk, random access storage device (RAM), read-only memory (ROM), erasable programmable read only memory
(EPROM or flash memory), optical fiber, portable compact disc read-only memory (CD-ROM), light storage device, magnetic memory device or
The above-mentioned any appropriate combination of person.In this application, can be any include computer readable storage medium or storage program
Tangible medium, the program can be commanded execution system, device either device use or it is in connection.And in this Shen
Please in, computer-readable signal media may include in a base band or as the data-signal that a carrier wave part is propagated,
In carry computer-readable program code.Diversified forms may be used in the data-signal of this propagation, including but not limited to
Electromagnetic signal, optical signal or above-mentioned any appropriate combination.Computer-readable signal media can also be computer-readable
Any computer-readable medium other than storage medium, the computer-readable medium can send, propagate or transmit for by
Instruction execution system, device either device use or program in connection.The journey for including on computer-readable medium
Sequence code can transmit with any suitable medium, including but not limited to：Wirelessly, electric wire, optical cable, RF etc. or above-mentioned
Any appropriate combination.
The calculating of the operation for executing the application can be write with one or more programming languages or combinations thereof
Machine program code, programming language include object oriented program language-such as Java, Smalltalk, C++, also
Such as including conventional procedural programming language-" C " language or similar programming language.Program code can be complete
It executes, partly executed on the user computer on the user computer entirely, being executed as an independent software package, part
Part executes or executes on a remote computer or server completely on the remote computer on the user computer.It is relating to
And in the situation of remote computer, remote computer can pass through the network of any kind --- including LAN (LAN) or extensively
Domain net (WAN)-be connected to subscriber computer, or, it may be connected to outer computer (such as provided using Internet service
Quotient is connected by internet).
Flow chart in attached drawing and block diagram, it is illustrated that according to the system of the various embodiments of the application, method and computer journey
The architecture, function and operation in the cards of sequence product.In this regard, each box in flowchart or block diagram can generation
A part for a part for one module, program segment, or code of table, the module, program segment, or code includes one or more uses
The executable instruction of the logic function as defined in realization.It should also be noted that in some implementations as replacements, being marked in box
The function of note can also occur in a different order than that indicated in the drawings.For example, two boxes succeedingly indicated are actually
It can be basically executed in parallel, they can also be executed in the opposite order sometimes, this is depended on the functions involved.Also it to note
Meaning, the combination of each box in block diagram and or flow chart and the box in block diagram and or flow chart can be with holding
The dedicated hardware based system of functions or operations as defined in row is realized, or can use specialized hardware and computer instruction
Combination realize.
Being described in unit involved in the embodiment of the present application can be realized by way of software, can also be by hard
The mode of part is realized.Described unit can also be arranged in the processor, for example, can be described as：A kind of processor packet
Include extraction unit, generation unit and evaluation unit.Wherein, the title of these units is not constituted under certain conditions to the unit
The restriction of itself, for example, extraction unit is also described as, " each image in the target image cluster obtained to cluster carries out
The unit of feature extraction ".
As on the other hand, present invention also provides a kind of computer-readable medium, which can be
Included in device described in above-described embodiment；Can also be individualism, and without be incorporated the device in.Above-mentioned calculating
Machine readable medium carries one or more program, when said one or multiple programs are executed by the device so that should
Device：Each image in the target image cluster obtained to cluster carries out feature extraction；Target is generated based on the feature extracted
The corresponding Eigen Covariance matrix of image cluster；Covariance matrix is inputted in cluster result evaluation model, target image cluster is obtained
Cluster result evaluation information；Wherein, sample image of the cluster result evaluation model based on marked cluster result evaluation information
Cluster training obtains.
Above description is only the preferred embodiment of the application and the explanation to institute's application technology principle.People in the art
Member should be appreciated that invention scope involved in the application, however it is not limited to technology made of the specific combination of above-mentioned technical characteristic
Scheme, while should also cover in the case where not departing from foregoing invention design, it is carried out by above-mentioned technical characteristic or its equivalent feature
Other technical solutions of arbitrary combination and formation.Such as features described above has similar work(with (but not limited to) disclosed herein
Can technical characteristic replaced mutually and the technical solution that is formed.
Claims (12)
1. a kind of image clustering evaluation of result method, including：
Each image in the target image cluster obtained to cluster carries out feature extraction；
The corresponding Eigen Covariance matrix of the target image cluster is generated based on the feature extracted；
By in the Eigen Covariance Input matrix cluster result evaluation model, the cluster result for obtaining the target image cluster is commented
Valence information；
Wherein, sample image cluster training of the cluster result evaluation model based on marked cluster result evaluation information obtains.
2. according to the method described in claim 1, wherein, the cluster result evaluation model is trained as follows obtains
's：
Obtain the label result of the cluster result evaluation information of multiple sample image clusters and each sample image cluster；
Feature extraction is carried out to each sample image cluster；
The sample characteristics covariance matrix of each sample image cluster is generated based on the feature extracted；
Using deep learning method, using the sample characteristics covariance matrix as the corresponding god of the cluster result evaluation model
Input through network, the label result of the cluster result evaluation information based on the sample image cluster and preset loss function instruction
Get the cluster result evaluation model, wherein cluster knot of the preset loss function based on the sample image cluster
The sample image cluster of the label result of fruit evaluation information neural network output corresponding with the cluster result evaluation model
Cluster result evaluation information prediction result between difference structure.
3. it is described to use deep learning method according to the method described in claim 2, wherein, by the sample characteristics covariance
Input of the matrix as the corresponding neural network of the cluster result evaluation model, the cluster result based on the sample image cluster
The label result of evaluation information and preset loss function train to obtain the cluster result evaluation model, including：
It determines the initial value of the parameter of the corresponding neural network of the cluster result evaluation model, executes and compare step；
The comparison step includes：The sample characteristics covariance matrix is inputted into the corresponding god of the cluster result evaluation model
Through network, the prediction result of the cluster result evaluation information of corresponding sample image cluster is obtained, judges the value of the loss function
Whether the preset condition of convergence is met；
If the judging result for comparing step is no, it is based on the loss function, the cluster is updated using gradient descent method
The corresponding parameter of the corresponding neural network of evaluation of result model, and execute the comparison step；
If the judging result for comparing step is yes, the ginseng of the corresponding neural network of the cluster result evaluation model is exported
Number.
4. according to the method described in claim 1, wherein, the described pair of each image clustered in obtained target image cluster carries out
Feature extraction, including：
Feature extraction is carried out to each image in the target image cluster using convolutional neural networks, obtains the target image
The feature set of graphs of each image in cluster；
It is described that the corresponding Eigen Covariance matrix of the target image cluster is generated based on the feature extracted, including：
Feature set of graphs based on each image in the target image cluster builds the eigenmatrix of the target image cluster；
Calculate the covariance matrix of the eigenmatrix of the target image cluster.
5. according to claim 1-4 any one of them methods, wherein the method further includes：
According to the reservation of the cluster result evaluation information of the target image cluster or the/deletion target image cluster.
6. a kind of image clustering evaluation of result device, including：
Extraction unit, each image in target image cluster for being obtained to cluster carry out feature extraction；
Generation unit, for generating the corresponding Eigen Covariance matrix of the target image cluster based on the feature extracted；
Evaluation unit, for by the Eigen Covariance Input matrix cluster result evaluation model, obtaining the target image
The cluster result evaluation information of cluster；
Wherein, sample image cluster training of the cluster result evaluation model based on marked cluster result evaluation information obtains.
7. device according to claim 6, wherein the cluster result evaluation model is trained as follows obtains
's：
Obtain the label result of the cluster result evaluation information of multiple sample image clusters and each sample image cluster；
Feature extraction is carried out to each sample image cluster；
The sample characteristics covariance matrix of each sample image cluster is generated based on the feature extracted；
Using deep learning method, using the sample characteristics covariance matrix as the corresponding god of the cluster result evaluation model
Input through network, the label result of the cluster result evaluation information based on the sample image cluster and preset loss function instruction
Get the cluster result evaluation model, wherein cluster knot of the preset loss function based on the sample image cluster
The sample image cluster of the label result of fruit evaluation information neural network output corresponding with the cluster result evaluation model
Cluster result evaluation information prediction result between difference structure.
8. device according to claim 7, wherein the cluster result evaluation model is to use deep learning method, will
Input of the sample characteristics covariance matrix as the corresponding neural network of the cluster result evaluation model is based on the sample
The label result of the cluster result evaluation information of this image cluster and preset loss function, trained further according to such as under type
It arrives：
It determines the initial value of the parameter of the corresponding neural network of the cluster result evaluation model, executes and compare step；
The comparison step includes：The sample characteristics covariance matrix is inputted into the corresponding god of the cluster result evaluation model
Through network, the prediction result of the cluster result evaluation information of corresponding sample image cluster is obtained, judges the value of the loss function
Whether the preset condition of convergence is met；
If the judging result for comparing step is no, it is based on the loss function, the cluster is updated using gradient descent method
The corresponding parameter of the corresponding neural network of evaluation of result model, and execute the comparison step；
If the judging result for comparing step is yes, the ginseng of the corresponding neural network of the cluster result evaluation model is exported
Number.
9. device according to claim 6, wherein the extraction unit is further used for as follows to clustering
To target image cluster in each image carry out feature extraction：
Feature extraction is carried out to each image in the target image cluster using convolutional neural networks, obtains the target image
The feature set of graphs of each image in cluster；
Above-mentioned generation unit is further used for generating the corresponding Eigen Covariance matrix of the target image cluster as follows：
Feature set of graphs based on each image in the target image cluster builds the eigenmatrix of the target image cluster；
Calculate the covariance matrix of the eigenmatrix of the target image cluster.
10. according to claim 6-9 any one of them devices, wherein described device further includes：
Processing unit, for according to the reservation of the cluster result evaluation information of the target image cluster or the/deletion target image
Cluster.
11. a kind of electronic equipment, including：
One or more processors；
Storage device, for storing one or more programs,
When one or more of programs are executed by one or more of processors so that one or more of processors are real
The now method as described in any in claim 1-5.
12. a kind of computer readable storage medium, is stored thereon with computer program, wherein described program is executed by processor
Methods of the Shi Shixian as described in any in claim 1-5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810007796.1A CN108564102A (en) | 2018-01-04 | 2018-01-04 | Image clustering evaluation of result method and apparatus |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810007796.1A CN108564102A (en) | 2018-01-04 | 2018-01-04 | Image clustering evaluation of result method and apparatus |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108564102A true CN108564102A (en) | 2018-09-21 |
Family
ID=63530706
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810007796.1A Pending CN108564102A (en) | 2018-01-04 | 2018-01-04 | Image clustering evaluation of result method and apparatus |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108564102A (en) |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109615075A (en) * | 2018-12-14 | 2019-04-12 | 大连海事大学 | A kind of resident's daily behavior recognition methods based on multi-level clustering model |
| CN110020022A (en) * | 2019-01-03 | 2019-07-16 | 阿里巴巴集团控股有限公司 | Data processing method, device, equipment and readable storage medium storing program for executing |
| CN110222590A (en) * | 2019-05-15 | 2019-09-10 | 北京字节跳动网络技术有限公司 | Image difference judgment method, device and electronic equipment |
| CN110400138A (en) * | 2019-07-30 | 2019-11-01 | 中国工商银行股份有限公司 | The method of mobile payment and device of illegal |
| CN110580510A (en) * | 2019-09-12 | 2019-12-17 | 深圳力维智联技术有限公司 | clustering result evaluation method and system |
| CN110826616A (en) * | 2019-10-31 | 2020-02-21 | Oppo广东移动通信有限公司 | Information processing method and device, electronic equipment and storage medium |
| CN111738319A (en) * | 2020-06-11 | 2020-10-02 | 佳都新太科技股份有限公司 | Clustering result evaluation method and device based on large-scale samples |
| CN112155729A (en) * | 2020-10-15 | 2021-01-01 | 中国科学院合肥物质科学研究院 | Intelligent automatic planning method and system for surgical puncture path and medical system |
| CN112232384A (en) * | 2020-09-27 | 2021-01-15 | 北京迈格威科技有限公司 | Model training method, image feature extraction method, target detection method and device |
| CN112906824A (en) * | 2021-03-29 | 2021-06-04 | 苏州科达科技股份有限公司 | Vehicle clustering method, system, device and storage medium |
| CN113011221A (en) * | 2019-12-19 | 2021-06-22 | 广州极飞科技股份有限公司 | Crop distribution information acquisition method and device and measurement system |
| WO2022121801A1 (en) * | 2020-12-07 | 2022-06-16 | 北京有竹居网络技术有限公司 | Information processing method and apparatus, and electronic device |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101334893A (en) * | 2008-08-01 | 2008-12-31 | 天津大学 | Fused image quality integrated evaluating method based on fuzzy neural network |
| US20140201208A1 (en) * | 2013-01-15 | 2014-07-17 | Corporation Symantec | Classifying Samples Using Clustering |
| CN106850333A (en) * | 2016-12-23 | 2017-06-13 | 中国科学院信息工程研究所 | A kind of network equipment recognition methods and system based on feedback cluster |
| CN106845331A (en) * | 2016-11-18 | 2017-06-13 | 深圳云天励飞技术有限公司 | A kind of image processing method and terminal |
-
2018
- 2018-01-04 CN CN201810007796.1A patent/CN108564102A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101334893A (en) * | 2008-08-01 | 2008-12-31 | 天津大学 | Fused image quality integrated evaluating method based on fuzzy neural network |
| US20140201208A1 (en) * | 2013-01-15 | 2014-07-17 | Corporation Symantec | Classifying Samples Using Clustering |
| CN106845331A (en) * | 2016-11-18 | 2017-06-13 | 深圳云天励飞技术有限公司 | A kind of image processing method and terminal |
| CN106850333A (en) * | 2016-12-23 | 2017-06-13 | 中国科学院信息工程研究所 | A kind of network equipment recognition methods and system based on feedback cluster |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109615075B (en) * | 2018-12-14 | 2022-08-19 | 大连海事大学 | Resident daily behavior identification method based on multilayer clustering model |
| CN109615075A (en) * | 2018-12-14 | 2019-04-12 | 大连海事大学 | A kind of resident's daily behavior recognition methods based on multi-level clustering model |
| CN110020022B (en) * | 2019-01-03 | 2023-08-01 | 创新先进技术有限公司 | Data processing method, device, equipment and readable storage medium |
| CN110020022A (en) * | 2019-01-03 | 2019-07-16 | 阿里巴巴集团控股有限公司 | Data processing method, device, equipment and readable storage medium storing program for executing |
| CN110222590A (en) * | 2019-05-15 | 2019-09-10 | 北京字节跳动网络技术有限公司 | Image difference judgment method, device and electronic equipment |
| CN110222590B (en) * | 2019-05-15 | 2021-05-25 | 北京字节跳动网络技术有限公司 | Image difference judgment method and device and electronic equipment |
| CN110400138A (en) * | 2019-07-30 | 2019-11-01 | 中国工商银行股份有限公司 | The method of mobile payment and device of illegal |
| CN110580510A (en) * | 2019-09-12 | 2019-12-17 | 深圳力维智联技术有限公司 | clustering result evaluation method and system |
| CN110580510B (en) * | 2019-09-12 | 2023-07-25 | 深圳力维智联技术有限公司 | Clustering result evaluation method and system |
| CN110826616A (en) * | 2019-10-31 | 2020-02-21 | Oppo广东移动通信有限公司 | Information processing method and device, electronic equipment and storage medium |
| CN110826616B (en) * | 2019-10-31 | 2023-06-30 | Oppo广东移动通信有限公司 | Information processing method and device, electronic equipment and storage medium |
| CN113011221A (en) * | 2019-12-19 | 2021-06-22 | 广州极飞科技股份有限公司 | Crop distribution information acquisition method and device and measurement system |
| CN111738319A (en) * | 2020-06-11 | 2020-10-02 | 佳都新太科技股份有限公司 | Clustering result evaluation method and device based on large-scale samples |
| CN112232384A (en) * | 2020-09-27 | 2021-01-15 | 北京迈格威科技有限公司 | Model training method, image feature extraction method, target detection method and device |
| CN112155729A (en) * | 2020-10-15 | 2021-01-01 | 中国科学院合肥物质科学研究院 | Intelligent automatic planning method and system for surgical puncture path and medical system |
| WO2022121801A1 (en) * | 2020-12-07 | 2022-06-16 | 北京有竹居网络技术有限公司 | Information processing method and apparatus, and electronic device |
| CN112906824B (en) * | 2021-03-29 | 2022-07-05 | 苏州科达科技股份有限公司 | Vehicle clustering method, system, device and storage medium |
| CN112906824A (en) * | 2021-03-29 | 2021-06-04 | 苏州科达科技股份有限公司 | Vehicle clustering method, system, device and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108564102A (en) | Image clustering evaluation of result method and apparatus | |
| US11195007B2 (en) | Classification of piping and instrumental diagram information using machine-learning | |
| US10762376B2 (en) | Method and apparatus for detecting text | |
| CN110362677B (en) | Text data category identification method and device, storage medium and computer equipment | |
| CN108280477B (en) | Method and apparatus for clustering images | |
| JP6209879B2 (en) | Convolutional neural network classifier system, training method, classification method and use thereof | |
| CN111461238B (en) | Model training method, character recognition method, device, equipment and storage medium | |
| CN108197592A (en) | Information acquisition method and device | |
| CN110222582B (en) | Image processing method and camera | |
| CN109145828A (en) | Method and apparatus for generating video classification detection model | |
| CN108256476A (en) | For identifying the method and apparatus of fruits and vegetables | |
| CN109711473A (en) | Item identification method, equipment and system | |
| CN111241992B (en) | Face recognition model construction method, recognition method, device, equipment and storage medium | |
| CN105894028A (en) | User identification method and device | |
| CN110457677A (en) | Entity-relationship recognition method and device, storage medium, computer equipment | |
| CN109871749A (en) | A kind of pedestrian based on depth Hash recognition methods and device, computer system again | |
| CN110781818B (en) | Video classification method, model training method, device and equipment | |
| CN108229680B (en) | Neural network system, remote sensing image recognition method, device, equipment and medium | |
| CN112699945A (en) | Data labeling method and device, storage medium and electronic device | |
| CN114170484B (en) | Picture attribute prediction method and device, electronic equipment and storage medium | |
| CN113140012B (en) | Image processing method, device, medium and electronic equipment | |
| CN112115996B (en) | Image data processing method, device, equipment and storage medium | |
| CN113657248A (en) | Training method and device for face recognition model and computer program product | |
| CN113627334A (en) | Object behavior identification method and device | |
| CN107948721A (en) | The method and apparatus of pushed information |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180921 |
|
| RJ01 | Rejection of invention patent application after publication |