CN105631404B - The method and device that photo is clustered - Google Patents
The method and device that photo is clustered Download PDFInfo
- Publication number
- CN105631404B CN105631404B CN201510955515.1A CN201510955515A CN105631404B CN 105631404 B CN105631404 B CN 105631404B CN 201510955515 A CN201510955515 A CN 201510955515A CN 105631404 B CN105631404 B CN 105631404B
- Authority
- CN
- China
- Prior art keywords
- face
- photograph album
- characteristic
- photo
- cloud
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/172—Classification, e.g. identification
-
- 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
- G06N3/088—Non-supervised learning, e.g. competitive learning
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Biomedical Technology (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The disclosure is directed to the method and devices that a kind of pair of photo is clustered.Method includes：Determine that number of pictures is lower than at least one sub- photograph album of the second face and corresponding first face characteristic of preset threshold from the first face photograph album of at least one of the first cloud photograph album；Determine that second user mark stores at least one third party's face photograph album and at least one corresponding second face characteristic of third party's face photograph album in the second cloud photograph album on Cloud Server；The similarity value for determining at least one corresponding first face characteristic of the sub- photograph album of the second face and at least one corresponding face characteristic of third party's face photograph album, obtains multiple similarity values；Human face photo identical at least one the second face corresponding face characteristic of sub- photograph album is determined from least one third party's face photograph album according to multiple similarity values.The disclosure can be such that the number of pictures in the first cloud photograph album is supplemented lower than the diversity of the photo in the sub- photograph album of face of preset threshold.
Description
Technical field
This disclosure relates to the method and device that image identification technical field more particularly to a kind of pair of photo are clustered.
Background technique
During being classified as unit of face to user's photograph album, it is necessary first to obtain the complete of user mobile phone upload
Portion's photo, to whole photos carry out Face datection extract face characteristic, by the face characteristic of extraction successively with classified people
Face carries out similarity measurement, and the photo with similarity feature is divided into the same face photograph album.Due in user's photograph album
Number of pictures it is limited, therefore face characteristic can be because the factors such as illumination, expression generate certain difference, and then cause using people
The knowledge of face feature cannot correctly cluster photo in the photograph album of the same face otherwise, to influence the recall rate of photograph album.
Summary of the invention
To overcome the problems in correlation technique, the embodiment of the present disclosure provides the side that a kind of pair of photo carries out classification storage
Method and device, to improve the recall rate of face photograph album.
According to the first aspect of the embodiments of the present disclosure, the method that a kind of pair of photo carries out classification storage is provided, including：
Determine from the first face photograph album of at least one of the first cloud photograph album number of pictures lower than preset threshold to
Few sub- photograph album of second face and at least one described corresponding first face characteristic of the sub- photograph album of the second face, it is described
First cloud photograph album corresponds to the first user identifier；
Determine that second user mark associated with first user identifier stores the second cloud on the cloud server
At least one third party's face photograph album and corresponding second people of at least one described third party's face photograph album in photograph album
Face feature；
Determine at least one described corresponding first face characteristic of the sub- photograph album of the second face and it is described at least one the
The similarity value of the corresponding face characteristic of the sub- photograph album of three faces, obtains multiple similarity values；
According to the multiple similarity value from least one described third party's face photograph album it is determining with it is described at least one
The identical human face photo of the corresponding face characteristic of the sub- photograph album of second face.
In one embodiment, the method may also include：
Determine communication list associated with first user identifier；
It searches in the communication list on the cloud server with the user identifier of cloud photograph album, by the communication list
In the user identifier with cloud photograph album be determined as second user mark associated with first user identifier.
In one embodiment, the method may also include：
Feature extraction is carried out to whole photos in the first cloud photograph album by the convolutional neural networks trained, is obtained at least
One third face characteristic；
It is described at least one that at least one described corresponding photo of third face characteristic is clustered as unit of face characteristic
A the first face photograph album.
In one embodiment, the method may also include：
Feature extraction is carried out to whole photos in the second cloud photograph album by the convolutional neural networks trained, is obtained
At least one the 4th face characteristic；
It is described at least one that at least one described corresponding photo of the 4th face characteristic is clustered as unit of face characteristic
A third party's face photograph album.
In one embodiment, the method may also include：
There is label face sample to be input to convolutional neural networks for set quantity, at least to the convolutional neural networks
One convolutional layer and at least one full articulamentum are trained；
When the optimal weight parameter of the connection in determining the convolutional neural networks between each node, described in deconditioning
Convolutional neural networks, the convolutional neural networks trained.
In one embodiment, the method may also include：
Determine the characteristic point in each photo in the first cloud photograph album about face；
The region of the face is determined from each photo according to the characteristic point of the face；
Picture material in the region of the face is subjected to affine transformation according to preset reference characteristic point and obtains face figure
Picture, the resolution ratio of the facial image are identical as the dimension of input layer of the convolutional neural networks trained；
Extract the face characteristic in the facial image by the convolutional neural networks trained, obtain it is described at least
One the first corresponding face characteristic of face photograph album.
In one embodiment, the method may also include：
Determine the characteristic point in each photo in the second cloud photograph album about face；
The region of the face is determined from each photo according to the characteristic point of the face；
Picture material in the region of the face is subjected to affine transformation according to preset reference characteristic point and obtains face figure
Picture, the resolution ratio of the facial image are identical as the dimension of input layer of the convolutional neural networks trained；
Extract the face characteristic in the facial image by the convolutional neural networks trained, obtain it is described at least
One corresponding second face characteristic of third party's face photograph album.
According to the second aspect of an embodiment of the present disclosure, the device that a kind of pair of photo is clustered is provided, including：
First determining module is configured as determining photo from the first face photograph album of at least one of the first cloud photograph album
Quantity is respectively right lower than at least one sub- photograph album of the second face of preset threshold and at least one described sub- photograph album of the second face
The first face characteristic answered, corresponding first user identifier of the first cloud photograph album；
Second determining module is configured to determine that second user mark associated with first user identifier described
At least one third party's face photograph album in the second cloud photograph album and at least one described third party's face are stored on Cloud Server
Corresponding second face characteristic of photograph album；
Third determining module is configured to determine that at least one described second face that first determining module determines
At least one described third party's face photograph album that corresponding first face characteristic of photograph album and second determining module determine
The similarity value of corresponding face characteristic obtains multiple similarity values；
4th determining module, the multiple similarity value for being configured as being determined according to the third determining module is from described
It is determined and at least one described corresponding face characteristic phase of the sub- photograph album of the second face at least one third party's face photograph album
Same human face photo.
In one embodiment, described device may also include：
5th determining module is configured to determine that communication list associated with first user identifier；
Searching module is configured as searching in the communication list that the 5th determining module determines in the cloud service
With the user identifier of cloud photograph album on device, the user identifier with cloud photograph album in the communication list is determined as and described first
The associated second user mark of user identifier.
In one embodiment, described device may also include：
First extraction module is configured as the convolutional neural networks by having trained to whole photos in the first cloud photograph album
Feature extraction is carried out, at least one third face characteristic is obtained；
First cluster module, be configured as extracting first extraction module described at least one third face it is special
Corresponding photo is levied to be clustered as unit of face characteristic as at least one described the first face photograph album.
In one embodiment, described device may also include：
Second extraction module is configured as the convolutional neural networks by having trained to the whole in the second cloud photograph album
Photo carries out feature extraction, obtains at least one the 4th face characteristic；
Second cluster module, be configured as extracting second extraction module described at least one the 4th face it is special
Corresponding photo is levied to be clustered as unit of face characteristic as at least one described third party's face photograph album.
In one embodiment, described device may also include：
First training module, be configured as setting quantity has label face sample to be input to convolutional neural networks, right
At least one convolutional layer of the convolutional neural networks and at least one full articulamentum are trained；
First control module is configured as the best weights of the connection in determining the convolutional neural networks between each node
When weight parameter, convolutional neural networks described in the first training module deconditioning, the convolutional Neural net trained are controlled
Network.
In one embodiment, described device may also include：
6th determining module is configured to determine that the feature in each photo in the first cloud photograph album about face
Point；
7th determining module is configured as the characteristic point of the face determined according to the 6th determining module from described
The region of the face is determined on each photo；
First conversion module is configured as in the image in the region for the face for determining the 7th determining module
Hold and facial image is obtained according to preset reference characteristic point progress affine transformation, the resolution ratio of the facial image has been trained with described
Convolutional neural networks input layer dimension it is identical；
Third extraction module is configured as extracting institute by the convolutional neural networks that first training module training obtains
The face characteristic in facial image is stated, at least one described the first corresponding face characteristic of face photograph album is obtained.
In one embodiment, described device may also include：
8th determining module is configured to determine that the feature in each photo in the second cloud photograph album about face
Point；
9th determining module is configured as the characteristic point of the face determined according to the 8th determining module from described
The region of the face is determined on each photo；
Second conversion module is configured as in the image in the region for the face for determining the 9th determining module
Hold and facial image is obtained according to preset reference characteristic point progress affine transformation, the resolution ratio of the facial image has been trained with described
Convolutional neural networks input layer dimension it is identical；
4th extraction module is configured as extracting by the convolutional neural networks that first training module training obtains
It is special to obtain corresponding second face of at least one third party's face photograph album for face characteristic in the facial image
Sign.
According to the third aspect of an embodiment of the present disclosure, the device that a kind of pair of photo is clustered is provided, including：
Processor；
Memory for storage processor executable instruction；
Wherein, the processor is configured to：
Determine from the first face photograph album of at least one of the first cloud photograph album number of pictures lower than preset threshold to
Few sub- photograph album of second face and at least one described corresponding first face characteristic of the sub- photograph album of the second face, it is described
First cloud photograph album corresponds to the first user identifier；
Determine that second user mark associated with first user identifier stores the second cloud on the cloud server
At least one third party's face photograph album and corresponding second people of at least one described third party's face photograph album in photograph album
Face feature；
Determine at least one described corresponding first face characteristic of the sub- photograph album of the second face and it is described at least one the
The similarity value of the corresponding face characteristic of the sub- photograph album of three faces, obtains multiple similarity values；
According to the multiple similarity value from least one described third party's face photograph album it is determining with it is described at least one
The identical human face photo of the corresponding face characteristic of the sub- photograph album of second face.
The technical scheme provided by this disclosed embodiment can include the following benefits：By from the first user identifier
The determining face with the first user identifier in the second cloud photograph album that associated second user mark stores on Cloud Server
The identical human face photo of the corresponding face characteristic of sub- photograph album, so as to so that photo in the first cloud photograph album of the first user identifier
Quantity is supplemented lower than the diversity of the photo in the sub- photograph album of face of preset threshold, when multifarious photo is enriched
Afterwards, the corresponding face characteristic of the sub- photograph album of the face that number of pictures can be enable less more accurately characterizes the face, Jin Er great
Recall rate of the photo at least one sub- photograph album of the second face during clustering recognition is improved greatly.
It should be understood that above general description and following detailed description be only it is exemplary and explanatory, not
The disclosure can be limited.
Detailed description of the invention
The drawings herein are incorporated into the specification and forms part of this specification, and shows and meets implementation of the invention
Example, and be used to explain the principle of the present invention together with specification.
Figure 1A is the flow chart of the method shown according to an exemplary embodiment clustered to photo.
Figure 1B is the schematic diagram of the first cloud photograph album shown according to an exemplary embodiment Yu the first face photograph album.
Fig. 2 is the flow chart according to the method clustered to photo shown in an exemplary embodiment one.
Fig. 3 A is the schematic diagram how to be trained to convolutional neural networks shown according to an exemplary embodiment two.
Fig. 3 B is the structural schematic diagram according to the convolutional neural networks shown in an exemplary embodiment two.
Fig. 4 is the flow chart according to the method clustered to photo shown in an exemplary embodiment three.
Fig. 5 is the flow chart according to the method clustered to photo shown in an exemplary embodiment four.
Fig. 6 is the block diagram for the device that a kind of pair of photo shown according to an exemplary embodiment is clustered.
Fig. 7 is the block diagram for the device that another kind shown according to an exemplary embodiment clusters photo.
Fig. 8 is the block diagram of another device clustered to photo shown according to an exemplary embodiment.
Fig. 9 is shown according to an exemplary embodiment a kind of suitable for the block diagram of the device clustered to photo.
Specific embodiment
Example embodiments are described in detail here, and the example is illustrated in the accompanying drawings.Following description is related to
When attached drawing, unless otherwise indicated, the same numbers in different drawings indicate the same or similar elements.Following exemplary embodiment
Described in embodiment do not represent all embodiments consistented with the present invention.On the contrary, they be only with it is such as appended
The example of device and method being described in detail in claims, some aspects of the invention are consistent.
Figure 1A is the flow chart of the method shown according to an exemplary embodiment clustered to photo, and Figure 1B is basis
The schematic diagram of the first cloud photograph album and the first face photograph album shown in one exemplary embodiment；The method that photo is clustered
Can apply on Cloud Server, Cloud Server to user upload onto the server on cloud photograph album obtained by disclosed method
The sub- photograph album of face as unit of face, as shown in Figure 1A, which includes the following steps S101-
S104：
In step s101, determine that number of pictures is lower than from the first face photograph album of at least one of the first cloud photograph album
At least one sub- photograph album of the second face of preset threshold and at least one corresponding first face of the sub- photograph album of the second face
Feature, corresponding first user identifier of the first cloud photograph album.
In step s 102, it determines that associated with the first user identifier second user is identified and stores the on Cloud Server
At least one third party's face photograph album and corresponding second people of at least one third party's face photograph album in two cloud photograph albums
Face feature.
In step s 103, at least one corresponding first face characteristic of the sub- photograph album of the second face and at least one is determined
The similarity value of a corresponding face characteristic of third party's face photograph album, obtains multiple similarity values.
In step S104, according to multiple similarity values from least one third party's face photograph album determine and at least one
The identical human face photo of the corresponding face characteristic of the sub- photograph album of second face.
In an exemplary scene, carried out so that the first user identifier is user A and second user is identified as user B as an example
Exemplary illustration, when user A is uploaded to together in its photo for storing (smart phone or tablet computer lamp) on an electronic device
After on Cloud Server, the photo of user A forms the first cloud photograph album 10 shown in Figure 1B on Cloud Server, and Cloud Server can be right
The photo of user A is clustered as unit of face, has obtained 8 people's face photograph albums shown in Figure 1B (i.e. described in the disclosure
At least one the first face photograph album), contain the sub- photograph album of the respective face of user A- user H in the first cloud photograph album 10,
In, user A11 indicates that the sub- photograph album of face of human face photo of the storage comprising user A, user B12 indicate that storage includes user B
Face photo the sub- photograph album of face, user C13 indicate storage comprising user C human face photo the sub- photograph album of face, user
D14 indicates that storage includes sub- photograph album of face of photo of face of user D, etc., it is possible thereby to by the same cloud photograph album
Photo stored as unit of face.It in one embodiment, can in order to which the identification for improving the sub- photograph album of face clusters recall rate
To carry out clustering processing to the face characteristic calculated by rank-order clustering algorithm, user A corresponding at least one is obtained
A the first face photograph album and at least one corresponding third party's face photograph album of user B.
In 8 people's face photograph albums, if user B and the sub- photograph album of the corresponding face of user C are (described in the disclosure at least
One sub- photograph album of the second face) in number of pictures (preset threshold can be set by experience lower than a preset threshold
It is fixed), user B and the corresponding face characteristic of the sub- photograph album of the corresponding face of user C cause to pass through phase since number of pictures is less at this time
Clustering algorithm in the technology of pass is difficult to flock together in the photo of user B and user C, causes user B and the corresponding people of user C
Face photograph album is more unused.For example, the photo that 200 include the face of user A is stored in the sub- photograph album of the face of user A,
The photo that 20 include the face of user B is stored in the sub- photograph album of the face of user B, is stored in the sub- photograph album of the face of user C
10 include user C face photo, store in the face of user D photograph album 100 include user D face
Photo, if defaulting in will big 30, the number of pictures in user B and the sub- photograph album of the corresponding face of user C be lower than 20.
In order to improve the accuracy of user B and the corresponding face characteristic of user C, need to increase user B's and user C
For the photo of recognition of face, at this point it is possible to be uploaded to the address list of Cloud Server by user A to search user B and user C
It, can be if determining that associated with user A second user is identified as user B after searching with the presence or absence of cloud photograph album
Determine that the second cloud photograph album of user B storage, Cloud Server can use and the first cloud the second cloud photograph album on Cloud Server
The identical method of photograph album carries out face cluster, and obtaining at least one third party's face photograph album (may include user A, user B, uses
Family M, user N etc.) and at least one corresponding second face characteristic of third party's face photograph album.
Determine corresponding first face characteristic of the sub- photograph album of the face of user B and the user A in the second cloud photograph album, user B, use
The similarity value (for example, respectively s1, s2, s3, s4) of family M, the corresponding face characteristic of user N, from the s1, s2, s3, s4
In determine the face characteristic most like with the face characteristic of the user B in the first cloud photograph album, for example, s2 be four similarity values
In maximum one, then the photo in the sub- photograph album of the corresponding face of s2 can be determined as with the user B's in the first cloud photograph album
The identical human face photo of the corresponding face characteristic of the sub- photograph album of face, at this time can be by face of the user B in the second cloud photograph album
The sub- photograph album of face of photograph album and the user B in the first cloud photograph album flocks together.Number of pictures in first cloud photograph album is lower than pre-
If details are not described herein if the sub- photograph album of other faces of threshold value improves the same foregoing description of the multifarious mode of photo.
In the present embodiment, pass through what is stored on Cloud Server from second user associated with the first user identifier mark
The identical human face photo of determination face characteristic corresponding with the sub- photograph album of face in the first user identifier in second cloud photograph album, thus
The number of pictures in the first cloud photograph album of the first user identifier can be made lower than the photo in the sub- photograph album of face of preset threshold
Diversity is supplemented, and after multifarious photo obtains abundant, the sub- photograph album of the face that number of pictures can be made less is corresponding
Face characteristic can more accurately characterize the face, and then greatly improve the photo in an at least sub- photograph album of the second face poly-
Recall rate in class identification process.
In one embodiment, method may also include：
Determine communication list associated with the first user identifier；
The user identifier on Cloud Server in communication list with cloud photograph album is searched, there will be cloud photograph album in communication list
User identifier be determined as associated with the first user identifier second user mark.
In one embodiment, method may also include：
Feature extraction is carried out to whole photos in the first cloud photograph album by the convolutional neural networks trained, is obtained at least
One third face characteristic；
At least one corresponding photo of third face characteristic is clustered as unit of face characteristic as at least one is the first
Face photograph album.
In one embodiment, method may also include：
Feature extraction is carried out to whole photos in the second cloud photograph album by the convolutional neural networks trained, is obtained at least
One the 4th face characteristic；
At least one corresponding photo of the 4th face characteristic is clustered as unit of face characteristic as at least one third party
Face photograph album.
In one embodiment, method may also include：
There is label face sample to be input to convolutional neural networks for set quantity, at least one of convolutional neural networks
Convolutional layer and at least one full articulamentum are trained；
When the optimal weight parameter of the connection in determining convolutional neural networks between each node, deconditioning convolutional Neural
Network, the convolutional neural networks trained.
In one embodiment, method may also include：
Determine the characteristic point in each photo in the first cloud photograph album about face；
The region of face is determined from each photo according to the characteristic point of face；
Picture material in the region of face is subjected to affine transformation according to preset reference characteristic point and obtains facial image, people
The resolution ratio of face image is identical as the dimension of the input layer for the convolutional neural networks trained；
The face characteristic in facial image is extracted by the convolutional neural networks trained, obtains at least one first face
Sub- corresponding first face characteristic of photograph album.
In one embodiment, method may also include：
Determine the characteristic point in each photo in the second cloud photograph album about face；
The region of face is determined from each photo according to the characteristic point of face；
Picture material in the region of face is subjected to affine transformation according to preset reference characteristic point and obtains facial image, people
The resolution ratio of face image is identical as the dimension of the input layer for the convolutional neural networks trained；
The face characteristic in facial image is extracted by the convolutional neural networks trained, obtains at least one third face
Sub- corresponding second face characteristic of photograph album.
Specifically how to realize and classification storage is carried out to photo, please refers to subsequent embodiment.
So far, the above method that the embodiment of the present disclosure provides, can make the photograph in the first cloud photograph album of the first user identifier
Piece quantity is supplemented lower than the diversity of the photo in the sub- photograph album of face of preset threshold, when multifarious photo is enriched
Afterwards, the corresponding face characteristic of the sub- photograph album of the face that number of pictures can be enable less more accurately characterizes the face, Jin Er great
Recall rate of the photo at least one sub- photograph album of the second face during clustering recognition is improved greatly.
The technical solution of embodiment of the present disclosure offer is provided below with specific embodiment.
Fig. 2 is the flow chart according to the method clustered to photo shown in an exemplary embodiment one；The present embodiment
The above method provided using the embodiment of the present disclosure is illustrated for how searching communication list, such as Fig. 2 institute
Show, includes the following steps：
In step s 201, determine that number of pictures is lower than from the first face photograph album of at least one of the first cloud photograph album
At least one sub- photograph album of the second face of preset threshold and at least one corresponding first face of the sub- photograph album of the second face
Feature, corresponding first user identifier of the first cloud photograph album.
The description of step S201 may refer to the associated description of above-mentioned Figure 1A embodiment, and this will not be detailed here.
In step S202, communication list associated with the first user identifier is determined.
In one embodiment, communication list can be the phone contact mode on the corresponding electronic equipment of the first user identifier
In address list or instant messenger in address list, above-mentioned address list can be uploaded to by cloud service by electronic equipment
Device, and bound with the user of the cloud photograph album on Cloud Server.
In step S203, the user identifier on Cloud Server in communication list with cloud photograph album is searched, name will be communicated
User identifier in list with cloud photograph album is determined as second user mark associated with the first user identifier.
For example, communication list is the address list in phone contact mode, it is corresponding in order to improve user B and user C
The accuracy of face characteristic needs to increase the photo for recognition of face of user B and user C, at this point it is possible to pass through user A
The address list of Cloud Server is uploaded to search user B and user C with the presence or absence of cloud photograph album, if determining and using after searching
A associated second user in family is identified as user B, then user B is confirmed as second user mark associated with user A.
In step S204, determines that associated with the first user identifier second user is identified and store the on Cloud Server
At least one third party's face photograph album and corresponding second people of at least one third party's face photograph album in two cloud photograph albums
Face feature.
In step S205, at least one corresponding first face characteristic of the sub- photograph album of the second face and at least one is determined
The similarity value of a corresponding face characteristic of third party's face photograph album, obtains multiple similarity values.
In step S206, according to multiple similarity values from least one third party's face photograph album determine and at least one
The identical human face photo of the corresponding face characteristic of the sub- photograph album of second face.
The description of step S204 to step S206 may refer to the associated description of above-mentioned Figure 1A embodiment, herein no longer in detail
It states.
The present embodiment on the basis of the advantageous effects with above-described embodiment, by with the first user identifier phase
The user identifier with cloud photograph album is searched in associated communication list, and the user identifier is associated with mark with first, from
And the number of pictures in the first cloud photograph album can be enriched lower than the face characteristic in the sub- photograph album of face of preset threshold, in face spy
When the diversity of sign must reache a certain level, the recall rate of photo can be greatly improved when carrying out clustering recognition again against piece.
Fig. 3 A is the schematic diagram how to be trained to convolutional neural networks shown according to an exemplary embodiment two, figure
3B is the structural schematic diagram according to the convolutional neural networks shown in an exemplary embodiment one；The present embodiment is implemented using the disclosure
The above method that example provides, by how exemplary by being carried out for thering is label face sample to be trained convolutional neural networks
Illustrate, as shown in Figure 3A, includes the following steps：
In step S301, there is label face sample to be input to convolutional neural networks for set quantity, to convolutional Neural
At least one convolutional layer of network and at least one full articulamentum are trained.
In step s 302, determine whether the connection in convolutional neural networks between each node reaches optimal weight parameter,
Between each node when being connected up to optimal weight parameter, execute step S302, connection between each node is not up to most
When good weight parameter, step S301 is continued to execute.
In step S303, deconditioning convolutional neural networks, the convolutional neural networks trained are controlled.
Before being trained to convolutional neural networks, need to prepare to set the face of quantity (can achieve ten thousand grades or more)
Sample (for example, 50000 face sample).Label (label) is carried out to the face sample of these magnanimity, for example, user E
The label of all face samples is all 1, and the label of all face samples of user F is all 2, etc., can prepare 1000 use
50000 facial images at family, the face sample of each user are 50, and the quantity of the face sample of magnanimity can achieve
50000, CNN is trained by this 50000 face sample.
The structure of CNN is referred to the signal of Fig. 3 B, as shown in Figure 3B, in the convolutional neural networks, including input layer,
First convolutional layer, the second convolutional layer, third convolutional layer, Volume Four lamination, the 5th convolutional layer, the first full articulamentum, second connect entirely
Connect layer and output layer.50000 above-mentioned face samples are input in the convolutional neural networks as training sample is instructed
The classification results practiced, and exported according to convolutional neural networks, constantly to the company between node in each base of the convolutional neural networks
The weight parameter connect is adjusted.During continuous adjustment, the convolutional neural networks the training sample based on input into
After row training, compared with the classification results that user demarcates, accuracy will be gradually increased the classification results of output.At the same time,
User can preset an accuracy threshold value, during continuous adjustment, if point of convolutional neural networks output
Class result is compared with the classification results that user demarcates, and after accuracy reaches pre-set accuracy threshold value, the convolution is refreshing at this time
Weight parameter through connecting between base level nodes each in network is optimal weight parameter, at this time it is considered that the convolutional Neural net
Network is trained to be finished.
In the present embodiment, by being trained to convolutional neural networks, there is mark to convolutional neural networks by classifier
Label face sample carries out classification calibration, the volume when there is the quantity of label face sample to reach certain rank, after can making training
Product neural network can recognize that the feature for being conducive to recognition of face in photo, it is ensured that the accuracy rate of recognition of face.
Fig. 4 is the flow chart according to the method clustered to photo shown in an exemplary embodiment three；The present embodiment
The above method provided using the embodiment of the present disclosure, is obtained at least with how to cluster in the first cloud photograph album of the first user identifier
It is illustrated for one the first face photograph album and in conjunction with Figure 1B, as shown in figure 4, including the following steps：
In step S401, feature is carried out to whole photos in the first cloud photograph album by the convolutional neural networks trained
It extracts, obtains at least one third face characteristic.
In step S402, at least one corresponding photo of third face characteristic is clustered as unit of face characteristic as extremely
Few people's face photograph album.
In one embodiment, it can be extracted in the first cloud photograph album 10 by the convolutional neural networks trained shown in Fig. 3 B
Whole photos third face characteristic, and whole photos are divided according to face characteristic, for example, special according to third face
Sign stores the photo of user A11 into the sub- photograph album of the corresponding face of user A11, by the photo storage of user A12 to user
In the sub- photograph album of the corresponding face of A12, etc., thus useful multiple sub- photograph albums of face in the first cloud photograph album 10.
Similar, Cloud Server can also extract the second cloud photograph album by the convolutional neural networks trained shown in Fig. 3 B
In whole photos the 4th face characteristic, and whole photos are divided according to the 4th face characteristic, for example, according to the 4th
Face characteristic stores the photo of user A into the sub- photograph album of the corresponding face of user A, by the photo storage of user B to user B
In the corresponding sub- photograph album of face, the photo of user M is stored into the sub- photograph album of the corresponding face of user M, the photo of user N is deposited
Store up in the sub- photograph album of the corresponding face of user N, etc..
The present embodiment on the basis of with above-described embodiment advantageous effects, by by the photo in cloud photograph album with people
Face is that unit is stored, and user is facilitated to arrange all photos in its cloud photograph album, improves the experience that user uses cloud photograph album.
Fig. 5 is the flow chart according to the method clustered to photo shown in an exemplary embodiment four；The present embodiment
The above method provided using the embodiment of the present disclosure, how to obtain at least one the first face photograph album corresponding first
It is illustrated for face characteristic, as shown in figure 5, including the following steps：
In step S501, the characteristic point in each photo in the first cloud photograph album about face is determined.
In step S502, the region of face is determined from each photo according to the characteristic point of face.
In one embodiment, each photograph in the first cloud photograph album can be obtained by human face detection tech in the related technology
About the characteristic point of face in piece, the area image of face is determined from photo according to characteristic point, for example, the resolution ratio of photo is
1000*1000, the eyes position that characteristic point is behaved on the face, then can determine the area of face with the edge of eyes position and face
Position and size of the area image on photo, for example, the region of face is 100*100.
In step S503, the picture material in the region of face is subjected to affine transformation according to preset reference characteristic point and is obtained
To facial image, the resolution ratio of facial image is identical as the dimension of the input layer for the convolutional neural networks trained.
In one embodiment, can establish include magnanimity face sample sample database, each face sample point
Resolution is identical as the dimension for the input layer for involving in neural network after scaling, carries out people to each face sample in sample database
Face detection, detects four characteristic points such as eyes central point, nose, mouth of face, passes through the eyes on the face sample of magnanimity
Central point, nose, mouth characteristic point obtain a preset reference characteristic point.
In one embodiment, since the size and the dimension of the input layer of CNN that are input to the area image of CNN may not phases
Together, therefore the area image that can also will test carries out affine transformation, so that it is guaranteed that different size of first area is by imitative
Penetrate transformation after it is identical as the dimension of the input layer of convolutional neural networks, for example, the size of the area image intercepted from photo is
100 × 100, the size obtained after being converted by affine transformation is 224 × 224, so as to so that area image and CNN
The dimension of input layer is identical, it is ensured that regional image information can accurately input the defeated of convolutional neural networks as shown in Figure 3B
Enter layer.
In step S504, extract the face characteristic in facial image by the convolutional neural networks trained, obtain to
A few the first corresponding face characteristic of face photograph album.
It is special for how to obtain corresponding second face of at least one third party's face photograph album in the second cloud photograph album
The description of sign may refer to the associated description of the present embodiment, and this will not be detailed here.
In the present embodiment, the area image of face is carried out according to preset reference characteristic point by affine transformation by affine transformation
Obtain to support the input layer of convolutional neural networks, it is ensured that the area image of face can accurately input the volume after training
The input layer of product neural network；Since the convolutional neural networks trained are that have label face sample training to obtain by magnanimity
, so as to so that the face characteristic that the convolutional neural networks trained extract can accurately indicate the face in photo,
Greatly improve the accuracy rate of later period recognition of face.
Fig. 6 is the block diagram for the device that a kind of pair of photo shown according to an exemplary embodiment is clustered, such as Fig. 6 institute
Show, the device clustered to photo includes：
First determining module 61 is configured as determining photograph from the first face photograph album of at least one of the first cloud photograph album
Piece quantity is respectively corresponded to lower than at least one sub- photograph album of the second face of preset threshold and at least one sub- photograph album of the second face
The first face characteristic, corresponding first user identifier of the first cloud photograph album；
Second determining module 62 is configured to determine that second user mark associated with the first user identifier in cloud service
At least one third party's face photograph album in the second cloud photograph album is stored on device and at least one third party's face photograph album is respectively right
The second face characteristic answered；
Third determining module 63 is configured to determine that at least one sub- photograph album of the second face that the first determining module 61 determines
Corresponding first face characteristic and at least one third party's face photograph album that the second determining module 62 determines are corresponding
The similarity value of face characteristic obtains multiple similarity values；
4th determining module 64, be configured as according to third determining module 63 determine multiple similarity values from least one
Human face photo identical at least one the second face corresponding face characteristic of sub- photograph album is determined in third party's face photograph album.
Fig. 7 is the block diagram for the device that another kind shown according to an exemplary embodiment clusters photo, such as Fig. 7 institute
Show, on the basis of above-mentioned embodiment illustrated in fig. 6, in one embodiment, device may also include：
5th determining module 65 is configured to determine that communication list associated with the first user identifier；
Searching module 66, being configured as searching has on Cloud Server in the communication list that the 5th determining module 65 determines
The user identifier of cloud photograph album, will communicate list in cloud photograph album user identifier be determined as it is associated with the first user identifier
Second user mark, the second determining module 62 determine that searching module 66 is found associated with the first user identifier second uses
Family mark stores at least one third party's face photograph album and at least one third party in the second cloud photograph album on Cloud Server
Corresponding second face characteristic of face photograph album.
In one embodiment, device may also include：
First extraction module 67 is configured as the convolutional neural networks by having trained and shines the whole in the first cloud photograph album
Piece carries out feature extraction, obtains at least one third face characteristic；
First cluster module 68 is configured as at least one the third face characteristic pair for extracting the first extraction module 67
The photo answered is clustered as unit of face characteristic as at least one the first face photograph album.
In one embodiment, device may also include：
Second extraction module 69 is configured as the convolutional neural networks by having trained and shines the whole in the second cloud photograph album
Piece carries out feature extraction, obtains at least one the 4th face characteristic；
Second cluster module 70 is configured as at least one the 4th face characteristic pair for extracting the second extraction module 69
The photo answered is clustered as unit of face characteristic as at least one third party's face photograph album.
In one embodiment, device may also include：
First training module 71, be configured as setting quantity has label face sample to be input to convolutional neural networks,
At least one convolutional layer of convolutional neural networks and at least one full articulamentum are trained；
First control module 72, is configured as the optimal weight of the connection in determining convolutional neural networks between each node
When parameter, 71 deconditioning convolutional neural networks of the first training module, the convolutional neural networks trained, for the are controlled
One extraction module 67 and the second extraction module 69 pass through the convolutional neural networks trained and shine the whole in respective cloud photograph album
Piece carries out feature extraction.
Fig. 8 is the block diagram of another device clustered to photo shown according to an exemplary embodiment, such as Fig. 8 institute
Show, on the basis of above-mentioned Fig. 6 or embodiment illustrated in fig. 7, in one embodiment, device may also include：
6th determining module 73 is configured to determine that the characteristic point in each photo in the first cloud photograph album about face；
7th determining module 74 is configured as the characteristic point of the face determined according to the 6th determining module 73 from each photo
The region of upper determining face；
First conversion module 75, the picture material root being configured as in the region for the face for determining the 7th determining module 74
Affine transformation, which is carried out, according to preset reference characteristic point obtains facial image, the resolution ratio of facial image and the convolutional Neural net trained
The dimension of the input layer of network is identical；
Third extraction module 76 is configured as extracting people by the convolutional neural networks that the training of the first training module 71 obtains
Face characteristic in face image obtains at least one the first corresponding face characteristic of face photograph album, the first determining module
61 determine that number of pictures is lower than at least one of preset threshold from the first face photograph album of at least one of the first cloud photograph album
The sub- photograph album of second face and at least one corresponding first face characteristic of the sub- photograph album of the second face.
In one embodiment, device may also include：
8th determining module 77 is configured to determine that the characteristic point in each photo in the second cloud photograph album about face；
9th determining module 78 is configured as the characteristic point of the face determined according to the 8th determining module 77 from each photo
The region of upper determining face；
Second conversion module 79, the picture material root being configured as in the region for the face for determining the 9th determining module 78
Affine transformation, which is carried out, according to preset reference characteristic point obtains facial image, the resolution ratio of facial image and the convolutional Neural net trained
The dimension of the input layer of network is identical；
4th extraction module 80 is configured as extracting people by the convolutional neural networks that the training of the first training module 71 obtains
Face characteristic in face image obtains at least one corresponding second face characteristic of third party's face photograph album, and second determines
Module 62 determines that second user mark associated with the first user identifier stores in the second cloud photograph album extremely on Cloud Server
Few third party's face photograph album and at least one corresponding second face characteristic of third party's face photograph album.
About the device in above-described embodiment, wherein modules execute the concrete mode of operation in related this method
Embodiment in be described in detail, no detailed explanation will be given here.
Fig. 9 is the block diagram for the device that a kind of pair of photo shown according to an exemplary embodiment is clustered.For example, device
900 may be provided as a server.Referring to Fig. 9, it further comprises one or more that device 900, which includes processing component 922,
Processor, and the memory resource as representated by memory 932, for store can by the instruction of the execution of processing component 922,
Such as application program.The application program stored in memory 932 may include it is one or more each correspond to one
The module of group instruction.In addition, processing component 922 is configured as executing instruction, to execute the above-mentioned method clustered to photo.
Device 900 can also include the power management that a power supply module 926 is configured as executive device 900, and one has
Line or radio network interface 950 are configured as device 900 being connected to network and input and output (I/O) interface 958.Dress
Setting 900 can operate based on the operating system for being stored in memory 932, such as Windows ServerTM, Mac OS XTM,
UnixTM, LinuxTM, FreeBSDTM or similar.
In the exemplary embodiment, a kind of non-transitorycomputer readable storage medium including instruction, example are additionally provided
It such as include the memory 932 of instruction, above-metioned instruction can be executed by the processing component 922 of device 900 to complete the above method.Example
Such as, the non-transitorycomputer readable storage medium can be ROM, random access memory (RAM), CD-ROM, tape, soft
Disk and optical data storage devices etc..
Those skilled in the art will readily occur to its of the disclosure after considering specification and practicing disclosure disclosed herein
Its embodiment.This application is intended to cover any variations, uses, or adaptations of the disclosure, these modifications, purposes or
Person's adaptive change follows the general principles of this disclosure and including the undocumented common knowledge in the art of the disclosure
Or conventional techniques.The description and examples are only to be considered as illustrative, and the true scope and spirit of the disclosure are by following
Claim is pointed out.
It should be understood that the present disclosure is not limited to the precise structures that have been described above and shown in the drawings, and
And various modifications and changes may be made without departing from the scope thereof.The scope of the present disclosure is only limited by the accompanying claims.
Claims (15)
1. the method that a kind of pair of photo is clustered, which is characterized in that the method includes：
Determine that number of pictures is lower than at least the one of preset threshold from the first face photograph album of at least one of the first cloud photograph album
A sub- photograph album of second face and at least one described corresponding first face characteristic of the sub- photograph album of the second face, described first
Cloud photograph album corresponds to the first user identifier；
Determine that second user mark associated with first user identifier is stored on Cloud Server in the second cloud photograph album
At least one third party's face photograph album and at least one described corresponding second face characteristic of third party's face photograph album；
Determine at least one described corresponding first face characteristic of the sub- photograph album of the second face and at least one described third party
The similarity value of the corresponding face characteristic of face photograph album, obtains multiple similarity values；
According to the multiple similarity value from least one described third party's face photograph album it is determining with it is described at least one second
The identical human face photo of the corresponding face characteristic of the sub- photograph album of face.
2. the method according to claim 1, wherein the method also includes：
Determine communication list associated with first user identifier；
It searches in the communication list on the cloud server with the user identifier of cloud photograph album, will have in the communication list
There is the user identifier of cloud photograph album to be determined as the second user mark associated with first user identifier.
3. the method according to claim 1, wherein the method also includes：
Feature extraction is carried out to whole photos in the first cloud photograph album by the convolutional neural networks trained, is obtained at least
One third face characteristic；
At least one described corresponding photo of third face characteristic cluster as unit of face characteristic be described at least one the
The sub- photograph album of one face.
4. the method according to claim 1, wherein the method also includes：
Feature extraction is carried out to whole photos in the second cloud photograph album by the convolutional neural networks trained, is obtained at least
One the 4th face characteristic；
At least one described corresponding photo of the 4th face characteristic cluster as unit of face characteristic be described at least one the
The sub- photograph album of three faces.
5. the method according to claim 3 or 4, which is characterized in that the method also includes：
There is label face sample to be input to convolutional neural networks for set quantity, at least one of the convolutional neural networks
Convolutional layer and at least one full articulamentum are trained；
When the optimal weight parameter of the connection in determining the convolutional neural networks between each node, convolution described in deconditioning
Neural network obtains the convolutional neural networks trained.
6. according to the method described in claim 5, it is characterized in that, the method also includes：
Determine the characteristic point in each photo in the first cloud photograph album about face；
The region of the face is determined from each photo according to the characteristic point of the face；
Picture material in the region of the face is subjected to affine transformation according to preset reference characteristic point and obtains facial image, institute
The resolution ratio for stating facial image is identical as the dimension of input layer of the convolutional neural networks trained；
Extract the face characteristic in the facial image by the convolutional neural networks trained, obtain it is described at least one
The first corresponding face characteristic of face photograph album.
7. according to the method described in claim 5, it is characterized in that, the method also includes：
Determine the characteristic point in each photo in the second cloud photograph album about face；
The region of the face is determined from each photo according to the characteristic point of the face；
Picture material in the region of the face is subjected to affine transformation according to preset reference characteristic point and obtains facial image, institute
The resolution ratio for stating facial image is identical as the dimension of input layer of the convolutional neural networks trained；
Extract the face characteristic in the facial image by the convolutional neural networks trained, obtain it is described at least one
Corresponding second face characteristic of third party's face photograph album.
8. the device that a kind of pair of photograph album is clustered, which is characterized in that described device includes：
First determining module is configured as determining number of pictures from the first face photograph album of at least one of the first cloud photograph album
It is corresponding lower than at least one sub- photograph album of the second face of preset threshold and at least one described sub- photograph album of the second face
First face characteristic, corresponding first user identifier of the first cloud photograph album；
Second determining module is configured to determine that second user mark associated with first user identifier in Cloud Server
At least one third party's face photograph album and at least one described third party's face photograph album in the second cloud photograph album of upper storage are respectively
Corresponding second face characteristic；
Third determining module is configured to determine that at least one described sub- photograph album of the second face that first determining module determines
At least one described third party's face photograph album that corresponding first face characteristic is determined with second determining module is respectively
The similarity value of corresponding face characteristic obtains multiple similarity values；
4th determining module, be configured as according to the third determining module determine the multiple similarity value from it is described at least
Determination is identical at least one described the second face corresponding face characteristic of sub- photograph album in one third party's face photograph album
Human face photo.
9. device according to claim 8, which is characterized in that described device further includes：
5th determining module is configured to determine that communication list associated with first user identifier；
Searching module is configured as searching in the communication list that the 5th determining module determines on the cloud server
User identifier with cloud photograph album in the communication list is determined as and first user by the user identifier with cloud photograph album
Identify the associated second user mark.
10. device according to claim 8, which is characterized in that described device further includes：
First extraction module is configured as the convolutional neural networks by having trained and carries out to whole photos in the first cloud photograph album
Feature extraction obtains at least one third face characteristic；
First cluster module, be configured as extracting first extraction module described at least one third face characteristic pair
The photo answered is clustered as unit of face characteristic as at least one described the first face photograph album.
11. device according to claim 10, which is characterized in that described device further includes：
Second extraction module is configured as the convolutional neural networks by having trained to whole photos in the second cloud photograph album
Feature extraction is carried out, at least one the 4th face characteristic is obtained；
Second cluster module, be configured as extracting second extraction module described at least one the 4th face characteristic pair
The photo answered is clustered as unit of face characteristic as at least one described third party's face photograph album.
12. device described in 0 or 11 according to claim 1, which is characterized in that described device further includes：
First training module, be configured as setting quantity has label face sample to be input to convolutional neural networks, to described
At least one convolutional layer of convolutional neural networks and at least one full articulamentum are trained；
First control module is configured as the optimal weight ginseng of the connection in determining the convolutional neural networks between each node
When number, convolutional neural networks described in the first training module deconditioning, the convolutional neural networks trained are controlled.
13. device according to claim 12, which is characterized in that described device further includes：
6th determining module is configured to determine that the characteristic point in each photo in the first cloud photograph album about face；
7th determining module is configured as the characteristic point of the face determined according to the 6th determining module from described each
The region of the face is determined on photo；
First conversion module, the picture material root being configured as in the region for the face for determining the 7th determining module
Affine transformation, which is carried out, according to preset reference characteristic point obtains facial image, the resolution ratio of the facial image and the volume trained
The dimension of the input layer of product neural network is identical；
Third extraction module is configured as extracting the people by the convolutional neural networks that first training module training obtains
Face characteristic in face image obtains at least one described the first corresponding face characteristic of face photograph album.
14. device according to claim 12, which is characterized in that described device further includes：
8th determining module is configured to determine that the characteristic point in each photo in the second cloud photograph album about face；
9th determining module is configured as the characteristic point of the face determined according to the 8th determining module from described each
The region of the face is determined on photo；
Second conversion module, the picture material root being configured as in the region for the face for determining the 9th determining module
Affine transformation, which is carried out, according to preset reference characteristic point obtains facial image, the resolution ratio of the facial image and the volume trained
The dimension of the input layer of product neural network is identical；
4th extraction module is configured as described in the convolutional neural networks extraction obtained by first training module training
Face characteristic in facial image obtains at least one described corresponding second face characteristic of third party's face photograph album.
15. the device that a kind of pair of photo is clustered, which is characterized in that described device includes：
Processor；
Memory for storage processor executable instruction；
Wherein, the processor is configured to：
Determine that number of pictures is lower than at least the one of preset threshold from the first face photograph album of at least one of the first cloud photograph album
A sub- photograph album of second face and at least one described corresponding first face characteristic of the sub- photograph album of the second face, described first
Cloud photograph album corresponds to the first user identifier；
Determine that second user mark associated with first user identifier is stored on Cloud Server in the second cloud photograph album
At least one third party's face photograph album and at least one described corresponding second face characteristic of third party's face photograph album；
Determine at least one described corresponding first face characteristic of the sub- photograph album of the second face and at least one described third party
The similarity value of the corresponding face characteristic of face photograph album, obtains multiple similarity values；
According to the multiple similarity value from least one described third party's face photograph album it is determining with it is described at least one second
The identical human face photo of the corresponding face characteristic of the sub- photograph album of face.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510955515.1A CN105631404B (en) | 2015-12-17 | 2015-12-17 | The method and device that photo is clustered |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510955515.1A CN105631404B (en) | 2015-12-17 | 2015-12-17 | The method and device that photo is clustered |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105631404A CN105631404A (en) | 2016-06-01 |
| CN105631404B true CN105631404B (en) | 2018-11-30 |
Family
ID=56046317
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510955515.1A Active CN105631404B (en) | 2015-12-17 | 2015-12-17 | The method and device that photo is clustered |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105631404B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106095876B (en) * | 2016-06-06 | 2021-11-09 | 北京小米移动软件有限公司 | Image processing method and device |
| CN106156347A (en) * | 2016-07-21 | 2016-11-23 | 北京奇虎科技有限公司 | Cloud photograph album classification methods of exhibiting, device and server |
| CN106339695B (en) * | 2016-09-20 | 2019-11-15 | 北京小米移动软件有限公司 | Face similarity detection method, device and terminal |
| CN107679532B (en) * | 2017-09-15 | 2019-12-03 | Oppo广东移动通信有限公司 | Data transmission method, device, mobile terminal and computer readable storage medium |
| CN108875654B (en) * | 2018-06-25 | 2021-03-05 | 深圳云天励飞技术有限公司 | Face feature acquisition method and device |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103399896A (en) * | 2013-07-19 | 2013-11-20 | 广州华多网络科技有限公司 | Method and system for recognizing association relationships among users |
| US8655028B2 (en) * | 2011-10-19 | 2014-02-18 | Primax Electronics Ltd. | Photo sharing system with face recognition function |
| CN103970830A (en) * | 2014-03-31 | 2014-08-06 | 小米科技有限责任公司 | Information recommendation method and device |
| CN104537380A (en) * | 2014-12-30 | 2015-04-22 | 小米科技有限责任公司 | Clustering method and device |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9195678B2 (en) * | 2011-01-24 | 2015-11-24 | T-Mobile Usa, Inc. | Automatic selection of digital images from a multi-sourced collection of digital images |
-
2015
- 2015-12-17 CN CN201510955515.1A patent/CN105631404B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8655028B2 (en) * | 2011-10-19 | 2014-02-18 | Primax Electronics Ltd. | Photo sharing system with face recognition function |
| CN103399896A (en) * | 2013-07-19 | 2013-11-20 | 广州华多网络科技有限公司 | Method and system for recognizing association relationships among users |
| CN103970830A (en) * | 2014-03-31 | 2014-08-06 | 小米科技有限责任公司 | Information recommendation method and device |
| CN104537380A (en) * | 2014-12-30 | 2015-04-22 | 小米科技有限责任公司 | Clustering method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105631404A (en) | 2016-06-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108304435B (en) | Information recommendation method and device, computer equipment and storage medium | |
| CN105631404B (en) | The method and device that photo is clustered | |
| WO2019200902A1 (en) | Image recognition method and device | |
| CN104915351B (en) | Picture sort method and terminal | |
| CN105095882B (en) | The recognition methods of gesture identification and device | |
| US8571332B2 (en) | Methods, systems, and media for automatically classifying face images | |
| US20150169978A1 (en) | Selection of representative images | |
| CN110941594A (en) | Splitting method and device of video file, electronic equipment and storage medium | |
| CN106575280B (en) | System and method for analyzing user-associated images to produce non-user generated labels and utilizing the generated labels | |
| CN109871821B (en) | Pedestrian re-identification method, device, equipment and storage medium of self-adaptive network | |
| CN107316035A (en) | Object identifying method and device based on deep learning neutral net | |
| CN107735795A (en) | Method and system for social relationships identification | |
| WO2021175071A1 (en) | Image processing method and apparatus, storage medium, and electronic device | |
| US11335127B2 (en) | Media processing method, related apparatus, and storage medium | |
| CN106897700B (en) | Single-sample face recognition method and system | |
| CN105956051A (en) | Information finding method, device and system | |
| CN116052218B (en) | Pedestrian re-identification method | |
| CN109993042A (en) | A kind of face identification method and its device | |
| TWI745818B (en) | Method and electronic equipment for visual positioning and computer readable storage medium thereof | |
| WO2023123923A1 (en) | Human body weight identification method, human body weight identification device, computer device, and medium | |
| CN114902299A (en) | Method, device, equipment and storage medium for detecting associated object in image | |
| CN111652260B (en) | Face clustering sample number selection method and system | |
| CN103020589A (en) | Face recognition method for single training sample | |
| CN108446688B (en) | Face image gender judgment method and device, computer equipment and storage medium | |
| CN107169093B (en) | target image acquisition method and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |