CN108491812A - The generation method and device of human face recognition model - Google Patents

The generation method and device of human face recognition model Download PDF

Info

Publication number
CN108491812A
CN108491812A CN201810268892.1A CN201810268892A CN108491812A CN 108491812 A CN108491812 A CN 108491812A CN 201810268892 A CN201810268892 A CN 201810268892A CN 108491812 A CN108491812 A CN 108491812A
Authority
CN
China
Prior art keywords
image
training sample
face
characteristic
target object
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN201810268892.1A
Other languages
Chinese (zh)
Other versions
CN108491812B (en
Inventor
张刚
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Baidu Netcom Science and Technology Co Ltd
Original Assignee
Beijing Baidu Netcom Science and Technology Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Beijing Baidu Netcom Science and Technology Co Ltd filed Critical Beijing Baidu Netcom Science and Technology Co Ltd
Priority to CN201810268892.1A priority Critical patent/CN108491812B/en
Publication of CN108491812A publication Critical patent/CN108491812A/en
Application granted granted Critical
Publication of CN108491812B publication Critical patent/CN108491812B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

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
    • 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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks

Abstract

The embodiment of the present application discloses the generation method and device of human face recognition model.One specific implementation mode of this method includes：Obtain training sample set, then each training sample in training sample set is input in Initial Face identification model and Initial Face identification model is trained, human face recognition model after being trained, human face recognition model are input to whether facial image therein meets predetermined genetic connection between corresponding object for identification.While the facial image of the facial image and relationship object that do not increase target object, expand trained characteristic pattern, reduces the manpower and materials cost and time cost for obtaining training facial image.Improve the efficiency of trained human face recognition model.

Description

The generation method and device of human face recognition model
Technical field
The invention relates to field of computer technology, and in particular to Internet technical field, more particularly, to people The generation method and device of face identification model.
Background technology
Recognition of face refers to comparing face visual signature information computer technology for identification using analysis.People Face identification product is widely used to the fields such as finance, safety check, medical treatment, public security.
In the process of recognition of face, the face characteristic of face to be identified can be matched with skin detection, root The identity information of face to be identified is predicted according to similarity.
In general, we can use training facial image to be trained human face recognition model so that human face recognition model It can carry out the processing such as the identification of facial image.
Invention content
The embodiment of the present application proposes a kind of generation method and device of human face recognition model.
In a first aspect, the embodiment of the present application provides a kind of generation method of human face recognition model, this method includes：It obtains Training sample set；Each training sample in training sample set is input in Initial Face identification model, Initial Face is known Other model is trained, and the human face recognition model after being trained, human face recognition model is input to face therein for identification Whether image meets predetermined genetic connection between corresponding object；Wherein, training sample set includes multiple training samples Image pair, at least one of training sample set training sample image are generated to being based on following steps：Obtain target object A training sample image of the face characteristic figure as training sample image centering；It obtains and is closed with predetermined blood relationship with target object The facial image of at least two relationship objects of system；Target image set is generated, target image set includes the face by relationship object The characteristic pattern that image generates and the characteristic pattern generated by the combination image of relationship object, wherein the combination image of relationship object For the default characteristic area of the facial image of the one of relationship object of interception, it is used in combination the default characteristic area intercepted to replace another The image that the corresponding characteristic area of facial image of one relationship object is generated；Arbitrary selection one is concentrated from target image Another training sample image of characteristic pattern as training sample image centering.
In some embodiments, training sample set further includes following at least one training sample image pair, the training sample This image to including target image face characteristic figure and do not have with target object the who object of predetermined genetic connection Face characteristic figure；And each training sample in training sample set is input in Initial Face identification model to Initial Face Identification model is trained, the human face recognition model after being trained, including：Each training sample in training sample set is defeated Enter and Initial Face identification model is trained into Initial Face identification model, the human face recognition model after being trained, with If make the image to be detected being input in human face recognition model to including target object to be detected face characteristic figure and wait for this Detected target object has the face characteristic figure of predetermined genetic connection, and it is default that the numerical value that human face recognition model is exported is more than first Threshold value, if image to be detected centering includes the face characteristic figure of target object and do not have predetermined genetic connection with the target object Face characteristic figure, the numerical value that human face recognition model is exported be less than the second predetermined threshold value；Wherein, the second predetermined threshold value is less than the One predetermined threshold value.
In some embodiments, obtaining training sample set includes：Obtain the facial image and target object of target object The respective facial image of at least two relationship objects with predetermined genetic connection and the face by least two relationship objects The combination image that image obtains；The facial image of target object, the facial image of relationship object, combination image are input in advance Trained face characteristic identification model, respectively obtain the face characteristic figure of target object, the face characteristic figure of relationship object and Combine the characteristic pattern of image；It is special from the face of relationship object using the face characteristic figure of target object as a training book image A characteristic pattern is arbitrarily chosen as another training sample image in the characteristic pattern of sign figure and combination image, obtains training sample set At least one of conjunction training sample image pair.
In some embodiments, it is inputted by the facial image of the facial image of target object, relationship object, combination image To face characteristic identification model trained in advance, the face characteristic figure of target object, the face characteristic of relationship object are respectively obtained Before the characteristic pattern of figure and combination image, this method further includes：By the face of the facial image of target object and relationship object Image, combination image carry out affine transformation, the relationship object face figure after target object facial image, transformation after being converted Combination image after picture, transformation；And the facial image of target object, the facial image of relationship object, combination image are inputted To face characteristic identification model trained in advance, the face characteristic figure of target object, the face characteristic of relationship object are respectively obtained Figure and the characteristic pattern for combining image further include：By the target object facial image after transformation, the relationship object face after transformation Combination image after image, transformation is separately input to the face characteristic identification model trained in advance, the target pair after being converted As the relationship object face characteristic figure after face characteristic figure, transformation and the combination image characteristic pattern after transformation；And by target pair The face characteristic figure of elephant is as a training book image, from appointing in the characteristic pattern of the face characteristic figure of relationship object and combination image Meaning chooses a characteristic pattern as another training sample image, obtains at least one of training sample set training sample image It is right, further include：Using the target object face characteristic figure after the face characteristic figure of target object or transformation as a training book figure Picture, from the target object face characteristic figure after transformation, the relationship object face characteristic figure after transformation and the combination image after transformation A characteristic pattern is arbitrarily chosen in characteristic pattern as another training sample image, obtains at least one of training sample set instruction Practice sample image pair.
In some embodiments, human face recognition model is convolutional neural networks model.
Second aspect, the embodiment of the present application provide a kind of generating means of human face recognition model, which includes：It obtains Unit is configured to obtain training sample set；Human face recognition model generation unit, being configured to will be in training sample set Each training sample is input in Initial Face identification model and is trained to Initial Face identification model, the face after being trained Identification model, human face recognition model be input to for identification facial image therein whether meet between corresponding object it is pre- Determine genetic connection；Wherein, training sample set includes multiple training sample images pair, at least one of training sample set instruction Practice sample image and is generated to being based on following steps：Obtain one of the face characteristic figure of target object as training sample image centering A training sample image；Obtain the facial image at least two relationship objects that there is predetermined genetic connection with target object；It is raw At target image set, target image set includes the characteristic pattern by the Face image synthesis of relationship object and the group by relationship object Close the characteristic pattern that image generates, wherein the combination image of relationship object is the facial image for intercepting one of relationship object Default characteristic area, is used in combination the default characteristic area intercepted to replace the corresponding feature of facial image of another relationship object The image that region is generated；From target image concentrate it is arbitrary selection one characteristic pattern as training sample image centering another Training sample image.
In some embodiments, training sample set further includes following at least one training sample image pair, the training sample This image to including target image face characteristic figure and do not have with target object the who object of predetermined genetic connection Face characteristic figure；And human face recognition model generation unit is further configured to：By each trained sample in training sample set Originally it is input in Initial Face identification model and Initial Face identification model is trained, the recognition of face mould after being trained Type, if so that the image to be detected being input in human face recognition model to including target object to be detected face characteristic figure and with There is the target object to be detected the face characteristic figure of predetermined genetic connection, the numerical value that human face recognition model is exported to be more than first Predetermined threshold value, if image to be detected centering includes the face characteristic figure of target object and do not have predetermined blood relationship with the target object The face characteristic figure of relationship, the numerical value that human face recognition model is exported are less than the second predetermined threshold value；Wherein, the second predetermined threshold value is small In the first predetermined threshold value.
In some embodiments, acquiring unit is further configured to：Obtain the facial image and target pair of target object As with predetermined genetic connection the respective facial image of at least two relationship objects and by the people of at least two relationship objects The combination image that face image obtains；The facial image of target object, the facial image of relationship object, combination image are input to pre- First trained face characteristic identification model, respectively obtain the face characteristic figure of target object, relationship object face characteristic figure with And the characteristic pattern of combination image；Using the face characteristic figure of target object as a training book image, from the face of relationship object A characteristic pattern is arbitrarily chosen as another training sample image in the characteristic pattern of characteristic pattern and combination image, obtains training sample At least one of set training sample image pair.
In some embodiments, acquiring unit is further configured to：By the facial image of target object, relationship object Facial image, combination image be input in advance trained face characteristic identification model, the face for respectively obtaining target object is special Before the characteristic pattern of sign figure, the face characteristic figure of relationship object and combination image, by the facial image and relationship of target object Facial image, the combination image of object carry out affine transformation, the relationship after target object facial image, transformation after being converted Combination image after object facial image, transformation；And by the target object facial image after transformation, the relationship object after transformation Combination image after facial image, transformation is separately input to the face characteristic identification model trained in advance, the mesh after being converted Mark the relationship object face characteristic figure after object face characteristic pattern, transformation and the combination image characteristic pattern after transformation；And by mesh Target object face characteristic figure after marking the face characteristic figure of object or converting is as a training book image, from the mesh after transformation It is arbitrarily chosen in the combination image characteristic pattern after relationship object face characteristic figure and transformation after mark object face characteristic pattern, transformation One characteristic pattern obtains at least one of training sample set training sample image pair as another training sample image.
In some embodiments, human face recognition model is convolutional neural networks model.
The third aspect, the embodiment of the present application provide a kind of server, which includes：One or more processors； Storage device, for storing one or more programs, when said one or multiple programs are held by said one or multiple processors When row so that said one or multiple processors realize the method as described in any realization method in first aspect.
Fourth aspect, the embodiment of the present application provide a kind of computer-readable medium, are stored thereon with computer program, In, the method as described in any realization method in first aspect is realized when which is executed by processor.
The generation method and device of human face recognition model provided by the embodiments of the present application, by the face by target object Characteristic pattern and have the face of any relationship object at least two relationship objects of predetermined genetic connection special with the target object Sign figure as training sample image to while, will also with target object have predetermined genetic connection at least two relationship objects Facial image any combination image characteristic pattern and target object face characteristic figure as training sample image to come pair Initial Face model is trained so that the human face recognition model after training, which can identify, is input to facial image therein to institute Whether predetermined genetic connection is met between corresponding object, to realize in the facial image and relationship for not increasing target object While the facial image of object, expand trained characteristic pattern, reduces the manpower and materials for obtaining training facial image Cost and time cost.Improve the efficiency of trained human face recognition model.
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 generation method of the human face recognition model of the application；
Fig. 3 is the generation schematic diagram of at least one training sample image pair in the training sample set according to the application；
Fig. 4 is the schematic diagram according to an application scenarios of the generation method of the human face recognition model of the application；
Fig. 5 is the flow chart according to another embodiment of the generation method of the human face recognition model of the application；
Fig. 6 is the structural schematic diagram according to one embodiment of the generating means of the human face recognition model of the application；
Fig. 7 is adapted for the structural schematic diagram of the computer system of the server 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 generation dress of the generation method or human face recognition model that can apply the human face recognition model of the application The exemplary system architecture 100 for the embodiment set.
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 is interacted by network 104 with server 105, to receive or send message etc..Terminal Equipment 101,102,103 can be various electronic equipments, including but not limited to smart mobile phone, tablet computer, portable meter on knee Calculation machine and desktop computer etc..
Server 105 can provide various services, for example, server 105 can by network 104 from terminal 101,102, 103 obtain input data, so as to realize the training to human face recognition model, and obtain the human face recognition model of training completion.
It should be noted that the generation method for the human face recognition model that the embodiment of the present application is provided is generally by server 105 execute, and correspondingly, the generating means of human face recognition model are generally positioned in server 105.
It should be noted that server can be hardware, can also be software.When server is hardware, may be implemented At the distributed server cluster that multiple servers form, individual server can also be implemented as.It, can when server is software It, can also to be implemented as multiple softwares or software module (such as providing the multiple softwares or software module of Distributed Services) It is implemented as single software or software module.It is not specifically limited herein.
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.
With continued reference to Fig. 2, it illustrates according to one embodiment of the generation method of the human face recognition model of the application Flow 200.The generation method of the human face recognition model, includes the following steps：
Step 201, training sample set is obtained.
In the present embodiment, the executive agent (such as server shown in FIG. 1) of the generation method of human face recognition model can To obtain training facial image from terminal device by wired connection mode or radio connection.Here facial image is trained It may include the facial image of target object.It is added to the identity of target object in advance in the facial image of above-mentioned target object Markup information.
After receiving the facial image of target object, above-mentioned executive agent can utilize various analysis methods to above-mentioned The facial image of target object does the analyzing processing for carrying out a step, to obtain training sample set.Above-mentioned training sample set Include multiple training sample images pair.At least one of training sample set training sample image is to that can be based on Fig. 3 institutes The step of showing generates.
Referring to FIG. 3, it illustrates the generation schematic diagrames of at least one training sample image pair in training sample set.
In the generation schematic diagram 300 of at least one of training sample set as shown in Figure 3 training sample image pair, instruction At least one training sample image by following steps to can be generated in white silk sample set.
Step 301, a training sample figure of the face characteristic figure as training sample image centering of target object is obtained Picture.
Above-mentioned executive agent can the facial image based on target object obtain target object using various analysis methods Face characteristic figure.The face characteristic figure of target object is such as determined using the method for image procossing.Such as it can be mended by light It repays, the processes such as image gray processing, Gaussian smoothing, similarity calculation and binaryzation determine the face characteristic figure of target object.
In the present embodiment, face characteristic figure refers to the facial color characteristic, textural characteristics, shape that can describe an object The image of the relative position relation feature at each position of shape feature and face.Here face characteristic figure can be two dimensional image.
It, can be by face for the facial image of target object in some optional realization methods of the present embodiment The region at each position of face is detected in image first.Then for each position of face, by various methods (such as scheme As the method for processing) extract the characteristic pattern at the position.The characteristic pattern of face is finally obtained by the characteristic pattern at each position of face.This In face position for example can be eyes, nose and face.
In the present embodiment, above-mentioned executive agent can be using the face characteristic figure of target object as training sample image pair In a training sample image.
Step 302, the facial image at least two relationship objects that there is predetermined genetic connection with target object is obtained.
Can also include at least two relationships that there is predetermined genetic connection with above-mentioned target object in above-mentioned training image The facial image of object.Here predetermined genetic connection for example can be nature direct line genetic connection, natural collateral line genetic connection Deng.Above-mentioned naturally lineal genetic connection includes parent/children, grand parents/grandchildren etc., and natural collateral line genetic connection includes brother Younger sister, sisters etc..
When predetermined genetic connection here be parent/children relation when, such as target object be children, above-mentioned at least two Relationship object may include two relationship objects namely the facial image of father and mother.
When predetermined genetic connection is lineal genetic connection, above-mentioned at least two passes with target object with predetermined relationship The facial image for being object can also include the facial image of grand parents of target object, grand parents facial image etc..
The markup information of the identity of relationship object, relationship here are added in the facial image of above-mentioned relation object in advance The markup information of the identity of object for example can be the information for being used to indicate the predetermined genetic connection met with target object.
Step 303, target image set is generated, target image set includes by the characteristic pattern of the Face image synthesis of relationship object And the characteristic pattern generated by the combination image of relationship object, wherein the combination image of relationship object, which is that interception is one of, to close It is the default characteristic area of the facial image of object, the default characteristic area intercepted is used in combination to replace the people of another relationship object The image that the corresponding characteristic area of face image is generated.
Face of the above-mentioned executive agent at least two relationship objects for obtaining that there is predetermined genetic connection with target object It, can be based on the Face image synthesis target image set of above-mentioned at least two relationship object after image.
First, above-mentioned executive agent can be by various analysis methods (such as method of image procossing) from above-mentioned at least two The corresponding face characteristic figure of above-mentioned at least two relationship object is respectively obtained in the facial image of a relationship object.
Then, by the combination image of the Face image synthesis relationship object at least two relationship objects.Specifically, above-mentioned Executive agent can intercept the default characteristic area of the facial image of any one relationship object in above-mentioned at least two relationship object Domain is used in combination the default characteristic area intercepted to replace the corresponding characteristic area generation group of facial image of another relationship object Close image.So multiple combination images can be obtained.Here default characteristic area for example can be eyes, nose and/ Or face.
That is, above-mentioned executive agent can be by the face figure of any one relationship object at least two relationship objects Eye areas as in intercepts out, replaces the corresponding eye areas of facial image of another relationship object, generates a group Close image；Alternatively, executive agent can be by the nose in the facial image of any one relationship object at least two relationship objects Subregion intercepts out, replaces the corresponding nasal area of facial image of another relationship object, generates a combination image；Or Person, executive agent can also cut the face region in the facial image of any one relationship object at least two relationship objects It takes out, replaces the corresponding face region of facial image of another relationship object, generate combination image, etc..When upper When the quantity for stating relationship object is N, when the quantity of above-mentioned default characteristic area is m, N can be obtainedmN number of combination image.Its Middle N >=2, and N is positive integer.M >=2, and m is positive integer.
Then, for each combination image, above-mentioned executive agent can obtain the constitutional diagram by various analysis methods The characteristic pattern of picture.Such as the characteristic pattern of group combination image can be obtained by the method for image procossing.
So, it is respectively corresponded to by the corresponding characteristic pattern of each combination image and above-mentioned at least two relationship object Face characteristic figure generate target image set.
Step 304, from target image concentrate it is arbitrary selection one characteristic pattern as training sample image centering another Training sample image.
In the present embodiment, above-mentioned executive agent can concentrate arbitrary selection from the target image obtained in step 303 Another training sample image of one characteristic pattern as training sample image centering.That is, a training sample image One training sample image of centering can be the face characteristic figure of target object, another training sample image can be it is above-mentioned extremely The corresponding face characteristic figure of any one relationship object in few two relationship objects, can also be each by least two relationship objects From Face image synthesis multiple combination images in any combination image characteristic pattern.
By step 301~step 304, above-mentioned executive agent can obtain multiple training sample images pair.Per a pair of of instruction Practice sample image to may include the corresponding face characteristic figure of target object, and concentrates any one chosen special by target image Sign figure.When above-mentioned target object is one, and the quantity of above-mentioned at least two relationship object is N, when above-mentioned default characteristic area For three eyes, nose, face regions when, N can be obtained3A training sample image pair.
So, the quantity of obtained training sample image pair is far more than only by the facial image of target image Any one in obtained face characteristic figure and above-mentioned at least two relationship objects with target object with predetermined genetic connection The face characteristic figure obtained in the facial image of relationship object combines the quantity for the training sample image pair to be formed.
It returns with continued reference to Fig. 2, the character image clustering method of the present embodiment further includes：
Step 202, each training sample in training sample set is input in Initial Face identification model to initial people Face identification model is trained, and the human face recognition model after being trained, human face recognition model is input to therein for identification Whether facial image meets predetermined genetic connection between corresponding object.
In the present embodiment, after obtaining training sample set in step 201, above-mentioned executive agent (such as shown in Fig. 1 Server) above-mentioned training sample set can be input in Initial Face identification model to Initial Face identification model carry out Training, the human face recognition model after being trained.
In the present embodiment, above-mentioned human face recognition model for example can be artificial nerve network model and supporting vector Other non-neural network models such as machine etc..
In some optional realization methods of the present embodiment, above-mentioned human face recognition model can be convolutional neural networks mould Type.Convolutional neural networks (Convolutional Neural Network, CNN) are one kind of depth artificial neural network, are led to Normal convolutional neural networks may include multiple feature extraction layers (also known as convolutional layer) and multiple down-sampling layers (also known as pond layer, Pooling layers).Wherein, feature extraction layer replaces connection with down-sampling layer.Each feature extraction layer may include at least one Convolution kernel.For a feature extraction layer, carries out convolution using a convolution kernel of this layer and the output of preceding layer and obtain one Characteristic pattern.Down-sampling layer is asked for the convolution results of the output to feature extraction layer connected to it at local average and dimensionality reduction Reason.Wherein, the convolution kernel in feature extraction layer includes multiple weights.Weights in convolution kernel can be obtained by multiple sample trainings It arrives.It is shared using local weight when each convolution kernel of convolutional neural networks is to image zooming-out characteristic pattern, god can be reduced Complexity through network model.
In some optional realization methods of the present embodiment, above-mentioned training sample set further includes following at least one training Sample image pair, the training sample image is to including the corresponding face characteristic figure of target object and being unsatisfactory for making a reservation for target object The face characteristic figure of the who object of genetic connection.
It is understood that in these optional realization methods, including the corresponding face characteristic figure of target object and with The training sample figure that target object is unsatisfactory for the face characteristic figure of the who object of predetermined genetic connection can be used as training relatively Negative sample in sample set.
Above-mentioned executive agent can will include that obtain training sample pair by step 301~304 corresponding with by target object The training sample that face characteristic figure and the face characteristic figure that the who object of predetermined genetic connection is unsatisfactory for target object form Each training sample image in the training sample set of image pair knows Initial Face to being input in Initial Face identification model Other model is trained, the human face recognition model after being trained.The human face recognition model after training is allow to identify input Whether meet predetermined genetic connection between corresponding object to image to be detected therein.
Specifically, if the image to be detected being input in human face recognition model is special to the face including target object to be detected Sign schemes and the face characteristic figure with object of the target object to be detected with predetermined genetic connection, human face recognition model output valve Can be the numerical value more than predetermined threshold value.If the image to be detected being input in human face recognition model is to including target pair to be detected The face characteristic figure of elephant and with the target object to be detected do not have predetermined genetic connection object face characteristic figure, understanding know Other model output value can be the numerical value less than above-mentioned predetermined threshold value.Here predetermined threshold value can be carried out according to specific application Setting, as an example, above-mentioned predetermined threshold value can be 0.5.
In some optional realization methods of the present embodiment, above-mentioned executive agent can also will pass through step shown in Fig. 3 Training sample pair is obtained in 301~304 and is closed without predetermined blood relationship by the face characteristic figure of target object and with target object The training sample image of the face characteristic figure composition of the object of system identifies mould to the training sample set constituted to Initial Face Type is trained.Human face recognition model after being trained, if so as to being input to image to be detected pair in human face recognition model Face characteristic figure including target object to be detected and the object face for meeting predetermined genetic connection with the target object to be detected Characteristic pattern, the then numerical value that the human face recognition model after training is exported are more than the first predetermined threshold value.If image to be detected centering packet It includes the face characteristic figure of target object and is unsatisfactory for the face characteristic figure of the object of predetermined genetic connection with the target object, then instruct The numerical value that human face recognition model after white silk is exported is less than the second predetermined threshold value.Here it is default that the second predetermined threshold value is less than first Threshold value.
The concrete numerical value of first predetermined threshold value and the second predetermined threshold value can be set as needed, not limited herein It is fixed.
It should be noted that above-mentioned artificial nerve network model, convolutional neural networks model and support vector machines etc. its His non-neural network model is the known technology studied and applied extensively at present, and details are not described herein.
Please further refer to Fig. 4, it illustrates an applications according to the generation method of the human face recognition model of the application The schematic diagram of scene.
In the application scenarios of Fig. 4, server 402 obtains training character image 403 from user terminal 401, wherein training Facial image includes the facial image of target object and has at least two relationship objects of predetermined genetic connection with target object Facial image；Later, according to training character image 403, server 402 can obtain training sample set 404；Training sample Set may include multiple training samples pair, and wherein at least one of training sample set training sample is to being according to following step Suddenly it obtains：A training sample image of the face characteristic figure of target object as training sample image centering is obtained first, Then, Face image synthesis based at least two relationship objects combines image, then obtain combination image characteristic pattern and The face characteristic figure of each relationship object will combine any feature figure in the face characteristic figure of the characteristic pattern and relationship image of image As another training sample image；Then each training sample in training sample set is input to initial people by server 402 405 are trained to Initial Face identification model in face identification model, the human face recognition model 406 after being trained is so that face Identification model, which can identify, is input to whether facial image therein meets predetermined genetic connection between corresponding object.
The method that above-described embodiment of the application provides by by target object face characteristic figure and with the target pair As the face characteristic figure of any relationship object at least two relationship objects with predetermined genetic connection is as training sample figure As to while, also by any group of facial image of at least two relationship objects with target object with predetermined genetic connection The face characteristic figure of the characteristic pattern and target object that close image is as training sample image to instructing to Initial Face model Practice so that training after human face recognition model can identify be input to facial image therein between corresponding object whether Meet predetermined genetic connection, to realize the face figure in the facial image and relationship object that do not increase training target object As while being trained to human face recognition model, having expanded trained characteristic pattern, reduces and obtain training facial image Manpower and materials cost and time cost.Improve the efficiency of trained human face recognition model.
With further reference to Fig. 5, it illustrates the flows 500 of another embodiment of the generation method of human face recognition model. The flow 500 of the generation method of the human face recognition model, includes the following steps：
Step 501, at least two passes of facial image and target object with predetermined genetic connection of target object are obtained The combination image for being the facial image of object and being combined by the facial image of at least two relationship objects.
In the present embodiment, the executive agent (such as server shown in FIG. 1) of the generation method of human face recognition model can The facial image and and target of target object are obtained from terminal device by wired connection mode or radio connection Object has the facial image of at least two relationship objects of predetermined genetic connection.It is pre- in the facial image of above-mentioned target object First it is added to the markup information of the identity of target object.In at least two relationships pair with target object with predetermined genetic connection The markup information of the identity of relationship object, the mark letter of the identity of relationship object here are added in the facial image of elephant in advance The information for the predetermined genetic connection that breath can for example be met for instruction with target object.
Above-mentioned executive agent can be according at least two relationship objects pair with target object with predetermined genetic connection Facial image obtains multiple combination images.The above-mentioned process for obtaining multiple combination images can refer in embodiment illustrated in fig. 3 Elaborating for step 303, does not repeat herein.
Step 502, the facial image of target object, the facial image of relationship object, combination image are input to advance instruction Experienced face characteristic identification model respectively obtains the face characteristic figure of target object, the face characteristic figure of relationship object and group Close the characteristic pattern of image.
Above-mentioned executive agent can have predetermined genetic connection at least two by the facial image of target object, with target object The facial image and combination image of a relationship object are separately input in the face characteristic identification model trained in advance, obtain mesh It marks the face characteristic figure of object, there is the respective face characteristic figure of at least two relationship object of predetermined genetic connection with target object And the characteristic pattern of combination image.
Above-mentioned default face characteristic identification model can be neural network model (such as artificial nerve network model, convolution Neural network model) and non-neural network model etc..
Step 503, special from the face of relationship object using the face characteristic figure of target object as a training sample image A characteristic pattern is arbitrarily chosen as another training sample image in the characteristic pattern of sign figure and combination image, obtains training sample At least one of this set training sample image pair.
Above-mentioned executive agent can be using the face characteristic figure of the target object obtained by face characteristic identification model as instruction Practice a training sample image of sample image centering, and it is pre- from being obtained having with target object by face characteristic identification model Determine to choose one in the respective face characteristic figure of at least two relationship objects of genetic connection and the characteristic pattern of any combination image Characteristic pattern generates the training sample pair during training sample set closes as another training sample image.So, it can obtain To multiple training samples pair in training sample set.
In some optional realization methods of the present embodiment, above-mentioned executive agent by the facial image of target object, with There is target object the facial image of at least two relationship object of predetermined genetic connection and combination image to be separately input in advance Before trained face characteristic identification model, there can also be predetermined blood relationship by the facial image of target object, with target object The facial images of at least two relationship objects of relationship and combination image respectively carry out after affine transformation respectively obtains transformation The facial image and the combination image after transformation of relationship object after the facial image of target object, transformation.Above-mentioned affine change It changes and may include：The processing such as zoom in and out, translate, rotating to facial image.
In these optional realization methods, above-mentioned executive agent can also be by the face figure of the target object after transformation The facial image of relationship object after picture, transformation and the combination image after transformation are input to above-mentioned face characteristic identification model and obtain The face characteristic figure and the constitutional diagram after transformation of relationship object after the face characteristic figure of target object after to transformation, transformation The characteristic pattern of picture.
Above-mentioned executive agent can also be by the target object face characteristic figure after the face characteristic figure of target object or transformation As a training book image, from the target object face characteristic figure after transformation, the relationship object face characteristic figure after transformation and A characteristic pattern is arbitrarily chosen in combination image characteristic pattern after transformation as another training sample image, obtains training sample set At least one of conjunction training sample image pair.
It, can be by the target object after the face characteristic figure of target object or transformation in these optional realization methods A training sample of the face characteristic figure as training sample centering, and by the face characteristic figure of the relationship object after transformation or Any feature figure in the characteristic pattern of combination image after transformation generates training sample set as another training sample image At least one of training sample image pair.It can be by these training sample images to being used in step 504 to Initial Face Identification model is trained, and can further expand the quantity of training sample image pair.
Step 504, each training sample in training sample set is input in Initial Face identification model to initial people Face identification model is trained, the human face recognition model after being trained.
Step 504 is identical as the step 202 of embodiment illustrated in fig. 2, does not repeat herein.
From figure 5 it can be seen that compared with the corresponding embodiments of Fig. 2, the generation of the human face recognition model in the present embodiment The flow 500 of method is highlighted combines facial image by the facial image of the facial image of target object, relationship object, is input to Face characteristic identification model trained in advance, obtain the face characteristic figure of target object, the face characteristic figure of relationship object and The step of combining the characteristic pattern of facial image.The scheme of the present embodiment description can be accelerated to obtain the speed of training data as a result,. Further, since the scheme of the present embodiment description is also using the target object face characteristic figure after transformation, the relationship object after transformation Combination image characteristic pattern after face characteristic figure and transformation is trained human face recognition model as training sample image, due to The relationship object face characteristic figure after target object face characteristic figure, transformation after above-mentioned transformation and the combination image after transformation are special Levying figure can be in the feature of different angle reflection target object facial image, the feature of relationship object face and combination image Feature, therefore using the target object face characteristic figure after transformation, the relationship object face characteristic figure after transformation and after converting Combination image characteristic pattern is trained Initial Face identification model the robust of the human face recognition model after can also improving training Property.
With further reference to Fig. 6, as the realization to method shown in above-mentioned each figure, this application provides a kind of recognition of face moulds One embodiment of the generating means of type, the device embodiment is corresponding with embodiment of the method shown in Fig. 2, which specifically may be used To be applied in various electronic equipments.
As shown in fig. 6, the generating means 600 of the human face recognition model of the present embodiment include：Acquiring unit 601, face are known Other model generation unit 602.Wherein, acquiring unit 601 are configured to obtain training sample set；Human face recognition model generates Unit 602 is configured to each training sample in training sample set being input in Initial Face identification model to initial people Face identification model is trained, and the human face recognition model after being trained, human face recognition model is input to therein for identification Whether facial image meets predetermined genetic connection between corresponding object；Wherein, training sample set includes multiple training Sample image pair, at least one of training sample set training sample image are generated to being based on following steps：Obtain target pair A training sample image of the face characteristic figure of elephant as training sample image centering；Obtain has predetermined blood with target object The facial image of at least two relationship objects of edge relationship；Target image set is generated, target image set includes by relationship object The characteristic pattern of Face image synthesis and the characteristic pattern generated by the combination image of relationship object, wherein the combination of relationship object Image is the default characteristic area for the facial image for intercepting one of relationship object, and the default characteristic area intercepted is used in combination to replace Change the image that the corresponding characteristic area of facial image of another relationship object is generated；Arbitrary selection is concentrated from target image Another training sample image of one characteristic pattern as training sample image centering.
In the present embodiment, the acquiring unit 601 of the generating means 600 of human face recognition model and human face recognition model generate The specific processing of unit 602 and its caused technique effect can be respectively with reference to step 201 and steps in 2 corresponding embodiment of figure 202 related description, details are not described herein.
In some optional realization methods of the present embodiment, training sample set further includes following at least one trained sample This image pair, the training sample image to including target image face characteristic figure and with target object do not have predetermined blood relationship The face characteristic figure of the who object of relationship；And human face recognition model generation unit is further configured to：By training sample Each training sample in set, which is input in Initial Face identification model, is trained Initial Face identification model, is trained Human face recognition model afterwards, if so that the image to be detected being input in human face recognition model is to including target object to be detected Face characteristic figure and the face characteristic figure with the target object to be detected with predetermined genetic connection, human face recognition model are exported Numerical value be more than the first predetermined threshold value, if image to be detected centering include target object face characteristic figure and with the target object Face characteristic figure without predetermined genetic connection, the numerical value that human face recognition model is exported are less than the second predetermined threshold value；Wherein, Second predetermined threshold value is less than the first predetermined threshold value.
In some optional realization methods of the present embodiment, acquiring unit is further configured to：Obtain target object Facial image has the respective facial image of at least two relationship objects of predetermined genetic connection with target object and by least The combination image that the facial image of two relationship objects obtains；By the facial image of target object, relationship object facial image, Combination image is input to face characteristic identification model trained in advance, respectively obtains face characteristic figure, the relationship pair of target object The face characteristic figure of elephant and the characteristic pattern for combining image；Using the face characteristic figure of target object as a training book image, From arbitrarily choosing a characteristic pattern as another training sample in the characteristic pattern of the face characteristic figure of relationship object and combination image Image obtains at least one of training sample set training sample image pair.
In some optional realization methods of the present embodiment, acquiring unit is further configured to：By target object Facial image, the facial image of relationship object, combination image are input to face characteristic identification model trained in advance, respectively obtain Before the characteristic pattern of the face characteristic figure of target object, the face characteristic figure of relationship object and combination image, by target object Facial image and relationship object facial image, combination image carry out affine transformation, the target object face after being converted The combination image after relationship object facial image, transformation after image, transformation；And by the target object face figure after transformation Relationship object facial image after picture, transformation, the combination image after transformation are separately input to the face characteristic trained in advance identification Model, the relationship object face characteristic figure after target object face characteristic figure, transformation after being converted and the combination after transformation Characteristics of image figure；And using the target object face characteristic figure after the face characteristic figure of target object or transformation as a training This image, from the target object face characteristic figure after transformation, the relationship object face characteristic figure after transformation and the combination after transformation A characteristic pattern is arbitrarily chosen in characteristics of image figure as another training sample image, obtains at least one in training sample set A training sample image pair.
In some optional realization methods of the present embodiment, above-mentioned human face recognition model is convolutional neural networks model.
Below with reference to Fig. 7, it illustrates the computer systems 700 suitable for the server for realizing the embodiment of the present application Structural schematic diagram.Server shown in Fig. 7 is only an example, should not be to the function and use scope band of the embodiment of the present application Carry out any restrictions.
As shown in fig. 7, computer system 700 includes central processing unit (CPU, Central Processing Unit) 701, it can be according to the program being stored in read-only memory (ROM, Read Only Memory) 702 or from storage section 708 programs being loaded into random access storage device (RAM, Random Access Memory) 703 and execute various appropriate Action and processing.In RAM 703, also it is stored with system 700 and operates required various programs and data.CPU 701、ROM 702 and RAM 703 is connected with each other by bus 704.Input/output (I/O, Input/Output) interface 705 is also connected to Bus 704.
It is connected to I/O interfaces 705 with lower component：Importation 706 including keyboard, mouse etc.；It is penetrated including such as cathode Spool (CRT, Cathode Ray Tube), liquid crystal display (LCD, Liquid Crystal Display) etc. and loud speaker Deng output par, c 707；Storage section 708 including hard disk etc.；And including such as LAN (LAN, Local Area Network) the communications portion 709 of the network interface card of card, modem etc..Communications portion 709 is via such as internet Network executes communication process.Driver 710 is also according to needing to be connected to I/O interfaces 705.Detachable media 711, such as disk, CD, magneto-optic disk, semiconductor memory etc. are mounted on driver 710 as needed, in order to from the calculating read thereon Machine program is mounted into storage section 708 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 709 from network, and/or from detachable media 711 are mounted.When the computer program is executed by central processing unit (CPU) 701, limited in execution the present processes Above-mentioned function.It should be noted that computer-readable medium described herein can be computer-readable signal media or Computer readable storage medium either the two arbitrarily combines.Computer readable storage medium for example can be --- but Be not limited to --- electricity, magnetic, optical, electromagnetic, infrared ray or semiconductor system, device or device, or arbitrary above combination. The more specific example of computer readable storage medium can include but is not limited to：Electrical connection with one or more conducting wires, Portable computer diskette, hard disk, random access storage device (RAM), read-only memory (ROM), erasable type may be programmed read-only deposit Reservoir (EPROM or flash memory), optical fiber, portable compact disc read-only memory (CD-ROM), light storage device, magnetic memory Part or above-mentioned any appropriate combination.In this application, computer readable storage medium can any be included or store The tangible medium of program, the program can be commanded the either device use or in connection of execution system, device.And In the application, computer-readable signal media may include the data letter propagated in a base band or as a carrier wave part Number, wherein carrying computer-readable program code.Diversified forms may be used in the data-signal of this propagation, including but not It is limited to electromagnetic signal, optical signal or above-mentioned any appropriate combination.Computer-readable signal media can also be computer Any computer-readable medium other than readable storage medium storing program for executing, the computer-readable medium can send, propagate or transmit use In by instruction execution system, device either device use or program in connection.Include on computer-readable medium Program code can transmit with any suitable medium, including but not limited to：Wirelessly, electric wire, optical cable, RF etc., Huo Zheshang Any appropriate combination stated.
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 Including conventional procedural programming language-such as " 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 acquiring unit and human face recognition model generation unit.Wherein, the title of these units is not constituted to this under certain conditions The restriction of unit itself, for example, acquiring unit is also described as " obtaining the unit of training sample set ".
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：Obtain training sample set；Each training sample in the training sample set is input to Initial Face identification model In the Initial Face identification model is trained, the human face recognition model after being trained, the human face recognition model is used It is input to whether facial image therein meets predetermined genetic connection between corresponding object in identification；Wherein, the instruction It includes multiple training sample images pair to practice sample set, and at least one of described training sample set training sample image is to base It is generated in following steps：Obtain a training sample figure of the face characteristic figure of target object as training sample image centering Picture；Obtain the facial image at least two relationship objects that there is the predetermined genetic connection with the target object；Generate mesh It marks on a map image set, the target image set includes by the characteristic pattern of the Face image synthesis of the relationship object and by the relationship The characteristic pattern that the combination image of object generates, wherein the combination image of the relationship object is to intercept one of relationship object Facial image default characteristic area, be used in combination the default characteristic area intercepted to replace the facial image of another relationship object The image that corresponding characteristic area is generated；Concentrate one characteristic pattern of arbitrary selection as training sample from the target image Another training sample image of image pair.
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 generation method of human face recognition model, including：
Obtain training sample set；
Each training sample in the training sample set is input in Initial Face identification model, the Initial Face is known Other model is trained, and the human face recognition model after being trained, the human face recognition model is input to therein for identification Whether facial image meets predetermined genetic connection between corresponding object；
Wherein, the training sample set includes multiple training sample images pair, at least one of described training sample set Training sample image is generated to being based on following steps：
Obtain a training sample image of the face characteristic figure of target object as training sample image centering；
Obtain the facial image at least two relationship objects that there is the predetermined genetic connection with the target object；
Generate target image set, the target image set include by the Face image synthesis of the relationship object characteristic pattern and The characteristic pattern generated by the combination image of the relationship object, wherein the combination image of the relationship object is interception wherein one The default characteristic area of the facial image of a relationship object is used in combination the default characteristic area intercepted to replace another relationship object The image that is generated of the corresponding characteristic area of facial image；
Another training sample of one characteristic pattern of arbitrary selection as training sample image centering is concentrated from the target image Image.
2. according to the method described in claim 1, wherein, the training sample set further includes following at least one training sample Image pair, the training sample image to including the target image face characteristic figure and do not have with the target object pre- Determine the face characteristic figure of the who object of genetic connection；And
Each training sample by the training sample set is input in Initial Face identification model to the initial people Face identification model is trained, the human face recognition model after being trained, including：
Each training sample in the training sample set is input in Initial Face identification model, the Initial Face is known Other model is trained, the human face recognition model after being trained, if so as to being input to the mapping to be checked in human face recognition model As to including target object to be detected face characteristic figure and with the target object to be detected have predetermined genetic connection face Characteristic pattern, the numerical value that the human face recognition model is exported is more than the first predetermined threshold value, if image to be detected centering includes target The face characteristic figure of object and the face characteristic figure for not having predetermined genetic connection with the target object, the human face recognition model The numerical value exported is less than the second predetermined threshold value；Wherein, second predetermined threshold value is less than first predetermined threshold value.
3. according to the method described in claim 1, wherein, the acquisition training sample set includes：
It obtains the facial image of target object, have at least two relationship objects of predetermined genetic connection each with the target object From facial image and the combination image that is obtained by the facial image of at least two relationship object；
The facial image of target object, the facial image of the relationship object, the combination image are input to training in advance Face characteristic identification model respectively obtains the face characteristic figure of target object, the face characteristic figure of relationship object and constitutional diagram The characteristic pattern of picture；
Using the face characteristic figure of target object as a training book image, from the face characteristic figure of the relationship object and combination A characteristic pattern is arbitrarily chosen in the characteristic pattern of image as another training sample image, is obtained in training sample set at least One training sample image pair.
4. according to the method described in claim 3, wherein, described by the facial image of target object, the relationship object Facial image, the combination image are input to face characteristic identification model trained in advance, respectively obtain the face of target object Before the characteristic pattern of characteristic pattern, the face characteristic figure of relationship object and combination image, the method further includes：
The facial image of the facial image of the target object and the relationship object, combination image are subjected to affine transformation, obtained The relationship object facial image after target object facial image, transformation, the combination image after transformation after to transformation；And
It is described that the facial image of target object, the facial image of the relationship object, the combination image are input to advance instruction Experienced face characteristic identification model respectively obtains the face characteristic figure of target object, the face characteristic figure of relationship object and group Close image characteristic pattern, further include：By the target object facial image after the transformation, the relationship object face figure after transformation Combination image after picture, transformation is separately input to the face characteristic identification model trained in advance, the target object after being converted The combination image characteristic pattern after relationship object face characteristic figure and transformation after face characteristic figure, transformation；And
It is described using the face characteristic figure of target object as a training book image, from the face characteristic figure of the relationship object and It combines in the characteristic pattern of image and arbitrarily chooses a characteristic pattern as another training sample image, obtain in training sample set At least one training sample image pair further includes：
Using the target object face characteristic figure after the face characteristic figure of target object or transformation as a training book image, from institute State the target object face characteristic figure after transformation, the relationship object face characteristic figure after transformation and the combination characteristics of image after transformation A characteristic pattern is arbitrarily chosen in figure as another training sample image, is obtained at least one of training sample set and is trained sample This image pair.
5. according to the method described in claim 1, wherein, the human face recognition model is convolutional neural networks model.
6. a kind of generating means of human face recognition model, including：
Acquiring unit is configured to obtain training sample set；
Human face recognition model generation unit is configured to each training sample in the training sample set being input to initial people The Initial Face identification model is trained in face identification model, the human face recognition model after being trained, the face Identification model is input to whether facial image therein meets predetermined genetic connection between corresponding object for identification；
Wherein, the training sample set includes multiple training sample images pair, at least one of described training sample set Training sample image is generated to being based on following steps：
Obtain a training sample image of the face characteristic figure of target object as training sample image centering；
Obtain the facial image at least two relationship objects that there is the predetermined genetic connection with the target object；
Generate target image set, the target image set include by the Face image synthesis of the relationship object characteristic pattern and The characteristic pattern generated by the combination image of the relationship object, wherein the combination image of the relationship object is interception wherein one The default characteristic area of the facial image of a relationship object is used in combination the default characteristic area intercepted to replace another relationship object The image that is generated of the corresponding characteristic area of facial image；
Another training sample of one characteristic pattern of arbitrary selection as training sample image centering is concentrated from the target image Image.
7. according to the method described in claim 6, wherein, the training sample set further includes following at least one training sample Image pair, the training sample image to including the target image face characteristic figure and do not have with the target object pre- Determine the face characteristic figure of the who object of genetic connection；And
The human face recognition model generation unit is further configured to：
Each training sample in the training sample set is input in Initial Face identification model, the Initial Face is known Other model is trained, the human face recognition model after being trained, if so as to being input to the mapping to be checked in human face recognition model As to including target object to be detected face characteristic figure and with the target object to be detected have predetermined genetic connection face Characteristic pattern, the numerical value that the human face recognition model is exported is more than the first predetermined threshold value, if image to be detected centering includes target The face characteristic figure of object and the face characteristic figure for not having predetermined genetic connection with the target object, the human face recognition model The numerical value exported is less than the second predetermined threshold value；Wherein, second predetermined threshold value is less than first predetermined threshold value.
8. device according to claim 6, wherein the acquiring unit is further configured to：Obtain target object Facial image, with the target object have predetermined genetic connection the respective facial image of at least two relationship objects and by The combination image that the facial image of at least two relationship object obtains；
The facial image of target object, the facial image of the relationship object, the combination image are input to training in advance Face characteristic identification model respectively obtains the face characteristic figure of target object, the face characteristic figure of relationship object and constitutional diagram The characteristic pattern of picture；
Using the face characteristic figure of target object as a training book image, from the face characteristic figure of the relationship object and combination A characteristic pattern is arbitrarily chosen in the characteristic pattern of image as another training sample image, is obtained in training sample set at least One training sample image pair.
9. device according to claim 8, wherein the acquiring unit is further configured to：
The facial image of target object, the facial image of the relationship object, the combination image are input in advance described Trained face characteristic identification model, respectively obtain the face characteristic figure of target object, the face characteristic figure of relationship object and Before the characteristic pattern for combining image, by the facial image of the facial image of the target object and the relationship object, constitutional diagram As carrying out affine transformation, the relationship object facial image after target object facial image, transformation after being converted, after transformation Combine image；And
By the target object facial image after the transformation, the relationship object facial image after transformation, the combination image after transformation It is separately input to the face characteristic identification model trained in advance, after the target object face characteristic figure, transformation after being converted Combination image characteristic pattern after relationship object face characteristic figure and transformation；And
Using the target object face characteristic figure after the face characteristic figure of target object or transformation as a training book image, from institute State the target object face characteristic figure after transformation, the relationship object face characteristic figure after transformation and the combination characteristics of image after transformation A characteristic pattern is arbitrarily chosen in figure as another training sample image, is obtained at least one of training sample set and is trained sample This image pair.
10. device according to claim 6, wherein the human face recognition model is convolutional neural networks model.
11. a kind of server, 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 medium, is stored thereon with computer program, wherein the program is realized when being executed by processor Method as described in any in claim 1-5.
CN201810268892.1A 2018-03-29 2018-03-29 Method and device for generating face recognition model Active CN108491812B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810268892.1A CN108491812B (en) 2018-03-29 2018-03-29 Method and device for generating face recognition model

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810268892.1A CN108491812B (en) 2018-03-29 2018-03-29 Method and device for generating face recognition model

Publications (2)

Publication Number Publication Date
CN108491812A true CN108491812A (en) 2018-09-04
CN108491812B CN108491812B (en) 2022-05-03

Family

ID=63317307

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810268892.1A Active CN108491812B (en) 2018-03-29 2018-03-29 Method and device for generating face recognition model

Country Status (1)

Country Link
CN (1) CN108491812B (en)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110163222A (en) * 2018-10-08 2019-08-23 腾讯科技（深圳）有限公司 A kind of method of image recognition, the method for model training and server
CN111368685A (en) * 2020-02-27 2020-07-03 北京字节跳动网络技术有限公司 Key point identification method and device, readable medium and electronic equipment
CN111860034A (en) * 2019-04-24 2020-10-30 广州煜煊信息科技有限公司 Household accident handling method
CN113393265A (en) * 2021-05-25 2021-09-14 浙江大华技术股份有限公司 Method for establishing database of feature library of passing object, electronic device and storage medium
CN113705276A (en) * 2020-05-20 2021-11-26 武汉Tcl集团工业研究院有限公司 Model construction method, model construction device, computer apparatus, and medium

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20000059431A (en) * 1999-03-03 2000-10-05 조성우 Family identification system from genetic database using DNA typing method and combination of markers for the system
CN105005774A (en) * 2015-07-28 2015-10-28 中国科学院自动化研究所 Face relative relation recognition method based on convolutional neural network and device thereof
CN105488463A (en) * 2015-11-25 2016-04-13 康佳集团股份有限公司 Lineal relationship recognizing method and system based on face biological features
CN106384116A (en) * 2016-08-29 2017-02-08 北京农业信息技术研究中心 Terahertz imaging based plant vein recognition method and device
CN106709482A (en) * 2017-03-17 2017-05-24 中国人民解放军国防科学技术大学 Method for identifying genetic relationship of figures based on self-encoder
CN106951858A (en) * 2017-03-17 2017-07-14 中国人民解放军国防科学技术大学 A kind of recognition methods of personage's affiliation and device based on depth convolutional network
CN107229902A (en) * 2017-04-12 2017-10-03 南京晓庄学院 A kind of twins' recognition methods based on improvement SVM

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20000059431A (en) * 1999-03-03 2000-10-05 조성우 Family identification system from genetic database using DNA typing method and combination of markers for the system
CN105005774A (en) * 2015-07-28 2015-10-28 中国科学院自动化研究所 Face relative relation recognition method based on convolutional neural network and device thereof
CN105488463A (en) * 2015-11-25 2016-04-13 康佳集团股份有限公司 Lineal relationship recognizing method and system based on face biological features
CN106384116A (en) * 2016-08-29 2017-02-08 北京农业信息技术研究中心 Terahertz imaging based plant vein recognition method and device
CN106709482A (en) * 2017-03-17 2017-05-24 中国人民解放军国防科学技术大学 Method for identifying genetic relationship of figures based on self-encoder
CN106951858A (en) * 2017-03-17 2017-07-14 中国人民解放军国防科学技术大学 A kind of recognition methods of personage's affiliation and device based on depth convolutional network
CN107229902A (en) * 2017-04-12 2017-10-03 南京晓庄学院 A kind of twins' recognition methods based on improvement SVM

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
A. BOTTINO 等: "A NEW PROBLEM IN FACE IMAGE ANALYSIS: FINDING KINSHIP CLUES FOR SIBLINGS PAIRS", 《PROCEEDINGS OF THE 1ST INTERNATIONAL CONFERENCE ON PATTERN RECOGNITION APPLICATIONS AND METHODS》 *
QINGYAN DUAN 等: "From Face Recognition to Kinship Verification: An Adaptation Approach", 《2017 IEEE INTERNATIONAL CONFERENCE ON COMPUTER VISION WORKSHOPS (ICCVW)》 *
李珏 等: "利用深度学习在人脸图像中识别亲缘关系", 《第十一届和谐人机环境联合会议论文集》 *

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110163222A (en) * 2018-10-08 2019-08-23 腾讯科技（深圳）有限公司 A kind of method of image recognition, the method for model training and server
CN111860034A (en) * 2019-04-24 2020-10-30 广州煜煊信息科技有限公司 Household accident handling method
CN111368685A (en) * 2020-02-27 2020-07-03 北京字节跳动网络技术有限公司 Key point identification method and device, readable medium and electronic equipment
CN111368685B (en) * 2020-02-27 2023-09-29 北京字节跳动网络技术有限公司 Method and device for identifying key points, readable medium and electronic equipment
CN113705276A (en) * 2020-05-20 2021-11-26 武汉Tcl集团工业研究院有限公司 Model construction method, model construction device, computer apparatus, and medium
CN113393265A (en) * 2021-05-25 2021-09-14 浙江大华技术股份有限公司 Method for establishing database of feature library of passing object, electronic device and storage medium

Also Published As

Publication number Publication date
CN108491812B (en) 2022-05-03

Similar Documents

Publication Publication Date Title
CN108509915A (en) The generation method and device of human face recognition model
CN107679490B (en) Method and apparatus for detection image quality
CN108898185A (en) Method and apparatus for generating image recognition model
CN108491812A (en) The generation method and device of human face recognition model
CN110288049A (en) Method and apparatus for generating image recognition model
CN108830235A (en) Method and apparatus for generating information
CN108898186A (en) Method and apparatus for extracting image
CN108446387A (en) Method and apparatus for updating face registration library
CN108229419A (en) For clustering the method and apparatus of image
CN108073910A (en) For generating the method and apparatus of face characteristic
CN110443222A (en) Method and apparatus for training face's critical point detection model
CN109034069A (en) Method and apparatus for generating information
CN108491823A (en) Method and apparatus for generating eye recognition model
CN108984399A (en) Detect method, electronic equipment and the computer-readable medium of interface difference
CN108509892A (en) Method and apparatus for generating near-infrared image
CN109344752A (en) Method and apparatus for handling mouth image
CN109241934A (en) Method and apparatus for generating information
CN109344762A (en) Image processing method and device
CN111931628B (en) Training method and device of face recognition model and related equipment
CN110009059A (en) Method and apparatus for generating model
CN108363999A (en) Operation based on recognition of face executes method and apparatus
CN108509921A (en) Method and apparatus for generating information
CN108446658A (en) The method and apparatus of facial image for identification
CN109145783A (en) Method and apparatus for generating information
CN109871791A (en) Image processing method and device

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant