CN109858445A - Method and apparatus for generating model - Google Patents
Method and apparatus for generating model Download PDFInfo
- Publication number
- CN109858445A CN109858445A CN201910099403.9A CN201910099403A CN109858445A CN 109858445 A CN109858445 A CN 109858445A CN 201910099403 A CN201910099403 A CN 201910099403A CN 109858445 A CN109858445 A CN 109858445A
- Authority
- CN
- China
- Prior art keywords
- key point
- face key
- point information
- training
- sample
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Landscapes
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
Embodiment of the disclosure discloses the method and apparatus for generating model.One specific embodiment of this method includes: acquisition training sample set；It is concentrated from training sample and chooses training sample, execute following training step: the sample facial image in training sample is inputted to the feature extraction layer of initial neural network, obtain characteristics of image；Characteristics of image is inputted to the first sub-network of initial neural network, generates face key point information；Face key point information and characteristics of image are inputted to the second sub-network of initial neural network, obtain the deviation of face key point information；Based on the sample face key point information in face key point information and training sample, the expectation deviation of face key point information is determined；Based on deviation and desired deviation, determine whether initial neural network trains completion；In response to determining that training is completed, the initial neural network that training is completed is determined as face key point identification model.The embodiment helps to realize more accurate face critical point detection.
Description
Technical field
Embodiment of the disclosure is related to field of computer technology, more particularly, to generates the method and apparatus of model.
Background technique
With the development of face recognition technology, face critical point detection technology is also widely applied, such as spy
Effect addition, human face three-dimensional model building etc..
Face key point refers to the point with the indexing of obvious semantic space in face.Currently, the stream of face critical point detection
Journey, which is generally, inputs a face critical point detection model trained in advance for facial image to be detected, obtains testing result.
Summary of the invention
Embodiment of the disclosure proposes the method and apparatus for generating model, and the side for handling facial image
Method and device.
In a first aspect, the embodiment of the present disclosure provides a kind of method for generating model, this method comprises: obtaining training
Sample set, wherein training sample includes that sample facial image and the sample face marked in advance for sample facial image are crucial
Point information；It is concentrated from training sample and chooses training sample, and execute following training step: will be in selected training sample
Sample facial image inputs the feature extraction layer of initial neural network, obtains characteristics of image；Characteristics of image obtained is inputted
First sub-network of initial neural network, generates the face key point information of sample facial image；Face generated is crucial
Point information and characteristics of image input the second sub-network of initial neural network, obtain deviation corresponding to face key point information；
Based on the sample face key point information in face key point information and training sample, face key point information generated is determined
Corresponding expectation deviation；Based on deviation corresponding to face key point information and desired deviation, determine that initial neural network is
No training is completed；In response to determining that training is completed, the initial neural network that training is completed is determined as face key point identification mould
Type.
In some embodiments, the second sub-network includes the first generation layer and the second generation layer；And by people generated
Face key point information and characteristics of image input the second sub-network of initial neural network, obtain corresponding to face key point information
Deviation, comprising: face key point information generated is inputted to the first generation layer of the second sub-network, obtains face key point letter
The corresponding hotspot graph of breath, wherein the image-region of hotspot graph includes numerical value set, for the numerical value in numerical value set, the number
Value is for characterizing face key point in the probability of the numerical value position；By hotspot graph obtained and characteristics of image input the
Second generation layer of two sub-networks obtains deviation corresponding to face key point information.
In some embodiments, this method further include: training is not completed in response to the initial neural network of determination, and adjustment is initial
Relevant parameter in neural network is concentrated from training sample and chooses training sample in unselected training sample, and uses
The initial neural network of the last time adjustment and the last training sample chosen, continue to execute training step.
Second aspect, embodiment of the disclosure provides a kind of method for handling facial image, this method comprises: obtaining
Take target facial image；Target facial image is inputted into the side using any embodiment in method described in above-mentioned first aspect
In the face key point identification model that method generates, generates face key point information corresponding to target facial image and face is crucial
Deviation corresponding to point information；Based on face key point information generated and deviation, generate corresponding to target facial image
As a result face key point information.
The third aspect, embodiment of the disclosure provide it is a kind of for generating the device of model, the device include: obtain it is single
Member is configured to obtain training sample set, wherein training sample include sample facial image and for sample facial image it is preparatory
The sample face key point information of mark；Training unit is configured to concentrate from training sample and chooses training sample, and executes
Sample facial image in selected training sample: being inputted the feature extraction layer of initial neural network by following training step,
Obtain characteristics of image；Characteristics of image obtained is inputted to the first sub-network of initial neural network, generates sample facial image
Face key point information；Face key point information generated and characteristics of image are inputted to the second subnet of initial neural network
Network obtains deviation corresponding to face key point information；It is closed based on the sample face in face key point information and training sample
Key point information, determines expectation deviation corresponding to face key point information generated；Based on corresponding to face key point information
Deviation and desired deviation, determine whether initial neural network trains completion；In response to determining that training is completed, training is completed
Initial neural network is determined as face key point identification model.
In some embodiments, the second sub-network includes the first generation layer and the second generation layer；And training unit is into one
Step is configured to: face key point information generated being inputted to the first generation layer of the second sub-network, obtains face key point
Hotspot graph corresponding to information, wherein the image-region of hotspot graph includes numerical value set, should for the numerical value in numerical value set
Numerical value is for characterizing face key point in the probability of the numerical value position；Hotspot graph obtained and characteristics of image are inputted
Second generation layer of the second sub-network obtains deviation corresponding to face key point information.
In some embodiments, device further include: adjustment unit is configured in response to determine initial neural network not
Training is completed, and is adjusted the relevant parameter in initial neural network, is concentrated from training sample and choose in unselected training sample
Training sample, and initial neural network and the last training sample chosen using the last adjustment, continue to execute
Training step.
Fourth aspect, embodiment of the disclosure provide a kind of for handling the device of facial image, which includes: figure
As acquiring unit, it is configured to obtain target facial image；First generation unit is configured to input target facial image and adopt
In the face key point identification model that the method for any embodiment generates in the method described in above-mentioned first aspect, mesh is generated
Mark deviation corresponding to face key point information corresponding to facial image and face key point information；Second generation unit, quilt
It is configured to generate result face key point corresponding to target facial image based on face key point information generated and deviation
Information.
5th aspect, embodiment of the disclosure provide a kind of electronic equipment, which includes: one or more places
Manage device；Storage device is stored thereon with one or more programs；When one or more programs are held by one or more processors
Row, so that one or more processors realize the side of any embodiment in the method as described in first aspect and second aspect
Method.
6th aspect, embodiment of the disclosure provide a kind of computer-readable medium, are stored thereon with computer program,
The program realizes any embodiment in method described in above-mentioned first aspect and second aspect method when being executed by processor.
The method and apparatus for generating model that embodiment of the disclosure provides, by obtaining training sample set, wherein
Training sample includes sample facial image and the sample face key point information that marks in advance for sample facial image, then from
Training sample, which is concentrated, chooses training sample, and executes following training step: by the sample face in selected training sample
Image inputs the feature extraction layer of initial neural network, obtains characteristics of image；Characteristics of image obtained is inputted into initial nerve
First sub-network of network generates the face key point information of sample facial image；By face key point information generated and
Characteristics of image inputs the second sub-network of initial neural network, obtains deviation corresponding to face key point information；Based on face
Sample face key point information in key point information and training sample, determines corresponding to face key point information generated
It is expected that deviation；Based on deviation corresponding to face key point information and desired deviation, determine whether initial neural network has trained
At；In response to determining that training is completed, the initial neural network that training is completed is determined as face key point identification model, thus institute
The face key point identification model of generation can predict the face key point information of facial image, and the face predicted simultaneously
The deviation of key point information facilitates face key point information and deviation using model prediction, generates more accurate result
Face key point information realizes more accurate face critical point detection.
Detailed description of the invention
By reading a detailed description of non-restrictive embodiments in the light of the attached drawings below, the disclosure is other
Feature, objects and advantages will become more apparent upon:
Fig. 1 is that one embodiment of the disclosure can be applied to exemplary system architecture figure therein；
Fig. 2 is the flow chart according to one embodiment of the method for generating model of the disclosure；
Fig. 3 is according to an embodiment of the present disclosure for generating the schematic diagram of an application scenarios of the method for model；
Fig. 4 is the flow chart according to another embodiment of the method for handling facial image of the disclosure；
Fig. 5 is the structural schematic diagram according to one embodiment of the device for generating model of the disclosure；
Fig. 6 is the structural schematic diagram according to one embodiment of the device for handling facial image of the disclosure；
Fig. 7 is adapted for the structural schematic diagram for the computer system for realizing the electronic equipment of embodiment of the disclosure.
Specific embodiment
The disclosure 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, part relevant to related invention is illustrated only in attached drawing.
It should be noted that in the absence of conflict, the feature in embodiment and embodiment in the disclosure can phase
Mutually combination.The disclosure is described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.
Fig. 1 shows the method for generating model that can apply the disclosure or the device for generating model, and
For handling the exemplary system architecture 100 of the method for facial image or the embodiment of the device for handling facial image.
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 to provide the medium of communication link.Network 104 can be with
Including various connection types, such as wired, wireless communication link or fiber optic cables etc..
User can be used terminal device 101,102,103 and be interacted by network 104 with server 105, to receive or send out
Send message etc..Various telecommunication customer end applications can be installed, such as image processing class is soft on terminal device 101,102,103
Part, web browser applications, searching class application, instant messaging tools, social platform software etc..
Terminal device 101,102,103 can be hardware, be also possible to software.When terminal device 101,102,103 is hard
When part, it can be various electronic equipments, including but not limited to smart phone, tablet computer, E-book reader, MP3 player
(Moving Picture Experts Group Audio Layer III, dynamic image expert's compression standard audio level 3),
MP4 (Moving Picture Experts Group Audio Layer IV, dynamic image expert's compression standard audio level
4) player, pocket computer on knee and desktop computer etc..It, can be with when terminal device 101,102,103 is software
It is mounted in above-mentioned cited electronic equipment.Multiple softwares or software module may be implemented into (such as providing distribution in it
The multiple softwares or software module of formula service), single software or software module also may be implemented into.It is not specifically limited herein.
Server 105 can be to provide the server of various services, such as can be and utilize terminal device 101,102,103
The training sample set of upload carries out the model training server of model training.Model training server can use the training of acquisition
Sample set carries out model training, generates face key point identification model.In addition, after training obtains face key point identification model,
Server can send face key point identification model to terminal device 101,102,103, also can use face key point
Identification model carries out the identification of face key point to facial image.
It should be noted that can be by server 105 for generating the method for model provided by embodiment of the disclosure
It executes, can also be executed by terminal device 101,102,103, correspondingly, the device for generating model can be set in service
In device 105, also it can be set in terminal device 101,102,103.In addition, for handling provided by embodiment of the disclosure
The method of facial image can be executed by server 105, can also be executed by terminal device 101,102,103, correspondingly, be used for
The device of processing facial image can be set in server 105, also can be set in terminal device 101,102,103.
It should be noted that server can be hardware, it is also possible to software.When server is hardware, may be implemented
At the distributed server cluster that multiple servers form, individual server also may be implemented into.It, can when server is software
It, can also be with to be implemented as multiple softwares or software module (such as providing multiple softwares of Distributed Services or software module)
It is implemented as single software or software module.It is not specifically limited herein.
It should be understood that the number of 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.Training sample set needed for training pattern is not required to from remote
Journey obtains or target facial image to be processed is not required in the case where long-range obtain, and above system framework can not include
Network, and only need server or terminal device.
With continued reference to Fig. 2, the process of one embodiment of the method for generating model according to the disclosure is shown
200.The method for being used to generate model, comprising the following steps:
Step 201, training sample set is obtained.
In the present embodiment, can lead to for generating the executing subject (such as server shown in FIG. 1) of the method for model
It crosses wired connection mode or radio connection obtains training sample set.Wherein, training sample include sample facial image and
The sample face key point information marked in advance for sample facial image.Sample facial image can be for the progress of sample face
Shoot image obtained.Sample face key point information is for characterizing position of the sample face key point in sample facial image
It sets, can include but is not limited at least one of following: number, text, symbol, image.For example, sample face key point information can
Think the key point coordinate for characterizing position of the sample face key point in sample facial image.Herein, key point is sat
Mark can be the coordinate under the coordinate system pre-established based on sample facial image.
In practice, face key point can be point crucial in face, specifically, can be influence face mask or five
The point of official's shape.As an example, face key point can be point corresponding to point, eyes corresponding to nose etc..
Specifically, above-mentioned executing subject is available to be pre-stored within local training sample set, also available communication
The training sample set that the electronic equipment (such as terminal device shown in FIG. 1) of connection is sent.
Step 202, it is concentrated from training sample and chooses training sample, and execute following training step: by selected instruction
Practice the feature extraction layer that the sample facial image in sample inputs initial neural network, obtains characteristics of image；By figure obtained
As feature inputs the first sub-network of initial neural network, the face key point information of generation sample facial image；It will be generated
Face key point information and characteristics of image input the second sub-network of initial neural network, it is right to obtain face key point information institute
The deviation answered；Based on the sample face key point information in face key point information and training sample, face generated is determined
Expectation deviation corresponding to key point information；Based on deviation corresponding to face key point information and desired deviation, determine initial
Whether neural network trains completion；In response to determining that training is completed, the initial neural network that training is completed is determined as face and is closed
Key point identification model.
In the present embodiment, based on training sample set obtained in step 201, above-mentioned executing subject can be from training sample
It concentrates and chooses training sample, and the following training step of execution (step 2021- step 2026):
Step 2021, the sample facial image in selected training sample is inputted to the feature extraction of initial neural network
Layer obtains characteristics of image.
Wherein, characteristics of image can be the features such as color, the shape of image.Initial neural network is predetermined, use
In the various neural networks (such as convolutional neural networks) for generating face key point identification model.Face key point identification model can
With face key point corresponding to facial image for identification.Herein, initial neural network can be unbred nerve
Network, or do not train by training but the neural network completed.Specifically, initial neural network includes feature extraction
Layer.Feature extraction layer is used to extract the characteristics of image of inputted facial image.Specifically, feature extraction layer includes that can extract
In addition the structure (such as convolutional layer) of characteristics of image also may include other structures (such as pond layer), herein with no restrictions.
Step 2022, characteristics of image obtained is inputted to the first sub-network of initial neural network, generates sample face
The face key point information of image.
In the present embodiment, initial neural network further includes the first sub-network.First sub-network is connect with feature extraction layer,
Characteristics of image for being exported based on feature extraction layer generates face key point information.Face key point information generated is
Face key point information corresponding to first sub-network predicts, sample facial image.
It is appreciated that often there is error between measured value and true value, so pre- using initial neural network in practice
The face key point information and practical face key point information measured are also typically present difference.
It should be noted that herein, the first sub-network may include for generating result (face key point information)
Structure (such as classifier, full articulamentum) can also include (such as defeated for exporting the structure of result (face key point information)
Layer out).
Step 2023, face key point information generated and characteristics of image are inputted to the second subnet of initial neural network
Network obtains deviation corresponding to face key point information.
In the present embodiment, initial neural network further includes the second sub-network, the second sub-network respectively with the first sub-network
It is connected with feature extraction layer, the characteristics of image for being exported based on feature extraction layer, determines that the face of the first sub-network output closes
Deviation corresponding to key point information.Deviation corresponding to face key point information is for characterizing face key point information generated
The difference of practical face key point information relative to sample facial image.Here, the second sub-network deviation generated is base
In the deviation that characteristics of image predicts.
It should be noted that herein, the second sub-network may include that (face key point information institute is right for generating result
The deviation answered) structure (such as classifier, full articulamentum), can also include for exporting result (face key point information institute
Corresponding deviation) structure (such as output layer).
In some optional implementations of the present embodiment, the second sub-network includes that the first generation layer and second generate
Layer；And above-mentioned executing subject can obtain deviation corresponding to face key point information by following steps:
Firstly, face key point information generated to be inputted to the first generation layer of the second sub-network, it is crucial to obtain face
Hotspot graph corresponding to point information.
Wherein, the first generation layer is connect with the first sub-network, the face key point letter for being exported based on the first sub-network
Breath generates hotspot graph corresponding to face key point information.Here, the image-region of hotspot graph includes numerical value set.For number
Numerical value in value set, the numerical value is for characterizing face key point in the probability of the numerical value position.It should be noted that
Hotspot graph is identical as the shape of sample facial image, size difference.And then the position where the numerical value in hotspot graph can be with sample
Position in this facial image is corresponding, and therefore, hotspot graph can serve to indicate that the face key point in sample facial image
Position.
It should be noted that hotspot graph may include at least one numerical value set, wherein at least one numerical value set
Each numerical value set can correspond to a face key point information.
Specifically, in, hotspot graph corresponding with the position in the sample facial image that face key point information is characterized
Position on numerical value can be 1.According to each position in hotspot graph with numerical value 1 corresponding at a distance from position, Ge Gewei
Setting corresponding numerical value can be gradually reduced.I.e. position corresponding to distance values 1 is remoter, and corresponding numerical value is smaller.
It should be noted that the position where numerical value in hotspot graph can be by the minimum rectangle for surrounding numerical value Lai really
It is fixed.Specifically, the center of above-mentioned minimum rectangle can be determined as numerical value position, alternatively, can be by minimum rectangle
Endpoint location be determined as numerical value position.
Then, the second generation layer that hotspot graph obtained and characteristics of image are inputted to the second sub-network obtains face and closes
Deviation corresponding to key point information.
Wherein, the second generation layer is connect with the first generation layer and feature extraction layer respectively, for defeated based on feature extraction layer
The hotspot graph of characteristics of image and the output of the first generation layer out, determines corresponding to the face key point information for inputting the first generation layer
Deviation.
In this implementation, hotspot graph can indicate the position range of face key point, believe compared to face key point
Breath, can more accurate, easily return out the position feature of face key point using hotspot graph, so can it is more quick,
Accurately generate deviation corresponding to face key point information.
Step 2024, based on the sample face key point information in face key point information and training sample, determination is given birth to
At face key point information corresponding to expectation deviation.
Herein, above-mentioned executing subject can determine face key point information that the first sub-network generates and mark in advance
The difference of sample face key point information, and then identified difference is determined as it is expected deviation.
As an example, the face key point information that the first sub-network generates is coordinate " (10,19) ".Sample face key point
Information is coordinate " (11,18) ".Then it is expected that deviation can be (- 1,1), wherein -1=10-11, for characterizing the difference of abscissa
It is different；1=19-18, for characterizing the difference of ordinate.
Step 2025, based on deviation corresponding to face key point information and desired deviation, determine that initial neural network is
No training is completed.
Herein, above-mentioned executing subject can use preset loss function and calculate obtained face key point information institute
Difference between corresponding deviation and expectation deviation, and then determine whether the difference being calculated is less than or equal to default difference threshold
Value determines that initial neural metwork training is completed in response to determining that the difference being calculated is less than or equal to default discrepancy threshold.Its
In, preset loss function can be various loss functions, for example, L2 norm or Euclidean distance etc..
Step 2026, in response to determining that training is completed, the initial neural network that training is completed is determined as face key point
Identification model.
In the present embodiment, above-mentioned executing subject can be completed in response to the initial neural metwork training of determination, and training is complete
At initial neural network be determined as face key point identification model.
In some optional implementations of the present embodiment, above-mentioned executing subject may also respond to determine initial nerve
Network not complete by training, adjusts the relevant parameter in initial neural network, unselected training sample is concentrated from training sample
Middle selection training sample, and initial neural network and the last training sample chosen using the last adjustment, after
It is continuous to execute above-mentioned training step (step 2021- step 2026).
Specifically, above-mentioned executing subject can training not be completed in response to the initial neural network of determination, obtained by calculating
Difference, adjust the relevant parameter of initial neural network.Here it is possible to adopt in various manners based on the discrepancy adjustment being calculated
The relevant parameter of initial neural network.For example, BP (Back Propagation, backpropagation) algorithm or SGD can be used
(Stochastic Gradient Descent, stochastic gradient descent) algorithm adjusts the relevant parameter of initial neural network.
With continued reference to the signal that Fig. 3, Fig. 3 are according to the application scenarios of the method for generating model of the present embodiment
Figure.In the application scenarios of Fig. 3, server 301 can obtain training sample set 302 first, wherein training sample includes sample
Facial image and the sample face key point information marked in advance for sample facial image.Sample face key point information is used for
Characterize the position of the sample face key point in sample facial image.For example, sample face key point information can be sample people
The coordinate of face key point.
Then, server 301 can choose training sample 3021 from training sample set 302, and execute following training
Step: the sample facial image 30211 in selected training sample 3021 is inputted to the feature extraction of initial neural network 303
Layer 3031 obtains characteristics of image 304；Characteristics of image 304 obtained is inputted to the first sub-network of initial neural network 303
3032, generate the face key point information 305 of sample facial image 30211；By face key point information 305 generated and figure
As feature 304 inputs the second sub-network 3033 of initial neural network 303, obtain inclined corresponding to face key point information 305
Poor 306；Based on the sample face key point information 30212 in face key point information 305 and training sample 3021, determination is given birth to
At face key point information 305 corresponding to expectation deviation 307；Based on deviation 306 corresponding to face key point information 305
With desired deviation 307, determine whether initial neural network 303 trains completion；In response to determining that training is completed, training is completed
Initial neural network 303 is determined as face key point identification model 308.
The method provided by the above embodiment of disclosure face key point identification model generated can predict people simultaneously
The face key point information of face image, and the deviation of face key point information predicted help to utilize model prediction
Face key point information and deviation generate more accurate result face key point information, realize that more accurate face is crucial
Point detection.
With further reference to Fig. 4, it illustrates the processes 400 of another embodiment of the method for handling facial image.
This is used to handle the process 400 of the method for facial image, comprising the following steps:
Step 401, target facial image is obtained.
It in the present embodiment, can for handling the executing subject (such as server shown in FIG. 1) of the method for facial image
To obtain target facial image by wired connection mode or radio connection.Wherein, target facial image can for
The facial image of face key point identification is carried out to it.Face key point is point crucial in face, specifically, can be influence
The point of face mask or face shape.
In the present embodiment, above-mentioned executing subject can obtain target facial image using various methods.Specifically, above-mentioned
Executing subject is available to be pre-stored within local target facial image or the available communication connection of above-mentioned executing subject
Electronic equipment (such as terminal device shown in FIG. 1) send target facial image.
Step 402, by target facial image input face key point identification model trained in advance, target face is generated
Deviation corresponding to face key point information and face key point information corresponding to image.
In the present embodiment, based on target facial image obtained in step 401, above-mentioned executing subject can be by target person
In face image input face key point identification model trained in advance, the letter of face key point corresponding to target facial image is generated
Deviation corresponding to breath and face key point information.Wherein, face key point information is for characterizing face key point in target person
Position in face image can include but is not limited at least one of following: text, number, symbol, image.Face key point letter
The corresponding deviation of breath, which can be used for characterizing, utilizes face corresponding to face key point identification model prediction target facial image
Prediction error when key point information.
In the present embodiment, face key point identification model is the method according to described in above-mentioned Fig. 2 corresponding embodiment
It generates, which is not described herein again.
Step 403, it is based on face key point information generated and deviation, generates result corresponding to target facial image
Face key point information.
In the present embodiment, based on the face key point information generated in step 402 and deviation, above-mentioned executing subject can be with
Generate result face key point information corresponding to target facial image.Wherein, as a result face key point information is to close to face
The face key point information of key point identification model output carries out the face key point information after error compensation.
It is appreciated that usually can all have prediction error when carrying out information prediction using model, to prediction result in practice
Carrying out error compensation can make compensated result more close to legitimate reading, help to improve the accuracy of information prediction.
As an example, the face key point information of face key point identification model output can be the coordinate of face key point
(10,19).The deviation of face key point identification model output can be (0.2, -0.5), wherein " 0.2 " can be used for characterizing people
The error of abscissa in the coordinate of the face key point of face key point identification model output；" -0.5 " can be used for characterizing face
The error of ordinate in the coordinate of the face key point of key point identification model output.In turn, above-mentioned executing subject can be right
The error " 0.2 " of abscissa " 10 " and abscissa in the coordinate of face key point carries out asking poor, the cross after obtaining error compensation
Coordinate " 9.8 "；The error " -0.5 " of ordinate " 19 " and ordinate in the coordinate of face key point is carried out asking poor, is missed
The compensated ordinate " 19.5 " of difference, finally, above-mentioned executing subject can use the abscissa " 9.8 " and vertical seat after error compensation
Mark " 19.5 " composition result face key point information " (9.8,19.5) " (coordinate of the face key point i.e. after error compensation).
The method that embodiment of the disclosure provides then is inputted target facial image by obtaining target facial image
In advance in trained face key point identification model, generates face key point information corresponding to target facial image and face closes
Deviation corresponding to key point information is finally based on face key point information generated and deviation, generates target facial image institute
Corresponding result face key point information, so as to the face key point information that is exported based on face key point identification model and
Deviation generates more accurate result face key point information, improves the accuracy of face key point identification.
With further reference to Fig. 5, as the realization to method shown in above-mentioned Fig. 2, present disclose provides one kind for generating mould
One embodiment of the device of type, the Installation practice is corresponding with embodiment of the method shown in Fig. 2, which can specifically answer
For in various electronic equipments.
As shown in figure 5, the present embodiment includes: acquiring unit 501 and training unit for generating the device 500 of model
502.Wherein, acquiring unit 501 is configured to obtain training sample set, wherein training sample includes sample facial image and needle
The sample face key point information that sample facial image is marked in advance；Training unit 502 is configured to concentrate from training sample
Training sample is chosen, and executes following training step: the sample facial image in selected training sample being inputted initial
The feature extraction layer of neural network obtains characteristics of image；Characteristics of image obtained is inputted to the first son of initial neural network
Network generates the face key point information of sample facial image；Face key point information generated and characteristics of image are inputted
Second sub-network of initial neural network, obtains deviation corresponding to face key point information；Based on face key point information and
Sample face key point information in training sample, determines expectation deviation corresponding to face key point information generated；Base
The deviation corresponding to face key point information and desired deviation, determine whether initial neural network trains completion；In response to true
Fixed training is completed, and the initial neural network that training is completed is determined as face key point identification model.
In the present embodiment, for generate model device 500 acquiring unit 501 can by wired connection mode or
Person's radio connection obtains training sample set.Wherein, training sample includes sample facial image and for sample facial image
The sample face key point information marked in advance.Sample facial image can be to carry out shooting figure obtained to sample face
Picture.Sample face key point information for characterizing position of the sample face key point in sample facial image, may include but
It is not limited at least one of following: number, text, symbol, image.
In practice, face key point can be point crucial in face, specifically, can be influence face mask or five
The point of official's shape.
In the present embodiment, the training sample set obtained based on acquiring unit 501, training unit 502 can be from training samples
This concentration selection training sample, and the following training step of execution (step 5021- step 5026):
Step 5021, the sample facial image in selected training sample is inputted to the feature extraction of initial neural network
Layer obtains characteristics of image.
Wherein, characteristics of image can be the features such as color, the shape of image.Initial neural network is predetermined, use
In the various neural networks (such as convolutional neural networks) for generating face key point identification model.Face key point identification model can
With face key point corresponding to facial image for identification.Herein, initial neural network can be unbred nerve
Network, or do not train by training but the neural network completed.Specifically, initial neural network includes feature extraction
Layer.Feature extraction layer is used to extract the characteristics of image of inputted facial image.
Step 5022, characteristics of image obtained is inputted to the first sub-network of initial neural network, generates sample face
The face key point information of image.
In the present embodiment, initial neural network further includes the first sub-network.First sub-network is connect with feature extraction layer,
Characteristics of image for being exported based on feature extraction layer generates face key point information.Face key point information generated is
Face key point information corresponding to first sub-network predicts, sample facial image.
Step 5023, face key point information generated and characteristics of image are inputted to the second subnet of initial neural network
Network obtains deviation corresponding to face key point information.
In the present embodiment, initial neural network further includes the second sub-network, the second sub-network respectively with the first sub-network
It is connected with feature extraction layer, the characteristics of image for being exported based on feature extraction layer, determines that the face of the first sub-network output closes
Deviation corresponding to key point information.Deviation corresponding to face key point information is for characterizing face key point information generated
The difference of practical face key point information relative to sample facial image.Here, the second sub-network deviation generated is base
In the deviation that characteristics of image predicts.
Step 5024, based on the sample face key point information in face key point information and training sample, determination is given birth to
At face key point information corresponding to expectation deviation.
Herein, training unit 502 can determine face key point information that the first sub-network generates and mark in advance
The difference of sample face key point information, and then identified difference is determined as it is expected deviation.
Step 5025, based on deviation corresponding to face key point information and desired deviation, determine that initial neural network is
No training is completed.
Herein, training unit 502 can use preset loss function and calculate obtained face key point information institute
Difference between corresponding deviation and expectation deviation, and then determine whether the difference being calculated is less than or equal to default difference threshold
Value determines that initial neural metwork training is completed in response to determining that the difference being calculated is less than or equal to default discrepancy threshold.
Step 5026, in response to determining that training is completed, the initial neural network that training is completed is determined as face key point
Identification model.
In the present embodiment, training unit 502 can be completed in response to the initial neural metwork training of determination, and training is completed
Initial neural network be determined as face key point identification model.
In some optional implementations of the present embodiment, the second sub-network includes that the first generation layer and second generate
Layer；And training unit 502 can be further configured to: face key point information generated is inputted the second sub-network
First generation layer obtains hotspot graph corresponding to face key point information, wherein the image-region of hotspot graph includes set of values
It closes, for the numerical value in numerical value set, the numerical value is for characterizing face key point in the probability of the numerical value position；By institute
The hotspot graph and characteristics of image of acquisition input the second generation layer of the second sub-network, obtain inclined corresponding to face key point information
Difference.
In some optional implementations of the present embodiment, device 500 can also include: that adjustment unit (does not show in figure
Out), it is configured in response to determine that initial neural network not complete by training, adjusts the relevant parameter in initial neural network, from
Training sample is concentrated and chooses training sample in unselected training sample, and uses the initial nerve net of the last adjustment
Network and the last training sample chosen, continue to execute training step.
It is understood that all units recorded in the device 500 and each step phase in the method with reference to Fig. 2 description
It is corresponding.As a result, above with respect to the operation of method description, the beneficial effect of feature and generation be equally applicable to device 500 and its
In include unit, details are not described herein.
The face key point identification model generated of device provided by the above embodiment 500 of the disclosure can be predicted simultaneously
The face key point information of facial image, and the deviation of face key point information predicted help to utilize model prediction
Face key point information and deviation, generate more accurate result face key point information, realize that more accurate face closes
The detection of key point.
With further reference to Fig. 6, as the realization to method shown in above-mentioned Fig. 4, present disclose provides one kind for handling people
One embodiment of the device of face image, the Installation practice is corresponding with embodiment of the method shown in Fig. 4, which specifically may be used
To be applied in various electronic equipments.
As shown in fig. 6, the device 600 for handling facial image of the present embodiment includes: image acquisition unit 601, the
One generation unit 602 and the second generation unit 603.Wherein, image acquisition unit 601 is configured to obtain target facial image；
First generation unit 602 is configured to input target facial image and be generated using method described in the corresponding embodiment of Fig. 2
Face key point identification model in, generate target facial image corresponding to face key point information and face key point information
Corresponding deviation；Second generation unit 603 is configured to generate target based on face key point information generated and deviation
Result face key point information corresponding to facial image.
It in the present embodiment, can be by wired company for handling the image acquisition unit 601 of the device 600 of facial image
It connects mode or radio connection obtains target facial image.Wherein, target facial image can be for carry out face to it
The facial image of key point identification.Face key point is crucial point in face, specifically, can for influence face mask or
The point of face shape.
In the present embodiment, the target facial image obtained based on image acquisition unit 601, the first generation unit 602 can
To generate people corresponding to target facial image in target facial image input face key point identification model trained in advance
Deviation corresponding to face key point information and face key point information.Wherein, face key point information is for characterizing face key
Position of the point in target facial image can include but is not limited at least one of following: text, number, symbol, image.People
Deviation corresponding to face key point information, which can be used for characterizing, predicts target facial image institute using face key point identification model
Prediction error when corresponding face key point information.
In the present embodiment, face key point identification model is the method according to described in above-mentioned Fig. 2 corresponding embodiment
It generates, which is not described herein again.
In the present embodiment, the face key point information and deviation generated based on the first generation unit 602, second generates list
Result face key point information corresponding to target facial image can be generated in member 603.Wherein, as a result face key point information is
Face key point information after carrying out error compensation to the face key point information of face key point identification model output.
It is understood that all units recorded in the device 600 and each step phase in the method with reference to Fig. 4 description
It is corresponding.As a result, above with respect to the operation of method description, the beneficial effect of feature and generation be equally applicable to device 600 and its
In include unit, details are not described herein.
The device provided by the above embodiment 600 of the disclosure is by obtaining target facial image, then by target face figure
As in input face key point identification model trained in advance, generate face key point information corresponding to target facial image and
Deviation corresponding to face key point information is finally based on face key point information generated and deviation, generates target face
Result face key point information corresponding to image, so as to the face key point exported based on face key point identification model
Information and deviation generate more accurate result face key point information, improve the accuracy of face key point identification.
Below with reference to Fig. 7, it illustrates the electronic equipment that is suitable for being used to realize embodiment of the disclosure, (example is as shown in figure 1
Server or terminal device) 700 structural schematic diagram.Terminal device in embodiment of the disclosure can include but is not limited to all
As mobile phone, laptop, digit broadcasting receiver, PDA (personal digital assistant), PAD (tablet computer), PMP are (portable
Formula multimedia player), the mobile terminal and such as number TV, desk-top meter of car-mounted terminal (such as vehicle mounted guidance terminal) etc.
The fixed terminal of calculation machine etc..Terminal device or server shown in Fig. 7 are only an example, should not be to the implementation of the disclosure
The function and use scope of example bring any restrictions.
As shown in fig. 7, electronic equipment 700 may include processing unit (such as central processing unit, graphics processor etc.)
701, random access can be loaded into according to the program being stored in read-only memory (ROM) 702 or from storage device 708
Program in memory (RAM) 703 and execute various movements appropriate and processing.In RAM 703, it is also stored with electronic equipment
Various programs and data needed for 700 operations.Processing unit 701, ROM 702 and RAM703 are connected with each other by bus 704.
Input/output (I/O) interface 705 is also connected to bus 704.
In general, following device can connect to I/O interface 705: including such as touch screen, touch tablet, keyboard, mouse, taking the photograph
As the input unit 706 of head, microphone, accelerometer, gyroscope etc.；Including such as liquid crystal display (LCD), loudspeaker, vibration
The output device 707 of dynamic device etc.；Storage device 708 including such as tape, hard disk etc.；And communication device 709.Communication device
709, which can permit electronic equipment 700, is wirelessly or non-wirelessly communicated with other equipment to exchange data.Although Fig. 7 shows tool
There is the electronic equipment 700 of various devices, it should be understood that being not required for implementing or having all devices shown.It can be with
Alternatively implement or have more or fewer devices.Each box shown in Fig. 7 can represent a device, can also root
According to needing to represent multiple devices.
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 from network by communication device 709, or from storage device 708
It is mounted, or is mounted from ROM 702.When the computer program is executed by processing unit 701, the implementation of the disclosure is executed
The above-mentioned function of being limited in the method for example.It should be noted that computer-readable medium described in embodiment of the disclosure can be with
It is computer-readable signal media or computer readable storage medium either the two any combination.It is computer-readable
Storage medium for example may be-but not limited to-the system of electricity, magnetic, optical, electromagnetic, infrared ray or semiconductor, device or
Device, or any above combination.The more specific example of computer readable storage medium can include but is not limited to: have
The electrical connection of one or more conducting wires, portable computer diskette, hard disk, random access storage device (RAM), read-only memory
(ROM), erasable programmable read only memory (EPROM or flash memory), optical fiber, portable compact disc read-only memory (CD-
ROM), light storage device, magnetic memory device or above-mentioned any appropriate combination.In embodiment of the disclosure, computer
Readable storage medium storing program for executing can be any tangible medium for including or store program, which can be commanded execution system, device
Either device use or in connection.And in embodiment of the disclosure, computer-readable signal media may include
In a base band or as the data-signal that carrier wave a part is propagated, wherein carrying computer-readable program code.It is this
The data-signal of propagation can take various forms, including but not limited to electromagnetic signal, optical signal or above-mentioned any appropriate
Combination.Computer-readable signal media can also be any computer-readable medium other than computer readable storage medium, should
Computer-readable signal media can send, propagate or transmit for by instruction execution system, device or device use or
Person's program in connection.The program code for including on computer-readable medium can transmit with any suitable medium,
Including but not limited to: electric wire, optical cable, RF (radio frequency) etc. or above-mentioned any appropriate combination.
Above-mentioned computer-readable medium can be included in above-mentioned electronic equipment；It is also possible to individualism, and not
It is fitted into the electronic equipment.Above-mentioned computer-readable medium carries one or more program, when said one or more
When a program is executed by the electronic equipment, so that the electronic equipment: obtaining training sample set, wherein training sample includes sample
Facial image and the sample face key point information marked in advance for sample facial image；It is concentrated from training sample and chooses training
Sample, and execute following training step: the sample facial image in selected training sample is inputted into initial neural network
Feature extraction layer, obtain characteristics of image；Characteristics of image obtained is inputted to the first sub-network of initial neural network, is generated
The face key point information of sample facial image；Face key point information generated and characteristics of image are inputted into initial nerve net
Second sub-network of network obtains deviation corresponding to face key point information；Based in face key point information and training sample
Sample face key point information, determine expectation deviation corresponding to face key point information generated；Based on face key
Deviation and desired deviation corresponding to point information, determine whether initial neural network trains completion；In response to determining that training is completed,
The initial neural network that training is completed is determined as face key point identification model.
In addition, when said one or multiple programs are executed by the electronic equipment, it is also possible that the electronic equipment: obtaining
Take target facial image；By in target facial image input face key point identification model trained in advance, target face is generated
Deviation corresponding to face key point information and face key point information corresponding to image；Based on face key point generated
Information and deviation generate result face key point information corresponding to target facial image.
The behaviour for executing embodiment of the disclosure can be write with one or more programming languages or combinations thereof
The computer program code of work, described program design language include object oriented program language-such as Java,
Smalltalk, C++ further include conventional procedural programming language-such as " C " language or similar program design language
Speech.Program code can be executed fully on the user computer, partly be executed on the user computer, as an independence
Software package execute, part on the user computer part execute on the remote computer or completely in remote computer or
It is executed on server.In situations involving remote computers, remote computer can pass through the network of any kind --- packet
It includes local area network (LAN) or wide area network (WAN)-is connected to subscriber computer, or, it may be connected to outer computer (such as benefit
It is connected with ISP by internet).
Flow chart and block diagram in attached drawing are illustrated according to the system of the various embodiments of the disclosure, 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 of one module, program segment or code of table, a part of the module, program segment or code include one or more use
The executable instruction of the logic function as defined in realizing.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, and this depends on the function involved.Also it to infuse
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 embodiment of the disclosure can be realized by way of software, can also be passed through
The mode of hardware is realized.Described unit also can be set in the processor, for example, can be described as: a kind of processor
Including acquiring unit and training unit.Wherein, the title of these units is not constituted to the unit itself under certain conditions
It limits, for example, acquiring unit is also described as " obtaining the unit of training sample set ".
Above description is only the preferred embodiment of the disclosure and the explanation to institute's application technology principle.Those skilled in the art
Member it should be appreciated that embodiment of the disclosure involved in invention scope, however it is not limited to the specific combination of above-mentioned technical characteristic and
At technical solution, while should also cover do not depart from foregoing invention design in the case where, by above-mentioned technical characteristic or its be equal
Feature carries out any combination and other technical solutions for being formed.Such as disclosed in features described above and embodiment of the disclosure (but
It is not limited to) technical characteristic with similar functions is replaced mutually and the technical solution that is formed.
Claims (10)
1. a kind of method for generating model, comprising:
Obtain training sample set, wherein training sample includes sample facial image and marks in advance for sample facial image
Sample face key point information；
It is concentrated from the training sample and chooses training sample, and execute following training step: will be in selected training sample
Sample facial image input the feature extraction layer of initial neural network, obtain characteristics of image；Characteristics of image obtained is defeated
Enter the first sub-network of initial neural network, generates the face key point information of sample facial image；Face generated is closed
Key point information and characteristics of image input the second sub-network of initial neural network, obtain inclined corresponding to face key point information
Difference；Based on the sample face key point information in face key point information and training sample, face key point generated is determined
Expectation deviation corresponding to information；Based on deviation corresponding to face key point information and desired deviation, initial nerve net is determined
Whether network trains completion；In response to determining that training is completed, the initial neural network that training is completed is determined as face key point and is known
Other model.
2. according to the method described in claim 1, wherein, the second sub-network includes the first generation layer and the second generation layer；And
Second sub-network that face key point information generated and characteristics of image are inputted to initial neural network obtains people
Deviation corresponding to face key point information, comprising:
Face key point information generated is inputted to the first generation layer of the second sub-network, it is right to obtain face key point information institute
The hotspot graph answered, wherein the image-region of hotspot graph includes numerical value set, and for the numerical value in numerical value set, which is used for
Face key point is characterized in the probability of the numerical value position；
The second generation layer that hotspot graph obtained and characteristics of image are inputted to the second sub-network, obtains face key point information institute
Corresponding deviation.
3. according to the method described in claim 1, wherein, the method also includes:
In response to the initial neural network of determination, training is not completed, and the relevant parameter in initial neural network is adjusted, from the training
Choose training sample in sample set in unselected training sample, and using the last adjustment initial neural network and
The training sample that the last time is chosen, continues to execute the training step.
4. a kind of method for handling facial image, comprising:
Obtain target facial image；
The face key point identification that target facial image input is generated using the method as described in one of claim 1-3
In model, generate inclined corresponding to face key point information corresponding to the target facial image and face key point information
Difference；
Based on face key point information generated and deviation, the key of result face corresponding to the target facial image is generated
Point information.
5. a kind of for generating the device of model, comprising:
Acquiring unit is configured to obtain training sample set, wherein training sample includes sample facial image and for sample people
The sample face key point information that face image marks in advance；
Training unit is configured to concentrate from the training sample and chooses training sample, and executes following training step: by institute
Sample facial image in the training sample of selection inputs the feature extraction layer of initial neural network, obtains characteristics of image；By institute
The characteristics of image of acquisition inputs the first sub-network of initial neural network, generates the face key point information of sample facial image；
Face key point information generated and characteristics of image are inputted to the second sub-network of initial neural network, obtain face key point
Deviation corresponding to information；Based on the sample face key point information in face key point information and training sample, determination is given birth to
At face key point information corresponding to expectation deviation；Based on deviation corresponding to face key point information and desired deviation,
Determine whether initial neural network trains completion；In response to determining that training is completed, the initial neural network that training is completed is determined
For face key point identification model.
6. device according to claim 5, wherein the second sub-network includes the first generation layer and the second generation layer；And
The training unit is further configured to:
Face key point information generated is inputted to the first generation layer of the second sub-network, it is right to obtain face key point information institute
The hotspot graph answered, wherein the image-region of hotspot graph includes numerical value set, and for the numerical value in numerical value set, which is used for
Face key point is characterized in the probability of the numerical value position；
The second generation layer that hotspot graph obtained and characteristics of image are inputted to the second sub-network, obtains face key point information institute
Corresponding deviation.
7. device according to claim 5, wherein described device further include:
Adjustment unit is configured in response to determine that initial neural network not complete by training, adjusts the phase in initial neural network
Parameter is closed, is concentrated from the training sample and chooses training sample in unselected training sample, and is adjusted using the last
Whole initial neural network and the last training sample chosen, continue to execute the training step.
8. a kind of for handling the device of facial image, comprising:
Image acquisition unit is configured to obtain target facial image；
First generation unit is configured to input the target facial image using the side as described in one of claim 1-3
In the face key point identification model that method generates, face key point information and face corresponding to the target facial image are generated
Deviation corresponding to key point information；
Second generation unit is configured to generate the target face figure based on face key point information generated and deviation
As corresponding result face key point information.
9. a kind of electronic equipment, comprising:
One or more processors；
Storage device is stored thereon with 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 1-4.
10. a kind of computer-readable medium, is stored thereon with computer program, wherein the realization when program is executed by processor
Method as described in any in claim 1-4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910099403.9A CN109858445B (en) | 2019-01-31 | 2019-01-31 | Method and apparatus for generating a model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910099403.9A CN109858445B (en) | 2019-01-31 | 2019-01-31 | Method and apparatus for generating a model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109858445A true CN109858445A (en) | 2019-06-07 |
| CN109858445B CN109858445B (en) | 2021-06-25 |
Family
ID=66897221
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910099403.9A Active CN109858445B (en) | 2019-01-31 | 2019-01-31 | Method and apparatus for generating a model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109858445B (en) |
Cited By (21)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110347134A (en) * | 2019-07-29 | 2019-10-18 | 南京图玩智能科技有限公司 | A kind of AI intelligence aquaculture specimen discerning method and cultivating system |
| CN110414570A (en) * | 2019-07-04 | 2019-11-05 | 北京迈格威科技有限公司 | Image classification model generating method, device, equipment and storage medium |
| CN110443222A (en) * | 2019-08-14 | 2019-11-12 | 北京百度网讯科技有限公司 | Method and apparatus for training face's critical point detection model |
| CN110738654A (en) * | 2019-10-18 | 2020-01-31 | 中国科学技术大学 | Key point extraction and bone age prediction method in hip joint image |
| CN110782408A (en) * | 2019-10-18 | 2020-02-11 | 杭州趣维科技有限公司 | Intelligent beautifying method and system based on convolutional neural network |
| CN110895809A (en) * | 2019-10-18 | 2020-03-20 | 中国科学技术大学 | Method for accurately extracting key points in hip joint image |
| CN110956127A (en) * | 2019-11-28 | 2020-04-03 | 重庆中星微人工智能芯片技术有限公司 | Method, apparatus, electronic device, and medium for generating feature vector |
| CN111275784A (en) * | 2020-01-20 | 2020-06-12 | 北京百度网讯科技有限公司 | Method and device for generating image |
| CN111524216A (en) * | 2020-04-10 | 2020-08-11 | 北京百度网讯科技有限公司 | Method and device for generating three-dimensional face data |
| CN111539903A (en) * | 2020-04-16 | 2020-08-14 | 北京百度网讯科技有限公司 | Method and device for training face image synthesis model |
| CN111582482A (en) * | 2020-05-11 | 2020-08-25 | 北京字节跳动网络技术有限公司 | Method, apparatus, device and medium for generating network model information |
| CN111860300A (en) * | 2020-07-17 | 2020-10-30 | 广州视源电子科技股份有限公司 | Key point detection method and device, terminal equipment and storage medium |
| CN112001372A (en) * | 2020-09-30 | 2020-11-27 | 苏州科达科技股份有限公司 | Face recognition model generation and face recognition method, system, device and medium |
| CN112102468A (en) * | 2020-08-07 | 2020-12-18 | 北京沃东天骏信息技术有限公司 | Model training method, virtual character image generation device, and storage medium |
| CN112241716A (en) * | 2020-10-23 | 2021-01-19 | 北京百度网讯科技有限公司 | Training sample generation method and device |
| CN112330781A (en) * | 2020-11-24 | 2021-02-05 | 北京百度网讯科技有限公司 | Method, device, equipment and storage medium for generating model and generating human face animation |
| CN112651490A (en) * | 2020-12-28 | 2021-04-13 | 深圳万兴软件有限公司 | Training method and device for face key point detection model and readable storage medium |
| CN113177606A (en) * | 2021-05-20 | 2021-07-27 | 上海商汤智能科技有限公司 | Image processing method, device, equipment and storage medium |
| CN113689527A (en) * | 2020-05-15 | 2021-11-23 | 武汉Tcl集团工业研究院有限公司 | Training method of face conversion model and face image conversion method |
| CN113762520A (en) * | 2020-06-04 | 2021-12-07 | 杭州海康威视数字技术股份有限公司 | Data processing method, device and equipment |
| CN113808010A (en) * | 2021-09-24 | 2021-12-17 | 深圳万兴软件有限公司 | Cartoon portrait generating method, device, equipment and medium without attribute deviation |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101751689A (en) * | 2009-09-28 | 2010-06-23 | 中国科学院自动化研究所 | Three-dimensional facial reconstruction method |
| CN106778531A (en) * | 2016-11-25 | 2017-05-31 | 北京小米移动软件有限公司 | Face detection method and device |
| CN108121951A (en) * | 2017-12-11 | 2018-06-05 | 北京小米移动软件有限公司 | Characteristic point positioning method and device |
| WO2018133034A1 (en) * | 2017-01-20 | 2018-07-26 | Intel Corporation | Dynamic emotion recognition in unconstrained scenarios |
| CN109101919A (en) * | 2018-08-03 | 2018-12-28 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating information |
| CN109214343A (en) * | 2018-09-14 | 2019-01-15 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating face critical point detection model |
| CN109726619A (en) * | 2017-10-31 | 2019-05-07 | 深圳市祈飞科技有限公司 | A kind of convolutional neural networks face identification method and system based on parameter sharing |
-
2019
- 2019-01-31 CN CN201910099403.9A patent/CN109858445B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101751689A (en) * | 2009-09-28 | 2010-06-23 | 中国科学院自动化研究所 | Three-dimensional facial reconstruction method |
| CN106778531A (en) * | 2016-11-25 | 2017-05-31 | 北京小米移动软件有限公司 | Face detection method and device |
| WO2018133034A1 (en) * | 2017-01-20 | 2018-07-26 | Intel Corporation | Dynamic emotion recognition in unconstrained scenarios |
| CN109726619A (en) * | 2017-10-31 | 2019-05-07 | 深圳市祈飞科技有限公司 | A kind of convolutional neural networks face identification method and system based on parameter sharing |
| CN108121951A (en) * | 2017-12-11 | 2018-06-05 | 北京小米移动软件有限公司 | Characteristic point positioning method and device |
| CN109101919A (en) * | 2018-08-03 | 2018-12-28 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating information |
| CN109214343A (en) * | 2018-09-14 | 2019-01-15 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating face critical point detection model |
Non-Patent Citations (4)
| Title |
|---|
| PENG Y 等: ""Designing a low-resolution system for long-range surveillance"", 《BIOMETRICS SPECIAL INTEREST GROUP》 * |
| YI SUN: ""Deep Convolutional network cascade for Facial Point Detection"", 《CVPR2013》 * |
| ZHOU Y 等: ""Survey of face detection on low-qualiyu images"", 《COMPUTER SOCIETY》 * |
| 汤一平 等: ""基于多任务卷积神经网络的舌象分类研究"", 《计算机科学》 * |
Cited By (33)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110414570B (en) * | 2019-07-04 | 2022-01-28 | 北京迈格威科技有限公司 | Image classification model generation method, device, equipment and storage medium |
| CN110414570A (en) * | 2019-07-04 | 2019-11-05 | 北京迈格威科技有限公司 | Image classification model generating method, device, equipment and storage medium |
| CN110347134A (en) * | 2019-07-29 | 2019-10-18 | 南京图玩智能科技有限公司 | A kind of AI intelligence aquaculture specimen discerning method and cultivating system |
| CN110443222A (en) * | 2019-08-14 | 2019-11-12 | 北京百度网讯科技有限公司 | Method and apparatus for training face's critical point detection model |
| CN110738654A (en) * | 2019-10-18 | 2020-01-31 | 中国科学技术大学 | Key point extraction and bone age prediction method in hip joint image |
| CN110782408A (en) * | 2019-10-18 | 2020-02-11 | 杭州趣维科技有限公司 | Intelligent beautifying method and system based on convolutional neural network |
| CN110895809A (en) * | 2019-10-18 | 2020-03-20 | 中国科学技术大学 | Method for accurately extracting key points in hip joint image |
| CN110738654B (en) * | 2019-10-18 | 2022-07-15 | 中国科学技术大学 | Key point extraction and bone age prediction method in hip joint image |
| CN110895809B (en) * | 2019-10-18 | 2022-07-15 | 中国科学技术大学 | Method for accurately extracting key points in hip joint image |
| CN110956127A (en) * | 2019-11-28 | 2020-04-03 | 重庆中星微人工智能芯片技术有限公司 | Method, apparatus, electronic device, and medium for generating feature vector |
| CN111275784A (en) * | 2020-01-20 | 2020-06-12 | 北京百度网讯科技有限公司 | Method and device for generating image |
| CN111275784B (en) * | 2020-01-20 | 2023-06-13 | 北京百度网讯科技有限公司 | Method and device for generating image |
| CN111524216A (en) * | 2020-04-10 | 2020-08-11 | 北京百度网讯科技有限公司 | Method and device for generating three-dimensional face data |
| CN111539903A (en) * | 2020-04-16 | 2020-08-14 | 北京百度网讯科技有限公司 | Method and device for training face image synthesis model |
| CN111539903B (en) * | 2020-04-16 | 2023-04-07 | 北京百度网讯科技有限公司 | Method and device for training face image synthesis model |
| CN111582482B (en) * | 2020-05-11 | 2023-12-15 | 抖音视界有限公司 | Method, apparatus, device and medium for generating network model information |
| CN111582482A (en) * | 2020-05-11 | 2020-08-25 | 北京字节跳动网络技术有限公司 | Method, apparatus, device and medium for generating network model information |
| CN113689527A (en) * | 2020-05-15 | 2021-11-23 | 武汉Tcl集团工业研究院有限公司 | Training method of face conversion model and face image conversion method |
| CN113762520A (en) * | 2020-06-04 | 2021-12-07 | 杭州海康威视数字技术股份有限公司 | Data processing method, device and equipment |
| CN111860300A (en) * | 2020-07-17 | 2020-10-30 | 广州视源电子科技股份有限公司 | Key point detection method and device, terminal equipment and storage medium |
| CN112102468B (en) * | 2020-08-07 | 2022-03-04 | 北京汇钧科技有限公司 | Model training method, virtual character image generation device, and storage medium |
| CN112102468A (en) * | 2020-08-07 | 2020-12-18 | 北京沃东天骏信息技术有限公司 | Model training method, virtual character image generation device, and storage medium |
| CN112001372A (en) * | 2020-09-30 | 2020-11-27 | 苏州科达科技股份有限公司 | Face recognition model generation and face recognition method, system, device and medium |
| CN112001372B (en) * | 2020-09-30 | 2023-02-03 | 苏州科达科技股份有限公司 | Face recognition model generation and face recognition method, system, device and medium |
| CN112241716A (en) * | 2020-10-23 | 2021-01-19 | 北京百度网讯科技有限公司 | Training sample generation method and device |
| CN112241716B (en) * | 2020-10-23 | 2023-06-20 | 北京百度网讯科技有限公司 | Training sample generation method and device |
| CN112330781A (en) * | 2020-11-24 | 2021-02-05 | 北京百度网讯科技有限公司 | Method, device, equipment and storage medium for generating model and generating human face animation |
| CN112651490A (en) * | 2020-12-28 | 2021-04-13 | 深圳万兴软件有限公司 | Training method and device for face key point detection model and readable storage medium |
| CN112651490B (en) * | 2020-12-28 | 2024-01-05 | 深圳万兴软件有限公司 | Training method and device for human face key point detection model and readable storage medium |
| CN113177606B (en) * | 2021-05-20 | 2023-11-28 | 上海商汤智能科技有限公司 | Image processing method, device, equipment and storage medium |
| CN113177606A (en) * | 2021-05-20 | 2021-07-27 | 上海商汤智能科技有限公司 | Image processing method, device, equipment and storage medium |
| CN113808010A (en) * | 2021-09-24 | 2021-12-17 | 深圳万兴软件有限公司 | Cartoon portrait generating method, device, equipment and medium without attribute deviation |
| CN113808010B (en) * | 2021-09-24 | 2023-08-11 | 深圳万兴软件有限公司 | Cartoon portrait generating method, device, equipment and medium without attribute deviation |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109858445B (en) | 2021-06-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109858445A (en) | Method and apparatus for generating model | |
| CN108898185A (en) | Method and apparatus for generating image recognition model | |
| CN108985257A (en) | Method and apparatus for generating information | |
| CN110288049A (en) | Method and apparatus for generating image recognition model | |
| CN109829432A (en) | Method and apparatus for generating information | |
| CN110009059B (en) | Method and apparatus for generating a model | |
| CN110110811A (en) | Method and apparatus for training pattern, the method and apparatus for predictive information | |
| CN109981787B (en) | Method and device for displaying information | |
| CN108416310A (en) | Method and apparatus for generating information | |
| CN109086719A (en) | Method and apparatus for output data | |
| CN107393541A (en) | Information Authentication method and apparatus | |
| CN109101919A (en) | Method and apparatus for generating information | |
| CN109815365A (en) | Method and apparatus for handling video | |
| CN109934191A (en) | Information processing method and device | |
| CN110348419A (en) | Method and apparatus for taking pictures | |
| CN109800730A (en) | The method and apparatus for generating model for generating head portrait | |
| CN109961141A (en) | Method and apparatus for generating quantization neural network | |
| CN110033423A (en) | Method and apparatus for handling image | |
| CN109918530A (en) | Method and apparatus for pushing image | |
| CN108491808A (en) | Method and device for obtaining information | |
| CN109117758A (en) | Method and apparatus for generating information | |
| CN110059623A (en) | Method and apparatus for generating information | |
| CN110032978A (en) | Method and apparatus for handling video | |
| CN110070076A (en) | Method and apparatus for choosing trained sample | |
| CN108521516A (en) | Control method and device for terminal 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 | ||
| CP01 | Change in the name or title of a patent holder |
Address after: 100041 B-0035, 2 floor, 3 building, 30 Shixing street, Shijingshan District, Beijing. Patentee after: Douyin Vision Co.,Ltd. Address before: 100041 B-0035, 2 floor, 3 building, 30 Shixing street, Shijingshan District, Beijing. Patentee before: Tiktok vision (Beijing) Co.,Ltd. Address after: 100041 B-0035, 2 floor, 3 building, 30 Shixing street, Shijingshan District, Beijing. Patentee after: Tiktok vision (Beijing) Co.,Ltd. Address before: 100041 B-0035, 2 floor, 3 building, 30 Shixing street, Shijingshan District, Beijing. Patentee before: BEIJING BYTEDANCE NETWORK TECHNOLOGY Co.,Ltd. |
|
| CP01 | Change in the name or title of a patent holder |