CN106529402B - The face character analysis method of convolutional neural networks based on multi-task learning - Google Patents
The face character analysis method of convolutional neural networks based on multi-task learning Download PDFInfo
- Publication number
- CN106529402B CN106529402B CN201610856231.1A CN201610856231A CN106529402B CN 106529402 B CN106529402 B CN 106529402B CN 201610856231 A CN201610856231 A CN 201610856231A CN 106529402 B CN106529402 B CN 106529402B
- Authority
- CN
- China
- Prior art keywords
- convolutional neural
- neural networks
- network
- multitask
- task
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 72
- 238000004458 analytical method Methods 0.000 title claims abstract description 23
- 230000001815 facial effect Effects 0.000 claims abstract description 37
- 238000000034 method Methods 0.000 claims abstract description 17
- 238000001514 detection method Methods 0.000 claims description 12
- 230000006870 function Effects 0.000 claims description 12
- 238000011478 gradient descent method Methods 0.000 claims description 3
- 230000001537 neural effect Effects 0.000 claims description 3
- 238000011176 pooling Methods 0.000 claims description 3
- 230000000452 restraining effect Effects 0.000 abstract description 3
- 238000003672 processing method Methods 0.000 abstract 1
- 238000010586 diagram Methods 0.000 description 6
- 238000013528 artificial neural network Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 206010040829 Skin discolouration Diseases 0.000 description 1
- 230000007935 neutral effect Effects 0.000 description 1
- 230000037303 wrinkles Effects 0.000 description 1
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
- G06V40/168—Feature extraction; Face representation
-
- 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
- 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
- G06V40/178—Human faces, e.g. facial parts, sketches or expressions estimating age from face image; using age information for improving recognition
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- General Health & Medical Sciences (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The face character analysis method of the invention discloses a kind of convolutional neural networks (CNN) based on multi-task learning.This method carries out age estimation, gender identification and species to facial image mainly based on convolutional neural networks, using the method for multi-task learning simultaneously.In traditional processing method, when face multi-attribute analysis, need to calculate several times, not only elapsed time, but also reduce the generalization ability of model.Then the present invention selects the shared part for restraining the weight initialization multitask network of most slow network, the independent sector of random initializtion multitask network by being trained respectively to three single task networks；Next multitask network is trained, obtains multitask CNN network model；Finally, the analysis at age, gender and ethnic three attributes can be carried out simultaneously using trained multitask CNN network model to the facial image of input, the time was not only saved but also had obtained higher accuracy.
Description
Technical field
The present invention relates to field of image recognition, and in particular to a kind of face of the convolutional neural networks based on multi-task learning
Property analysis method.
Background technique
Traditional facial image analytical technology is often just for individual task, such as age estimation, gender identification, species
Deng when for face multi-attribute analysis, needing to calculate several times, very elapsed time, be extremely difficult to actual demand.In addition, single
The facial image analytical technology of task ignores the connection between each information, cannot make full use of and to be contained in facial image
Information.The facial characteristics of face in different sexes, it is not agnate between be it is different, as between men and women, between black and white ethnic group
Skin exquisiteness degree, the colour of skin, skin Lightening degree etc. have differences, and the meeting such as the bright degree of skin, color, wrinkle texture
Corresponding variation occurs with age, pace of change is different also with sex, race.It can be seen that each face letter
It is closely connected between breath, each task progress independent study can be lost into many useful information to a certain extent, from
And reduce the generalization ability of model.
Summary of the invention
In order to solve the above problem in the prior art, the present invention proposes a kind of convolutional Neural net based on multi-task learning
The face character analysis method of network, the generalization ability of calculating speed and model when improving face multi-attribute analysis.
The face character analysis method of convolutional neural networks proposed by the present invention based on multi-task learning, including single task
Model analysis, multi task model training and face character judge three parts；
Single task model analysis:
The original sample of each age facial image is carried out face critical point detection, after pedestrian's face of going forward side by side alignment by step A1
It is cut according to pre-set dimension and generates the new samples comprising facial image；
Age estimation network, gender identification network, race is respectively trained in step A2, the new samples generated using step A1
Three single task convolutional neural networks of sorter network, the convergence rate of more each network obtain a most slow list of convergence rate
The weight of task convolutional neural networks；
Multi task model training:
Step B1 constructs multitask convolutional neural networks, and there are three tasks to export altogether for the network, respectively corresponds the age and estimates
It calculates, gender identification and species, three tasks all use softmax loss function as objective function；The multitask volume
Product neural network includes shared part for data sharing in multi-task learning and information exchange and for calculating above-mentioned three
The independent sector of a task output；Utilize the weight initialization multitask convolution of the step A2 single task convolutional neural networks obtained
The shared part of neural network, the multitask convolutional neural networks after forming initialization；
Step B2, using the new samples generated in step A1, training multitask convolutional neural networks are obtained trained more
Task convolutional neural networks model；
Face character judgement:
Step C1 carries out Face datection to inputted picture, judges whether comprising facial image, as schemed comprising if to input
As carrying out face critical point detection, then pedestrian's face of going forward side by side alignment cuts generation according to pre-set dimension and includes the new of facial image
Picture；
New picture obtained by step C1 is input to the multitask convolutional neural networks model that step B2 is obtained by step C2, into
The estimation of row age, gender identification and species.
Preferably, step A1 specifically includes the following contents:
Step A11 chooses each age facial image as original sample；
Step A12 carries out face critical point detection to selected original sample, obtains two key points；
Step A13, the alignment according to the position and its line of two key points to original sample progress facial image are described
The alignment of facial image includes rotation, scaling, translation to original sample；
Sample after being aligned in step A13 is cut according to pre-set dimension and generates the new sample comprising facial image by step A14
This.
Preferably, age estimation network, gender identification three network, species network lists is being respectively trained in step A2
When task convolutional neural networks, the conditions such as learning rate, learning strategy are completely the same.
Preferably, floor portions of the shared part of multitask convolutional neural networks for multitask network, packet in step B1
Include data input and convolutional layer and pond layer.
Preferably, in step B1 multitask convolutional neural networks independent sector are as follows: the age estimation, gender identification, race
Classify three independent network structures of task, each separate network structure possesses independent convolutional layer, pooling layers and full connection
Layer.
Preferably, multitask convolutional neural networks total losses function are as follows:
lmulti-task=α lage+β·lgender+γ·lrace
Wherein lmulti-taskFor the total losses of multitask network, α, β, γ are respectively the weight system of preset three tasks
Number, lage、lgender、lraceAge estimation loss, gender identification loss, Zhong Zufen respectively in multitask convolutional neural networks
Class loss.
Preferably, the method for multitask convolutional neural networks is trained in step B2 are as follows:
Step B21 randomly selects m images from the new samples that step A1 is generated, is input to and constructs in step bl is determined simultaneously
Initialized multitask convolutional neural networks carry out multitask and synchronize training；
Step B22 calculates separately age estimation loss l to transmitting before multitask convolutional neural networksage, gender identification damage
Lose lgenderL is lost with speciesrace；
Step B23 calculates the total losses l of multitask convolutional neural networksmulti-task；
Step B24: judging whether network training restrains, and deconditioning and obtains multitask convolutional neural networks if convergence
Model, it is no to then follow the steps B25；
Step B25: calculating each parameter gradients of network using back-propagation algorithm, updates network using stochastic gradient descent method
Parameter weight；Return step B21.
Preferably, step C1 specifically includes the following contents:
Step C11 detects whether it includes face to the picture inputted, abandons the picture if not including face,
Otherwise C12 is entered step；
Step C12 carries out face critical point detection to the picture inputted, obtains two key points；
Step C13 carries out the alignment of facial image according to the position and its line of above-mentioned key point to original image, described
The alignment of facial image includes rotation, scaling, translation to original image；
Picture after being aligned in step C13 is cut according to pre-set dimension and generates the new figure comprising facial image by step C14
Piece.
Preferably, step C2 specifically includes the following contents:
Step C21, the new picture that step C1 is obtained are input to trained multitask convolutional neural networks, carry out
Face multi-attribute analysis；
Step C22 obtains age Probability p 1 (i), and final age estimation result is the mathematical expectation at each age,Wherein output node number is k+1, and i is output node serial number, a (i)
For the corresponding age numerical value of output node i；
Step C23 obtains gender class probability p2 (i), and final gender recognition result is the gender of maximum probability,
Gender_pre=argmaxiP2 (i), wherein i is gender classification sequence number；
Step C24, obtains ethnic class probability p3 (i), and final species result is the race of maximum probability, race_
Pre=argmaxiP3 (i), wherein i is ethnic classification sequence number.
Preferably, two key points described in step A12 and step C12 are two central points and upper lip central point.
Then the present invention utilizes trained multi-task learning by single task model analysis, multi task model training
Convolutional neural networks judge a variety of attributes of face (age, sex, race), improve meter when face multi-attribute analysis
Calculate the generalization ability of speed and model.
Detailed description of the invention
Fig. 1 is the step A1 flow diagram of the present embodiment；
Fig. 2 is the step A2 flow diagram of the present embodiment；
Fig. 3 is the multitask convolutional neural networks block schematic illustration of the present embodiment；
Fig. 4 is the step B2 flow diagram of the present embodiment；
Fig. 5 is the step C1 flow diagram of the present embodiment；
Fig. 6 is the step C2 flow diagram of the present embodiment.
Specific embodiment
The preferred embodiment of the present invention described with reference to the accompanying drawings.It will be apparent to a skilled person that this
A little embodiments are used only for explaining technical principle of the invention, it is not intended that limit the scope of the invention.
The present invention is based on the face character analysis methods of the convolutional neural networks of multi-task learning, first carry out single task instruction
Practice, finds out and restrain most slow network；The weight of the most slow network model of trained convergence is assigned to multitask convolutional Neural again
The shared part of network, then carry out multitask and synchronize training；The work that this step assigns weight can be such that multitask training becomes easy,
Multitask network can largely be reduced for the training difficulty for restraining most slow task, make the convergence step base of each task
This is consistent.
The present invention includes that single task model analysis, multi task model training and face character judge three parts；
Single task model analysis:
The original sample of each age facial image is carried out face critical point detection, after pedestrian's face of going forward side by side alignment by step A1
It is cut according to pre-set dimension and generates the new samples comprising facial image；
Age estimation network, gender identification network, race is respectively trained in step A2, the new samples generated using step A1
Three single task convolutional neural networks of sorter network, the convergence rate of more each network obtain a most slow list of convergence rate
The weight of task convolutional neural networks；
Multi task model training:
Step B1 constructs multitask convolutional neural networks, and there are three tasks to export altogether for the network, respectively corresponds the age and estimates
It calculates, gender identification and species, three tasks all use softmax loss function as objective function；The multitask volume
Product neural network includes shared part for data sharing in multi-task learning and information exchange and for calculating above-mentioned three
The independent sector of a task output；Utilize the weight initialization multitask convolution of the step A2 single task convolutional neural networks obtained
The shared part of neural network, the multitask convolutional neural networks after forming initialization；
Step B2, using the new samples generated in step A1, training multitask convolutional neural networks are obtained trained more
Task convolutional neural networks model；
Face character judgement:
Step C1 carries out Face datection to inputted picture, judges whether comprising facial image, as schemed comprising if to input
As carrying out face critical point detection, then pedestrian's face of going forward side by side alignment cuts generation according to pre-set dimension and includes the new of facial image
Picture；
New picture obtained by step C1 is input to the multitask convolutional neural networks model that step B2 is obtained by step C2, into
The estimation of row age, gender identification and species.
As shown in Figure 1, step A1 specifically includes the following contents in the present embodiment:
Step A11 chooses each age facial image as original sample；
Step A12 carries out face critical point detection to selected original sample, obtains two key points；
Step A13, the alignment according to the position and its line of two key points to original sample progress facial image are described
The alignment of facial image includes rotation, scaling, translation to original sample；
Sample after being aligned in step A13 is cut according to pre-set dimension and generates the new sample comprising facial image by step A14
This.
As shown in Fig. 2, age estimation network, gender identification network, species is respectively trained in step A2 in the present embodiment
When three single task convolutional neural networks of network, the conditions such as learning rate, learning strategy are completely the same.Three are obtained in the present embodiment
The convergence rate of task is most slow for age estimation tasks, gender identification mission and species task convergence rate substantially phase
Together, so obtaining the weight for restraining most slow age estimation tasks network model.
It is illustrated in figure 3 the present embodiment multitask convolutional neural networks frame diagram.
In the present embodiment, the shared part of multitask convolutional neural networks is the bottom portion of multitask network in step B1
Point, including data input and convolutional layer and pond layer；Realize data sharing, information exchange in multi-task learning in shared part；
Three mission bit streams are shared, help to improve the generalization ability of model.As shown in figure 3, the inclusion layer of the present embodiment include input,
Convolutional layer 1, pond layer 1, convolutional layer 2, pond layer 2, convolutional layer 3, convolutional layer 4.
In the present embodiment, the independent sector of multitask convolutional neural networks in step B1 are as follows: the age is estimated, gender identifies,
Three independent network structures of task of species, each separate network structure possess layer (pond independent convolutional layer, pooling
Change layer) and full articulamentum, for training more single-minded feature.As shown in figure 3, the independent sector of the present embodiment is three points
Branch is respectively used to age estimation, gender identification, the calculating and output of species, and corresponding age estimation branch includes the age
Convolutional layer 5, age pond layer 5, age full articulamentum 6, gender estimate that branch includes gender convolutional layer 5, gender pond layer 5, property
Infull articulamentum 6, race's estimation branch include ethnic convolutional layer 5, ethnic pond layer 5, the full articulamentum 6 of race.
In the present embodiment, single task network structure can also regard a part of multitask network as, only other
The independent sector of business eliminates, such as the structure of single task age network is by other and ethnic two tasks of multitask net neutral
Independent sector remove.
In the present embodiment, shown in multitask convolutional neural networks total losses function such as formula (1):
lmulti-task=α lage+β·lgender+γ·lrace (1)
Wherein lmulti-taskFor the total losses of multitask network, α, β, γ are respectively the weight system of preset three tasks
Number, lage、lgender、lraceAge estimation loss, gender identification loss, Zhong Zufen respectively in multitask convolutional neural networks
Class loss.
In the present embodiment, the method for training multitask convolutional neural networks is as shown in Figure 4 in step B2, comprising:
Step B21 randomly selects m images from the new samples that step A1 is generated, is input to and constructs in step bl is determined simultaneously
Initialized multitask convolutional neural networks carry out multitask and synchronize training；
Step B22 calculates separately age estimation loss l to transmitting before multitask convolutional neural networksage, gender identification damage
Lose lgenderL is lost with speciesrace；
Step B23 calculates the total losses l of multitask convolutional neural networksmulti-task；
Step B24: judging whether network training restrains, and deconditioning and obtains multitask convolutional neural networks if convergence
Model, it is no to then follow the steps B25；
Step B25: each parameter gradients of network are calculated using back-propagation algorithm, using stochastic gradient descent method
(Stochastic Gradient Descent, SGD) updates network parameter weight；Return step B21.
In the present embodiment, step C1 is specifically as shown in Figure 5, comprising:
Step C11 detects whether it includes face to the picture inputted, abandons the picture if not including face,
Otherwise C12 is entered step；
Step C12 carries out face critical point detection to the picture inputted, obtains two key points；
Step C13 carries out the alignment of facial image according to the position and its line of above-mentioned key point to original image, described
The alignment of facial image includes rotation, scaling, translation to original image；
Picture after being aligned in step C13 is cut according to pre-set dimension and generates the new figure comprising facial image by step C14
Piece.
In the present embodiment, step C2 is specifically as shown in Figure 6, comprising:
Step C21, the new picture that step C1 is obtained are input to trained multitask convolutional neural networks, forward direction
Transmitting carries out face multi-attribute analysis；
Step C22 obtains age Probability p 1 (i), and final age estimation result is the mathematical expectation at each age,Wherein output node number is k+1, and i is output node serial number, and a (i) is defeated
The corresponding age numerical value of egress i；
Step C23 obtains gender class probability p2 (i), and final gender recognition result is the gender of maximum probability,
Gender_pre=argmaxiP2 (i), wherein i is gender classification sequence number；
Step C24, obtains ethnic class probability p3 (i), and final species result is the race of maximum probability, race_
Pre=argmaxiP3 (i), wherein i is ethnic classification sequence number.
In the present embodiment, two key points described in step A12 and step C12 are two central points and upper lip center
Point, the resolution ratio of pre-set dimension described in step A14 or C14 are 224*224.
Those skilled in the art should be able to recognize that, side described in conjunction with the examples disclosed in the embodiments of the present disclosure
Method step, can be realized with electronic hardware, computer software, or a combination of the two, in order to clearly demonstrate electronic hardware and
The interchangeability of software generally describes each exemplary composition and step according to function in the above description.These
Function is executed actually with electronic hardware or software mode, specific application and design constraint depending on technical solution.
Those skilled in the art can use different methods to achieve the described function each specific application, but this reality
Now it should not be considered as beyond the scope of the present invention.
So far, it has been combined preferred embodiment shown in the drawings and describes technical solution of the present invention, still, this field
Technical staff is it is easily understood that protection scope of the present invention is expressly not limited to these specific embodiments.Without departing from this
Under the premise of the principle of invention, those skilled in the art can make equivalent change or replacement to the relevant technologies feature, these
Technical solution after change or replacement will fall within the scope of protection of the present invention.
Claims (10)
1. a kind of face character analysis method of the convolutional neural networks based on multi-task learning, which is characterized in that including single
Business model analysis, multi task model training and face character judge three parts；
Single task model analysis:
The original sample of each age facial image is carried out face critical point detection by step A1, after pedestrian's face of going forward side by side alignment according to
Pre-set dimension, which is cut, generates the new samples comprising facial image；
Age estimation network, gender identification network, species is respectively trained in step A2, the new samples generated using step A1
Three single task convolutional neural networks of network, the convergence rate of more each network obtain a most slow single task of convergence rate
The weight of convolutional neural networks；
Multi task model training:
Step B1 constructs multitask convolutional neural networks, and there are three tasks to export altogether for the network, respectively corresponds age estimation, property
Not Shi Bie and species, three tasks all use softmax loss function as objective function；The multitask convolutional Neural
Network includes shared part for data sharing in multi-task learning and information exchange and for calculating above three task
The independent sector of output；Utilize the weight initialization multitask convolutional Neural net of the step A2 single task convolutional neural networks obtained
The shared part of network, the multitask convolutional neural networks after forming initialization；
Step B2, using the new samples generated in step A1, training multitask convolutional neural networks obtain trained multitask
Convolutional neural networks model；
Face character judgement:
Step C1 carries out Face datection to inputted picture, judges whether comprising facial image, as comprising if to input picture into
Pedestrian's face critical point detection, pedestrian's face of going forward side by side alignment, then cuts according to pre-set dimension and generates the new picture comprising facial image；
New picture obtained by step C1 is input to the multitask convolutional neural networks model that step B2 is obtained by step C2, carries out year
Age estimation, gender identification and species.
2. the method according to claim 1, wherein step A1 specifically includes the following contents:
Step A11 chooses each age facial image as original sample；
Step A12 carries out face critical point detection to selected original sample, obtains two key points；
Step A13, the alignment according to the position and its line of two key points to original sample progress facial image, the face
The alignment of image includes rotation, scaling and/or translation to original sample；
Sample after being aligned in step A13 is cut according to pre-set dimension and generates the new samples comprising facial image by step A14.
3. the method according to claim 1, wherein age estimation network, gender is being respectively trained in step A2
When identifying three network, species network single task convolutional neural networks, learning rate and learning strategy are completely the same.
4. the method according to claim 1, wherein in step B1 multitask convolutional neural networks shared part
For the floor portions of multitask network, including data input, convolutional layer and pond layer.
5. according to the method described in claim 4, it is characterized in that, in step B1 multitask convolutional neural networks independent sector
Are as follows: age estimation, three gender identification, species independent network structures of task, each separate network structure possess independence
Convolutional layer, pooling layers and full articulamentum.
6. according to the method described in claim 5, it is characterized in that, multitask convolutional neural networks total losses function are as follows:
lmulti-task=α lage+β·lgender+γ·lrace
Wherein lmulti-taskFor the total losses of multitask network, α, β, γ are respectively the weight coefficient of preset three tasks, lage、
lgender、lraceAge estimation loss, gender identification loss, species loss respectively in multitask convolutional neural networks.
7. according to the method described in claim 6, it is characterized in that, the method for training multitask convolutional neural networks in step B2
Are as follows:
Step B21 randomly selects m images from the new samples that step A1 is generated, be input to building in step bl is determined and at the beginning of
The multitask convolutional neural networks of beginningization carry out multitask and synchronize training；
Step B22 calculates separately age estimation loss l to transmitting before multitask convolutional neural networksage, gender identification loss
lgenderL is lost with speciesrace；
Step B23 calculates the total losses l of multitask convolutional neural networksmulti-task；
Step B24: judging whether network training restrains, and deconditioning and obtains multitask convolutional neural networks mould if convergence
Type, it is no to then follow the steps B25；
Step B25: calculating each parameter gradients of network using back-propagation algorithm, updates network parameter using stochastic gradient descent method
Weight；Return step B21.
8. the method according to claim 1, wherein step C1 specifically includes the following contents:
Step C11 detects whether it includes face to the picture inputted, abandons the picture if not including face, otherwise
Enter step C12；
Step C12 carries out face critical point detection to the picture inputted, obtains two key points；
Step C13 carries out the alignment of facial image, the face according to the position and its line of above-mentioned key point to original image
The alignment of image includes rotation, scaling and/or translation to original image；
Picture after being aligned in step C13 is cut according to pre-set dimension and generates the new picture comprising facial image by step C14.
9. the method according to claim 1, wherein step C2 specifically includes the following contents:
Step C21, the new picture that step C1 is obtained are input to trained multitask convolutional neural networks, carry out face
Multi-attribute analysis；
Step C22 obtains age Probability p 1 (i), and final age estimation result is the mathematical expectation at each age,Wherein output node number is k+1, and i is output node serial number, and a (i) is defeated
The corresponding age numerical value of egress i；
Step C23, obtains gender class probability p2 (i), and final gender recognition result is the gender of maximum probability, gender_pre
=argmaxiP2 (i), wherein i is gender classification sequence number；
Step C24, obtains ethnic class probability p3 (i), and final species result is the race of maximum probability, race_pre=
argmaxiP3 (i), wherein i is ethnic classification sequence number.
10. the method according to claim 2 or 8, which is characterized in that two key points are two central points and upper
Lip central point.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610856231.1A CN106529402B (en) | 2016-09-27 | 2016-09-27 | The face character analysis method of convolutional neural networks based on multi-task learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610856231.1A CN106529402B (en) | 2016-09-27 | 2016-09-27 | The face character analysis method of convolutional neural networks based on multi-task learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106529402A CN106529402A (en) | 2017-03-22 |
| CN106529402B true CN106529402B (en) | 2019-05-28 |
Family
ID=58344469
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610856231.1A Active CN106529402B (en) | 2016-09-27 | 2016-09-27 | The face character analysis method of convolutional neural networks based on multi-task learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106529402B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11776323B2 (en) | 2022-02-15 | 2023-10-03 | Ford Global Technologies, Llc | Biometric task network |
Families Citing this family (110)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107133948B (en) * | 2017-05-09 | 2020-05-08 | 电子科技大学 | Image blurring and noise evaluation method based on multitask convolution neural network |
| CN110709856B (en) * | 2017-05-31 | 2023-11-28 | 宝洁公司 | System and method for determining apparent skin age |
| CN107247947B (en) * | 2017-07-07 | 2021-02-09 | 智慧眼科技股份有限公司 | Face attribute identification method and device |
| CN107463888B (en) * | 2017-07-21 | 2020-05-19 | 竹间智能科技（上海）有限公司 | Face emotion analysis method and system based on multi-task learning and deep learning |
| CN109299487B (en) * | 2017-07-25 | 2023-01-06 | 展讯通信（上海）有限公司 | Neural network system, accelerator, modeling method and device, medium and system |
| CN109359499A (en) | 2017-07-26 | 2019-02-19 | 虹软科技股份有限公司 | A kind of method and apparatus for face classifier |
| CN107545243A (en) * | 2017-08-07 | 2018-01-05 | 南京信息工程大学 | Yellow race's face identification method based on depth convolution model |
| CN107590460B (en) * | 2017-09-12 | 2019-05-03 | 北京达佳互联信息技术有限公司 | Face classification method, apparatus and intelligent terminal |
| CN109583277B (en) * | 2017-09-29 | 2021-04-20 | 大连恒锐科技股份有限公司 | Gender determination method of barefoot footprint based on CNN |
| US10962979B2 (en) * | 2017-09-30 | 2021-03-30 | Tusimple, Inc. | System and method for multitask processing for autonomous vehicle computation and control |
| CN107742107B (en) * | 2017-10-20 | 2019-03-01 | 北京达佳互联信息技术有限公司 | Facial image classification method, device and server |
| CN109697395A (en) * | 2017-10-24 | 2019-04-30 | 华为技术有限公司 | Thingness prediction technique, device and equipment |
| CN108021863B (en) * | 2017-11-01 | 2022-05-06 | 平安科技（深圳）有限公司 | Electronic device, age classification method based on image and storage medium |
| CN108052862B (en) * | 2017-11-09 | 2019-12-06 | 北京达佳互联信息技术有限公司 | Age estimation method and device |
| CN107886073B (en) * | 2017-11-10 | 2021-07-27 | 重庆邮电大学 | Fine-grained vehicle multi-attribute identification method based on convolutional neural network |
| CN108197525B (en) * | 2017-11-20 | 2020-08-11 | 中国科学院自动化研究所 | Face image generation method and device |
| CN108171254A (en) * | 2017-11-22 | 2018-06-15 | 北京达佳互联信息技术有限公司 | Image tag determines method, apparatus and terminal |
| CN110444214B (en) * | 2017-11-24 | 2021-08-17 | 深圳市腾讯计算机***有限公司 | Speech signal processing model training method and device, electronic equipment and storage medium |
| CN107844781A (en) * | 2017-11-28 | 2018-03-27 | 腾讯科技（深圳）有限公司 | Face character recognition methods and device, electronic equipment and storage medium |
| CN108564097B (en) * | 2017-12-05 | 2020-09-22 | 华南理工大学 | Multi-scale target detection method based on deep convolutional neural network |
| CN107977633B (en) * | 2017-12-06 | 2019-04-09 | 平安科技（深圳）有限公司 | Age recognition methods, device and the storage medium of facial image |
| CN107844784A (en) * | 2017-12-08 | 2018-03-27 | 广东美的智能机器人有限公司 | Face identification method, device, computer equipment and readable storage medium storing program for executing |
| CN107958444A (en) * | 2017-12-28 | 2018-04-24 | 江西高创保安服务技术有限公司 | A kind of face super-resolution reconstruction method based on deep learning |
| CN108345832A (en) * | 2017-12-28 | 2018-07-31 | 新智数字科技有限公司 | A kind of method, apparatus and equipment of Face datection |
| CN108229503A (en) * | 2018-01-04 | 2018-06-29 | 浙江大学 | A kind of feature extracting method for clothes photo |
| CN108256482B (en) * | 2018-01-18 | 2021-06-25 | 中科视拓（北京）科技有限公司 | Face age estimation method for distributed learning based on convolutional neural network |
| CN108875932A (en) * | 2018-02-27 | 2018-11-23 | 北京旷视科技有限公司 | Image-recognizing method, device and system and storage medium |
| CN108416327B (en) * | 2018-03-28 | 2022-04-29 | 京东方科技集团股份有限公司 | Target detection method and device, computer equipment and readable storage medium |
| CN108509041A (en) * | 2018-03-29 | 2018-09-07 | 百度在线网络技术（北京）有限公司 | Method and apparatus for executing operation |
| CN108537193A (en) * | 2018-04-17 | 2018-09-14 | 厦门美图之家科技有限公司 | Ethnic attribute recognition approach and mobile terminal in a kind of face character |
| CN110395257B (en) * | 2018-04-20 | 2021-04-23 | 北京图森未来科技有限公司 | Lane line example detection method and device and automatic driving vehicle |
| CN108665065B (en) * | 2018-04-25 | 2020-08-04 | 清华大学 | Method, device and equipment for processing task data and storage medium |
| CN108764048B (en) * | 2018-04-28 | 2021-03-16 | 中国科学院自动化研究所 | Face key point detection method and device |
| CN110516512B (en) * | 2018-05-21 | 2023-08-25 | 北京中科奥森数据科技有限公司 | Training method of pedestrian attribute analysis model, pedestrian attribute identification method and device |
| CN108734146A (en) * | 2018-05-28 | 2018-11-02 | 北京达佳互联信息技术有限公司 | Facial image Age estimation method, apparatus, computer equipment and storage medium |
| CN108960283B (en) * | 2018-05-30 | 2022-01-11 | 北京市商汤科技开发有限公司 | Classification task increment processing method and device, electronic equipment and storage medium |
| CN108764203A (en) * | 2018-06-06 | 2018-11-06 | 四川大学 | A kind of pedestrian's quantitative analysis and display systems towards urban planning |
| CN108764207B (en) * | 2018-06-07 | 2021-10-19 | 厦门大学 | Face expression recognition method based on multitask convolutional neural network |
| CN109325398B (en) * | 2018-06-30 | 2020-10-09 | 东南大学 | Human face attribute analysis method based on transfer learning |
| CN109063759A (en) * | 2018-07-20 | 2018-12-21 | 浙江大学 | A kind of neural network structure searching method applied to the more attribute forecasts of picture |
| CN109145798B (en) * | 2018-08-13 | 2021-10-22 | 浙江零跑科技股份有限公司 | Driving scene target identification and travelable region segmentation integration method |
| US10747989B2 (en) * | 2018-08-21 | 2020-08-18 | Software Ag | Systems and/or methods for accelerating facial feature vector matching with supervised machine learning |
| CN109145987B (en) * | 2018-08-21 | 2020-10-20 | 厦门美图之家科技有限公司 | Model generation method and device |
| CN109242516A (en) * | 2018-09-06 | 2019-01-18 | 北京京东尚科信息技术有限公司 | The single method and apparatus of processing service |
| CN109472193A (en) * | 2018-09-21 | 2019-03-15 | 北京飞搜科技有限公司 | Method for detecting human face and device |
| CN109359562A (en) * | 2018-09-29 | 2019-02-19 | 佳都新太科技股份有限公司 | Target identification method, device, target identification equipment and storage medium |
| CN111079479A (en) | 2018-10-19 | 2020-04-28 | 北京市商汤科技开发有限公司 | Child state analysis method and device, vehicle, electronic device and storage medium |
| CN109086837A (en) * | 2018-10-24 | 2018-12-25 | 高嵩 | User property classification method, storage medium, device and electronic equipment based on convolutional neural networks |
| CN109508654B (en) * | 2018-10-26 | 2021-01-05 | 中国地质大学（武汉） | Face analysis method and system fusing multitask and multi-scale convolutional neural network |
| CN109522824A (en) * | 2018-10-30 | 2019-03-26 | 平安科技（深圳）有限公司 | Face character recognition methods, device, computer installation and storage medium |
| CN109522872A (en) * | 2018-12-04 | 2019-03-26 | 西安电子科技大学 | A kind of face identification method, device, computer equipment and storage medium |
| CN111291253B (en) * | 2018-12-06 | 2023-09-19 | 北京嘀嘀无限科技发展有限公司 | Model training method, consultation recommendation method and device and electronic equipment |
| CN111291604A (en) * | 2018-12-07 | 2020-06-16 | 深圳光启空间技术有限公司 | Face attribute identification method, device, storage medium and processor |
| CN109727071A (en) * | 2018-12-28 | 2019-05-07 | 中国科学院半导体研究所 | Method and system for advertisement recommendation |
| CN109583431A (en) * | 2019-01-02 | 2019-04-05 | 上海极链网络科技有限公司 | A kind of face Emotion identification model, method and its electronic device |
| CN110427795A (en) * | 2019-01-28 | 2019-11-08 | 厦门瑞为信息技术有限公司 | A kind of property analysis method based on head photo, system and computer equipment |
| CN109886154A (en) * | 2019-01-30 | 2019-06-14 | 电子科技大学 | Most pedestrian's appearance attribute recognition methods according to collection joint training based on Inception V3 |
| CN111598600A (en) * | 2019-02-21 | 2020-08-28 | 虹软科技股份有限公司 | Multimedia information pushing method and system and terminal equipment |
| CN109871814B (en) * | 2019-02-22 | 2022-06-21 | 成都旷视金智科技有限公司 | Age estimation method and device, electronic equipment and computer storage medium |
| CN111626087A (en) * | 2019-02-28 | 2020-09-04 | 北京市商汤科技开发有限公司 | Neural network training and eye opening and closing state detection method, device and equipment |
| CN109947960B (en) * | 2019-03-08 | 2024-01-23 | 南京信息工程大学 | Face multi-attribute joint estimation model construction method based on depth convolution |
| CN109893095A (en) * | 2019-03-11 | 2019-06-18 | 常州市贝叶斯智能科技有限公司 | A kind of intelligent robot system of body composition detection and analysis |
| CN110069994B (en) * | 2019-03-18 | 2021-03-23 | 中国科学院自动化研究所 | Face attribute recognition system and method based on face multiple regions |
| CN109993102B (en) * | 2019-03-28 | 2021-09-17 | 北京达佳互联信息技术有限公司 | Similar face retrieval method, device and storage medium |
| JP7231464B2 (en) | 2019-04-09 | 2023-03-01 | 株式会社日立製作所 | Object recognition system and object recognition method |
| CN110046571B (en) * | 2019-04-15 | 2020-05-15 | 北京字节跳动网络技术有限公司 | Method and device for identifying age |
| CN110084216B (en) * | 2019-05-06 | 2021-11-09 | 苏州科达科技股份有限公司 | Face recognition model training and face recognition method, system, device and medium |
| CN110163269A (en) * | 2019-05-09 | 2019-08-23 | 北京迈格威科技有限公司 | Model generating method, device and computer equipment based on deep learning |
| CN110111246B (en) * | 2019-05-15 | 2022-02-25 | 北京市商汤科技开发有限公司 | Virtual head portrait generation method and device and storage medium |
| CN110334735B (en) * | 2019-05-31 | 2022-07-08 | 北京奇艺世纪科技有限公司 | Multitask network generation method and device, computer equipment and storage medium |
| CN110232352B (en) * | 2019-06-12 | 2022-11-25 | 东北大学 | Improvement method of multi-task cascade convolution neural network model for face recognition |
| CN110287836B (en) * | 2019-06-14 | 2021-10-15 | 北京迈格威科技有限公司 | Image classification method and device, computer equipment and storage medium |
| CN112149449A (en) * | 2019-06-26 | 2020-12-29 | 北京华捷艾米科技有限公司 | Face attribute recognition method and system based on deep learning |
| CN110263756A (en) * | 2019-06-28 | 2019-09-20 | 东北大学 | A kind of human face super-resolution reconstructing system based on joint multi-task learning |
| CN110348416A (en) * | 2019-07-17 | 2019-10-18 | 北方工业大学 | Multi-task face recognition method based on multi-scale feature fusion convolutional neural network |
| CN112333545B (en) * | 2019-07-31 | 2022-03-22 | Tcl科技集团股份有限公司 | Television content recommendation method, system, storage medium and smart television |
| CN110414489A (en) * | 2019-08-21 | 2019-11-05 | 五邑大学 | A kind of face beauty prediction technique based on multi-task learning |
| CN110503072B (en) * | 2019-08-29 | 2023-05-26 | 南京信息工程大学 | Face age estimation method based on multi-branch CNN architecture |
| CN110584654A (en) * | 2019-10-09 | 2019-12-20 | 中山大学 | Multi-mode convolutional neural network-based electrocardiosignal classification method |
| CN112825122A (en) * | 2019-11-20 | 2021-05-21 | 北京眼神智能科技有限公司 | Ethnicity judgment method, ethnicity judgment device, ethnicity judgment medium and ethnicity judgment equipment based on two-dimensional face image |
| CN112825117A (en) * | 2019-11-20 | 2021-05-21 | 北京眼神智能科技有限公司 | Behavior attribute judgment method, behavior attribute judgment device, behavior attribute judgment medium and behavior attribute judgment equipment based on head features |
| CN112825119A (en) * | 2019-11-20 | 2021-05-21 | 北京眼神智能科技有限公司 | Face attribute judgment method and device, computer readable storage medium and equipment |
| CN111191675B (en) * | 2019-12-03 | 2023-10-24 | 深圳市华尊科技股份有限公司 | Pedestrian attribute identification model realization method and related device |
| CN111339818B (en) * | 2019-12-18 | 2023-06-06 | 中国人民解放军第四军医大学 | Face multi-attribute recognition system |
| CN110874587B (en) * | 2019-12-26 | 2020-07-28 | 浙江大学 | Face characteristic parameter extraction system |
| CN111325108B (en) * | 2020-01-22 | 2023-05-26 | 中能国际高新科技研究院有限公司 | Multitasking network model, using method, device and storage medium |
| CN111369615B (en) * | 2020-02-21 | 2023-05-23 | 苏州优纳医疗器械有限公司 | Nuclear central point detection method based on multitasking convolutional neural network |
| CN111539452B (en) * | 2020-03-26 | 2024-03-26 | 深圳云天励飞技术有限公司 | Image recognition method and device for multi-task attribute, electronic equipment and storage medium |
| CN111401294B (en) * | 2020-03-27 | 2022-07-15 | 山东财经大学 | Multi-task face attribute classification method and system based on adaptive feature fusion |
| CN111507263B (en) * | 2020-04-17 | 2022-08-05 | 电子科技大学 | Face multi-attribute recognition method based on multi-source data |
| CN111626115A (en) * | 2020-04-20 | 2020-09-04 | 北京市西城区培智中心学校 | Face attribute identification method and device |
| CN111680547B (en) * | 2020-04-27 | 2024-01-12 | 阿波罗智能技术(北京)有限公司 | Traffic countdown sign recognition method and device, electronic equipment and storage medium |
| CN111598000A (en) * | 2020-05-18 | 2020-08-28 | 中移（杭州）信息技术有限公司 | Face recognition method, device, server and readable storage medium based on multiple tasks |
| WO2021237727A1 (en) * | 2020-05-29 | 2021-12-02 | Siemens Aktiengesellschaft | Method and apparatus of image processing |
| CN111626303B (en) * | 2020-05-29 | 2021-04-13 | 南京甄视智能科技有限公司 | Sex and age identification method, sex and age identification device, storage medium and server |
| CN112052734B (en) * | 2020-07-31 | 2024-05-17 | 北京奇艺世纪科技有限公司 | Character characteristic recognition network generation method, character characteristic recognition network generation program, character characteristic recognition apparatus, device and medium |
| US20220327361A1 (en) * | 2020-09-11 | 2022-10-13 | Boe Technology Group Co., Ltd. | Method for Training Joint Model, Object Information Processing Method, Apparatus, and System |
| US20230290134A1 (en) * | 2020-09-25 | 2023-09-14 | Intel Corporation | Method and system of multiple facial attributes recognition using highly efficient neural networks |
| CN112085540A (en) * | 2020-09-27 | 2020-12-15 | 湖北科技学院 | Intelligent advertisement pushing system and method based on artificial intelligence technology |
| CN112163676B (en) * | 2020-10-13 | 2024-04-05 | 北京百度网讯科技有限公司 | Method, device, equipment and storage medium for training multitasking service prediction model |
| US11675876B2 (en) | 2020-10-28 | 2023-06-13 | International Business Machines Corporation | Training robust machine learning models |
| CN114529731A (en) * | 2020-10-30 | 2022-05-24 | 北京眼神智能科技有限公司 | Face feature point positioning and attribute analysis method and device, storage medium and equipment |
| CN112784916B (en) * | 2021-01-29 | 2022-03-04 | 西安电子科技大学 | Air target micro-motion parameter real-time extraction method based on multitask convolutional network |
| CN112906668B (en) * | 2021-04-07 | 2023-08-25 | 上海应用技术大学 | Face information identification method based on convolutional neural network |
| CN113095211B (en) * | 2021-04-08 | 2023-08-25 | 第六镜科技(成都)有限公司 | Image processing method, system and electronic equipment |
| CN113420768A (en) * | 2021-08-24 | 2021-09-21 | 深圳市信润富联数字科技有限公司 | Core category determination method and device, electronic equipment and storage medium |
| CN114005166A (en) * | 2021-11-25 | 2022-02-01 | 浩云科技股份有限公司 | Face recognition method and device based on multi-task learning |
| CN114332914A (en) * | 2021-11-29 | 2022-04-12 | 中国电子科技集团公司电子科学研究院 | Personnel feature identification method, device and computer-readable storage medium |
| CN114821538A (en) * | 2022-05-19 | 2022-07-29 | 北京地平线机器人技术研发有限公司 | Training method and device of multi-task model |
| CN117079337B (en) * | 2023-10-17 | 2024-02-06 | 成都信息工程大学 | High-precision face attribute feature recognition device and method |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103824054A (en) * | 2014-02-17 | 2014-05-28 | 北京旷视科技有限公司 | Cascaded depth neural network-based face attribute recognition method |
| CN105469041A (en) * | 2015-11-19 | 2016-04-06 | 上海交通大学 | Facial point detection system based on multi-task regularization and layer-by-layer supervision neural networ |
| CN105760859A (en) * | 2016-03-22 | 2016-07-13 | 中国科学院自动化研究所 | Method and device for identifying reticulate pattern face image based on multi-task convolutional neural network |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106062774B (en) * | 2014-11-15 | 2020-01-03 | 北京旷视科技有限公司 | Face detection using machine learning |
-
2016
- 2016-09-27 CN CN201610856231.1A patent/CN106529402B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103824054A (en) * | 2014-02-17 | 2014-05-28 | 北京旷视科技有限公司 | Cascaded depth neural network-based face attribute recognition method |
| CN105469041A (en) * | 2015-11-19 | 2016-04-06 | 上海交通大学 | Facial point detection system based on multi-task regularization and layer-by-layer supervision neural networ |
| CN105760859A (en) * | 2016-03-22 | 2016-07-13 | 中国科学院自动化研究所 | Method and device for identifying reticulate pattern face image based on multi-task convolutional neural network |
Non-Patent Citations (1)
| Title |
|---|
| 多任务学习及卷积神经网络在人脸识别中的应用;邵蔚元 等;《计算机工程与应用》;20160701;第52卷(第13期);第32-37,88页 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11776323B2 (en) | 2022-02-15 | 2023-10-03 | Ford Global Technologies, Llc | Biometric task network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106529402A (en) | 2017-03-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106529402B (en) | The face character analysis method of convolutional neural networks based on multi-task learning | |
| CN106503623B (en) | Facial image age estimation method based on convolutional neural networks | |
| WO2022111236A1 (en) | Facial expression recognition method and system combined with attention mechanism | |
| CN104392228B (en) | Unmanned plane image object class detection method based on conditional random field models | |
| CN106897714A (en) | A kind of video actions detection method based on convolutional neural networks | |
| CN113344976A (en) | Visual tracking method based on target object characterization point estimation | |
| CN108805083A (en) | The video behavior detection method of single phase | |
| CN107436950A (en) | A kind of itinerary recommends method and system | |
| CN109376747A (en) | A kind of video flame detecting method based on double-current convolutional neural networks | |
| CN105447529A (en) | Costume detection and attribute value identification method and system | |
| CN110309732A (en) | Activity recognition method based on skeleton video | |
| CN107292950A (en) | A kind of end-to-end three-dimensional facial reconstruction method based on deep neural network | |
| CN110097029B (en) | Identity authentication method based on high way network multi-view gait recognition | |
| CN106203430A (en) | A kind of significance object detecting method based on foreground focused degree and background priori | |
| CN113408455A (en) | Action identification method, system and storage medium based on multi-stream information enhanced graph convolution network | |
| CN109543602A (en) | A kind of recognition methods again of the pedestrian based on multi-view image feature decomposition | |
| CN111985405A (en) | Face age synthesis method and system | |
| CN101441717A (en) | Method and system for detecting eroticism video | |
| CN110246148A (en) | The conspicuousness detection method of multi-modal depth information fusion and attention study | |
| CN101710418A (en) | Interactive mode image partitioning method based on geodesic distance | |
| CN114582030A (en) | Behavior recognition method based on service robot | |
| CN109255382A (en) | For the nerve network system of picture match positioning, method and device | |
| CN113610046B (en) | Behavior recognition method based on depth video linkage characteristics | |
| CN110006444A (en) | A kind of anti-interference visual odometry construction method based on optimization mixed Gauss model | |
| CN115359366A (en) | Remote sensing image target detection method based on parameter optimization |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | 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 |