CN105512620B - The training method and device of convolutional neural networks for recognition of face - Google Patents
The training method and device of convolutional neural networks for recognition of face Download PDFInfo
- Publication number
- CN105512620B CN105512620B CN201510857317.1A CN201510857317A CN105512620B CN 105512620 B CN105512620 B CN 105512620B CN 201510857317 A CN201510857317 A CN 201510857317A CN 105512620 B CN105512620 B CN 105512620B
- Authority
- CN
- China
- Prior art keywords
- sample
- training
- facial image
- convolutional neural
- neural networks
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses the training methods and device of a kind of convolutional neural networks for recognition of face, belong to field of face identification, this method comprises: building sample training library；Use sample training library training convolutional neural networks；The feature vector of all samples in sample training library is extracted using the convolutional neural networks after training；Calculate the distance between every two feature vector；Sample is constructed to training library；Sample is combined into training library by the sample of face images sample to collection, sample to set include foreign peoples's sample to similar sample pair, foreign peoples's sample is constituted to by facial image sample and the facial image sample for being less than certain value with the facial image sample distance, and similar sample is constituted to by facial image sample and the facial image sample for being greater than certain value with the facial image sample distance；Using sample to training library training convolutional neural networks.This method, which is effectively avoided, identifies mistake as caused by makeup and external environment influence.
Description
Technical field
The invention belongs to field of face identification, particularly relate to a kind of training side of convolutional neural networks for recognition of face
Method and device.
Background technique
With going deep into for the rise of deep learning, the especially research of depth convolutional neural networks, largely based on convolution mind
Network model through network (Convolutional Neural Network, CNN) is applied to image procossing and image recognition
Etc., the achievement to attract people's attention is especially achieved in field of face identification.
It would generally be had problems that in recognition of face and field of authentication, for example, due to makeup and external environment influence
It is possible that the photo of two different peoples is much like, two photos of same person differ greatly.This kind of exceptional sample is to cause
Identify the major reason of mistake.
Summary of the invention
The present invention provides the training method and device of a kind of convolutional neural networks for recognition of face, and this method is effective
It avoids and identifies mistake as caused by makeup and external environment influence.
In order to solve the above technical problems, present invention offer technical solution is as follows:
A kind of training method of the convolutional neural networks for recognition of face, comprising:
Construct sample training library；Wherein: the sample training library includes multiple sample classes, includes multiple in each sample class
Facial image sample；Wherein, the facial image sample of the same person forms a sample class, and each facial image sample standard deviation is corresponding
There is a class label, the class label of multiple facial image samples in a sample class is identical；
Use sample training library training convolutional neural networks；
The feature vector of all samples in the sample training library is extracted using the convolutional neural networks after training；
Calculate the distance between every two feature vector；
Sample is constructed to training library；Wherein: the sample is to training library by the sample of face images sample to set
Composition, the sample to set include foreign peoples's sample to similar sample pair, foreign peoples's sample to by facial image sample and
It is less than certain value from the facial image sample distance and belongs to the facial image sample of different sample classes with the facial image sample
Constitute, the similar sample to by facial image sample and with facial image sample distance be greater than certain value and with the face figure
The decent facial image sample for belonging to identical sample class is constituted；
Using the sample to training library training convolutional neural networks.
A kind of method of recognition of face, comprising:
Acquire facial image；
The feature vector of facial image is extracted using convolutional neural networks, the convolutional neural networks pass through above-mentioned method
Training obtains；
Recognition of face is carried out using described eigenvector.
A kind of training device of the convolutional neural networks for recognition of face, comprising:
First construction unit, for constructing sample training library；Wherein: the sample training library includes multiple sample classes, often
It include multiple facial image samples in a sample class；Wherein, the facial image sample of the same person forms a sample class, each
Facial image sample standard deviation is corresponding with a class label, the class label phase of multiple facial image samples in a sample class
Together；
First training unit, for using sample training library training convolutional neural networks；
Extraction unit, for extracting all samples in the sample training library using the convolutional neural networks after training
Feature vector；
Computing unit, for calculating the distance between every two feature vector；
Second construction unit, for constructing sample to training library；Wherein: the sample is to training library by face images
The sample of sample is combined into collection, the sample to set include foreign peoples's sample to similar sample pair, foreign peoples's sample pair
By facial image sample and it is less than certain value from the facial image sample distance and belongs to different samples with the facial image sample
The facial image sample of class is constituted, and the similar sample is greater than one to by facial image sample and with the facial image sample distance
Definite value and the facial image sample composition for belonging to identical sample class with the facial image sample；
Second training unit, for using the sample to training library training convolutional neural networks.
A kind of device of recognition of face, comprising:
Acquisition module, for acquiring facial image；
Extraction module, for using convolutional neural networks to extract the feature vector of facial image, the convolutional neural networks
It is obtained by above-mentioned device training；
Identification module, for carrying out recognition of face using described eigenvector.
The invention has the following advantages:
The present invention uses sample training library training convolutional neural networks first, the convolutional Neural net then obtained using training
The distance between network extracts feature, and calculate two feature vectors；Sample is being constructed to training library, then according to range information
Using sample to training convolutional neural networks again, the parameter of convolutional neural networks is further adjusted.
Due to sample to training library by foreign peoples's sample of face images sample to similar sample to constituting, foreign peoples's sample
This is constituted to by facial image sample and the facial image sample for being less than certain value with the facial image sample distance, similar sample
It is constituted to by facial image sample and the facial image sample for being greater than certain value with the facial image sample distance, uses sample pair
Training library training convolutional neural networks again, i.e., so that convolutional neural networks to these similar samples to and foreign peoples's sample difference
It is more sensitive, i.e., it is " reinforcing otherness comparison " above-mentioned, effectively avoid the identification as caused by makeup and external environment influence
Mistake.
Therefore, the present invention, which is effectively avoided, identifies mistake as caused by makeup and external environment influence.
Detailed description of the invention
Fig. 1 is the process of one embodiment of the training method of the convolutional neural networks for recognition of face of the invention
Figure；
Fig. 2 is the schematic diagram of one embodiment of the convolutional neural networks in the present invention；
Fig. 3 is the signal of one embodiment of the training device of the convolutional neural networks for recognition of face of the invention
Figure.
Specific embodiment
To keep the technical problem to be solved in the present invention, technical solution and advantage clearer, below in conjunction with attached drawing and tool
Body embodiment is described in detail.
On the one hand, the embodiment of the present invention provides a kind of training method of convolutional neural networks for recognition of face, such as Fig. 1
It is shown, comprising:
Step 101: building sample training library；Wherein: sample training library includes multiple sample classes, includes in each sample class
Multiple facial image samples.Sample training library is to arrange the sample complete or collected works to be formed by pretreated facial image sample.Specifically
, these facial image samples are divided into k sample class (the facial image sample of the same person forms a sample class), each sample
Class includes multiple facial image samples.And each facial image sample standard deviation is corresponding with a class label, in a sample class
Facial image sample class label it is identical.
Step 102: using sample training library training convolutional neural networks.When training, need behind convolutional neural networks
A sorter network is added, which can be divided into facial image sample k classification, and (k is the number of sample class above-mentioned
Amount), the parameters value of convolutional neural networks is then obtained by backpropagation.
Step 103: the feature vector of all samples in sample training library is extracted using the convolutional neural networks after training.
Each facial image sample is passed through into above-mentioned trained convolutional neural networks, obtains the feature vector of a fixed dimension.
Step 104: calculating the distance between every two feature vector.Distance is the various distances under generalized distance defines,
Such as Euclidean distance or mahalanobis distance.
Step 105: building sample is to training library；Wherein: sample is to training library by the sample pair of face images sample
Collection is combined into, sample to set include foreign peoples's sample to similar sample pair, foreign peoples's sample to by facial image sample and with this
Facial image sample distance is less than certain value and the facial image sample for belonging to different sample classes from the facial image sample is constituted,
Similar sample to by facial image sample and with facial image sample distance be greater than certain value and with the facial image sample category
It is constituted in the facial image sample of identical sample class.
Step 106: using sample to training library training convolutional neural networks.When training, need after convolutional neural networks
Face adds a sorter network, which can be divided into the result of facial image sample pair in a 2 classifications (i.e. face sample pair
It is the same person or different people), the parameters value of convolutional neural networks is then obtained by backpropagation.
Characteristics of cognition based on people: by constantly expanding congener body sensing range and reinforcing otherness comparison, Ke Yiti
The identification of high identity.For example, wrong identification can often occur for strange twins, however then for known twins
It can differentiate rapidly.For another example, even if with heavy makeup for known star, can also be easy to be recognized.
In conjunction with this discovery, in the training algorithm of analogy to convolutional neural networks, the embodiment of the present invention uses sample first
Then training library training convolutional neural networks extract feature using the convolutional neural networks that training obtains, and calculate two spies
Levy the distance between vector；Sample is being constructed to training library, then using sample to the mind of training convolutional again according to range information
Through network, the parameter of convolutional neural networks is further adjusted.
Due to sample to training library by foreign peoples's sample of face images sample to similar sample to constituting, foreign peoples's sample
This is constituted to by facial image sample and the facial image sample for being less than certain value with the facial image sample distance, similar sample
It is constituted to by facial image sample and the facial image sample for being greater than certain value with the facial image sample distance, uses sample pair
Training library training convolutional neural networks again, i.e., so that convolutional neural networks to these similar samples to and foreign peoples's sample difference
It is more sensitive, i.e., it is " reinforcing otherness comparison " above-mentioned, effectively avoid the identification as caused by makeup and external environment influence
Mistake.
Therefore, the embodiment of the present invention, which is effectively avoided, identifies mistake as caused by makeup and external environment influence.
In the embodiment of the present invention, various method training convolutional neural networks can be used, it is preferred that use in step 102
The method of sample training library training convolutional neural networks can be with are as follows:
Using sample training library, and pass through softmax classifier training convolutional neural networks, point of softmax classifier
Class quantity is identical as the quantity of sample class.
In the present embodiment, in training, the feature vector that convolutional neural networks extract facial image sample is first passed through, so
Classified afterwards using softmax classifier, the classification quantity of softmax classifier at this time is identical as the quantity of sample class
(i.e. sample class has several, and facial image sample is just divided into several classes by softmax classifier), then backpropagation is carried out, so repeatedly
In generation, repeatedly, obtains the parameter of convolutional neural networks.The embodiment of the present invention is trained by softmax sorter network, is avoided
Gradient disperse problem.
In above-mentioned each embodiment, distance is the various distances under generalized distance defines, that is, meets orthotropicity, symmetry with
And all measures of triangle inequality.For convenience of calculation, it is preferred that distance is Euclidean distance or mahalanobis distance.
It is further, further to training library training convolutional neural networks using sample in step 106 are as follows:
Using sample to training library, and by softmax classifier training convolutional neural networks, softmax classifier is
Two classifiers.
In the present embodiment, in training, the feature vector that convolutional neural networks extract facial image sample is first passed through, so
Classified afterwards using softmax classifier, softmax classifier at this time is that (i.e. softmax classifier is to sample for two classifiers
This judges whether it is same class to handling), then backpropagation is carried out, such iteration is multiple, obtains convolutional neural networks
Parameter.The embodiment of the present invention is trained by softmax sorter network, simple and convenient, avoids gradient disperse problem.
Moreover, as shown in Fig. 2, above-mentioned convolutional neural networks include:
Convolution operation is carried out to facial image sample, obtains convolution characteristic pattern；
Activation operation is carried out to convolution characteristic pattern, obtains activation characteristic pattern；
Down-sampling operation is carried out to activation characteristic pattern, obtains sampling characteristic pattern；
Above-mentioned steps are repeated several times to sampling characteristic pattern；
Vectorization operation is carried out, facial image sampling feature vectors are obtained.
The present invention is illustrated with a preferred embodiment below:
1. the training convolutional neural networks first under classification task are passing through for the structure of CNN+softmax classifier
Pretreated sample training library([n] indicates set { 1,2 ..., n }, x hereiIndicate i-th of facial image sample
This,For the class label of corresponding individual, that is, indicate which people i-th of sample is) on the training CNN network.
2. extracting face images sample using the convolutional neural networks after training as face picture feature extractor
Feature vector { (fi)}i∈[n].Here all feature vector dimensions are all the same.
3. then calculating the distance between the feature vector of combination of two, distance here is each under generalized distance defines
Kind distance, that is, meet orthotropicity, all measures of symmetry and triangle inequality.Such as Euclidean distance, mahalanobis distance
Etc..
4. the distance in step 3 between calculated feature vector two-by-two is combined, to each facial image sample architecture sample
To set.Sample includes the two or more foreign peoples samples nearest with the sample, foreign peoples's sample pair formed therewith to set.With
And two or more similar sample similar samples pair formed therewith with the sample farthest.Then new sample is formed to training
Library:
HereIt is the similar sample characteristics remote with ith feature vector distance m.It is with ith feature to span
The foreign peoples sample characteristics close from m.ysIndicate the bivector [1,0] of same person ', ydIndicate different people bivector [0,
1]’。
5. by the CNN network after step 1 training in conjunction with bis- classifier of softmax, on sample is to training library
With BP algorithm training CNN network, final convolutional neural networks are obtained.
On the other hand, the embodiment of the present invention provides a kind of method of recognition of face, comprising:
Acquire facial image；
The feature vector of facial image is extracted using convolutional neural networks, convolutional neural networks are trained by the above method
It arrives；
Recognition of face is carried out using feature vector.
The embodiment of the present invention, which is effectively avoided, identifies mistake as caused by makeup and external environment influence.
In another aspect, the embodiment of the present invention provides a kind of training device of convolutional neural networks for recognition of face, such as
Shown in Fig. 3, comprising:
First construction unit 11, for constructing sample training library；Wherein: sample training library includes multiple sample classes, each
It include multiple facial image samples in sample class；Wherein, the facial image sample of the same person forms a sample class, everyone
Face image sample standard deviation is corresponding with a class label, and the class label of multiple facial image samples in a sample class is identical；
First training unit 12, for using sample training library training convolutional neural networks；
Extraction unit 13, for using the convolutional neural networks after training to extract the spy of all samples in sample training library
Levy vector；
Computing unit 14, for calculating the distance between every two feature vector；
Second construction unit 15, for constructing sample to training library；Wherein: sample is to training library by face images sample
This sample is combined into collection, sample to set include foreign peoples's sample to similar sample pair, foreign peoples's sample is to by facial image
Sample and the face figure for belonging to different sample classes from the facial image sample distance less than certain value and with the facial image sample
Decent composition, similar sample to by facial image sample and with facial image sample distance be greater than certain value and with the face
The facial image sample that image pattern belongs to identical sample class is constituted；
Second training unit 16, for using sample to training library training convolutional neural networks.
The embodiment of the present invention, which is effectively avoided, identifies mistake as caused by makeup and external environment influence.
In the embodiment of the present invention, various method training convolutional neural networks can be used, it is preferred that
First training unit is further used for:
Using sample training library, and pass through softmax classifier training convolutional neural networks, point of softmax classifier
Class quantity is identical as the quantity of sample class.The embodiment of the present invention is trained by softmax sorter network, avoids gradient more
The problem of dissipating.
In above-mentioned each embodiment, distance is the various distances under generalized distance defines, that is, meets orthotropicity, symmetry with
And all measures of triangle inequality.For convenience of calculation, it is preferred that distance is Euclidean distance or mahalanobis distance.
Further, the second training unit is further used for:
Using sample to training library, and by softmax classifier training convolutional neural networks, softmax classifier is
Two classifiers.The embodiment of the present invention is trained by softmax sorter network, simple and convenient, avoids gradient disperse problem.
Moreover, above-mentioned convolutional neural networks include:
Convolution unit obtains convolution characteristic pattern for carrying out convolution operation to facial image sample；
Unit is activated, for carrying out activation operation to convolution characteristic pattern, obtains activation characteristic pattern；
Downsampling unit obtains sampling characteristic pattern for carrying out down-sampling operation to activation characteristic pattern；
Above-mentioned steps are repeated several times to sampling characteristic pattern；
Vectorization unit obtains facial image sampling feature vectors for carrying out vectorization operation.
In another aspect, the embodiment of the present invention provides a kind of device of recognition of face, comprising:
Acquisition module, for acquiring facial image；
Extraction module, for using convolutional neural networks to extract the feature vector of facial image, convolutional neural networks pass through
Above-mentioned apparatus training obtains；
Identification module, for carrying out recognition of face using feature vector.
The embodiment of the present invention, which is effectively avoided, identifies mistake as caused by makeup and external environment influence.
The above is a preferred embodiment of the present invention, it is noted that for those skilled in the art
For, without departing from the principles of the present invention, it can also make several improvements and retouch, these improvements and modifications
It should be regarded as protection scope of the present invention.
Claims (10)
1. a kind of training method of the convolutional neural networks for recognition of face characterized by comprising
Construct sample training library；Wherein: the sample training library includes multiple sample classes, includes multiple faces in each sample class
Image pattern；Wherein, the facial image sample of the same person forms a sample class, and each facial image sample standard deviation is corresponding with one
The class label of a class label, multiple facial image samples in a sample class is identical；
Use sample training library training convolutional neural networks；
The feature vector of all samples in the sample training library is extracted using the convolutional neural networks after training；
Calculate the distance between every two feature vector；
Sample is constructed to training library；Wherein: the sample is combined into training library by the sample of face images sample to collection,
The sample to set include foreign peoples's sample to similar sample pair, foreign peoples's sample is to by facial image sample and and the people
Face image sample distance is less than certain value and the facial image sample for belonging to different sample classes from the facial image sample is constituted, institute
State similar sample to by facial image sample and with facial image sample distance be greater than certain value and with the facial image sample
The facial image sample for belonging to identical sample class is constituted；
Using the sample to training library training convolutional neural networks.
2. the training method of the convolutional neural networks according to claim 1 for recognition of face, which is characterized in that described
It is further using sample training library training convolutional neural networks are as follows:
Using the sample training library, and pass through softmax classifier training convolutional neural networks, the softmax classifier
Classification quantity it is identical as the quantity of the sample class.
3. the training method of the convolutional neural networks according to claim 1 for recognition of face, which is characterized in that described
Distance is Euclidean distance or mahalanobis distance.
4. the training method of the convolutional neural networks according to claim 1 for recognition of face, which is characterized in that described
It is further to training library training convolutional neural networks using the sample are as follows:
Using the sample to training library, and pass through softmax classifier training convolutional neural networks, the softmax classification
Device is two classifiers.
5. a kind of method of recognition of face characterized by comprising
Acquire facial image；
The feature vector of facial image is extracted using convolutional neural networks, the convolutional neural networks are appointed by claim 1-4
The training of method described in one obtains；
Recognition of face is carried out using described eigenvector.
6. a kind of training device of the convolutional neural networks for recognition of face characterized by comprising
First construction unit, for constructing sample training library；Wherein: the sample training library includes multiple sample classes, each sample
It include multiple facial image samples in this class；Wherein, the facial image sample of the same person forms a sample class, each face
Image pattern is corresponding with a class label, and the class label of multiple facial image samples in a sample class is identical；
First training unit, for using sample training library training convolutional neural networks；
Extraction unit, for extracting the feature of all samples in the sample training library using the convolutional neural networks after training
Vector；
Computing unit, for calculating the distance between every two feature vector；
Second construction unit, for constructing sample to training library；Wherein: the sample is to training library by face images sample
Sample collection is combined into, the sample to set include foreign peoples's sample to similar sample pair, foreign peoples's sample is to by people
Face image sample and it is less than certain value from facial image sample distance and belongs to different sample classes with the facial image sample
Facial image sample is constituted, and the similar sample is greater than certain value to by facial image sample and with the facial image sample distance
And belong to the facial image sample composition of identical sample class with the facial image sample；
Second training unit, for using the sample to training library training convolutional neural networks.
7. the training device of the convolutional neural networks according to claim 6 for recognition of face, which is characterized in that described
First training unit is further used for:
Using the sample training library, and pass through softmax classifier training convolutional neural networks, the softmax classifier
Classification quantity it is identical as the quantity of the sample class.
8. the training device of the convolutional neural networks according to claim 6 for recognition of face, which is characterized in that described
Distance is Euclidean distance or mahalanobis distance.
9. the training device of the convolutional neural networks according to claim 6 for recognition of face, which is characterized in that described
Second training unit is further used for:
Using the sample to training library, and pass through softmax classifier training convolutional neural networks, the softmax classification
Device is two classifiers.
10. a kind of device of recognition of face characterized by comprising
Acquisition module, for acquiring facial image；
Extraction module, for using convolutional neural networks to extract the feature vector of facial image, the convolutional neural networks pass through
Any device training of claim 6-9 obtains；
Identification module, for carrying out recognition of face using described eigenvector.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510857317.1A CN105512620B (en) | 2015-11-30 | 2015-11-30 | The training method and device of convolutional neural networks for recognition of face |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510857317.1A CN105512620B (en) | 2015-11-30 | 2015-11-30 | The training method and device of convolutional neural networks for recognition of face |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105512620A CN105512620A (en) | 2016-04-20 |
| CN105512620B true CN105512620B (en) | 2019-07-26 |
Family
ID=55720587
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510857317.1A Active CN105512620B (en) | 2015-11-30 | 2015-11-30 | The training method and device of convolutional neural networks for recognition of face |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105512620B (en) |
Families Citing this family (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106021442B (en) * | 2016-05-16 | 2019-10-01 | 江苏大学 | A kind of Internet news summary extracting method |
| CN106126579B (en) * | 2016-06-17 | 2020-04-28 | 北京市商汤科技开发有限公司 | Object identification method and device, data processing device and terminal equipment |
| CN106022317A (en) * | 2016-06-27 | 2016-10-12 | 北京小米移动软件有限公司 | Face identification method and apparatus |
| CN106372581B (en) * | 2016-08-25 | 2020-09-04 | 中国传媒大学 | Method for constructing and training face recognition feature extraction network |
| CN108960001B (en) * | 2017-05-17 | 2021-12-24 | 富士通株式会社 | Method and device for training image processing device for face recognition |
| CN107578029B (en) * | 2017-09-21 | 2020-03-27 | 北京邮电大学 | Computer-aided picture authentication method and device, electronic equipment and storage medium |
| CN108154092B (en) * | 2017-12-13 | 2022-02-22 | 北京小米移动软件有限公司 | Face feature prediction method and device |
| CN108446617B (en) * | 2018-03-09 | 2022-04-22 | 华南理工大学 | Side face interference resistant rapid human face detection method |
| CN108446674A (en) * | 2018-04-28 | 2018-08-24 | 平安科技（深圳）有限公司 | Electronic device, personal identification method and storage medium based on facial image and voiceprint |
| WO2020034126A1 (en) * | 2018-08-15 | 2020-02-20 | 深圳先进技术研究院 | Sample training method, classification method, identification method, device, medium, and system |
| CN109376679A (en) * | 2018-11-05 | 2019-02-22 | 绍兴文理学院 | A kind of face identification system and method based on deep learning |
| CN109635778B (en) * | 2018-12-25 | 2020-01-03 | 北京心法科技有限公司 | Risk behavior monitoring and early warning method and system suitable for special population |
| CN111652260B (en) * | 2019-04-30 | 2023-06-20 | 上海铼锶信息技术有限公司 | Face clustering sample number selection method and system |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102831389A (en) * | 2012-06-28 | 2012-12-19 | 北京工业大学 | Facial expression recognition algorithm based on discriminative component analysis |
| US8582807B2 (en) * | 2010-03-15 | 2013-11-12 | Nec Laboratories America, Inc. | Systems and methods for determining personal characteristics |
| CN104899579A (en) * | 2015-06-29 | 2015-09-09 | 小米科技有限责任公司 | Face recognition method and face recognition device |
-
2015
- 2015-11-30 CN CN201510857317.1A patent/CN105512620B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8582807B2 (en) * | 2010-03-15 | 2013-11-12 | Nec Laboratories America, Inc. | Systems and methods for determining personal characteristics |
| CN102831389A (en) * | 2012-06-28 | 2012-12-19 | 北京工业大学 | Facial expression recognition algorithm based on discriminative component analysis |
| CN104899579A (en) * | 2015-06-29 | 2015-09-09 | 小米科技有限责任公司 | Face recognition method and face recognition device |
Non-Patent Citations (1)
| Title |
|---|
| 基于深度神经网络的动态人脸识别方法及应用;陈国平 等;《微型电脑应用》;20150920;第31卷(第9期);第39-41页 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105512620A (en) | 2016-04-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105512620B (en) | The training method and device of convolutional neural networks for recognition of face | |
| Haque et al. | Two-handed bangla sign language recognition using principal component analysis (PCA) and KNN algorithm | |
| Angelou et al. | Graph-based multimodal fusion with metric learning for multimodal classification | |
| Alom et al. | Digit recognition in sign language based on convolutional neural network and support vector machine | |
| Mukhopadhyay et al. | Real time facial expression and emotion recognition using eigen faces, LBPH and fisher algorithms | |
| Agbo-Ajala et al. | Face-based age and gender classification using deep learning model | |
| Mahmud et al. | Face detection and recognition system | |
| Padhi et al. | Hand gesture recognition using densenet201-mediapipe hybrid modelling | |
| Zahid et al. | A Multi Stage Approach for Object and Face Detection using CNN | |
| Devi et al. | Content based feature combination method for face image retrieval using neural network and SVM classifier for face recognition | |
| Chihaoui et al. | Face recognition using HMM-LBP | |
| Nahar et al. | Twins and Similar Faces Recognition Using Geometric and Photometric Features with Transfer Learning | |
| Cotter et al. | A hardware accelerated multilevel visual classifier for embedded visual-assist systems | |
| Mohammed et al. | Face Recognition Based on Viola-Jones Face Detection Method and Principle Component Analysis (PCA) | |
| Bingöl et al. | Facial action unit intensity estimation using rotation invariant features and regression analysis | |
| Yang et al. | Face recognition based on Weber symmetrical local graph structure | |
| Siyal et al. | Human Action Recognition using ConvLSTM with GAN and transfer learning | |
| Balazia et al. | Walker-independent features for gait recognition from motion capture data | |
| Arshad | Detection Of Human Gender From Eyes Images Using Dnn Approach | |
| Li et al. | Boosting constrained mutual subspace method for robust image-set based object recognition | |
| Shah et al. | Face recognition using 2DPCA and ANFIS classifier | |
| Al-Ghanim et al. | Face Identification Under Disguise and Makeup Based on Hybrid Deep Learning | |
| Ramachandran et al. | Facial expression classification system with emotional back propagation neural network | |
| Tambi et al. | Person-dependent face recognition using histogram of oriented gradients (HOG) and convolution neural network (CNN) | |
| AlDahoul et al. | Feature fusion: H-ELM based learned features and hand-crafted features for human activity recognition |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB02 | Change of applicant information |
Address after: 100085, 1 floor 8, 1 Street, ten Street, Haidian District, Beijing. Applicant after: Beijing eye Intelligence Technology Co., Ltd. Address before: 100085, 1 floor 8, 1 Street, ten Street, Haidian District, Beijing. Applicant before: Beijing Techshino Technology Co., Ltd. |
|
| CB02 | Change of applicant information | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20211213 Address after: 071800 Beijing Tianjin talent home (Xincheng community), West District, Xiongxian Economic Development Zone, Baoding City, Hebei Province Patentee after: BEIJING EYECOOL TECHNOLOGY Co.,Ltd. Patentee after: Beijing Eye Intelligent Technology Co., Ltd Address before: 100085, 1 floor 8, 1 Street, ten Street, Haidian District, Beijing. Patentee before: Beijing Eyes Intelligent Technology Co.,Ltd. |
|
| TR01 | Transfer of patent right | ||
| PE01 | Entry into force of the registration of the contract for pledge of patent right |
Denomination of invention: Training method and device of convolutional neural network for face recognition Effective date of registration: 20220228 Granted publication date: 20190726 Pledgee: China Construction Bank Corporation Xiongxian sub branch Pledgor: BEIJING EYECOOL TECHNOLOGY Co.,Ltd. Registration number: Y2022990000113 |
|
| PE01 | Entry into force of the registration of the contract for pledge of patent right |