CN104866829B - A cross-age face verification method based on feature learning - Google Patents
A cross-age face verification method based on feature learning Download PDFInfo
- Publication number
- CN104866829B CN104866829B CN201510270145.8A CN201510270145A CN104866829B CN 104866829 B CN104866829 B CN 104866829B CN 201510270145 A CN201510270145 A CN 201510270145A CN 104866829 B CN104866829 B CN 104866829B
- Authority
- CN
- China
- Prior art keywords
- feature
- image
- layer
- face verification
- lbp
- 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
- 238000000034 method Methods 0.000 title claims abstract description 57
- 238000012795 verification Methods 0.000 title claims abstract description 26
- 239000013598 vector Substances 0.000 claims abstract description 28
- 238000000605 extraction Methods 0.000 claims abstract description 11
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 10
- 230000001815 facial effect Effects 0.000 claims description 28
- 238000004364 calculation method Methods 0.000 claims description 9
- 238000003475 lamination Methods 0.000 claims description 3
- 230000004927 fusion Effects 0.000 claims description 2
- 239000010410 layer Substances 0.000 description 47
- 238000002474 experimental method Methods 0.000 description 8
- 238000013135 deep learning Methods 0.000 description 6
- 238000012549 training Methods 0.000 description 4
- 238000001514 detection method Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 238000013480 data collection Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000008921 facial expression Effects 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 230000001537 neural effect Effects 0.000 description 2
- 241001269238 Data Species 0.000 description 1
- 210000004556 brain Anatomy 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 210000004709 eyebrow Anatomy 0.000 description 1
- 230000009647 facial growth Effects 0.000 description 1
- 238000000556 factor analysis Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 210000004209 hair Anatomy 0.000 description 1
- 210000003128 head Anatomy 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 239000011229 interlayer Substances 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 238000007500 overflow downdraw method Methods 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000001105 regulatory effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000000007 visual effect 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
-
- 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
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
本发明公开了一种基于特征学习的跨年龄人脸验证方法，包括如下步骤：(1)获取待对比的两幅人脸图像；(2)利用人脸特征点定位的方法对两幅人脸图像进行对齐操作；(3)分别对每幅图像进行特征提取，方法为：①通过深度卷积神经网络自动提取高层语义特征；②计算图像的LBP直方图特征；③将①和②中获得的特征进行融合，表达为特征向量；(4)采用余弦相似度方法计算步骤(3)获得的两幅图像的特征向量之间的距离，据此判断两幅图像是否来自同一人。本发明首次将深度网络应用到跨年龄人脸验证，同时创造性地将手工设计的LBP直方图特征与深度网络自主学习的特征进行融合，实现高层语义特征与低层特征的互补，具有更好的准确率。
The invention discloses a cross-age face verification method based on feature learning, comprising the following steps: (1) acquiring two face images to be compared; Image alignment operation; (3) Feature extraction for each image separately, the methods are: ① Automatically extract high-level semantic features through deep convolutional neural network; ② Calculate the LBP histogram feature of the image; The features are fused and expressed as feature vectors; (4) the cosine similarity method is used to calculate the distance between the feature vectors of the two images obtained in step (3), and based on this, it is judged whether the two images are from the same person. The invention applies the deep network to cross-age face verification for the first time, and at the same time creatively fuses the hand-designed LBP histogram features with the features learned by the deep network autonomously, realizing the complementation of high-level semantic features and low-level features, and has better accuracy Rate.
Description
Technical field
The present invention relates to a kind of face verification methods, and in particular to a kind of face verification method across the age, especially base
In across the age face verification method of feature learning.
Background technique
Face is widely used in the identification of various occasions as one most significant region of people of identification.Generally
For, the recognition methods of face includes four steps: man face image acquiring and detection, facial image pretreatment, facial image spy
Sign is extracted, face matches and verifying.Usually using some Feature Descriptors manually set, such as LBP, SIFT and Gabor etc.,
It indicates human face data, measures the similarity of a pair of of image using COS distance, to realize judgement verifying.
But with advancing age, the face of people can be inevitably generated variation.In some occasions, an only people
The photo of different age group, such as the photo only before the more than ten years need to carry out in the head portrait of alternative personnel and existing clue
Comparison, to achieve the goal, this requires carry out across age face verification.So-called across age face verification is exactly given one
The facial image of a little different age groups, determines whether these facial images belong to the same person.If face verification method can
The variation that reply face generates with advancing age, in archive management system, the criminal of security authentication systems, public security system
The fields such as identification, the monitoring of bank and customs, will have broad application prospects.
In order to realize across age verification, most of traditional methods were modeled to the age, passed through design face growth
Model carries out the face verification across the age.However, such methods generally require rely on priori, such as individual practical year
Age, and not all data set can provide age information.
Deep learning method simulates the Gradation processing structure of human brain, and the inherence of data rich is portrayed with succinct expression way
Information, it is a kind of model of nonlinearity, and data capability of fitting and learning ability with super strength, ability to express is stronger,
The internal information of data rich can more be portrayed.Depth network can learn from data to feature, this mode unsupervisedly
The feature practised also complies with the mechanism in the human perception world, and often has one by the feature that deep learning method learns
Fixed semantic feature.Chinese invention patent application CN104573679A discloses the people based on deep learning under a kind of monitoring scene
Face identifying system, including video acquisition unit, Face datection unit, matching display unit, storage unit, wherein detection unit
In be provided with face difference module, neural network model that face difference module is established using deep learning module carries out face and sentences
Not, which is 5 layers of neural net layer.This method attempts to use the quick knowledge that deep learning method realizes face
Not, it but is difficult to reach target using 5 layers of neural net layer, therefore, in the disclosure fails to provide identification accurately
The assessment data of rate.
For the step feature extraction of most critical in face verification, it is primarily present two problems at present:
1, the monotonicity of facial image.The a large amount of human face datas being currently known are concentrated, and facial image is often more dull
, and most methods are done on single scale at present, the feature extracted in this way is often not abundant enough, is not enough to characterize
Face.
2, the problem of another merits attention is exactly the acquisition of feature.What traditional face verification used is all hand-designed
Feature, this feature specific aim is relatively high, but typically low-level feature, does not often include semantic information, and extensive
It is indifferent.With the arrival of big data era, data volume is also increasing, and how automatically to obtain feature is worth as one
The project of research.
Summary of the invention
Goal of the invention of the invention is to provide a kind of across age face verification method based on feature learning, to improve across year
The accuracy rate of age face verification.
To achieve the above object of the invention, the technical solution adopted by the present invention is that: a kind of across age people based on feature learning
Face verification method, includes the following steps:
(1) two width facial images to be compared are obtained；
(2) alignment operation is carried out to two width facial images using the method for facial modeling；
(3) feature extraction, the feature extraction are carried out to each image respectively are as follows:
1. automatically extracting high-level semantics features by depth convolutional neural networks；
2. calculating the LBP histogram feature of image；
3. obtaining the feature of image by 1. and 2. the middle feature obtained merges, being expressed as feature vector；Here it uses
Fusion method be that two-part feature vector is connected into a feature vector.
(4) the distance between the feature vector of two images that step (3) obtain is calculated using cosine similarity method, according to
This judges whether two images come from same people.
Preferred technical solution in step (2), carries out alignment operation using Flandmark method.
In above-mentioned technical proposal, the depth convolutional neural networks are successively made of following each layer: input layer, the first convolution
Layer, the first maximum pond layer, the second convolutional layer, the second maximum pond layer, third convolutional layer, third maximum pond layer, Volume Four
Lamination, full articulamentum, output layer.
Wherein, it is 4 × 4 wave devices that first convolutional layer, which is equipped with 20 convolution kernel sizes, the first maximum pond layer, pond
Step-length is 2, and the second convolutional layer is equipped with the filter that 40 convolution kernel sizes are 3 × 3, the second maximum pond layer, and pond step-length is
2, third convolutional layer is equipped with the filter that 60 convolution kernel sizes are 4 × 4, third maximum pond layer, and pond step-length is the 2, the 4th
Convolutional layer is equipped with the filter that 60 convolution kernel sizes are 3 × 3.
In above-mentioned technical proposal, the softmax classifier of a K class is provided in the output layer, K is the number to be classified
Mesh.
The calculation method of LBP histogram feature is to give piece image I, the pyramid representation of image I are as follows:, whereinIt is Gauss
Core, k are the pyramidal numbers of plies；The image pyramid that G is indicated, (x, y) indicate the position of pixel.
It is divided into 8 × 8 block, each piece of statistics LBP histogram, by each piece of LBP histogram for pyramidal each layer
Figure connects into a vector, the LBP pyramid representation of image are as follows:
, wherein L
It (I) is the statement mapped image.
Cosine similarity calculation method is,
The feature vector of image i is,
The feature vector of image j is,
In formula, the number that n is characterized,For n-th of feature of image i,For n-th of feature of image j；
Cosine similarity calculation formula between image i and the feature vector of image j are as follows:
, k is feature in feature vector
Serial number.
Due to the above technical solutions, the present invention has the following advantages over the prior art:
1, the present invention automatically obtains the high-level semantics features of facial image using nine layer depth convolutional neural networks, it
The shared complexity for reducing network model of weight, this is also for the first time by depth network application to across age face verification.
2, the present invention creatively carries out the feature of the LBP histogram feature of hand-designed and depth Web-based Self-regulated Learning
Fusion realizes that high-level semantics features are complementary with low-level feature, the experimental results showed that it is with better accuracy rate.
Detailed description of the invention
Fig. 1 is the frame composition of method in the embodiment of the present invention；
Fig. 2 is depth convolutional neural networks architecture diagram used in embodiment；
Fig. 3 is CACD data set exemplary diagram used in embodiment；
Fig. 4 is the ROC curve figure of result in embodiment.
Specific embodiment
The invention will be further described with reference to the accompanying drawings and embodiments:
Embodiment one:
A kind of across age face verification method based on feature learning, it is shown in Figure 1, include the following steps: that (1) obtains
Two width facial images to be compared；
(2) alignment operation is carried out to two width facial images using the method for facial modeling；
Under free environmental condition, facial image inevitably will receive facial expression, illumination or block
Influence, when the face part of two images without sufficiently be aligned when, this influence can be amplified.
The purpose of facial modeling is to further determine that facial feature points (eyes, eyebrow on the basis of Face datection
Hair, nose, mouth, face's outer profile) position.The basic ideas of location algorithm are: the textural characteristics of face and each feature
Position constraint between point combines.The facial characteristics point location of early stage is concentrated mainly in the positioning of several key points, such as
The center of eyes and mouth.Later, researchers had found that more feature points constraints, which are added, can effectively improve accuracy rate, increase
Stiff stability.In the present embodiment, face alignment is carried out using Flandmark method.It is the facial key point of a detection
Open Source Code library, and real-time alignment operation may be implemented.
(3) feature extraction, the feature extraction are carried out to each image respectively are as follows:
1. automatically extracting high-level semantics features by depth convolutional neural networks (DCNN)；
In the present embodiment, the visual signature extracted in image for level is operated using convolution operation and pondization, from office
Portion's low-level feature is to global high-level characteristic.Wherein include 4 convolutional layers, a maximum pond layer is connect behind each convolutional layer.?
Each convolutional layer, the weight of neuron is shared in layer, and there is no shared for interlayer weight.It is finally two full connections
The output of layer, layer second from the bottom is obtained coarse feature, and the last layer is output layer, and output is every width facial image
The corresponding ID number of the facial image of corresponding maximum probability, that is, the corresponding classification of every piece image.
Specifically, referring to shown in attached drawing 2, in the overall structure of DCNN, process registration process that a size is 55 × 47
Facial image be input to first convolutional layer of convolutional neural networks, it is 4 × 4 filtering that first layer, which has 20 convolution kernel sizes,
Device.20 characteristic patterns are obtained, then these characteristic patterns are input to a maximum pond layer, and pond step-length is 2.Then pond layer
Input of the output as next convolutional layer, this convolutional layer have 40 convolution kernels for 3 × 3 filter.This three layers main
Work when extract image low-level feature such as simple edge feature.When image local has small variation, most
Great Chiization is so that the obtained result of our convolutional layers is more robust.And the 2D face alignment mentioned before, so that whole network pair
There is better adaptability in some subtle adjustment of face.It is 4 × 4 that our third convolutional layer, which is equipped with 60 convolution kernel sizes,
Filter, then carry out the maximum pond of third time, pond step-length is 2, and the 4th layer of last convolutional layer is equipped with 60 convolution kernels
The filter that size is 3 × 3.By the layer-by-layer extraction of DCNN, a series of face characteristic has been obtained.By the process of feature extraction
It is defined as f=Conv(x, q), wherein Conv () indicates that the feature extraction function of convolutional neural networks, x indicate the image of input,
The feature vector that f is represented, what q was indicated is the parameter for needing to learn in DCNN.
The output of the last one full articulamentum will be sent in the softmax classifier of a K class, and what K was indicated will classify
Number, this classifier generate a class label distribution.A given input, it is assumed that wherein k-th of output is, then
The activation primitive of softmax classifier is are as follows:
Trained purpose is the probability for maximizing every width facial image and being assigned to correct face_id, and this problem can
To be converted into the cross entropy loss function for minimizing each training sample.If k is the correct label value of training sample,
So this loss function can be with is defined as:Letter can be lost to this using the method for stochastic gradient descent
Number carries out minimum calculating, and gradient declines using most classic BP algorithm.In a convolutional layer, upper one layer of feature maps
Convolution is carried out by a convolution kernel that can learn, then passes through an activation primitive, so that it may output feature map is obtained, it is each
A output map may be the value of multiple input map.
2. calculating the LBP histogram feature of image；
The calculation method of LBP histogram feature is to give piece image I, the pyramid representation of image I are as follows:, whereinIt is Gauss
Core, s are the pyramidal numbers of plies；
It is divided into 8 × 8 block, each piece of statistics LBP histogram, by each piece of LBP histogram for pyramidal each layer
Figure connects into a vector, the LBP pyramid representation of image are as follows:
3. obtaining the feature of image by 1. and 2. the middle feature obtained merges, being expressed as feature vector；
(4) the distance between the feature vector of two images that step (3) obtain is calculated using cosine similarity method, according to
This judges whether two images come from same people.
Cosine similarity calculation method is,
The feature vector of image i is,
The feature vector of image j is,
In formula, the number that n is characterized,For n-th of feature of image i,For n-th of feature of image j；
Cosine similarity calculation formula between image i and the feature vector of image j are as follows:
, k is feature in feature vector
Serial number.
By being trained on the data set of given label, the parameter of classifier is finely adjusted, this reality is enabled to
The classifier performance for applying example is more preferable.
Illustrate effect of the invention by specifically comparing further below.
Experiment carries out on CACD (Cross-Age Celebrity Dataset) age data collection and LFW data set.
CACD age data collection comprises more than 160000 facial images from 2000 different people, everyone has multiple not the same years
The image of age grade section, age range are 16 to 62 years old.Since in LFW data set, most people all only has a picture, therefore
It is less feasible for being trained with this data set, therefore, carries out model using this data set of CelebFaces+
Training, this data set includes 202599 pictures from 10177 different people, and LFW and CelebFaces data set
In people it is essentially identical, therefore can train and model and then be tested on LFW data set on this.Experimental Hardware
Environment: Window 7, Core i7 processor, dominant frequency 3.4G inside save as 8G.Code running environment is: Matlab 2013a.
In an experiment, experiment is measured with True Positive Rate (TPR)-False Positive Rate (TPR)
Effect.TPR and FPR are defined as follows:
1, the experiment on CACD data set
Data set example is referring to shown in attached drawing 3.
For different method for measuring similarity, the results are shown in Table 1.
The comparison of the different method for measuring similarity of table 1
| Measure | Accuracy rate (%) |
| Distance | 86.3 |
| Euclidean distance | 81.4 |
| Pasteur's distance | 82.1 |
| Cosine similarity | 89.5 |
Algorithm of the invention is assessed with ROC curve, and is compared with other methods.Compare experiment
Method has: gradient direction pyramid (GOP),Norm, Bayesian+PointFive Face (PFF).In this experiment, it surveys
Examination collection contains 2000 pairs of positive samples and 2000 pairs of negative samples.The ROC curve of experimental result is as shown in Figure 4.
From fig. 4, it can be seen that using method (DCNN+LBPH) of the invention compared to other methods for, as a result on have
A degree of raising.
In the CACD data set subset set, also the accuracy rate of verifying is assessed, and the ratio of other methods
Relatively as shown in table 2.
Table 2:
| Method | The accuracy rate (%) of verifying |
| High-Dimensional LBP | 81.6 |
| Hidden Factor Analysis | 84.4 |
| Cross-Age Reference Coding | 87.6 |
| DCNN+LBPH(the present embodiment) | 89.5 |
| Manually, average value | 85.7 |
| Manually, ballot method | 94.2 |
2, the experiment on LFW data set
LFW data set is the standard data set of a field of face identification.In this huge data set, it is added
Limitation, for example after the variation at age, intense light irradiation or expression, any slight progress for surmounting current algorithm is all not allow very much
Easy.
After running through training set, the work that the present embodiment is verified in this subset of LFW view2, this subset is
List all same facial image pair and non-same facial image pair.It takes same with the experiment on CACD data set
Method, the first step are to obtain an accurate face expression, and it is same to verify whether to belong to then to calculate their similarity
Individual.Method of the invention introduces deep learning frame, and compared with other algorithms, experimental result is as shown in table 3.As it can be seen that this
Other methods before the method for invention is an advantage in accuracy rate.
Table 3
| Method | Accuracy rate (%) |
| PLDA | 90.07 |
| Joint Bayesian | 90.90 |
| Linear rectified units | 80.73 |
| GSML | 84.18 |
| OSS, TSS, full | 86.83 |
| DCNN+LBPH(the present embodiment) | 91.40 |
Claims (7)
1. a kind of across age face verification method based on feature learning, which comprises the steps of:
(1) two width facial images to be compared are obtained；
(2) alignment operation is carried out to two width facial images using the method for facial modeling；
(3) feature extraction, the feature extraction are carried out to each image respectively are as follows:
1. automatically extracting high-level semantics features by depth convolutional neural networks；
2. calculating the LBP histogram feature of image；
The calculation method of LBP histogram feature is to give piece image I, the pyramid representation of image I are as follows:, whereinIt is Gaussian kernel,
K is the pyramidal number of plies；The image pyramid that G is indicated, (x, y) indicate the position of pixel；
It is divided into 8 × 8 block for pyramidal each layer, each piece of statistics LBP histogram connects each piece of LBP histogram
It is connected into a vector, the LBP pyramid representation of image are as follows:
L( I ) = [LBP(G 0( I )); LBP(G 1( I )); …; LBP(G S ( I))], wherein L(I) it is to image
The statement mapped；
3. obtaining the feature of image by 1. and 2. the middle feature obtained merges, being expressed as feature vector；
(4) the distance between the feature vector of two images that step (3) obtain is calculated using cosine similarity method, sentenced accordingly
Whether disconnected two images come from same people.
2. across the age face verification method according to claim 1 based on feature learning, it is characterised in that: step (2)
In, alignment operation is carried out using Flandmark method.
3. across the age face verification method according to claim 1 based on feature learning, it is characterised in that: the depth
Convolutional neural networks are successively made of following each layer: input layer, the first convolutional layer, the first maximum pond layer, the second convolutional layer, the
Two maximum pond layers, third convolutional layer, third maximum pond layer, Volume Four lamination, full articulamentum, output layer.
4. across the age face verification method according to claim 3 based on feature learning, it is characterised in that: described first
Convolutional layer is equipped with the filter that 20 convolution kernel sizes are 4 × 4, and the first maximum pond layer, the step-length in pond is 2, the second convolution
Layer is equipped with the filter that 40 convolution kernel sizes are 3 × 3, and the second maximum pond layer, pond step-length is 2, and third convolutional layer is equipped with
The filter that 60 convolution kernel sizes are 4 × 4, third maximum pond layer, pond step-length are 2, and Volume Four lamination is equipped with 60 volumes
The filter that product core size is 3 × 3.
5. across the age face verification method according to claim 3 based on feature learning, it is characterised in that: the output
The softmax classifier of a K class is provided in layer, K is the number to be classified.
6. across the age face verification method according to claim 1 based on feature learning, it is characterised in that: step (3)
3. in, the method for fusion is that two-part feature vector is connected into a feature vector.
7. across the age face verification method according to claim 1 based on feature learning, it is characterised in that: cosine is similar
Spending calculation method is,
The feature vector of image i isD i = (W i1, W i2,…,W ik , …, W in ),
The feature vector of image j isD j = (W j1, W j2,…,W jk , …, W jn ),
In formula,nThe number being characterized,W ik It is the of image ikA feature,W jk It is the of image jkA feature；
Cosine similarity calculation formula between image i and the feature vector of image j are as follows:
, k is the serial number of feature in feature vector.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510270145.8A CN104866829B (en) | 2015-05-25 | 2015-05-25 | A cross-age face verification method based on feature learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510270145.8A CN104866829B (en) | 2015-05-25 | 2015-05-25 | A cross-age face verification method based on feature learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104866829A CN104866829A (en) | 2015-08-26 |
| CN104866829B true CN104866829B (en) | 2019-02-19 |
Family
ID=53912650
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510270145.8A Active CN104866829B (en) | 2015-05-25 | 2015-05-25 | A cross-age face verification method based on feature learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104866829B (en) |
Families Citing this family (49)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CA2979560C (en) | 2015-03-05 | 2023-11-07 | Magic Leap, Inc. | Systems and methods for augmented reality |
| US10838207B2 (en) | 2015-03-05 | 2020-11-17 | Magic Leap, Inc. | Systems and methods for augmented reality |
| EP3384468A4 (en) * | 2015-12-04 | 2019-01-30 | Magic Leap, Inc. | Relocalization systems and methods |
| CN105550658A (en) * | 2015-12-24 | 2016-05-04 | 蔡叶荷 | Face comparison method based on high-dimensional LBP (Local Binary Patterns) and convolutional neural network feature fusion |
| CN105844227B (en) * | 2016-03-21 | 2019-06-04 | 湖南君士德赛科技发展有限公司 | A driver identity authentication method for school bus safety |
| CN105956572A (en) * | 2016-05-15 | 2016-09-21 | 北京工业大学 | In vivo face detection method based on convolutional neural network |
| CN106022380A (en) * | 2016-05-25 | 2016-10-12 | 中国科学院自动化研究所 | Individual identity identification method based on deep learning |
| CN106022287A (en) * | 2016-05-27 | 2016-10-12 | 广东顺德中山大学卡内基梅隆大学国际联合研究院 | Over-age face verification method based on deep learning and dictionary representation |
| CN107480575B (en) * | 2016-06-07 | 2020-09-22 | 深圳市商汤科技有限公司 | Model training method, trans-age face recognition method and corresponding device |
| CN106096535B (en) * | 2016-06-07 | 2020-10-23 | 广东顺德中山大学卡内基梅隆大学国际联合研究院 | A face verification method based on bilinear joint CNN |
| CN106227718A (en) * | 2016-07-18 | 2016-12-14 | 中国民航大学 | Land based on CNN sky call semantic consistency method of calibration |
| CN106203533B (en) * | 2016-07-26 | 2019-09-20 | 厦门大学 | Deep learning face verification method based on hybrid training |
| US10649211B2 (en) | 2016-08-02 | 2020-05-12 | Magic Leap, Inc. | Fixed-distance virtual and augmented reality systems and methods |
| CN106326868A (en) * | 2016-08-26 | 2017-01-11 | 江苏华通晟云科技有限公司 | Face identification method based on cosine similarity measure learning |
| CN106650573B (en) * | 2016-09-13 | 2019-07-16 | 华南理工大学 | A cross-age face verification method and system |
| CN107871099A (en) * | 2016-09-23 | 2018-04-03 | 北京眼神科技有限公司 | Face detection method and apparatus |
| CN106485235B (en) * | 2016-10-24 | 2019-05-03 | 厦门美图之家科技有限公司 | A kind of convolutional neural networks generation method, age recognition methods and relevant apparatus |
| CN106557743B (en) * | 2016-10-26 | 2020-04-24 | 桂林电子科技大学 | Facial feature extraction system and method based on FECNN |
| CN107053168A (en) * | 2016-12-09 | 2017-08-18 | 南京理工大学 | A kind of target identification method and hot line robot based on deep learning network |
| CN106650650B (en) * | 2016-12-14 | 2020-04-24 | 广东顺德中山大学卡内基梅隆大学国际联合研究院 | Cross-age face recognition method |
| CN106599926A (en) * | 2016-12-20 | 2017-04-26 | 上海寒武纪信息科技有限公司 | Expression picture pushing method and system |
| CN106709448A (en) * | 2016-12-21 | 2017-05-24 | 华南理工大学 | Face convolutional neural network-based traditional Chinese medicine constitution recognition method |
| CN108229269A (en) * | 2016-12-31 | 2018-06-29 | 深圳市商汤科技有限公司 | Method for detecting human face, device and electronic equipment |
| US10812936B2 (en) | 2017-01-23 | 2020-10-20 | Magic Leap, Inc. | Localization determination for mixed reality systems |
| JP7055815B2 (en) | 2017-03-17 | 2022-04-18 | マジック リープ， インコーポレイテッド | A mixed reality system that involves warping virtual content and how to use it to generate virtual content |
| AU2018234921B2 (en) | 2017-03-17 | 2021-10-07 | Magic Leap, Inc. | Mixed reality system with color virtual content warping and method of generating virtual content using same |
| WO2018170409A1 (en) | 2017-03-17 | 2018-09-20 | Magic Leap, Inc. | Mixed reality system with multi-source virtual content compositing and method of generating virtual content using same |
| CN106980834A (en) * | 2017-03-24 | 2017-07-25 | 中山大学 | A kind of automatic classification method of facial image |
| CN107341463B (en) * | 2017-06-28 | 2020-06-05 | 苏州飞搜科技有限公司 | Face feature recognition method combining image quality analysis and metric learning |
| CN107622225B (en) * | 2017-07-27 | 2020-04-03 | 成都信息工程大学 | Face recognition method based on independent component analysis network |
| CN107507286B (en) * | 2017-08-02 | 2020-09-29 | 五邑大学 | Bimodal biological characteristic sign-in system based on face and handwritten signature |
| CN107590449A (en) * | 2017-08-31 | 2018-01-16 | 电子科技大学 | A kind of gesture detecting method based on weighted feature spectrum fusion |
| CN107590478A (en) * | 2017-09-26 | 2018-01-16 | 四川长虹电器股份有限公司 | A kind of age estimation method based on deep learning |
| CN107818314B (en) * | 2017-11-22 | 2019-03-19 | 北京达佳互联信息技术有限公司 | Face image processing method, device and server |
| CN108647668A (en) * | 2018-05-21 | 2018-10-12 | 北京亮亮视野科技有限公司 | The construction method of multiple dimensioned lightweight Face datection model and the method for detecting human face based on the model |
| CN108830211A (en) * | 2018-06-11 | 2018-11-16 | 厦门中控智慧信息技术有限公司 | Face identification method and Related product based on deep learning |
| CN108960412B (en) * | 2018-06-29 | 2022-09-30 | 北京京东尚科信息技术有限公司 | Image recognition method, device and computer readable storage medium |
| CN112513712B (en) | 2018-07-23 | 2023-05-09 | 奇跃公司 | Mixed reality system with virtual content warping and method of generating virtual content using the system |
| CN109255289B (en) * | 2018-07-27 | 2021-10-26 | 电子科技大学 | Cross-aging face recognition method based on unified generation model |
| CN109344774A (en) * | 2018-10-08 | 2019-02-15 | 国网经济技术研究院有限公司 | Thermal power station target identification method in remote sensing image |
| CN109815864B (en) * | 2019-01-11 | 2021-01-01 | 浙江工业大学 | Facial image age identification method based on transfer learning |
| CN109766998A (en) * | 2019-01-14 | 2019-05-17 | 温岭市非普电气有限公司 | A kind of feature pyramid network structure accelerated based on MobileNet |
| CN109886121B (en) * | 2019-01-23 | 2021-04-06 | 浙江大学 | Human face key point positioning method for shielding robustness |
| CN110322509B (en) * | 2019-06-26 | 2021-11-12 | 重庆邮电大学 | Target positioning method, system and computer equipment based on hierarchical class activation graph |
| CN111881721B (en) * | 2020-06-10 | 2021-03-30 | 广东芯盾微电子科技有限公司 | Cross-age face recognition method, system, device and storage medium |
| CN111710424A (en) * | 2020-06-19 | 2020-09-25 | 浙江新芮信息科技有限公司 | Catering personnel health monitoring method, device and computer-readable storage medium |
| CN114155567A (en) * | 2020-09-08 | 2022-03-08 | 京东方科技集团股份有限公司 | Target detection method and device, storage medium and electronic device |
| CN112287965A (en) * | 2020-09-21 | 2021-01-29 | 卓尔智联(武汉)研究院有限公司 | Image quality detection model training method and device and computer equipment |
| CN112766157B (en) * | 2021-01-20 | 2022-08-30 | 乐山师范学院 | Cross-age face image recognition method based on disentanglement representation learning |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110103695A1 (en) * | 2009-11-04 | 2011-05-05 | Canon Kabushiki Kaisha | Image processing apparatus and image processing method |
| CN102663413A (en) * | 2012-03-09 | 2012-09-12 | 中盾信安科技(江苏)有限公司 | Multi-gesture and cross-age oriented face image authentication method |
| CN103902964A (en) * | 2012-12-31 | 2014-07-02 | 深圳先进技术研究院 | Face recognition method |
-
2015
- 2015-05-25 CN CN201510270145.8A patent/CN104866829B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110103695A1 (en) * | 2009-11-04 | 2011-05-05 | Canon Kabushiki Kaisha | Image processing apparatus and image processing method |
| CN102663413A (en) * | 2012-03-09 | 2012-09-12 | 中盾信安科技(江苏)有限公司 | Multi-gesture and cross-age oriented face image authentication method |
| CN103902964A (en) * | 2012-12-31 | 2014-07-02 | 深圳先进技术研究院 | Face recognition method |
Non-Patent Citations (6)
| Title |
|---|
| Cross-Age Reference Coding for Age-Invariant Face Recognition and Retrieval;Bor-Chun Chen etal;《European Conference on Computer Vision》;20141231;第768-783页 * |
| DeepFace: Closing the Gap to Human-Level Performance in Face Verification;Yaniv Taigman etal;《2014 IEEE Conference on Computer Vision and Pattern Recognition》;20141231;第1701-1708页 * |
| Face Recognition: A Literature Survey;W. ZHAO etal;《ACM Computing Surveys》;20031231;第35卷(第4期);第399-458页 * |
| 基于LBP的人脸识别研究;黄非非;《中国优秀硕士学位论文全文数据库 信息科技辑》;20091215;第I138-644页 * |
| 基于二元模式的人脸识别与表情识别研究;付晓峰;《中国博士学位论文全文数据库 信息科技辑》;20090615;第I138-57页 * |
| 基于局部二值模式和深度学习的人脸识别;张雯 等;《计算机应用》;20150510;第35卷(第5期);第1474-1478页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104866829A (en) | 2015-08-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104866829B (en) | A cross-age face verification method based on feature learning | |
| CN107145842B (en) | Face recognition method combining LBP characteristic graph and convolutional neural network | |
| CN103679158B (en) | Face authentication method and device | |
| CN102938065B (en) | Face feature extraction method and face identification method based on large-scale image data | |
| CN107967458A (en) | A kind of face identification method | |
| CN100440246C (en) | Positioning method for human face characteristic point | |
| CN109117744A (en) | A kind of twin neural network training method for face verification | |
| CN105373777B (en) | A kind of method and device for recognition of face | |
| CN107133612A (en) | Based on image procossing and the intelligent ward of speech recognition technology and its operation method | |
| CN102629320B (en) | Ordinal measurement statistical description face recognition method based on feature level | |
| CN106650693A (en) | Multi-feature fusion identification algorithm used for human face comparison | |
| CN104143079A (en) | Method and system for face attribute recognition | |
| CN103605972A (en) | Non-restricted environment face verification method based on block depth neural network | |
| CN107220598B (en) | Iris image classification method based on deep learning features and Fisher Vector coding model | |
| CN103020602B (en) | Based on the face identification method of neural network | |
| CN103714326B (en) | One-sample face identification method | |
| CN108681725A (en) | A kind of weighting sparse representation face identification method | |
| CN105844221A (en) | Human face expression identification method based on Vadaboost screening characteristic block | |
| CN105138968A (en) | Face authentication method and device | |
| CN107977439A (en) | A kind of facial image base construction method | |
| CN108960201A (en) | A kind of expression recognition method extracted based on face key point and sparse expression is classified | |
| CN106250825A (en) | A kind of at the medical insurance adaptive face identification system of applications fields scape | |
| CN106096517A (en) | A kind of face identification method based on low-rank matrix Yu eigenface | |
| CN109977887A (en) | A kind of face identification method of anti-age interference | |
| CN107992783A (en) | Face image processing process and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| EXSB | Decision made by sipo to initiate substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |