CN105512624A - Smile face recognition method and device for human face image - Google Patents
Smile face recognition method and device for human face image Download PDFInfo
- Publication number
- CN105512624A CN105512624A CN201510868158.5A CN201510868158A CN105512624A CN 105512624 A CN105512624 A CN 105512624A CN 201510868158 A CN201510868158 A CN 201510868158A CN 105512624 A CN105512624 A CN 105512624A
- Authority
- CN
- China
- Prior art keywords
- face
- neural networks
- convolutional neural
- label information
- expressive features
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 48
- 238000012549 training Methods 0.000 claims abstract description 35
- 238000007781 pre-processing Methods 0.000 claims abstract description 22
- 238000000605 extraction Methods 0.000 claims abstract description 14
- 230000001815 facial effect Effects 0.000 claims description 107
- 238000013527 convolutional neural network Methods 0.000 claims description 101
- 230000014509 gene expression Effects 0.000 claims description 53
- 230000008921 facial expression Effects 0.000 claims description 28
- 230000008569 process Effects 0.000 claims description 19
- 230000001105 regulatory effect Effects 0.000 claims description 9
- 238000006243 chemical reaction Methods 0.000 claims description 6
- 238000004148 unit process Methods 0.000 claims description 2
- 238000013528 artificial neural network Methods 0.000 abstract 1
- 230000000694 effects Effects 0.000 description 16
- 238000010586 diagram Methods 0.000 description 14
- 238000005516 engineering process Methods 0.000 description 11
- 230000004913 activation Effects 0.000 description 9
- 238000011478 gradient descent method Methods 0.000 description 6
- 210000004709 eyebrow Anatomy 0.000 description 4
- 238000004513 sizing Methods 0.000 description 4
- 210000000887 face Anatomy 0.000 description 3
- 238000012544 monitoring process Methods 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 230000003292 diminished effect Effects 0.000 description 2
- 239000004744 fabric Substances 0.000 description 2
- 238000007726 management method Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000011017 operating method Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 238000005267 amalgamation Methods 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 239000004615 ingredient Substances 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 238000011160 research Methods 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
- G06V40/174—Facial expression recognition
- G06V40/175—Static expression
-
- 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
-
- 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)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Human Computer Interaction (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Biomedical Technology (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Evolutionary Computation (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a smile face recognition device for a human face image, and the device comprises an image recognition and extraction unit which is used for detecting the positions of human faces in a plurality of to-be-recognized human face images, extracting the human face images from the plurality of to-be-recognized human face images, and transmitting the human face images to an image preprocessing unit; the image preprocessing unit which is connected with the image recognition and extraction unit; a network building unit which is used for building a convolution neural network; a network training unit which is connected with the network building unit and the image preprocessing unit; and a recognition and judgment unit which is connected with the network training unit and the image preprocessing unit, and is used for the recognition of human faces. In addition, the invention also discloses a smile face recognition method for the human face image. According to the invention, the device and method can quickly and effectively carry out the accurate recognition and judgment of smile faces in a large number of human face images while guaranteeing the high-quality smile face recognition of the human face images, meet the requirements of a user for a function of smile face recognition, improves the work efficiency of the user, and saves the precious time for the user.
Description
Technical field
The present invention relates to the technical field such as pattern-recognition and computer vision, particularly relate to a kind of smiling face's recognition methods and device thereof of facial image.
Background technology
At present, along with the development of human sciences's technology, face recognition technology is more and more universal in people's daily life, no matter in artificial intelligence study or public safety applications, face recognition technology is a forward position, hot technology always, has very important status.
Wherein, smiling face in face recognition technology identifies, it is a very important content in technical field of computer vision, along with the growth of the application demands such as such as face payment, sentiment analysis, medical monitoring, smiling face identifies the important component part as finishing man-machine interaction, obtain the concern of more and more people, this just impels global scientific research personnel to go into overdrive smiling face's recognition technology.
At present, some traditional methods first extract the multiple low-level features of face, then multiple low-level features merged by complicated amalgamation mode, finally sends in sorter to carry out smiling face and to classify judgement.But, the low-level features of these hand-designed is expressed, cannot well give expression to the expression information contained in face, and recognition speed is low, need to spend the more time, and recognition accuracy is poor, is therefore unfavorable for that realizing smiling face identifies, cannot meet the demand of user to smiling face's recognition function, the serious use sense reducing user is subject to.
Therefore, at present in the urgent need to developing a kind of technology, it can while guarantee carries out high-quality smiling face identification to face picture, carry out identification to smiling face in a large amount of face picture quickly and efficiently to judge, meet the requirement of user to smiling face's recognition function, improve the work efficiency of user, save the time of people's preciousness.
Summary of the invention
In view of this, the object of this invention is to provide a kind of smiling face's recognition methods and device thereof of facial image, it can while guarantee carries out high-quality smiling face identification to face picture, quickly and efficiently accurately judgement is identified to smiling face in a large amount of face picture, meet the requirement of user to smiling face's recognition function, improve the work efficiency of user, save the time of people's preciousness, the product use sense being conducive to improving user is subject to, and is of great practical significance.
For this reason, the invention provides a kind of smiling face's recognition methods of facial image, comprise step:
The first step: multiple character images identified for needing smiling face, detects the position of face, and identifies the facial image extracted wherein;
Second step: the facial image extracted facial image being scaled to pre-set dimension size, and conversion process becomes gray-scale map, and give the expression label information of pre-set categories for often opening described facial image;
3rd step: set up convolutional neural networks, described convolutional neural networks comprises successively to input layer, default multiple convolutional layer, default multiple full articulamentum and output layer that inputted facial image processes;
4th step: described convolutional neural networks is trained, expand the expressive features otherness between multiple facial images with different classes of expression label information, reduce the expressive features otherness between multiple facial images with identical category expression label information simultaneously;
5th step: by processed one-tenth gray-scale map and convergent-divergent pre-set dimension, treat that the facial image of often opening that smiling face identifies all has been input in the convolutional neural networks of training, by described convolutional neural networks extract this facial image expressive features value and be sent to sorter carry out smiling face judge classification, realize smiling face's identifying operation.
Wherein, in second step, the human face expression label information of described pre-set categories comprises smile label information and non-smile label information.
Wherein, described convolutional neural networks comprises an input layer, four convolutional layers, a full articulamentum and output layers.
Wherein, in described 4th step, the step that described convolutional neural networks is trained is specially:
The human face expression label information of any two facial images and correspondence thereof is input to the input layer of described convolutional neural networks, extracted the expressive features value of these two facial images by the convolutional layer of described convolutional neural networks and full articulamentum, then export from output layer;
The expressive features value of these two face picture is sent into sorter classify, according to the human face expression label information that these two face picture have, calculate the first-loss value of the expressive features value obtaining these two face picture;
Relatively whether the expressive features value of these two face picture, have the human face expression label information of identical category according to these two face picture, calculates the second penalty values of the expressive features value obtaining these two face picture;
The all weights using the first-loss value of the expressive features value of these two face picture and the second penalty values to come together oppositely to regulate in described convolutional neural networks, complete the training to described convolutional neural networks.
In addition, present invention also offers a kind of smiling face's recognition device of facial image, comprising:
Image recognition extraction unit, for for multiple character images needing smiling face to identify, detects the position of face, and identifies the facial image extracted wherein, then send to image pre-processing unit;
Image pre-processing unit, be connected with image recognition extraction unit, for extracted facial image being scaled to the facial image of pre-set dimension size, and conversion process becomes gray-scale map, give the expression label information of pre-set categories simultaneously for often opening described facial image, and export to network training unit and identify judging unit;
Network sets up unit, and for setting up convolutional neural networks, described convolutional neural networks comprises successively to input layer, default multiple convolutional layer, default multiple full articulamentum and output layer that inputted facial image processes;
Network training unit, set up unit with network respectively, image pre-processing unit is connected, for training described convolutional neural networks, expand the expressive features otherness between multiple facial images with different classes of expression label information, reduce the expressive features otherness between multiple facial images with identical category expression label information simultaneously;
Identify judging unit, be connected with network training unit, image pre-processing unit respectively, for the facial image of often opening through image pre-processing unit process has all been input in the convolutional neural networks of training, by described convolutional neural networks extract this facial image expressive features value and be sent to sorter carry out smiling face judge classification, realize smiling face's identifying operation.
Wherein, the human face expression label information of described pre-set categories comprises smile label information and non-smile label information.
Wherein, described convolutional neural networks comprises an input layer, four convolutional layers, a full articulamentum and output layers.
Wherein, described network training unit comprises characteristic extracting module, first-loss value acquisition module, the second penalty values acquisition module and reverse adjustment module, wherein:
Characteristic extracting module, for the human face expression label information of any two facial images and correspondence thereof being input to the input layer of described convolutional neural networks, extracted the expressive features value of these two facial images by the convolutional layer of described convolutional neural networks and full articulamentum, then export from output layer;
First-loss value acquisition module, be connected with characteristic extracting module, classify for the expressive features value of these two face picture is sent into sorter, according to the human face expression label information that these two face picture have, calculate the first-loss value of the expressive features value obtaining these two face picture;
Second penalty values acquisition module, be connected with characteristic extracting module, for comparing the expressive features value of these two face picture, whether there is according to these two face picture the human face expression label information of identical category, calculating the second penalty values of the expressive features value obtaining these two face picture;
Reverse adjustment module, be connected with the second penalty values acquisition module with first-loss value acquisition module respectively, first-loss value and the second penalty values for using the expressive features value of these two face picture are come together all weights oppositely regulated in described convolutional neural networks, namely complete the training to described convolutional neural networks.
From above technical scheme provided by the invention, compared with prior art, the invention provides a kind of smiling face's recognition methods and device thereof of character image, its utilize constructed by convolutional neural networks, the expressive features extracting face completes smiling face's identification, can while guarantee carries out high-quality smiling face identification to face picture, fast, effectively accurately judgement is identified to smiling face in a large amount of face picture, meet the requirement of user to smiling face's recognition function, improve the work efficiency of user, save the time of people's preciousness, the product use sense being conducive to improving user is subject to, be of great practical significance.
Accompanying drawing explanation
Fig. 1 is the process flow diagram of smiling face's recognition methods of a kind of facial image provided by the invention;
Fig. 2 is in smiling face's recognition methods of a kind of facial image provided by the invention, the schematic diagram of the face picture of expression of smiling;
Fig. 3 is in smiling face's recognition methods of a kind of facial image provided by the invention, the schematic diagram of the face picture of the normal expression of input (non-expression of smiling);
Fig. 4 is in smiling face's recognition methods of a kind of facial image provided by the invention, a kind of example structure schematic diagram of each ingredient in constructed convolutional neural networks;
Fig. 5 is the smiling face's recognition methods utilizing a kind of facial image provided by the invention, is judged as the schematic diagram of the facial image of smile expressive features;
Fig. 6 is the smiling face's recognition methods utilizing a kind of facial image provided by the invention, is judged as the schematic diagram of the facial image of normal expression (non-smile expression) feature;
Fig. 7 is the block diagram of smiling face's recognition device of a kind of facial image provided by the invention.
Embodiment
In order to make those skilled in the art person understand the present invention program better, below in conjunction with drawings and embodiments, the present invention is described in further detail.
Fig. 1 is the process flow diagram of smiling face's recognition methods of a kind of facial image provided by the invention;
See Fig. 1, smiling face's recognition methods of a kind of facial image provided by the invention, comprises the following steps:
Step S101: multiple character images identified for needing smiling face, detects the position of face, and identifies the facial image extracted wherein;
It should be noted that, current existing face recognition technology mainly judges to identify face according to eyes and the relative position of mouth and the approximate shape of face.Current is that core defines multiple face identification system with face recognition module, is provided with and applies widely: recognition of face access management system, recognition of face access control and attendance system and face recognition video monitoring system etc.
In the present invention, in specific implementation, can using two, ear, nose, the face such as eyebrow and mouth are as key point, namely eyespot, auriculare, nose point, eyebrow point and stomion are set, detect the position determining face, and determine shape and the profile of face in character image, the then corresponding facial image extracted in character image.
Step S102: the facial image extracted facial image being scaled to pre-set dimension size, and conversion process becomes gray-scale map;
Step S103: the expression label information giving pre-set categories for often opening described facial image;
In the present invention, described pre-set dimension size can according to user need arrange in advance, can be such as the arbitrary dimension between 48 × 48 pixels to 256 × 256 pixels, be preferably the size of 90 × 90 pixels.
In the present invention, the human face expression label information of described pre-set categories comprises smile label information and normal (non-smile) label information, smile label information and normal (non-smile) label information is given respectively see the face picture in Fig. 2 and Fig. 3, Fig. 2 and Fig. 3.
Step S104: set up convolutional neural networks (ConvolutionalNeuralNetwork, CNN), described convolutional neural networks comprises successively to input layer, default multiple convolutional layer, default multiple full articulamentum and output layer that inputted facial image processes, see Fig. 4;
Step S105: described convolutional neural networks is trained, expand the expressive features otherness between multiple facial images with different classes of expression label information, reduce the expressive features otherness between multiple facial images with identical category expression label information simultaneously;
Step S106: by processed one-tenth gray-scale map and convergent-divergent pre-set dimension, treat that the facial image of often opening that smiling face identifies all has been input in the convolutional neural networks of training, by described convolutional neural networks extract this facial image expressive features value and be sent to sorter carry out smiling face judge classification, whether the final face judged in facial image is smiling face, realizes smiling face's identifying operation.
It should be noted that, for the present invention, the entitlement weight average random initializtion of described convolutional neural networks.Wherein, see Fig. 4, the pre-sizing of number of described convolutional layer, between 3 to 7, is preferably 4; The pre-sizing of number of described full articulamentum is between 1 to 3, is preferably 1.
For the present invention, in described convolutional neural networks, the activation function of described convolutional layer preferably uses ReLU function, and the number of the sub-size of the step-length of each convolutional layer, convolution, convolution all can freely be arranged, as shown in Figure 4, the text description of Fig. 4 is specifically shown in embodiment below to network structure.The facial image of described gray scale is as input picture, after the input of each convolutional layer and the multiplied by weight of this layer, a numerical value can be obtained, the principle of ReLU function is exactly, if this numerical value is greater than 0, so output valve just preserves this calculated value, if this calculated value is less than 0, so output valve just preserves into 0.Certainly, ReLU function also can change other activation function into.
For the present invention, in described convolutional neural networks, full articulamentum preferably uses sigmoid activation function, certainly, also can use other activation functions.Described full articulamentum is for extracting the expressive features value of face picture.
For the present invention, it should be noted that, the layer that described convolutional neural networks contains convolution operation by some is interconnected the network structure formed, and his Main Function is the feature in order to extract picture, the expressive features value of such as facial image.
For described convolutional neural networks, the effect of the input layer wherein had is in order to image data (as facial image) is sent into convolutional neural networks (network structure), so that subsequent treatment; The effect of convolutional layer is the local features extracting picture; The effect of full articulamentum is to extract the feature having more distinction from the output of last layer; The effect of output layer is to judge whether face is smile; By the input of last time and and lower one deck between weight, obtain corresponding output valve.
For the present invention, it should be noted that, the step S105 that described convolutional neural networks is trained be specially:
Step S1051: the input layer human face expression label information of any two facial images and correspondence thereof being input to described convolutional neural networks, extracted the expressive features value of these two facial images by the convolutional layer of described convolutional neural networks and full articulamentum, then export from output layer;
Step S1052: the expressive features value of these two face picture is sent into sorter and classifies, according to the human face expression label information that these two face picture have, calculates the first-loss value of the expressive features value obtaining these two face picture;
It should be noted that, for calculating the first-loss function of the first-loss value of the expressive features value obtaining these two face picture being:
Wherein, x is the expressive features value extracting facial image in step S1051, and p (x) is real classification situation, and q (x) is the probability of prediction.First-loss value the second penalty values in integrating step 1053 can oppositely regulate whole weights in convolutional neural networks together.
For the present invention, it should be noted that, the effect of sorter is the feature according to convolutional neural networks extraction above, classifies to the expression classification of face.In specific implementation, the present invention can adopt softmax sorter.
For softmax sorter, it can calculate the probability distribution of different expression, judges which kind of expression face inputs according to different probability distribution.Concrete operating process is the output of front one deck be one is eigenwert, and then the present invention is normalized by these eigenwerts are multiplied by different weights, can obtain the probability distribution of different expression.
For the present invention, it should be noted that, the effect of first penalty values of the expressive features value of described two face picture calculates according to identifying information, be exactly specifically, if the probability results judged last and legitimate reading are relatively, then penalty values is less, if last result differs comparatively far away with real result, then penalty values is larger; Two kinds of faces are sent in convolutional neural networks of the present invention respectively, two first-loss values that different face is corresponding can be obtained, after being combined with the second penalty values that subsequent step obtains, be used for together adjusting network weight.
For the present invention, about first-loss function formula, it should be noted that, p (x) is known real table mutual affection cloth, and q (x) is the caluclate table feelings distribution probability calculated according to softmax.
It should be noted that, for the present invention, the expression information of human face expression is exactly, if this expression laughs at, the probability so laughed at is exactly 1, if do not laughed at, the probability do not laughed at is exactly 1, and another probability is exactly 0, namely obtains known real table feelings distribution p (x) above; And the probable value judged is the number between 0 to 1, represent the probability of certain expression, caluclate table feelings distribution probability q (x) namely.
Whether step S1053: the expressive features value comparing these two face picture, have the human face expression label information of identical category according to these two face picture, calculate the second penalty values of the expressive features value obtaining these two face picture;
It should be noted that, for calculating the second loss function of the second penalty values of the expressive features value obtaining these two face picture being:
Wherein, x
iand y
ibe represent two photos being input to sorter, their expressive features value is respectively f (x
i) and f (y
j), c
ij=1 represents that two input photos are same expression, c
ij=0 represents that two input photos are different expressions.The object of this second loss function is: if two photos are same expressions, then reduce the difference of two photo eigen, if two photos are different expressions, then increase the difference of two photo eigen, the loss function in integrating step S1052 oppositely regulates whole weights of convolutional neural networks together.
It should be noted that, for the present invention, the effect of described second penalty values is diminished at the feature gap of similar expression, widened by the characteristic distance that difference is expressed one's feelings simultaneously; First calculate the expressive features value of two face picture, namely the output of last full articulamentum, then by the computing formula of the second loss function above, obtain the second penalty values, the f (x of the inside
i) and f (y
j) represent two face picture respectively.
About the computing formula of the second loss function, it should be noted that, f (x
i) and f (y
j) be the output of last full articulamentum of convolutional neural networks, see Fig. 5, Fig. 6, be visual human face expression feature schematic diagram.
Step S1054: all weights using the first-loss value of the expressive features value of these two face picture and the second penalty values to come together oppositely to regulate in described convolutional neural networks, namely complete the training to described convolutional neural networks.
For the present invention, it should be noted that, first-loss value and these two penalty values of the second penalty values are added and just obtain final penalty values; The process of weight is regulated to use existing known gradient descent method.
It should be noted that, the present invention is in order to the weight in regulating networks to the object that convolutional neural networks is trained; Particular by calculating first-loss value and the second penalty values respectively, obtain final penalty values, then by gradient descent method regulating networks weight.
It should be noted that, above step S1051 to S1054, is train described convolutional neural networks based on gradient descent method and back-propagation algorithm.
By operating procedure S106, see Fig. 5, Fig. 6, be respectively the smiling face's recognition methods utilizing a kind of facial image provided by the invention, be judged as the facial image schematic diagram of smile expressive features and the schematic diagram of normal expression (non-expression of smiling) feature.
It should be noted that, for the present invention, it utilizes degree of depth convolutional neural networks to extract in picture characteristic information to complete smiling face's identification.The method uses two kinds of labels as supervision message training network, the otherness of different classes of characteristics of image can be expanded simultaneously, increase the similarity of the characteristics of image of identical category simultaneously, therefore can extract and there is distinctive feature more, realize better smiling face's recognition function, be conducive to solving smiling face's identification problem.Present invention utilizes convolutional neural networks and there is powerful extraction feature capabilities, be extracted the expressive features in facial image, ensure that the accuracy rate that final smiling face identifies.Successful of the present invention is due to traditional smiling face's recognition effect.
In order to describe specific embodiment of the invention method in detail, with certain smiling face's identification database for embodiment is further detailed the inventive method.This database comprises 4000 photos, comprises different scene, as daytime, night, indoor, outdoor etc., also comprises different faces, as the male sex, women, youth, year wait for a long time.In an embodiment of the present invention, run above-mentioned steps S101 to S104 successively, set up the convolutional neural networks with 4 convolutional layers and 1 full articulamentum, see Fig. 4, the entitlement weight average random initializtion of this convolutional neural networks.Wherein the activation function of convolutional layer is ReLU function, and the facial image of feeding is the picture of 90 × 90 pixel sizes, ground floor convolutional layer adopt 32 be of a size of 11 × 11 × 1 convolution son; Second layer convolutional layer adopt 96 be of a size of 5 × 5 × 32 convolution son; Third layer convolutional layer adopt 128 be of a size of 2 × 1 × 96 convolution son; 4th layer of convolutional layer adopt 96 be of a size of 2 × 1 × 128 convolution; The dimension of the full articulamentum connected below is 160, as shown in Figure 4.Then operating procedure S105, S106 successively, carries out smiling face to the facial image often opened in photo in certain smiling face's identification database and identifies judgement, and judge face whether smiling face, final realization identifies the smiling face of photos all in smiling face's identification database.
Based on smiling face's recognition methods of a kind of facial image that the invention described above provides, see Fig. 7, smiling face's recognition device of a kind of facial image provided by the invention, comprising:
Image recognition extraction unit 701, for for multiple character images needing smiling face to identify, detects the position of face, and identifies the facial image extracted wherein, then send to image pre-processing unit;
Image pre-processing unit 702, be connected with image recognition extraction unit, for extracted facial image being scaled to the facial image of pre-set dimension size, and conversion process becomes gray-scale map, give the expression label information of pre-set categories simultaneously for often opening described facial image, and export to network training unit 704 and identify judging unit 705;
Network sets up unit 703, and for setting up convolutional neural networks (CNN), described convolutional neural networks comprises successively to input layer, default multiple convolutional layer, default multiple full articulamentum and output layer that inputted facial image processes, see Fig. 4;
Network training unit 704, set up unit 703 with network respectively, image pre-processing unit 702 is connected, for training described convolutional neural networks, expand the expressive features otherness between multiple facial images with different classes of expression label information, reduce the expressive features otherness between multiple facial images with identical category expression label information simultaneously;
Identify judging unit 705, be connected with network training unit 704, image pre-processing unit 702 respectively, for treating that the facial image of often opening that smiling face identifies all has been input in the convolutional neural networks of training by what process through image pre-processing unit 702, by described convolutional neural networks extract this facial image expressive features value and be sent to sorter carry out smiling face judge classification, whether the final face judged in facial image is smiling face, realizes smiling face's identifying operation.
In the present invention, described image recognition extraction unit 701 can be any one face recognition module existing.
In the present invention, described image pre-processing unit 702, network are set up unit 703, network training unit 704 and are identified central processor CPU, digital signal processor DSP or single-chip microprocessor MCU that judging unit 705 can be respectively apparatus of the present invention and installs.Described image pre-processing unit 702, network are set up unit 703, network training unit 704 and are identified that judging unit 705 for the device arranged separately, also can integratedly can be set together.
In the present invention, it should be noted that, current existing face recognition technology mainly judges to identify face according to eyes and the relative position of mouth and the approximate shape of face.Current is that core defines multiple face identification system with face recognition module, is provided with and applies widely: recognition of face access management system, recognition of face access control and attendance system and face recognition video monitoring system etc.
In the present invention, in specific implementation, for image recognition extraction unit, can using two, ear, nose, the face such as eyebrow and mouth are as key point, namely eyespot, auriculare, nose point, eyebrow point and stomion are set, detect the position determining face, and determine shape and the profile of face in character image, then the corresponding facial image extracted in character image.
In the present invention, for image pre-processing unit, described pre-set dimension size can according to user need arrange in advance, can be such as the arbitrary dimension between 48 × 48 pixels to 256 × 256 pixels, be preferably the size of 90 × 90 pixels.
In the present invention, the human face expression label information of described pre-set categories comprises smile label information and normal (non-smile) label information, smile label information and normal (non-smile) label information is given respectively see the face picture in Fig. 2 and Fig. 3, Fig. 2 and Fig. 3.
It should be noted that, for the present invention, network is set up to the convolutional neural networks of unit foundation, the entitlement weight average random initializtion of described convolutional neural networks.Wherein, see Fig. 4, the pre-sizing of number of described convolutional layer, between 3 to 7, is preferably 4; The pre-sizing of number of described full articulamentum is between 1 to 3, is preferably 1.
For the present invention, in described convolutional neural networks, the activation function of described convolutional layer preferably uses ReLU function, and the number of the sub-size of the step-length of each convolutional layer, convolution, convolution all can freely be arranged, as shown in Figure 4, the text description of Fig. 4 is specifically shown in embodiment below to network structure.The facial image of described gray scale is as input picture, after the input of each convolutional layer and the multiplied by weight of this layer, a numerical value can be obtained, the principle of ReLU function is exactly, if this numerical value is greater than 0, so output valve just preserves this calculated value, if this calculated value is less than 0, so output valve just preserves into 0.Certainly, ReLU function also can change other activation function into.
For the present invention, in described convolutional neural networks, full articulamentum preferably uses sigmoid activation function, certainly, also can use other activation functions.Described full articulamentum is for extracting the expressive features value of face picture.
For the present invention, it should be noted that, the layer that described convolutional neural networks contains convolution operation by some is interconnected the network structure formed, and his Main Function is the feature in order to extract picture, the expressive features value of such as facial image.
For described convolutional neural networks, the effect of the input layer wherein had is in order to image data (as facial image) is sent into convolutional neural networks (network structure), so that subsequent treatment; The effect of convolutional layer is the local features extracting picture; The effect of full articulamentum is to extract the feature having more distinction from the output of last layer; The effect of output layer is to judge whether face is smile; By the input of last time and and lower one deck between weight, obtain corresponding output valve.
For the present invention, it should be noted that, described network training unit 704, for training described convolutional neural networks, specifically comprises characteristic extracting module, first-loss value acquisition module, the second penalty values acquisition module and reverse adjustment module, wherein:
Characteristic extracting module, for the human face expression label information of any two facial images and correspondence thereof being input to the input layer of described convolutional neural networks, extracted the expressive features value of these two facial images by the convolutional layer of described convolutional neural networks and full articulamentum, then export from output layer;
First-loss value acquisition module, be connected with characteristic extracting module, classify for the expressive features value of these two face picture is sent into sorter, according to the human face expression label information that these two face picture have, calculate the first-loss value of the expressive features value obtaining these two face picture;
It should be noted that, for calculating the first-loss function of the first-loss value of the expressive features value obtaining these two face picture being:
Wherein, x is the expressive features value extracting facial image in step S1051, and p (x) is real classification situation, and q (x) is the probability of prediction.First-loss value the second penalty values in integrating step 1053 can oppositely regulate whole weights in convolutional neural networks together.
For the present invention, it should be noted that, the effect of sorter is the feature according to convolutional neural networks extraction above, classifies to the expression classification of face.In specific implementation, the present invention can adopt softmax sorter.
For softmax sorter, it can calculate the probability distribution of different expression, judges which kind of expression face inputs according to different probability distribution.Concrete operating process is the output of front one deck be one is eigenwert, is then normalized by these eigenwerts are multiplied by different weights, can obtain the probability distribution of different expression.
For the present invention, it should be noted that, the effect of first penalty values of the expressive features value of described two face picture calculates according to identifying information, be exactly specifically, if the probability results judged last and legitimate reading are relatively, then penalty values is less, if last result differs comparatively far away with real result, then penalty values is larger; Two kinds of faces are sent in convolutional neural networks of the present invention respectively, two first-loss values that different face is corresponding can be obtained, after being combined with the second penalty values that subsequent step obtains, be used for together adjusting network weight.
For the present invention, about first-loss function formula, it should be noted that, p (x) is known real table mutual affection cloth, and q (x) is the caluclate table feelings distribution probability calculated according to softmax.
It should be noted that, for the present invention, the expression information of human face expression is exactly, if this expression laughs at, the probability so laughed at is exactly 1, if do not laughed at, the probability do not laughed at is exactly 1, and another probability is exactly 0, namely obtains known real table feelings distribution p (x) above; And the probable value judged is the number between 0 to 1, represent the probability of certain expression, caluclate table feelings distribution probability q (x) namely.
Second penalty values acquisition module, be connected with characteristic extracting module, for comparing the expressive features value of these two face picture, whether there is according to these two face picture the human face expression label information of identical category, calculating the second penalty values of the expressive features value obtaining these two face picture;
It should be noted that, for calculating the second loss function of the second penalty values of the expressive features value obtaining these two face picture being:
Wherein, x
iand y
ibe represent two photos being input to sorter, their expressive features value is respectively f (x
i) and f (y
j), c
ij=1 represents that two input photos are same expression, c
ij=0 represents that two input photos are different expressions.The object of this second loss function is: if two photos are same expressions, then reduce the difference of two photo eigen, if two photos are different expressions, then increase the difference of two photo eigen, the loss function in integrating step S1052 oppositely regulates whole weights of convolutional neural networks together.
It should be noted that, for the present invention, the effect of described second penalty values is diminished at the feature gap of similar expression, widened by the characteristic distance that difference is expressed one's feelings simultaneously; First calculate the expressive features value of two face picture, namely the output of last full articulamentum, then by the computing formula of the second loss function above, obtain the second penalty values, the f (x of the inside
i) and f (y
j) represent two face picture respectively, see Fig. 5, Fig. 6, be visual human face expression feature schematic diagram.
About the computing formula of the second loss function, it should be noted that, f (x
i) and f (y
j) be the output of last full articulamentum of convolutional neural networks.
Reverse adjustment module, be connected with the second penalty values acquisition module with first-loss value acquisition module respectively, first-loss value and the second penalty values for using the expressive features value of these two face picture are come together all weights oppositely regulated in described convolutional neural networks, namely complete the training to described convolutional neural networks.
For the present invention, it should be noted that, first-loss value and these two penalty values of the second penalty values are added and just obtain final penalty values; The process of weight is regulated to use existing known gradient descent method.
The present invention is in order to the weight in regulating networks to the object that convolutional neural networks is trained; Particular by calculating first-loss value and the second penalty values respectively, obtain final penalty values, then by gradient descent method regulating networks weight.
It should be noted that, described network training unit 704 runs the above step first step to the 4th step, is train described convolutional neural networks based on gradient descent method and back-propagation algorithm.
See Fig. 5, Fig. 6, be respectively the smiling face's recognition methods utilizing a kind of facial image provided by the invention, be judged as the facial image schematic diagram of smile expressive features and the schematic diagram of normal expression (non-expression of smiling) feature.
It should be noted that, for the present invention, it utilizes degree of depth convolutional neural networks to extract in picture characteristic information to complete smiling face's identification.The method uses two kinds of labels as supervision message training network, the otherness of different classes of characteristics of image can be expanded simultaneously, increase the similarity of the characteristics of image of identical category simultaneously, therefore can extract and there is distinctive feature more, realize better smiling face's recognition function, be conducive to solving smiling face's identification problem.Present invention utilizes convolutional neural networks and there is powerful extraction feature capabilities, be extracted the expressive features in facial image, ensure that the accuracy rate that final smiling face identifies.Successful of the present invention is due to traditional smiling face's recognition effect.
In sum, compared with prior art, the invention provides a kind of smiling face's recognition methods and device thereof of facial image, its utilize constructed by convolutional neural networks, the expressive features extracting face completes smiling face's identification, can while guarantee carries out high-quality smiling face identification to face picture, fast, effectively accurately judgement is identified to smiling face in a large amount of face picture, meet the requirement of user to smiling face's recognition function, improve the work efficiency of user, save the time of people's preciousness, the product use sense being conducive to improving user is subject to, be of great practical significance.
The technology that the application of the invention provides, can make people work and the convenience of living is greatly improved, and drastically increases the living standard of people.
The above is only the preferred embodiment of the present invention; it should be pointed out that for those skilled in the art, under the premise without departing from the principles of the invention; can also make some improvements and modifications, these improvements and modifications also should be considered as protection scope of the present invention.
Claims (8)
1. smiling face's recognition methods of facial image, is characterized in that, comprise step:
The first step: multiple character images identified for needing smiling face, detects the position of face, and identifies the facial image extracted wherein;
Second step: the facial image extracted facial image being scaled to pre-set dimension size, and conversion process becomes gray-scale map, and give the expression label information of pre-set categories for often opening described facial image;
3rd step: set up convolutional neural networks, described convolutional neural networks comprises successively to input layer, default multiple convolutional layer, default multiple full articulamentum and output layer that inputted facial image processes;
4th step: described convolutional neural networks is trained, expand the expressive features otherness between multiple facial images with different classes of expression label information, reduce the expressive features otherness between multiple facial images with identical category expression label information simultaneously;
5th step: by processed one-tenth gray-scale map and convergent-divergent pre-set dimension, treat that the facial image of often opening that smiling face identifies all has been input in the convolutional neural networks of training, by described convolutional neural networks extract this facial image expressive features value and be sent to sorter carry out smiling face judge classification, realize smiling face's identifying operation.
2. the method for claim 1, is characterized in that, in second step, the human face expression label information of described pre-set categories comprises smile label information and non-smile label information.
3. the method for claim 1, is characterized in that, described convolutional neural networks comprises an input layer, four convolutional layers, a full articulamentum and output layers.
4. the method for claim 1, is characterized in that, in described 4th step, is specially the step that described convolutional neural networks is trained:
The human face expression label information of any two facial images and correspondence thereof is input to the input layer of described convolutional neural networks, extracted the expressive features value of these two facial images by the convolutional layer of described convolutional neural networks and full articulamentum, then export from output layer;
The expressive features value of these two face picture is sent into sorter classify, according to the human face expression label information that these two face picture have, calculate the first-loss value of the expressive features value obtaining these two face picture;
Relatively whether the expressive features value of these two face picture, have the human face expression label information of identical category according to these two face picture, calculates the second penalty values of the expressive features value obtaining these two face picture;
The all weights using the first-loss value of the expressive features value of these two face picture and the second penalty values to come together oppositely to regulate in described convolutional neural networks, complete the training to described convolutional neural networks.
5. smiling face's recognition device of facial image, is characterized in that, comprising:
Image recognition extraction unit, for for multiple character images needing smiling face to identify, detects the position of face, and identifies the facial image extracted wherein, then send to image pre-processing unit;
Image pre-processing unit, be connected with image recognition extraction unit, for extracted facial image being scaled to the facial image of pre-set dimension size, and conversion process becomes gray-scale map, give the expression label information of pre-set categories simultaneously for often opening described facial image, and export to network training unit and identify judging unit;
Network sets up unit, and for setting up convolutional neural networks, described convolutional neural networks comprises successively to input layer, default multiple convolutional layer, default multiple full articulamentum and output layer that inputted facial image processes;
Network training unit, set up unit with network respectively, image pre-processing unit is connected, for training described convolutional neural networks, expand the expressive features otherness between multiple facial images with different classes of expression label information, reduce the expressive features otherness between multiple facial images with identical category expression label information simultaneously;
Identify judging unit, be connected with network training unit, image pre-processing unit respectively, for the facial image of often opening through image pre-processing unit process has all been input in the convolutional neural networks of training, by described convolutional neural networks extract this facial image expressive features value and be sent to sorter carry out smiling face judge classification, realize smiling face's identifying operation.
6. device as claimed in claim 5, it is characterized in that, the human face expression label information of described pre-set categories comprises smile label information and non-smile label information.
7. device as claimed in claim 5, it is characterized in that, described convolutional neural networks comprises an input layer, four convolutional layers, a full articulamentum and output layers.
8. device as claimed in claim 5, is characterized in that, described network training unit comprises characteristic extracting module, first-loss value acquisition module, the second penalty values acquisition module and reverse adjustment module, wherein:
Characteristic extracting module, for the human face expression label information of any two facial images and correspondence thereof being input to the input layer of described convolutional neural networks, extracted the expressive features value of these two facial images by the convolutional layer of described convolutional neural networks and full articulamentum, then export from output layer;
First-loss value acquisition module, be connected with characteristic extracting module, classify for the expressive features value of these two face picture is sent into sorter, according to the human face expression label information that these two face picture have, calculate the first-loss value of the expressive features value obtaining these two face picture;
Second penalty values acquisition module, be connected with characteristic extracting module, for comparing the expressive features value of these two face picture, whether there is according to these two face picture the human face expression label information of identical category, calculating the second penalty values of the expressive features value obtaining these two face picture;
Reverse adjustment module, be connected with the second penalty values acquisition module with first-loss value acquisition module respectively, first-loss value and the second penalty values for using the expressive features value of these two face picture are come together all weights oppositely regulated in described convolutional neural networks, namely complete the training to described convolutional neural networks.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510868158.5A CN105512624B (en) | 2015-12-01 | 2015-12-01 | A kind of smiling face's recognition methods of facial image and its device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510868158.5A CN105512624B (en) | 2015-12-01 | 2015-12-01 | A kind of smiling face's recognition methods of facial image and its device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105512624A true CN105512624A (en) | 2016-04-20 |
| CN105512624B CN105512624B (en) | 2019-06-21 |
Family
ID=55720591
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510868158.5A Active CN105512624B (en) | 2015-12-01 | 2015-12-01 | A kind of smiling face's recognition methods of facial image and its device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105512624B (en) |
Cited By (29)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105956570A (en) * | 2016-05-11 | 2016-09-21 | 电子科技大学 | Lip characteristic and deep learning based smiling face recognition method |
| CN106295678A (en) * | 2016-07-27 | 2017-01-04 | 北京旷视科技有限公司 | Neural metwork training and construction method and device and object detection method and device |
| CN106341724A (en) * | 2016-08-29 | 2017-01-18 | 刘永娜 | Expression image marking method and system |
| CN106372651A (en) * | 2016-08-22 | 2017-02-01 | 平安科技（深圳）有限公司 | Picture quality detection method and device |
| CN106529504A (en) * | 2016-12-02 | 2017-03-22 | 合肥工业大学 | Dual-mode video emotion recognition method with composite spatial-temporal characteristic |
| CN106548159A (en) * | 2016-11-08 | 2017-03-29 | 中国科学院自动化研究所 | Reticulate pattern facial image recognition method and device based on full convolutional neural networks |
| CN106778664A (en) * | 2016-12-29 | 2017-05-31 | 天津中科智能识别产业技术研究院有限公司 | The dividing method and its device of iris region in a kind of iris image |
| CN107133578A (en) * | 2017-04-19 | 2017-09-05 | 华南理工大学 | A kind of facial expression recognizing method transmitted based on file and system |
| CN107358169A (en) * | 2017-06-21 | 2017-11-17 | 厦门中控智慧信息技术有限公司 | A kind of facial expression recognizing method and expression recognition device |
| CN107506722A (en) * | 2017-08-18 | 2017-12-22 | 中国地质大学（武汉） | One kind is based on depth sparse convolution neutral net face emotion identification method |
| CN107590457A (en) * | 2017-09-07 | 2018-01-16 | 竹间智能科技（上海）有限公司 | Emotion identification method and system based on the cascade change network architecture |
| CN107871099A (en) * | 2016-09-23 | 2018-04-03 | 北京眼神科技有限公司 | Face detection method and apparatus |
| CN107871098A (en) * | 2016-09-23 | 2018-04-03 | 北京眼神科技有限公司 | Method and device for acquiring human face characteristic points |
| CN107944458A (en) * | 2017-12-08 | 2018-04-20 | 北京维大成科技有限公司 | A kind of image-recognizing method and device based on convolutional neural networks |
| CN108154222A (en) * | 2016-12-02 | 2018-06-12 | 北京市商汤科技开发有限公司 | Deep neural network training method and system, electronic equipment |
| CN109034264A (en) * | 2018-08-15 | 2018-12-18 | 云南大学 | Traffic accident seriousness predicts CSP-CNN model and its modeling method |
| CN109360183A (en) * | 2018-08-20 | 2019-02-19 | 中国电子进出口有限公司 | A kind of quality of human face image appraisal procedure and system based on convolutional neural networks |
| WO2019033573A1 (en) * | 2017-08-17 | 2019-02-21 | 平安科技（深圳）有限公司 | Facial emotion identification method, apparatus and storage medium |
| CN109684911A (en) * | 2018-10-30 | 2019-04-26 | 百度在线网络技术（北京）有限公司 | Expression recognition method, device, electronic equipment and storage medium |
| CN110059625A (en) * | 2019-04-18 | 2019-07-26 | 重庆大学 | A kind of face training and recognition methods based on mixup |
| CN110175583A (en) * | 2019-05-30 | 2019-08-27 | 重庆跃途科技有限公司 | It is a kind of in the campus universe security monitoring analysis method based on video AI |
| CN110458021A (en) * | 2019-07-10 | 2019-11-15 | 上海交通大学 | A kind of face moving cell detection method based on physical characteristic and distribution character |
| CN110472509A (en) * | 2019-07-15 | 2019-11-19 | 中国平安人寿保险股份有限公司 | Fat or thin recognition methods and device, electronic equipment based on facial image |
| CN110866962A (en) * | 2019-11-20 | 2020-03-06 | 成都威爱新经济技术研究院有限公司 | Virtual portrait and expression synchronization method based on convolutional neural network |
| CN111079712A (en) * | 2019-12-31 | 2020-04-28 | 中国银行股份有限公司 | Authority management method and device based on face recognition |
| CN111160453A (en) * | 2019-12-27 | 2020-05-15 | 联想(北京)有限公司 | Information processing method and device and computer readable storage medium |
| CN111314760A (en) * | 2020-01-19 | 2020-06-19 | 深圳市爱深盈通信息技术有限公司 | Television and smiling face shooting method thereof |
| CN111652242A (en) * | 2020-04-20 | 2020-09-11 | 北京迈格威科技有限公司 | Image processing method, image processing device, electronic equipment and storage medium |
| US11062453B2 (en) | 2016-12-02 | 2021-07-13 | Beijing Sensetime Technology Development Co., Ltd. | Method and system for scene parsing and storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120155718A1 (en) * | 2010-12-21 | 2012-06-21 | Samsung Electronics Co. Ltd. | Face recognition apparatus and method |
| CN103793718A (en) * | 2013-12-11 | 2014-05-14 | 台州学院 | Deep study-based facial expression recognition method |
| CN105005774A (en) * | 2015-07-28 | 2015-10-28 | 中国科学院自动化研究所 | Face relative relation recognition method based on convolutional neural network and device thereof |
-
2015
- 2015-12-01 CN CN201510868158.5A patent/CN105512624B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120155718A1 (en) * | 2010-12-21 | 2012-06-21 | Samsung Electronics Co. Ltd. | Face recognition apparatus and method |
| CN103793718A (en) * | 2013-12-11 | 2014-05-14 | 台州学院 | Deep study-based facial expression recognition method |
| CN105005774A (en) * | 2015-07-28 | 2015-10-28 | 中国科学院自动化研究所 | Face relative relation recognition method based on convolutional neural network and device thereof |
Non-Patent Citations (1)
| Title |
|---|
| 徐奔: "基于DWT和BP神经网络的人脸识别方法", 《电脑知识与技术》 * |
Cited By (44)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105956570A (en) * | 2016-05-11 | 2016-09-21 | 电子科技大学 | Lip characteristic and deep learning based smiling face recognition method |
| CN105956570B (en) * | 2016-05-11 | 2019-08-20 | 电子科技大学 | Smiling face's recognition methods based on lip feature and deep learning |
| CN106295678A (en) * | 2016-07-27 | 2017-01-04 | 北京旷视科技有限公司 | Neural metwork training and construction method and device and object detection method and device |
| US10769493B2 (en) | 2016-07-27 | 2020-09-08 | Beijing Kuangshi Technology Co., Ltd. | Method and apparatus for neural network training and construction and method and apparatus for object detection |
| CN106372651B (en) * | 2016-08-22 | 2018-03-06 | 平安科技（深圳）有限公司 | The detection method and device of picture quality |
| CN106372651A (en) * | 2016-08-22 | 2017-02-01 | 平安科技（深圳）有限公司 | Picture quality detection method and device |
| CN106341724A (en) * | 2016-08-29 | 2017-01-18 | 刘永娜 | Expression image marking 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 |
| CN107871098A (en) * | 2016-09-23 | 2018-04-03 | 北京眼神科技有限公司 | Method and device for acquiring human face characteristic points |
| CN106548159A (en) * | 2016-11-08 | 2017-03-29 | 中国科学院自动化研究所 | Reticulate pattern facial image recognition method and device based on full convolutional neural networks |
| CN108154222A (en) * | 2016-12-02 | 2018-06-12 | 北京市商汤科技开发有限公司 | Deep neural network training method and system, electronic equipment |
| CN108154222B (en) * | 2016-12-02 | 2020-08-11 | 北京市商汤科技开发有限公司 | Deep neural network training method and system and electronic equipment |
| CN106529504B (en) * | 2016-12-02 | 2019-05-31 | 合肥工业大学 | A kind of bimodal video feeling recognition methods of compound space-time characteristic |
| US11062453B2 (en) | 2016-12-02 | 2021-07-13 | Beijing Sensetime Technology Development Co., Ltd. | Method and system for scene parsing and storage medium |
| CN106529504A (en) * | 2016-12-02 | 2017-03-22 | 合肥工业大学 | Dual-mode video emotion recognition method with composite spatial-temporal characteristic |
| CN106778664B (en) * | 2016-12-29 | 2020-12-15 | 天津中科智能识别产业技术研究院有限公司 | Iris image iris area segmentation method and device |
| CN106778664A (en) * | 2016-12-29 | 2017-05-31 | 天津中科智能识别产业技术研究院有限公司 | The dividing method and its device of iris region in a kind of iris image |
| CN107133578B (en) * | 2017-04-19 | 2020-05-22 | 华南理工大学 | Facial expression recognition method and system based on file transmission |
| CN107133578A (en) * | 2017-04-19 | 2017-09-05 | 华南理工大学 | A kind of facial expression recognizing method transmitted based on file and system |
| CN107358169A (en) * | 2017-06-21 | 2017-11-17 | 厦门中控智慧信息技术有限公司 | A kind of facial expression recognizing method and expression recognition device |
| WO2019033573A1 (en) * | 2017-08-17 | 2019-02-21 | 平安科技（深圳）有限公司 | Facial emotion identification method, apparatus and storage medium |
| CN107506722A (en) * | 2017-08-18 | 2017-12-22 | 中国地质大学（武汉） | One kind is based on depth sparse convolution neutral net face emotion identification method |
| CN107590457A (en) * | 2017-09-07 | 2018-01-16 | 竹间智能科技（上海）有限公司 | Emotion identification method and system based on the cascade change network architecture |
| CN107944458A (en) * | 2017-12-08 | 2018-04-20 | 北京维大成科技有限公司 | A kind of image-recognizing method and device based on convolutional neural networks |
| CN109034264B (en) * | 2018-08-15 | 2021-11-19 | 云南大学 | CSP-CNN model for predicting severity of traffic accident and modeling method thereof |
| CN109034264A (en) * | 2018-08-15 | 2018-12-18 | 云南大学 | Traffic accident seriousness predicts CSP-CNN model and its modeling method |
| CN109360183A (en) * | 2018-08-20 | 2019-02-19 | 中国电子进出口有限公司 | A kind of quality of human face image appraisal procedure and system based on convolutional neural networks |
| CN109684911A (en) * | 2018-10-30 | 2019-04-26 | 百度在线网络技术（北京）有限公司 | Expression recognition method, device, electronic equipment and storage medium |
| US11151363B2 (en) | 2018-10-30 | 2021-10-19 | Baidu Online Network Technology (Beijing) Co., Ltd. | Expression recognition method, apparatus, electronic device, and storage medium |
| CN110059625B (en) * | 2019-04-18 | 2023-04-07 | 重庆大学 | Face training and recognition method based on mixup |
| CN110059625A (en) * | 2019-04-18 | 2019-07-26 | 重庆大学 | A kind of face training and recognition methods based on mixup |
| CN110175583A (en) * | 2019-05-30 | 2019-08-27 | 重庆跃途科技有限公司 | It is a kind of in the campus universe security monitoring analysis method based on video AI |
| CN110458021A (en) * | 2019-07-10 | 2019-11-15 | 上海交通大学 | A kind of face moving cell detection method based on physical characteristic and distribution character |
| CN110472509A (en) * | 2019-07-15 | 2019-11-19 | 中国平安人寿保险股份有限公司 | Fat or thin recognition methods and device, electronic equipment based on facial image |
| CN110472509B (en) * | 2019-07-15 | 2024-04-26 | 中国平安人寿保险股份有限公司 | Fat-lean recognition method and device based on face image and electronic equipment |
| CN110866962A (en) * | 2019-11-20 | 2020-03-06 | 成都威爱新经济技术研究院有限公司 | Virtual portrait and expression synchronization method based on convolutional neural network |
| CN110866962B (en) * | 2019-11-20 | 2023-06-16 | 成都威爱新经济技术研究院有限公司 | Virtual portrait and expression synchronization method based on convolutional neural network |
| CN111160453A (en) * | 2019-12-27 | 2020-05-15 | 联想(北京)有限公司 | Information processing method and device and computer readable storage medium |
| CN111079712A (en) * | 2019-12-31 | 2020-04-28 | 中国银行股份有限公司 | Authority management method and device based on face recognition |
| CN111079712B (en) * | 2019-12-31 | 2023-04-21 | 中国银行股份有限公司 | Permission management method and device based on face recognition |
| CN111314760B (en) * | 2020-01-19 | 2021-07-02 | 深圳市爱深盈通信息技术有限公司 | Television and smiling face shooting method thereof |
| CN111314760A (en) * | 2020-01-19 | 2020-06-19 | 深圳市爱深盈通信息技术有限公司 | Television and smiling face shooting method thereof |
| CN111652242A (en) * | 2020-04-20 | 2020-09-11 | 北京迈格威科技有限公司 | Image processing method, image processing device, electronic equipment and storage medium |
| CN111652242B (en) * | 2020-04-20 | 2023-07-04 | 北京迈格威科技有限公司 | Image processing method, device, electronic equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105512624B (en) | 2019-06-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105512624A (en) | Smile face recognition method and device for human face image | |
| CN104992167B (en) | A kind of method for detecting human face and device based on convolutional neural networks | |
| CN109657592B (en) | Face recognition method of intelligent excavator | |
| CN105005774B (en) | A kind of recognition methods of face kinship and device based on convolutional neural networks | |
| WO2022111236A1 (en) | Facial expression recognition method and system combined with attention mechanism | |
| CN102637251B (en) | Face recognition method based on reference features | |
| CN111723786B (en) | Method and device for detecting wearing of safety helmet based on single model prediction | |
| WO2020114118A1 (en) | Facial attribute identification method and device, storage medium and processor | |
| CN112131978B (en) | Video classification method and device, electronic equipment and storage medium | |
| CN106803069A (en) | Crowd's level of happiness recognition methods based on deep learning | |
| CN102938065B (en) | Face feature extraction method and face identification method based on large-scale image data | |
| CN109635727A (en) | A kind of facial expression recognizing method and device | |
| CN110969087B (en) | Gait recognition method and system | |
| CN107145857A (en) | Face character recognition methods, device and method for establishing model | |
| CN110796018B (en) | Hand motion recognition method based on depth image and color image | |
| CN104361316B (en) | Dimension emotion recognition method based on multi-scale time sequence modeling | |
| CN111597955A (en) | Smart home control method and device based on expression emotion recognition of deep learning | |
| CN105956570B (en) | Smiling face's recognition methods based on lip feature and deep learning | |
| CN102567716B (en) | Face synthetic system and implementation method | |
| CN104091176A (en) | Technology for applying figure and head portrait comparison to videos | |
| CN111368672A (en) | Construction method and device for genetic disease facial recognition model | |
| CN108427740B (en) | Image emotion classification and retrieval algorithm based on depth metric learning | |
| CN110543906B (en) | Automatic skin recognition method based on Mask R-CNN model | |
| CN101853397A (en) | Bionic human face detection method based on human visual characteristics | |
| CN109101108A (en) | Method and system based on three decision optimization intelligence cockpit human-computer interaction interfaces |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CP03 | Change of name, title or address | ||
| CP03 | Change of name, title or address |
Address after: 300457 unit 1001, block 1, msd-g1, TEDA, No.57, 2nd Street, Binhai New Area Economic and Technological Development Zone, Tianjin Patentee after: Tianjin Zhongke intelligent identification Co.,Ltd. Address before: 300457 No. 57, Second Avenue, Economic and Technological Development Zone, Binhai New Area, Tianjin Patentee before: TIANJIN ZHONGKE INTELLIGENT IDENTIFICATION INDUSTRY TECHNOLOGY RESEARCH INSTITUTE Co.,Ltd. |