CN107301406A - Fast face angle recognition method based on deep learning - Google Patents
Fast face angle recognition method based on deep learning Download PDFInfo
- Publication number
- CN107301406A CN107301406A CN201710571930.6A CN201710571930A CN107301406A CN 107301406 A CN107301406 A CN 107301406A CN 201710571930 A CN201710571930 A CN 201710571930A CN 107301406 A CN107301406 A CN 107301406A
- Authority
- CN
- China
- Prior art keywords
- angle
- deep learning
- face
- negative
- positive
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/161—Detection; Localisation; Normalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
Abstract
Fast face angle recognition method based on deep learning.Facial angle identification is related to head anglec of rotation around three axles in three-dimensional perpendicular coordinate system, because of facial angle change, can cause the excalation of face feature information.The present invention comprises the following steps:(1)Facial image when rotating 0 °, positive and negative 15 °, positive and negative 30 ° and positive and negative 45 ° is marked to set up database;(2)Angle signature is subjected to feature extraction by five layers of convolution module, by two layers of full connection, the characteristic vector of 2048 dimensions is converted into, the input for layer of classifying as softmax carries out classification to set up convolutional neural networks structure;(3)Based on the operating systems of ubuntu 16.04, under GPU1080, convolutional neural networks are trained under deep learning framework CAFFE, facial angle identification model is obtained.The present invention is used for the Real time identification of the face three-dimensional perspective based on deep learning.
Description
Technical field:
The present invention relates to field of face identification, more particularly to a kind of fast face angle recognition method based on deep learning.
Background technology:
Facial angle identification technology is every image containing face for collecting, and rapidly and accurately can be obtained in image
The three-dimensional perspective information of face, it is possible to use these information do further recognition of face, and facial angle identification is that computer is regarded
Feel field and an important research direction of field of face identification, are a vital steps in face identification system, with wide
General application value and good market prospects, existing facial angle identification technology have the method based on template, based on face
Method of set feature etc., but these methods have some limitations, and are only capable of recognizing the face on the single direction of small range
Angle change, the robustness for influence factors such as illumination variation, partial occlusion and expression shape changes is not high, for the multi-party of complexity
It can not be determined to the facial angle of change simultaneously.
The content of the invention:
The invention aims to overcome above-mentioned the deficiencies in the prior art can be rapidly while carrying out three dimensions there is provided one kind
The fast face angle recognition method based on deep learning of the facial angle identification of degree.
Above-mentioned purpose is realized by following technical scheme:
A kind of fast face angle recognition method based on deep learning, this method comprises the following steps:(1)By 0 ° of rotation, just
Facial image at minus 15 °, positive and negative 30 ° and positive and negative 45 ° is marked to set up database;(2)Angle signature is passed through five layers
Convolution module carries out feature extraction, by two layers of full connection, is converted into the characteristic vector of 2048 dimensions, is used as softmax classification layers
Input, carry out classification so as to setting up convolutional neural networks structure;(3)Based on the operating systems of ubuntu 16.04, GPU1080
Under, convolutional neural networks are trained under deep learning framework CAFFE, facial angle identification model is obtained.
The described fast face angle recognition method based on deep learning, the described detailed process for setting up database
For:The database is the facial image of 300 different people, totally 7000, and the image in the database includes head in three-dimensional
In vertical coordinate system around each axle rotate 0 °, positive and negative 15 °, positive and negative 30 ° and positive and negative 45 ° when facial image, the anglec of rotation is returned
One turns to value between 0-1 as label value, and the facial angle in every image is marked, and the image after mark is used for nerve
The training of network.
The described fast face angle recognition method based on deep learning, described convolutional neural networks structure of setting up
Detailed process is:Input layer receives input data, obtains view data and its respective labels value, and it is comprising three to set up data set
The angle signature of label value, the respectively each axle rotation of corresponding three-dimensional coordinate system, then carries out feature by five layers of convolution module and carries
Take, each convolution module includes convolutional layer and pond layer, and the characteristic vector extracted is input into full articulamentum, complete by two layers
Connection, characteristic pattern is converted into the characteristic vector of 2048 dimensions, the input for layer of classifying as softmax, is classified, three labels
The classification layer arranged side by side of correspondence three, each classification layer obtains the angle that face rotates in three-dimensional system of coordinate around each axle.
The described fast face angle recognition method based on deep learning, the described facial angle identification model that obtains
Detailed process is:Based on the operating systems of ubuntu 16.04, under GPU1080, CAFFE deep learning frameworks are built, training is used
Data set is input to convolutional neural networks, by 100000 training, obtains facial angle identification model, calls the model to carry out
Facial angle recognizes test experiments, and every face of identification was taken within 0.02 second, the face three-dimensional perspective error identified
Within 3 °.
Beneficial effect:
1. the present invention includes the two dimensional image of face according to what is collected, while the three-dimensional perspective of face in image is obtained, to enter
The recognition of face of one step provides convenient, and then improves the face recognition accuracy rate under actual environment nature.
Foundation of the present invention including database, convolutional neural networks structure design, using the data set of foundation in depth
Practise and train neutral net under framework, obtain facial angle identification model, call the model to realize the real-time of face three-dimensional perspective
Identification.
The present invention every face of identification was taken within 0.02 second, the face three-dimensional perspective error identified 3 ° with
It is interior, it is possible to achieve fast and accurately facial angle is recognized.
Brief description of the drawings:
Accompanying drawing 1 is the structural representation of the present invention.
Accompanying drawing 2 is the convolutional neural networks structure chart of the present invention.
Embodiment:
Embodiment 1:
A kind of fast face angle recognition method based on deep learning, this method comprises the following steps:(1)By 0 ° of rotation, just
Facial image at minus 15 °, positive and negative 30 ° and positive and negative 45 ° is marked to set up database;(2)Angle signature is passed through five layers
Convolution module carries out feature extraction, by two layers of full connection, is converted into the characteristic vector of 2048 dimensions, is used as softmax classification layers
Input, carry out classification so as to setting up convolutional neural networks structure;(3)Based on the operating systems of ubuntu 16.04, GPU1080
Under, convolutional neural networks are trained under deep learning framework CAFFE, facial angle identification model is obtained.
Embodiment 2:
The fast face angle recognition method based on deep learning according to embodiment 1, the described tool for setting up database
Body process is:The database is the facial image of 300 different people, totally 7000, and the image in the database includes head
In three-dimensional perpendicular coordinate system around each axle rotate 0 °, positive and negative 15 °, positive and negative 30 ° and positive and negative 45 ° when facial image, will rotate
Angle is normalized to value between 0-1 as label value, and the facial angle in every image is marked, and the image after mark is used
In the training of neutral net.
Embodiment 3:
The fast face angle recognition method based on deep learning according to embodiment 1 or 2, described sets up convolutional Neural
The detailed process of network structure is:Input layer receives input data, obtains view data and its respective labels value, sets up data set
To include the angle signature of three label values, the respectively each axle rotation of corresponding three-dimensional coordinate system, then by five layers of convolution module
Feature extraction is carried out, each convolution module includes convolutional layer and pond layer, the characteristic vector extracted is input into full articulamentum,
By two layers of full connection, characteristic pattern is converted into the characteristic vector of 2048 dimensions, the input for layer of classifying as softmax is divided
Class, three label correspondences three classification layer arranged side by side, each classification layer obtains face and rotated in three-dimensional system of coordinate around each axle
Angle.
Embodiment 4:
The fast face angle recognition method based on deep learning according to embodiment 1 or 2 or 3, described obtains face
The detailed process of angle recognition model is:Based on the operating systems of ubuntu 16.04, under GPU1080, CAFFE depth is built
Framework is practised, training is input to convolutional neural networks with data set, by 100000 training, facial angle identification mould is obtained
Type, calls the model to carry out facial angle identification test experiments, and taking within 0.02 second for every face of identification is identified
Face three-dimensional perspective error is within 3 °.
Claims (4)
1. a kind of fast face angle recognition method based on deep learning, it is characterized in that:This method comprises the following steps:(1)
Facial image when rotating 0 °, positive and negative 15 °, positive and negative 30 ° and positive and negative 45 ° is marked to set up database;(2)By angle mark
Label carry out feature extraction by five layers of convolution module, by two layers of full connection, are converted into the characteristic vector of 2048 dimensions, as
The input of softmax classification layers, carries out classification to set up convolutional neural networks structure;(3)Operated based on ubuntu 16.04
Under system, GPU1080, convolutional neural networks are trained under deep learning framework CAFFE, facial angle identification mould is obtained
Type.
2. the fast face angle recognition method according to claim 1 based on deep learning, it is characterized in that:Described builds
The detailed process of vertical database is:The database is the facial image of 300 different people, totally 7000, in the database
Image include head in three-dimensional perpendicular coordinate system around each axle rotate 0 °, positive and negative 15 °, positive and negative 30 ° and positive and negative 45 ° when face
Image, the value between the anglec of rotation is normalized into 0-1 is marked as label value to the facial angle in every image, mark
Image after note is used for the training of neutral net.
3. the fast face angle recognition method according to claim 1 or 2 based on deep learning, it is characterized in that:It is described
The detailed process for setting up convolutional neural networks structure be:Input layer receives input data, obtains view data and its corresponding mark
Label value, it is the angle signature for including three label values, the respectively each axle rotation of corresponding three-dimensional coordinate system, Ran Houjing to set up data set
Cross five layers of convolution module and carry out feature extraction, each convolution module includes convolutional layer and pond layer, by the characteristic vector extracted
Full articulamentum is input to, by two layers of full connection, characteristic pattern is converted into the characteristic vector of 2048 dimensions, classified as softmax
The input of layer, is classified, and three label correspondences three classification layer arranged side by side, each classification layer obtains face in three-dimensional system of coordinate
In around each axle rotate angle.
4. the fast face angle recognition method based on deep learning according to claim 1 or 2 or 3, it is characterized in that:Institute
That states, which obtains the detailed process of facial angle identification model, is:Based on the operating systems of ubuntu 16.04, under GPU1080, build
CAFFE deep learning frameworks, convolutional neural networks are input to by training with data set, by 100000 training, obtain face
Angle recognition model, calls the model to carry out facial angle identification test experiments, every face of identification take 0.02 second with
Interior, the face three-dimensional perspective error identified is within 3 °.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710571930.6A CN107301406A (en) | 2017-07-13 | 2017-07-13 | Fast face angle recognition method based on deep learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710571930.6A CN107301406A (en) | 2017-07-13 | 2017-07-13 | Fast face angle recognition method based on deep learning |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107301406A true CN107301406A (en) | 2017-10-27 |
Family
ID=60133908
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710571930.6A Pending CN107301406A (en) | 2017-07-13 | 2017-07-13 | Fast face angle recognition method based on deep learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107301406A (en) |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107765858A (en) * | 2017-11-06 | 2018-03-06 | 广东欧珀移动通信有限公司 | Determine the method, apparatus, terminal and storage medium of facial angle |
CN107909095A (en) * | 2017-11-07 | 2018-04-13 | 江苏大学 | A kind of image-recognizing method based on deep learning |
CN108596089A (en) * | 2018-04-24 | 2018-09-28 | 北京达佳互联信息技术有限公司 | Human face posture detection method, device, computer equipment and storage medium |
CN108960093A (en) * | 2018-06-21 | 2018-12-07 | 阿里体育有限公司 | The recognition methods and equipment of face's rotational angle |
CN109740492A (en) * | 2018-12-27 | 2019-05-10 | 郑州云海信息技术有限公司 | A kind of identity identifying method and device |
CN109784202A (en) * | 2018-12-24 | 2019-05-21 | 珠海格力电器股份有限公司 | Recognition methods, device and the washing machine and computer readable storage medium of object are left in article to be washed |
CN109948510A (en) * | 2019-03-14 | 2019-06-28 | 北京易道博识科技有限公司 | A kind of file and picture example dividing method and device |
CN109993150A (en) * | 2019-04-15 | 2019-07-09 | 北京字节跳动网络技术有限公司 | The method and apparatus at age for identification |
CN110110712A (en) * | 2019-06-06 | 2019-08-09 | 江苏尚飞光电科技股份有限公司 | A kind of novel Activity recognition screening machine |
CN110175572A (en) * | 2019-05-28 | 2019-08-27 | 深圳市商汤科技有限公司 | Face image processing process and device, electronic equipment and storage medium |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102799901A (en) * | 2012-07-10 | 2012-11-28 | 辉路科技(北京)有限公司 | Method for multi-angle face detection |
CN103310179A (en) * | 2012-03-06 | 2013-09-18 | 上海骏聿数码科技有限公司 | Method and system for optimal attitude detection based on face recognition technology |
CN103793693A (en) * | 2014-02-08 | 2014-05-14 | 厦门美图网科技有限公司 | Method for detecting face turning and facial form optimizing method with method for detecting face turning |
CN105718868A (en) * | 2016-01-18 | 2016-06-29 | 中国科学院计算技术研究所 | Face detection system and method for multi-pose faces |
CN105760836A (en) * | 2016-02-17 | 2016-07-13 | 厦门美图之家科技有限公司 | Multi-angle face alignment method based on deep learning and system thereof and photographing terminal |
CN105787478A (en) * | 2016-04-14 | 2016-07-20 | 中南大学 | Face direction change recognition method based on neural network and sensitivity parameter |
CN106372630A (en) * | 2016-11-23 | 2017-02-01 | 华南理工大学 | Face direction detection method based on deep learning |
CN106503687A (en) * | 2016-11-09 | 2017-03-15 | 合肥工业大学 | The monitor video system for identifying figures of fusion face multi-angle feature and its method |
-
2017
- 2017-07-13 CN CN201710571930.6A patent/CN107301406A/en active Pending
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103310179A (en) * | 2012-03-06 | 2013-09-18 | 上海骏聿数码科技有限公司 | Method and system for optimal attitude detection based on face recognition technology |
CN102799901A (en) * | 2012-07-10 | 2012-11-28 | 辉路科技(北京)有限公司 | Method for multi-angle face detection |
CN103793693A (en) * | 2014-02-08 | 2014-05-14 | 厦门美图网科技有限公司 | Method for detecting face turning and facial form optimizing method with method for detecting face turning |
CN105718868A (en) * | 2016-01-18 | 2016-06-29 | 中国科学院计算技术研究所 | Face detection system and method for multi-pose faces |
CN105760836A (en) * | 2016-02-17 | 2016-07-13 | 厦门美图之家科技有限公司 | Multi-angle face alignment method based on deep learning and system thereof and photographing terminal |
CN105787478A (en) * | 2016-04-14 | 2016-07-20 | 中南大学 | Face direction change recognition method based on neural network and sensitivity parameter |
CN106503687A (en) * | 2016-11-09 | 2017-03-15 | 合肥工业大学 | The monitor video system for identifying figures of fusion face multi-angle feature and its method |
CN106372630A (en) * | 2016-11-23 | 2017-02-01 | 华南理工大学 | Face direction detection method based on deep learning |
Non-Patent Citations (2)
Title |
---|
IACOPO MASI等: "Pose-Aware Face Recognition in the Wild", 《2016 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION》 * |
邓宗平 等: "基于深度学习的人脸姿态分类方法", 《计算机技术与发展》 * |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107765858A (en) * | 2017-11-06 | 2018-03-06 | 广东欧珀移动通信有限公司 | Determine the method, apparatus, terminal and storage medium of facial angle |
CN107765858B (en) * | 2017-11-06 | 2019-12-31 | Oppo广东移动通信有限公司 | Method, device, terminal and storage medium for determining face angle |
CN107909095A (en) * | 2017-11-07 | 2018-04-13 | 江苏大学 | A kind of image-recognizing method based on deep learning |
CN108596089A (en) * | 2018-04-24 | 2018-09-28 | 北京达佳互联信息技术有限公司 | Human face posture detection method, device, computer equipment and storage medium |
CN108960093A (en) * | 2018-06-21 | 2018-12-07 | 阿里体育有限公司 | The recognition methods and equipment of face's rotational angle |
CN109784202A (en) * | 2018-12-24 | 2019-05-21 | 珠海格力电器股份有限公司 | Recognition methods, device and the washing machine and computer readable storage medium of object are left in article to be washed |
CN109740492A (en) * | 2018-12-27 | 2019-05-10 | 郑州云海信息技术有限公司 | A kind of identity identifying method and device |
CN109948510A (en) * | 2019-03-14 | 2019-06-28 | 北京易道博识科技有限公司 | A kind of file and picture example dividing method and device |
CN109993150A (en) * | 2019-04-15 | 2019-07-09 | 北京字节跳动网络技术有限公司 | The method and apparatus at age for identification |
CN110175572A (en) * | 2019-05-28 | 2019-08-27 | 深圳市商汤科技有限公司 | Face image processing process and device, electronic equipment and storage medium |
CN110110712A (en) * | 2019-06-06 | 2019-08-09 | 江苏尚飞光电科技股份有限公司 | A kind of novel Activity recognition screening machine |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107301406A (en) | Fast face angle recognition method based on deep learning | |
CN105956560B (en) | A kind of model recognizing method based on the multiple dimensioned depth convolution feature of pondization | |
CN105975931B (en) | A kind of convolutional neural networks face identification method based on multiple dimensioned pond | |
CN107679522B (en) | Multi-stream LSTM-based action identification method | |
CN104866829B (en) | A kind of across age face verification method based on feature learning | |
CN103761536B (en) | Human face beautifying method based on non-supervision optimal beauty features and depth evaluation model | |
CN107506722A (en) | One kind is based on depth sparse convolution neutral net face emotion identification method | |
CN105139004A (en) | Face expression identification method based on video sequences | |
CN105205449B (en) | Sign Language Recognition Method based on deep learning | |
CN104021375B (en) | A kind of model recognizing method based on machine learning | |
CN106384094A (en) | Chinese word stock automatic generation method based on writing style modeling | |
CN110135277B (en) | Human behavior recognition method based on convolutional neural network | |
CN1710593A (en) | Hand-characteristic mix-together identifying method based on characteristic relation measure | |
CN109241995B (en) | Image identification method based on improved ArcFace loss function | |
CN103500340B (en) | Human body behavior identification method based on thematic knowledge transfer | |
CN110188669B (en) | Air handwritten character track recovery method based on attention mechanism | |
CN103226713B (en) | A kind of various visual angles Activity recognition method | |
CN108875819B (en) | Object and component joint detection method based on long-term and short-term memory network | |
CN102651072A (en) | Classification method for three-dimensional human motion data | |
CN107729890A (en) | Face identification method based on LBP and deep learning | |
CN104408470A (en) | Gender detection method based on average face preliminary learning | |
CN106599810A (en) | Head pose estimation method based on stacked auto-encoding | |
CN110942110A (en) | Feature extraction method and device of three-dimensional model | |
CN111259950B (en) | Method for training YOLO neural network based on 3D model | |
CN103927555A (en) | Static sign language letter recognition system and method based on Kinect sensor |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20171027 |
|
WD01 | Invention patent application deemed withdrawn after publication |