CN107844782A - A kind of face identification method based on the serial depth network of multitask - Google Patents
A kind of face identification method based on the serial depth network of multitask Download PDFInfo
- Publication number
- CN107844782A CN107844782A CN201711225933.0A CN201711225933A CN107844782A CN 107844782 A CN107844782 A CN 107844782A CN 201711225933 A CN201711225933 A CN 201711225933A CN 107844782 A CN107844782 A CN 107844782A
- Authority
- CN
- China
- Prior art keywords
- face
- positive
- image
- multitask
- side face
- 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
- G06V40/172—Classification, e.g. identification
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Multimedia (AREA)
- Human Computer Interaction (AREA)
- Oral & Maxillofacial Surgery (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
- Collating Specific Patterns (AREA)
Abstract
The present invention discloses a kind of face identification method based on the serial depth network of multitask, is related to computer identification field；The present invention carries out recognition of face using the identification model established：Gather image and carry out Face datection, the facial image detected is input in multitask serial network, the facial image of input classified using diversity classification task network, if classification results are positive face, exports recognition result；If classification results are side face, side face is then converted into positive face, the characteristics of recognition result is directed to non-controllable natural environment servant face image using the invention described above method is exported again, design effective face identification method, efficiently face can correctly be identified, available for fields such as financial authentication, intelligent security guard, ecommerce, product competitiveness is improved simultaneously, more contribute to company to obtain good economic benefit and social benefit.
Description
Technical field
The present invention discloses a kind of face identification method, is related to computer identification field, specifically a kind of to be based on more
It is engaged in the face identification method of serial depth network.
Background technology
With the development of science and technology, the improvement of people's living standards, recognition of face is in fields such as finance, ecommerce, security protections
Application it is more and more wider, its market scale reaches nearly hundred billion ranks.Existing face identification method under controllable environment
Higher accuracy of identification can be reached, but under non-controllable natural environment, recognition performance still has much room for improvement.Different from controllable ring
Variation is presented in the image gathered under border, the facial image collected under non-controllable environment.For example, collected under natural environment
Facial image multiple angles be present.The facial image of multi-angle often result in it is similar between larger diversity, so as to reduce
The accuracy of identification of conventional method.Therefore, the present invention provides a kind of face identification method based on the serial depth network of multitask,
The characteristics of for non-controllable natural environment servant face image, effective face identification method is designed, can be efficiently to people
Face is correctly identified, available for fields such as financial authentication, intelligent security guard, ecommerce, improves product competitiveness simultaneously,
Company is more contributed to obtain good economic benefit and social benefit.
The content of the invention
The present invention provides a kind of face identification method based on the serial depth network of multitask, it is possible to increase the identification of side face
Precision, so as to improve the recognition of face precision under non-controllable natural scene, have broad application prospects.
Concrete scheme proposed by the present invention is：
A kind of face identification method based on the serial depth network of multitask：
Using the image collected as training set, the facial image detected is obtained, rower is entered according to the facial angle of training set
Note, labeled as positive face and side face, align face and side face carry out classification based training,
Side face is converted into positive face, so as to which the positive face after conversion is identified,
By the face image of all samples, it is trained using human face recognition model, establishes identification model,
Identification model is recycled to carry out recognition of face：Gather image and carry out Face datection, the facial image detected is input to
In multitask serial network, the facial image of input is classified using diversity classification task network, if classification results are
Positive face, then export recognition result；If classification results are side face, side face is converted into positive face, then export recognition result.
Described method carries out Face datection using MTCNN to the image collected.
Described method carries out classification instruction as face classification task model using Alexnet structures to the face of training set
Practice.
Side face is converted into positive face by described method using TP-GAN, so as to which the positive face after conversion is identified.
Described method is trained by the use of facenet as human face recognition model, establishes identification model.
Usefulness of the present invention is：
The present invention provides a kind of face identification method based on the serial depth network of multitask：
Using the image collected as training set, the facial image detected is obtained, rower is entered according to the facial angle of training set
Note, labeled as positive face and side face, align face and side face carry out classification based training,
Side face is converted into positive face, so as to which the positive face after conversion is identified,
By the face image of all samples, it is trained using human face recognition model, establishes identification model,
Identification model is recycled to carry out recognition of face：Gather image and carry out Face datection, the facial image detected is input to
In multitask serial network, the facial image of input is classified using diversity classification task network, if classification results are
Positive face, then export recognition result；If classification results are side face, side face is converted into positive face, then export recognition result.
The characteristics of being directed to non-controllable natural environment servant face image using the invention described above method, design effective face and know
Other method, efficiently face correctly can be identified, available for financial authentication, intelligent security guard, ecommerce
Deng field, product competitiveness is improved simultaneously, more contribute to company to obtain good economic benefit and social benefit.
Brief description of the drawings
Fig. 1 face identification process figures of the present invention；
Fig. 2 the inventive method flow charts.
Embodiment
The present invention provides a kind of face identification method based on the serial depth network of multitask：
Using the image collected as training set, the facial image detected is obtained, rower is entered according to the facial angle of training set
Note, labeled as positive face and side face, align face and side face carry out classification based training,
Side face is converted into positive face, so as to which the positive face after conversion is identified,
By the face image of all samples, it is trained using human face recognition model, establishes identification model,
Identification model is recycled to carry out recognition of face：Gather image and carry out Face datection, the facial image detected is input to
In multitask serial network, the facial image of input is classified using diversity classification task network, if classification results are
Positive face, then export recognition result；If classification results are side face, side face is converted into positive face, then export recognition result.
The present invention is further described with reference to accompanying drawing.
Using the inventive method,
The training stage of network is carried out first：
Diversity classification task network training：Using the image collected as training set, the image collected is entered using MTCNN
Row Face datection, the facial image detected is obtained, be marked according to the facial angle of training set, labeled as positive face and side
Face, positive face are labeled as 0, and side face is labeled as 1, by the use of Alexnet structures as face classification task model to the positive face of face with
Side face carries out classification based training,
Side face processing Task Network training：It is higher for the accuracy of identification of positive face, it is relatively low for the accuracy of identification of side face, in order to more
Good identification side face, positive face is converted into using TP-GAN by side face, so as to which the positive face after conversion is identified,
By the face image of all samples, it is trained by the use of facenet as human face recognition model, establishes the identification of the present invention
Model,
Identification model is recycled to carry out recognition of face：Face datection, the face that will be detected are carried out to collection image using MTCNN
Image is input in multitask serial network, and the facial image of input is classified using diversity classification task network, if
Classification results are positive face, then positive face are input in identification model, export recognition result；If classification results are side face, by side
Face is converted into positive face, the positive face of generation is input in identification model, then export recognition result.
Claims (5)
1. a kind of face identification method based on the serial depth network of multitask, it is characterized in that
Using the image collected as training set, the facial image detected is obtained, rower is entered according to the facial angle of training set
Note, labeled as positive face and side face, align face and side face carry out classification based training,
Side face is converted into positive face, so as to which the positive face after conversion is identified,
By the face image of all samples, it is trained using human face recognition model, establishes identification model,
Identification model is recycled to carry out recognition of face：Gather image and carry out Face datection, the facial image detected is input to
In multitask serial network, the facial image of input is classified using diversity classification task network, if classification results are
Positive face, then export recognition result；If classification results are side face, side face is converted into positive face, then export recognition result.
2. according to the method for claim 1, it is characterized in that carrying out Face datection to the image collected using MTCNN.
3. method according to claim 1 or 2, it is characterized in that being used as face classification task model by the use of Alexnet structures
Classification based training is carried out to the face of training set.
4. according to the method for claim 3, it is characterized in that side face is converted into positive face using TP-GAN, after to conversion
Positive face be identified.
5. according to the method for claim 4, it is characterized in that being trained by the use of facenet as human face recognition model, build
Vertical identification model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711225933.0A CN107844782A (en) | 2017-11-29 | 2017-11-29 | A kind of face identification method based on the serial depth network of multitask |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711225933.0A CN107844782A (en) | 2017-11-29 | 2017-11-29 | A kind of face identification method based on the serial depth network of multitask |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107844782A true CN107844782A (en) | 2018-03-27 |
Family
ID=61680485
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711225933.0A Pending CN107844782A (en) | 2017-11-29 | 2017-11-29 | A kind of face identification method based on the serial depth network of multitask |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107844782A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109784243A (en) * | 2018-12-29 | 2019-05-21 | 网易（杭州）网络有限公司 | Identity determines method and device, neural network training method and device, medium |
| CN110197125A (en) * | 2019-05-05 | 2019-09-03 | 上海资汇信息科技有限公司 | Face identification method under unconfined condition |
| CN110399811A (en) * | 2019-07-08 | 2019-11-01 | 厦门市美亚柏科信息股份有限公司 | A kind of face identification method, device and storage medium |
| CN110751009A (en) * | 2018-12-20 | 2020-02-04 | 北京嘀嘀无限科技发展有限公司 | Face recognition method, target recognition device and electronic equipment |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105138956A (en) * | 2015-07-22 | 2015-12-09 | 小米科技有限责任公司 | Face detection method and device |
| US20160148080A1 (en) * | 2014-11-24 | 2016-05-26 | Samsung Electronics Co., Ltd. | Method and apparatus for recognizing object, and method and apparatus for training recognizer |
| CN105654049A (en) * | 2015-12-29 | 2016-06-08 | 中国科学院深圳先进技术研究院 | Facial expression recognition method and device |
| CN106203395A (en) * | 2016-07-26 | 2016-12-07 | 厦门大学 | Face character recognition methods based on the study of the multitask degree of depth |
| CN106650558A (en) * | 2015-11-04 | 2017-05-10 | 上海市公安局刑事侦查总队 | Facial recognition method and device |
| CN107392118A (en) * | 2017-07-04 | 2017-11-24 | 竹间智能科技（上海）有限公司 | The recognition methods of reinforcing face character and the system of generation network are resisted based on multitask |
-
2017
- 2017-11-29 CN CN201711225933.0A patent/CN107844782A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160148080A1 (en) * | 2014-11-24 | 2016-05-26 | Samsung Electronics Co., Ltd. | Method and apparatus for recognizing object, and method and apparatus for training recognizer |
| CN105138956A (en) * | 2015-07-22 | 2015-12-09 | 小米科技有限责任公司 | Face detection method and device |
| CN106650558A (en) * | 2015-11-04 | 2017-05-10 | 上海市公安局刑事侦查总队 | Facial recognition method and device |
| CN105654049A (en) * | 2015-12-29 | 2016-06-08 | 中国科学院深圳先进技术研究院 | Facial expression recognition method and device |
| CN106203395A (en) * | 2016-07-26 | 2016-12-07 | 厦门大学 | Face character recognition methods based on the study of the multitask degree of depth |
| CN107392118A (en) * | 2017-07-04 | 2017-11-24 | 竹间智能科技（上海）有限公司 | The recognition methods of reinforcing face character and the system of generation network are resisted based on multitask |
Non-Patent Citations (2)
| Title |
|---|
| RUI HUANG, SHU ZHANG, TIANYU LI, RAN HE: "Beyond Face Rotation: Global and Local Perception GAN for Photorealistic and Identity Preserving Frontal View Synthesis", 《2017 ICCV》 * |
| XI YIN, XIAOMING LIU: "Multi-Task Convolutional Neural Network for Pose-Invariant Face Recognition", 《COMPUTER VISION AND PATTERN RECOGNITION》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110751009A (en) * | 2018-12-20 | 2020-02-04 | 北京嘀嘀无限科技发展有限公司 | Face recognition method, target recognition device and electronic equipment |
| CN109784243A (en) * | 2018-12-29 | 2019-05-21 | 网易（杭州）网络有限公司 | Identity determines method and device, neural network training method and device, medium |
| CN109784243B (en) * | 2018-12-29 | 2021-07-09 | 网易（杭州）网络有限公司 | Identity determination method and device, neural network training method and device, and medium |
| CN110197125A (en) * | 2019-05-05 | 2019-09-03 | 上海资汇信息科技有限公司 | Face identification method under unconfined condition |
| CN110399811A (en) * | 2019-07-08 | 2019-11-01 | 厦门市美亚柏科信息股份有限公司 | A kind of face identification method, device and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107844782A (en) | A kind of face identification method based on the serial depth network of multitask | |
| CN105718889B (en) | Based on GB (2D)2The face personal identification method of PCANet depth convolution model | |
| CN106056082B (en) | A kind of video actions recognition methods based on sparse low-rank coding | |
| Schaefer et al. | Multi-dimensional local binary pattern descriptors for improved texture analysis | |
| CN110472652B (en) | Small sample classification method based on semantic guidance | |
| CN107122375A (en) | The recognition methods of image subject based on characteristics of image | |
| CN104156707A (en) | Fingerprint identification method and fingerprint identification device | |
| CN104134061A (en) | Number gesture recognition method for support vector machine based on feature fusion | |
| CN105512620A (en) | Convolutional neural network training method and apparatus for face recognition | |
| Zhang et al. | Structured weak semantic space construction for visual categorization | |
| CN101996328A (en) | Wood identification method | |
| Ramya et al. | Leaf disease detection and classification using neural networks | |
| CN104966101A (en) | Solar cell classification method based on LabVIEW | |
| CN106651988A (en) | Automatic drawing system for face line paint | |
| Lu et al. | Hand motion with eyes-free interaction for authentication in virtual reality | |
| Vithlani et al. | Structural and statistical feature extraction methods for character and digit recognition | |
| Sainin et al. | Malaysian medicinal plant leaf shape identification and classification | |
| Azimi et al. | Performance evalution of 3D keypoint detectors and descriptors for plants health classification | |
| CN108764116A (en) | A kind of face identification system and method based on depth cost-sensitive | |
| Lin et al. | Masked face recognition via self-attention based local consistency regularization | |
| CN104298975B (en) | A kind of distributed image recognition methods | |
| Deufemia et al. | Segmentation and recognition of petroglyphs using generic Fourier descriptors | |
| Kim et al. | A Study on Face Recognition using Support Vector Machine | |
| CN110032662A (en) | Video image distortion effect model building method based on improved dice loss function | |
| Li et al. | Analysis of AdaBoost-based face detection algorithm |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180327 |
|
| RJ01 | Rejection of invention patent application after publication |