CN108805094A - Data enhancement methods based on artificial face - Google Patents
Data enhancement methods based on artificial face Download PDFInfo
- Publication number
- CN108805094A CN108805094A CN201810629048.7A CN201810629048A CN108805094A CN 108805094 A CN108805094 A CN 108805094A CN 201810629048 A CN201810629048 A CN 201810629048A CN 108805094 A CN108805094 A CN 108805094A
- Authority
- CN
- China
- Prior art keywords
- image
- face
- data
- data set
- artificial
- 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
- G06V40/168—Feature extraction; Face representation
- G06V40/171—Local features and components; Facial parts ; Occluding parts, e.g. glasses; Geometrical relationships
-
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Oral & Maxillofacial Surgery (AREA)
- General Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Human Computer Interaction (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The embodiment of the present invention discloses a kind of data enhancement methods based on artificial face, can more effectively extract expressive features, improves generalization ability.This method includes:Obtain facial expression image data set;Image data in the data set is expanded using artificial face data enhancing mechanism, pretreatment operation is carried out to image data in the data set;Image face characteristic in the data set is trained on depth convolutional neural networks using the method based on ROI;According to the face structure of face, single image different zones in the data set are divided, different interest regions is set, image in the data set is expanded.
Description
Technical field
The present invention relates to field of image recognition more particularly to a kind of data enhancement methods based on artificial face.
Background technology
Facial expression recognition is the hot research problem of computer vision, pattern-recognition, artificial intelligence field, face
Expression can convey very abundant emotion information, and as computer technology is in the universal of people's daily life, human face expression is known
Application prospect not in fields such as human-computer interaction, home entertaining, public safeties even medical treatment is more extensive.
Traditional facial expression recognition extracts the feature of sample to be tested first, then with training sample carry out pattern classification and
Matching, recognition effect depend on the quality of feature, and people spend great effort to find the Expressive Features that can have discrimination, but
Be but encounter bottleneck at many aspects, such as the environmental changes such as description manually extracted is illuminated by the light, deformation, angle influence compared with
Greatly, anti-interference ability is weak, and effect differs in different identification missions, and transplantability is poor.
Invention content
The embodiment of the present invention provides a kind of data enhancement methods based on artificial face, can more effectively extract expression spy
Sign improves generalization ability.
The embodiment of the present invention adopts the following technical scheme that:
A kind of data enhancement methods based on artificial face, including:
Obtain facial expression image data set;
Image data in the data set is expanded using artificial face data enhancing mechanism, to scheming in the data set
As data carry out pretreatment operation;
Image face characteristic in the data set is trained on depth convolutional neural networks using the method based on ROI;
According to the face structure of face, single image different zones in the data set are divided, are arranged different
Image in the data set is expanded in interest region.
Optionally, the facial expression image data set is the CK+ data sets comprising multiple facial images, the acquisition face
Portion's facial expression image data set includes:
By CK+ data sets altogether according to 6:1:3 ratio cut partition training set verifies collection, test set, and ensures in each set
Piece identity is not overlapped.
Optionally, the use artificial face data enhancing mechanism, which to image data in the data set expand, includes:
To all image zooming-out face key points in the CK+ training sets, 68 spies of face are detected using the libraries dlib
Point is levied, every image can be indicated with the label matrix of a 68*2.
Optionally, the use artificial face data enhancing mechanism, which to image data in the data set expand, includes:
Image is divided into multiple expression classifications in the data set, and in same expression classification, artificial face number is used to image A
Expanded according to Enhancement Method;
Arbitrarily choose an image B makes the corresponding label matrix Q of image B using operations such as rotation, scaling and scales
The corresponding label matrix P of image A are adapted to as far as possible, and image B is mapped on image A;
By the Gaussian Blur with image B divided by image B, the Gaussian Blur for being then multiplied by image A realizes the color of image B
It balances and matches with image A.
Optionally, the use artificial face data enhancing mechanism, which to image data in the data set expand, includes:
Three types shade is generated using label matrix, polygonal region that respectively eyes+nose surrounds, face+under
Bar polygonal region surrounded and full face region;
Three kinds of local features of image B are mixed into figure by the final display portion that image A and image B is selected using shade
As in A, thus generating three corresponding artificial faces.
Optionally, the use artificial face data enhancing mechanism, which to image data in the data set expand, includes:
The eyebrow of another face is merged on a face to nasal portion;
Merge face+chin portion of another face;
68 characteristic points for merging another pictures constitute whole human face regions, can training set data be extended to N*
N*3-2N。
Optionally, described to include to image data progress pretreatment operation in the data set:
Face datection is carried out to every facial image by the Adaboost method for detecting human face based on Haar features, is cut
Human face region removes background influence;
Spatial normalization is carried out to image using opencv visions library, the line adjusted between face two is allowed to keep water
It is flat, face is snapped into same position;
Histogram equalization is carried out to all images, enhances the contrast of image, brightness of image caused by weakening illumination is poor
Influence;
By all image normalizations to 256*256 pixels.
Optionally, the application trains image in the data set based on the method for ROI on depth convolutional neural networks
Face characteristic includes:
According to face face structure, training set and verification collection facial image are divided into 7 ROI interest regions, respectively
Left eye, right eye, nose, face, eyes+nose, nose+face, full face, cutting scheme pay close attention to a nose mouth in different expressions
In difference.
Optionally, further include:
Use the AlexNet (exclusive noun) of the pre-training on ImageNet (computer vision system identification project name)
Convolutional neural networks model (AlexNet-CNN), is finely adjusted parameter on training set, and initial learning rate is 0.001, instruction
White silk integrates the corresponding class label of ROI image as the label of original image.
Optionally, further include:
Test image is divided into identical 7 ROI regions when test, uses trained AlexNet-CNN models pair 7
ROI image is opened to be differentiated;
The most differentiation result of number get tickets as finally to the recognition result of the test image.
The data enhancement methods based on artificial face based on the above-mentioned technical proposal obtain facial expression image data set, make
Manually face data enhancing mechanism concentrates image data to expand data, concentrates image data to carry out pretreatment behaviour data
Make, image face characteristic in the data set is trained on depth convolutional neural networks using the method based on ROI, according to face
Face structure, to data concentrate single image different zones divide, different interest regions is set, to data concentrate scheme
As being expanded, expressive features can be more effectively extracted to realize, improve generalization ability.
It should be understood that above general description and following detailed description is only exemplary and explanatory, not
The disclosure can be limited.
Description of the drawings
The drawings herein are incorporated into the specification and forms part of this specification, and shows the implementation for meeting the present invention
Example, and be used to explain the principle of the present invention together with specification.
Fig. 1 is the flow chart of the data enhancement methods based on artificial face shown in the embodiment of the present invention.
Specific implementation mode
Example embodiments are described in detail here, and the example is illustrated in the accompanying drawings.Following description is related to
When attached drawing, unless otherwise indicated, the same numbers in different drawings indicate the same or similar elements.Following exemplary embodiment
Described in embodiment do not represent and the consistent all embodiments of the present invention.On the contrary, they be only with it is such as appended
The example of the consistent device and method of some aspects being described in detail in claims, of the invention.
The embodiment of the present invention is insufficient for current public data collection data scale, and deep learning method is in limited trained number
According to the upper feature extraction scarce capacity of collection, lead to that generalization ability is poor in practice, robustness is low, proposes that one kind is novel
Data based on artificial face enhance mechanism, by carrying out Data expansion to presently disclosed standard database, CNN is made to count greatly
Model training and arameter optimization are more effectively carried out on the basis of;Extract a kind of improved convolutional Neural net based on ROI region
Network model introduces region of interest ROI and voting mechanism based on face structure priori, and structure is a kind of for face
The deep learning of expression classification trains improvement project, and the test method on the depth convolutional neural networks model of several mainstreams
Validity.
Embodiment 1
As shown in Figure 1, the embodiment of the present invention provides a kind of data enhancement methods based on artificial face, including:
11, facial expression image data set is obtained;
12, image data in the data set is expanded using artificial face data enhancing mechanism, to the data set
Middle image data carries out pretreatment operation;
13, using the method based on ROI (Region of interest, interest region) on depth convolutional neural networks
Image face characteristic in the training data set;
14, according to the face structure of face, single image different zones in the data set are divided, setting is different
Interest region, image in the data set is expanded.
Optionally, the facial expression image data set is CK+ (the The Extended comprising multiple facial images
Cohn-Kanade Database) data set, the acquisition facial expression image data set includes:
By CK+ (The Extended Cohn-Kanade Database) data set altogether according to 6:1:3 ratio cut partition instruction
Practice collection, verification collection, test set, and ensures that piece identity is not overlapped in each set.
Optionally, the use artificial face data enhancing mechanism, which to image data in the data set expand, includes:
To all image zooming-out face key points in the CK+ training sets, 68 spies of face are detected using the libraries dlib
Point is levied, every image can be indicated with the label matrix of a 68*2.
Optionally, the use artificial face data enhancing mechanism, which to image data in the data set expand, includes:
Image is divided into multiple expression classifications in the data set, and in same expression classification, artificial face number is used to image A
Expanded according to Enhancement Method;
Arbitrarily choose an image B makes the corresponding label matrix Q of image B using operations such as rotation, scaling and scales
The corresponding label matrix P of image A are adapted to as far as possible, and image B is mapped on image A;
By the Gaussian Blur with image B divided by image B, the Gaussian Blur for being then multiplied by image A realizes the color of image B
It balances and matches with image A.
Optionally, the use artificial face data enhancing mechanism, which to image data in the data set expand, includes:
Three types shade, respectively eyes are generated using label matrix (the label matrix that arbitrary spy's key point is surrounded)
The polygonal region and full face region that polygonal region that+nose surrounds, face+chin surround;
Three kinds of local features of image B are mixed into figure by the final display portion that image A and image B is selected using shade
As in A, thus generating three corresponding artificial faces.
Optionally, the use artificial face data enhancing mechanism, which to image data in the data set expand, includes:
The eyebrow of another face is merged on a face to nasal portion;
Merge face+chin portion of another face;
68 characteristic points for merging another pictures constitute whole human face regions, can training set data be extended to N*
(data set can be extended to N*N to N*3-2N by a type of shade, and data set can be extended to N* by two kinds of shade
N*2-N, similarly the shade of K types data set can be extended to K*N*N- (K-1) * N).
Optionally, described to include to image data progress pretreatment operation in the data set:
By being based on Adaboost (exclusive noun) method for detecting human face of Haar (exclusive noun) feature to every face
Image carries out Face datection, cuts human face region, removes background influence;
Spatial normalization is carried out to image using opencv visions library, the line adjusted between face two is allowed to keep water
It is flat, face is snapped into same position;
Histogram equalization is carried out to all images, enhances the contrast of image, brightness of image caused by weakening illumination is poor
Influence;
By all image normalizations to 256*256 pixels.
Optionally, the application trains image in the data set based on the method for ROI on depth convolutional neural networks
Face characteristic includes:
According to face face structure, training set and verification collection facial image are divided into 7 ROI interest regions, respectively
Left eye, right eye, nose, face, eyes+nose, nose+face, full face, cutting scheme pay close attention to a nose mouth in different expressions
In difference.
Optionally, further include:
Using the AlexNet convolutional neural networks model (AlexNet-CNN) of the pre-training on ImageNet, in training set
On parameter is finely adjusted, initial learning rate is 0.001, and training set ROI image corresponding class label is original image
Label.
Optionally, further include:
Test image is divided into identical 7 ROI regions when test, uses trained AlexNet-CNN models pair 7
ROI image is opened to be differentiated;
The most differentiation result of number get tickets as finally to the recognition result of the test image.
Data enhancement methods of the embodiment of the present invention based on artificial face obtain facial expression image data set, using artificial
Face data enhance mechanism and concentrate image data to expand data, concentrate image data to carry out pretreatment operation data, answer
Image face characteristic in the data set is trained on depth convolutional neural networks with the method based on ROI, according to the face of face
Portion's structure, to data concentrate single image different zones divide, different interest regions is set, to data concentrate image into
Row expands, and can more effectively extract expressive features to realize, improve generalization ability.
The embodiment of the present invention and general different, the present invention based on the data enhancement methods of operations such as rotating, blocking, translate
A kind of artificial face data enhancing mechanism for being absorbed in expressive features is proposed, by the characteristic area pair for merging different facial images
Data set is expanded, and effective solution leads to deep neural network due to presently disclosed human face expression data set deficiency
Model cannot effectively extract expressive features, and model overfitting problem is serious, in actual use to the generalization ability of new data
The problems such as poor.By exploring a kind of rationally effective data enhancement methods, meets convolutional neural networks and amount of training data is wanted
It asks, to the robustness and generalization ability of lift scheme.
Several interest regions are arranged according to face face structure, by facial image in the embodiment of the present invention, use ROI image
Training convolutional neural networks model, actively guide CNN concern with the relevant characteristic area of expression shape change, excavate ROI region between
Distributed expression it is special, help to enhance the reliability to predicting target, while ROI methods can also regard a kind of data enhancing as
Mode, such method also can lift scheme to a certain extent robustness and generalization ability.
The multiplication of ROI data of the embodiment of the present invention is directed to the training stage, and test phase most straightforward approach is to test image
Directly differentiate, but such method can waste the distributed expression characteristic about ROI region remembered in model, the present invention is knowing
It is also improved in other method, proposes a kind of method of discrimination based on ballot detection, it is identical by being divided to test image
ROI interest region votes to the differentiation result in interest region using model, and the classification for selecting number of votes obtained most is as final
Recognition result.
Embodiment 2
The data enhancement methods based on artificial face that the present embodiment the present invention will be described in detail embodiment provides, this method include
Following steps:
201, CK+ data sets totally 510 facial images are obtained, by data set according to 6:1:3 ratio cut partition training set is tested
Card collection, test set, and ensure that piece identity is not overlapped in each set.
202, to all image zooming-out face key points in CK+ training sets, 68 spies of face are detected using the libraries dlib
Point is levied, every image can be indicated with the label matrix of a 68*2.
203, in same expression classification, image A is expanded using artificial face data enhancement methods, arbitrarily chooses one
Image B is opened, using operations such as rotation, scaling and scales, so that the corresponding label matrix Q of image B is adapted to image A as far as possible and corresponds to
Label matrix P, image B is mapped on image A.
204, the color balance of adjustment image B, is allowed to match with image A, can be by with image B's divided by image B
Gaussian Blur, the Gaussian Blur for being then multiplied by image A are realized.
205, three types shade, polygonal region that respectively eyes+nose surrounds, face are generated using label matrix
The polygonal region and full face region that+chin surrounds.
206, using shade select image A and image B which should be partly the image finally shown, can will scheme
As three kinds of local features of B are mixed into image A, three corresponding artificial faces are thus generated.
207, it concentrates all images to expand artificial face using the method data, another face is merged on a face
Eyebrow to nasal portion;Merge face+chin portion of another face;Merge 68 characteristic point institute structures of another pictures
At whole human face regions, training set data can be extended to N*N*3-2N.
208, all images concentrated to data pre-process, and pass through the Adaboost faces based on Haar features first
Detection method carries out Face datection to every face picture, cuts human face region, removes background influence;Use opencv visions library
Spatial normalization is carried out to image, the line between face two is adjusted and is allowed to holding level, face is snapped into same position;
Histogram equalization is carried out to all images, enhances the contrast of image, the influence of brightness of image difference caused by weakening illumination;Most
Afterwards by all image normalizations to 256*256 pixels.
209, according to face face structure, training set and verification collection facial image are divided into 7 ROI interest regions, point
Not Wei left eye, right eye, nose, face, eyes+nose, nose+face, full face, cutting scheme pays close attention to a nose mouth in difference
Difference in expression, ROI methods allow training data to expand 7 times again.
210, it using the AlexNet convolutional neural networks model (AlexNet-CNN) of the pre-training on ImageNet, is instructing
Practice and parameter is finely adjusted on collection, initial learning rate is 0.001, and the corresponding class label of training set ROI image is original graph
The label of picture.
211, test phase, is equally divided into identical 7 ROI regions by test image, and use is trained
7 ROI images of AlexNet-CNN models pair differentiate that several most differentiation results of getting tickets are as finally to the test image
Recognition result.
Data enhancement methods of the embodiment of the present invention based on artificial face obtain facial expression image data set, using artificial
Face data enhance mechanism and concentrate image data to expand data, concentrate image data to carry out pretreatment operation data, answer
Image face characteristic in the data set is trained on depth convolutional neural networks with the method based on ROI, according to the face of face
Portion's structure, to data concentrate single image different zones divide, different interest regions is set, to data concentrate image into
Row expands, and can more effectively extract expressive features to realize, improve generalization ability.
The embodiment of the present invention and general different, the present invention based on the data enhancement methods of operations such as rotating, blocking, translate
A kind of artificial face data enhancing mechanism for being absorbed in expressive features is proposed, by the characteristic area pair for merging different facial images
Data set is expanded, and effective solution leads to deep neural network due to presently disclosed human face expression data set deficiency
Model cannot effectively extract expressive features, and model overfitting problem is serious, in actual use to the generalization ability of new data
The problems such as poor.By exploring a kind of rationally effective data enhancement methods, meets convolutional neural networks and amount of training data is wanted
It asks, to the robustness and generalization ability of lift scheme.
Several interest regions are arranged according to face face structure, by facial image in the embodiment of the present invention, use ROI image
Training convolutional neural networks model, actively guide CNN concern with the relevant characteristic area of expression shape change, excavate ROI region between
Distributed expression it is special, help to enhance the reliability to predicting target, while ROI methods can also regard a kind of data enhancing as
Mode, such method also can lift scheme to a certain extent robustness and generalization ability.
The multiplication of ROI data of the embodiment of the present invention is directed to the training stage, and test phase most straightforward approach is to test image
Directly differentiate, but such method can waste the distributed expression characteristic about ROI region remembered in model, the present invention is knowing
It is also improved in other method, proposes a kind of method of discrimination based on ballot detection, it is identical by being divided to test image
ROI interest region votes to the differentiation result in interest region using model, and the classification for selecting number of votes obtained most is as final
Recognition result.
The embodiment of the present invention expands CK+ training sets using artificial face data enhancement methods, can put forward data volume
42K are raised to, while CNN is trained using the training method based on ROI, data volume is promoted to 284K, enhanced data set
Performance on trained convolutional neural networks is better than raw data set, test set is identified by ROI voting mechanisms, in CK
Discrimination can be promoted 5% or so on+data set;The generalization ability for introducing cross datasets experimental verification model simultaneously, passes through
The face picture for largely carrying affective characteristics is collected in internet, is manually marked, and structure one is different from regular face
The Wild static faces expression data collection of nature, the recognition capability for testing CNN models to random new data test table
It is bright on Wild data sets, artificial face data enhancement methods and the neural network model based on ROI can promote recognition effect
About 7.5%.
Various embodiments of the present invention are described above, above description is exemplary, and non-exclusive, and
It is not limited to disclosed each embodiment.Without departing from the scope and spirit of illustrated each embodiment, for this skill
Many modifications and changes will be apparent from for the those of ordinary skill in art field.The selection of term used herein, purport
In the principle, practical application or improvement to the technology in market for best explaining each embodiment, or make the art
Other those of ordinary skill can understand each embodiment disclosed herein.
Those skilled in the art will readily occur to its of the disclosure after considering specification and putting into practice disclosure disclosed herein
Its embodiment.This application is intended to cover any variations, uses, or adaptations of the disclosure, these modifications, purposes or
Person's adaptive change follows the general principles of this disclosure and includes the undocumented common knowledge in the art of the disclosure
Or conventional techniques.
Claims (10)
1. a kind of data enhancement methods based on artificial face, which is characterized in that including:
Obtain facial expression image data set;
Image data in the data set is expanded using artificial face data enhancing mechanism, to picture number in the data set
According to progress pretreatment operation;
Using the method based on interest region ROI, image face is special in the training data set on depth convolutional neural networks
Sign;
According to the face structure of face, single image different zones in the data set are divided, different interest is set
Image in the data set is expanded in region.
2. according to the method described in claim 1, it is characterized in that, the facial expression image data set is to include multiple faces
The CK+ data sets of image, the acquisition facial expression image data set include:
By CK+ data sets altogether according to 6:1:3 ratio cut partition training set verifies collection, test set, and ensures personage in each set
Identity is not overlapped.
3. according to the method described in claim 1, it is characterized in that, described enhance mechanism to the data using artificial face data
Concentration image data expand:
To all image zooming-out face key points in the CK+ training sets, 68 characteristic points of face are detected using the libraries dlib,
Every image can be indicated with the label matrix of a 68*2.
4. according to the method described in claim 3, it is characterized in that, described enhance mechanism to the data using artificial face data
Concentration image data expand:
Image is divided into multiple expression classifications in the data set, in same expression classification, is increased using artificial face data to image A
Strong method is expanded;
Arbitrarily choose an image B makes the corresponding label matrix Q of image B to the greatest extent may be used using operations such as rotation, scaling and scales
The corresponding label matrix P of image A are adapted to, image B is mapped on image A;
By the Gaussian Blur with image B divided by image B, the Gaussian Blur for being then multiplied by image A realizes the color balance of image B
And match with image A.
5. according to the method described in claim 4, it is characterized in that, described enhance mechanism to the data using artificial face data
Concentration image data expand:
Three types shade is generated using label matrix, polygonal region that respectively eyes+nose surrounds, face+chin enclose
At polygonal region and full face region;
Three kinds of local features of image B are mixed into image A by the final display portion that image A and image B is selected using shade
In, thus generate three corresponding artificial faces.
6. according to the method described in claim 5, it is characterized in that, described enhance mechanism to the data using artificial face data
Concentration image data expand:
The eyebrow of another face is merged on a face to nasal portion;
Merge face+chin portion of another face;
68 characteristic points for merging another pictures constitute whole human face regions, can training set data be extended to N*N*3-
2N。
7. according to the method described in claim 1, it is characterized in that, described pre-process image data in the data set
Operation includes:
Face datection is carried out to every facial image by the Adaboost method for detecting human face based on Haar features, cuts face
Region removes background influence;
Spatial normalization is carried out to image using opencv visions library, the line between face two is adjusted and is allowed to holding level,
Face is snapped into same position;
Histogram equalization is carried out to all images, enhances the contrast of image, the shadow of brightness of image difference caused by weakening illumination
It rings;
By all image normalizations to 256*256 pixels.
8. according to the method described in claim 1, it is characterized in that, the application based on the method for ROI in depth convolutional Neural
Image face characteristic includes in the training data set on network:
According to face face structure, training set and verification collection facial image are divided into 7 ROI interest regions, respectively left eye,
Right eye, nose, face, eyes+nose, nose+face, full face, cutting scheme pay close attention to a nose mouth in different expressions
Difference.
9. method according to any one of claim 1 to 8, which is characterized in that further include:
It is right on training set using the AlexNet convolutional neural networks model (AlexNet-CNN) of the pre-training on ImageNet
Parameter is finely adjusted, and initial learning rate is 0.001, and the corresponding class label of training set ROI image is the label of original image.
10. according to the method described in claim 9, it is characterized in that, further including:
Test image is divided into identical 7 ROI regions when test, uses 7, trained AlexNet-CNN models pair
ROI image is differentiated;
The most differentiation result of number get tickets as finally to the recognition result of the test image.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810629048.7A CN108805094A (en) | 2018-06-19 | 2018-06-19 | Data enhancement methods based on artificial face |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810629048.7A CN108805094A (en) | 2018-06-19 | 2018-06-19 | Data enhancement methods based on artificial face |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108805094A true CN108805094A (en) | 2018-11-13 |
Family
ID=64083271
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810629048.7A Pending CN108805094A (en) | 2018-06-19 | 2018-06-19 | Data enhancement methods based on artificial face |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108805094A (en) |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110008972A (en) * | 2018-11-15 | 2019-07-12 | 阿里巴巴集团控股有限公司 | Method and apparatus for data enhancing |
CN110210401A (en) * | 2019-06-03 | 2019-09-06 | 多维协同人工智能技术研究院(重庆)有限公司 | A kind of Intelligent target detection method under dim light |
CN110309349A (en) * | 2019-04-08 | 2019-10-08 | 浙江工业大学 | A kind of music generating method based on human facial expression recognition and Recognition with Recurrent Neural Network |
CN111159150A (en) * | 2019-12-19 | 2020-05-15 | 北京文安智能技术股份有限公司 | Data expansion method and device |
CN111178337A (en) * | 2020-01-07 | 2020-05-19 | 南京甄视智能科技有限公司 | Human face key point data enhancement method, device and system and model training method |
CN111666911A (en) * | 2020-06-13 | 2020-09-15 | 天津大学 | Micro-expression data expansion method and device |
CN111881720A (en) * | 2020-06-09 | 2020-11-03 | 山东大学 | Data automatic enhancement expansion method, data automatic enhancement identification method and data automatic enhancement expansion system for deep learning |
CN113112410A (en) * | 2020-01-10 | 2021-07-13 | 华为技术有限公司 | Data enhancement method and device, computing equipment, chip and computer storage medium |
CN113610049A (en) * | 2021-08-25 | 2021-11-05 | 云南电网有限责任公司电力科学研究院 | Mobile terminal face detection method |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8873813B2 (en) * | 2012-09-17 | 2014-10-28 | Z Advanced Computing, Inc. | Application of Z-webs and Z-factors to analytics, search engine, learning, recognition, natural language, and other utilities |
CN106570471A (en) * | 2016-10-26 | 2017-04-19 | 武汉科技大学 | Scale adaptive multi-attitude face tracking method based on compressive tracking algorithm |
CN107153816A (en) * | 2017-04-16 | 2017-09-12 | 五邑大学 | A kind of data enhancement methods recognized for robust human face |
CN107679526A (en) * | 2017-11-14 | 2018-02-09 | 北京科技大学 | A kind of micro- expression recognition method of face |
CN107767335A (en) * | 2017-11-14 | 2018-03-06 | 上海易络客网络技术有限公司 | A kind of image interfusion method and system based on face recognition features' point location |
CN108108677A (en) * | 2017-12-12 | 2018-06-01 | 重庆邮电大学 | One kind is based on improved CNN facial expression recognizing methods |
-
2018
- 2018-06-19 CN CN201810629048.7A patent/CN108805094A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8873813B2 (en) * | 2012-09-17 | 2014-10-28 | Z Advanced Computing, Inc. | Application of Z-webs and Z-factors to analytics, search engine, learning, recognition, natural language, and other utilities |
CN106570471A (en) * | 2016-10-26 | 2017-04-19 | 武汉科技大学 | Scale adaptive multi-attitude face tracking method based on compressive tracking algorithm |
CN107153816A (en) * | 2017-04-16 | 2017-09-12 | 五邑大学 | A kind of data enhancement methods recognized for robust human face |
CN107679526A (en) * | 2017-11-14 | 2018-02-09 | 北京科技大学 | A kind of micro- expression recognition method of face |
CN107767335A (en) * | 2017-11-14 | 2018-03-06 | 上海易络客网络技术有限公司 | A kind of image interfusion method and system based on face recognition features' point location |
CN108108677A (en) * | 2017-12-12 | 2018-06-01 | 重庆邮电大学 | One kind is based on improved CNN facial expression recognizing methods |
Non-Patent Citations (3)
Title |
---|
XIAO SUN等: "Improved facial expression recognition method based on ROI deep convolutional neutral network", 《2017 SEVENTH INTERNATIONAL CONFERENCE ON AFFECTIVE COMPUTING AND INTELLIGENT INTERACTION (ACII)》 * |
孙晓等: "基于ROI-KNN卷积神经网络的面部表情识别", 《自动化学报》 * |
董海鹰: "《智能控制理论及应用》", 30 September 2016 * |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110008972A (en) * | 2018-11-15 | 2019-07-12 | 阿里巴巴集团控股有限公司 | Method and apparatus for data enhancing |
CN110008972B (en) * | 2018-11-15 | 2023-06-06 | 创新先进技术有限公司 | Method and apparatus for data enhancement |
CN110309349A (en) * | 2019-04-08 | 2019-10-08 | 浙江工业大学 | A kind of music generating method based on human facial expression recognition and Recognition with Recurrent Neural Network |
CN110210401A (en) * | 2019-06-03 | 2019-09-06 | 多维协同人工智能技术研究院(重庆)有限公司 | A kind of Intelligent target detection method under dim light |
CN110210401B (en) * | 2019-06-03 | 2022-10-21 | 多维协同人工智能技术研究院(重庆)有限公司 | Intelligent target detection method under weak light |
CN111159150A (en) * | 2019-12-19 | 2020-05-15 | 北京文安智能技术股份有限公司 | Data expansion method and device |
CN111178337A (en) * | 2020-01-07 | 2020-05-19 | 南京甄视智能科技有限公司 | Human face key point data enhancement method, device and system and model training method |
CN113112410A (en) * | 2020-01-10 | 2021-07-13 | 华为技术有限公司 | Data enhancement method and device, computing equipment, chip and computer storage medium |
CN111881720A (en) * | 2020-06-09 | 2020-11-03 | 山东大学 | Data automatic enhancement expansion method, data automatic enhancement identification method and data automatic enhancement expansion system for deep learning |
CN111881720B (en) * | 2020-06-09 | 2024-01-16 | 山东大学 | Automatic enhancement and expansion method, recognition method and system for data for deep learning |
CN111666911A (en) * | 2020-06-13 | 2020-09-15 | 天津大学 | Micro-expression data expansion method and device |
CN113610049A (en) * | 2021-08-25 | 2021-11-05 | 云南电网有限责任公司电力科学研究院 | Mobile terminal face detection method |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108805094A (en) | Data enhancement methods based on artificial face | |
Lin | Face detection in complicated backgrounds and different illumination conditions by using YCbCr color space and neural network | |
Kramer et al. | Robust social categorization emerges from learning the identities of very few faces. | |
CN109815801A (en) | Face identification method and device based on deep learning | |
CN107194347A (en) | A kind of method that micro- expression detection is carried out based on Facial Action Coding System | |
CN106326874A (en) | Method and device for recognizing iris in human eye images | |
Deshmukh et al. | Facial emotion recognition system through machine learning approach | |
Hebbale et al. | Real time COVID-19 facemask detection using deep learning | |
Wang et al. | Forgerynir: deep face forgery and detection in near-infrared scenario | |
Singh et al. | Facial emotion recognition using convolutional neural network | |
CN109614927A (en) | Micro- Expression Recognition based on front and back frame difference and Feature Dimension Reduction | |
Gorbova et al. | Going deeper in hidden sadness recognition using spontaneous micro expressions database | |
Zhang et al. | Facial action unit detection with local key facial sub-region based multi-label classification for micro-expression analysis | |
Volokitin et al. | Predicting when saliency maps are accurate and eye fixations consistent | |
Faria et al. | Interface framework to drive an intelligent wheelchair using facial expressions | |
Sharma et al. | Indian face age database: A database for face recognition with age variation | |
Schwaninger et al. | Computational modeling of face recognition based on psychophysical experiments | |
Pan et al. | Automated drawing psychoanalysis via house-tree-person test | |
Wu et al. | Spontaneous versus posed smile recognition via region-specific texture descriptor and geometric facial dynamics | |
Raiturkar et al. | Identifying computer-generated portraits: an eye tracking study | |
Musa | Facial Emotion Detection for Educational Purpose Using Image Processing Technique | |
Wang et al. | Visual attention is attracted by text features even in scenes without text | |
Sharma et al. | A state-of-the-art real-time face detection, tracing and recognition system | |
Strezoski et al. | OmniEyes: Analysis and synthesis of artistically painted eyes | |
Tsank | Face Perception: The Interaction of Eye Movements with Internal Face Representations |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181113 |