CN107729796A - A kind of face picture detection method and device - Google Patents
A kind of face picture detection method and device Download PDFInfo
- Publication number
- CN107729796A CN107729796A CN201610663327.6A CN201610663327A CN107729796A CN 107729796 A CN107729796 A CN 107729796A CN 201610663327 A CN201610663327 A CN 201610663327A CN 107729796 A CN107729796 A CN 107729796A
- Authority
- CN
- China
- Prior art keywords
- picture
- graders
- face
- cnn
- detected
- 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
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- 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)
- Image Processing (AREA)
Abstract
The present invention relates to technical field of image processing, more particularly to a kind of face picture detection method and device, including:According to cascade graders, judge whether picture to be detected is face picture, if determine that picture to be detected is face picture according to cascade graders, then according to CNN graders, judge whether picture to be detected is face picture, if determine that picture to be detected is face picture according to CNN graders, it is determined that picture to be detected is face picture.The embodiment of the present invention filters out non-face picture by cascade graders first, therefore reduces the picture number of CNN detection of classifier, can effectively improve the bulk velocity of CNN graders;And detected before CNN detection of classifier by cascade graders, thus the degree of accuracy of detection can be effectively improved.
Description
Technical field
The present invention relates to technical field of image processing, more particularly to a kind of face picture detection method and device.
Background technology
Face picture detection is a key link in face identification system.The face picture detection research of early stage is main
For the face picture (such as picture without background) with compared with Condition of Strong Constraint, often assume that face location obtains always or easily
, therefore face picture test problems are not taken seriously.With the development of the applications such as ecommerce, recognition of face, which turns into, most to be had
The biometric verification of identity means of potentiality, it is certain that this application background requires that Automatic face recognition system can have to general picture
Recognition capability.
There are a variety of methods to realize that face picture detects at present, wherein with deep learning algorithm, particularly CNN
(Convolutional Neural Network, convolutional neural networks) grader is the most popular and effectively, when being detected
Picture amount when being not very big, CNN graders can complete the detection classification of face picture well.But because CNN classifies
Device face picture detection application in be primarily present the problem of be:When the picture amount detected is bigger, it will cause to examine
Degree of testing the speed declines;And simple carries out face picture detection using CNN graders, and it is not very that can have recognition accuracy yet
It is high.
In summary, it is not fast enough and accurate to there is speed when carrying out recognition of face using CNN graders in prior art
Degree is not extra high technical problem.
The content of the invention
The present invention provides a kind of face picture detection method and device, to solve to use CNN present in prior art
Speed is not fast enough during grader progress recognition of face and the degree of accuracy is not extra high technical problem.
On the one hand, the embodiment of the present invention provides a kind of face picture detection method, including:
According to cascade graders, judge whether picture to be detected is face picture, the cascade graders are roots
Train what is obtained according to previously selected face samples pictures and non-face samples pictures;
If determine that the picture to be detected is face picture according to the cascade graders, according to convolutional Neural net
Network CNN graders, judge whether the picture to be detected is face picture, the CNN graders are according to the selected people
The samples pictures correctly classified by the cascade graders in face samples pictures and the non-face samples pictures train to obtain
's;
If determine that the picture to be detected is face picture according to the CNN graders, it is determined that the picture to be detected
For face picture.
Alternatively, it is described judge whether picture to be detected is face picture after, in addition to:If according to the cascade
Grader determines that the picture to be detected is non-face picture, it is determined that the picture to be detected is non-face picture.
Alternatively, it is described according to CNN graders, after judging whether the picture to be detected is face picture, in addition to:
If determine that the picture to be detected is non-face picture according to the CNN graders, it is determined that the picture to be detected is inhuman
Face picture.
Alternatively, trained to obtain the CNN graders according to following method:
According to the previously selected face samples pictures and non-face samples pictures, the CNN graders are instructed
Practice, obtain the first CNN graders, the first CNN graders include Softmax determinants;
The Softmax determinants in the first CNN graders are replaced with into support vector machines determinant;
Classified according in the previously selected face samples pictures and the non-face samples pictures by the cascade
The samples pictures that device is correctly classified, the first CNN graders are trained, obtain the 2nd CNN graders;
The CNN graders that the 2nd CNN graders are obtained as training.
Alternatively, the determinant in the CNN graders is following any:Softmax determinants, SVM determinants.
On the other hand, the embodiment of the present invention provides a kind of face picture detection means, including:
First facial image detection unit, for according to cascade graders, judging whether picture to be detected is face figure
Piece, the cascade graders train to obtain according to previously selected face samples pictures and non-face samples pictures;
Second facial image detection unit, if for determining that the picture to be detected is according to the cascade graders
Face picture, then according to convolutional neural networks CNN graders, judge whether the picture to be detected is face picture, the CNN
Grader be according in the selected face samples pictures and the non-face samples pictures by the cascade graders just
The samples pictures really classified train what is obtained;
Face picture determining unit, if for determining that the picture to be detected is face picture according to the CNN graders,
It is face picture then to determine the picture to be detected.
Alternatively, the face picture determining unit, is additionally operable to:If treated according to determining the cascade graders
Detection picture is non-face picture, it is determined that the picture to be detected is non-face picture.
Alternatively, the face picture determining unit, is additionally operable to:If determined according to the CNN graders described to be detected
Picture is non-face picture, it is determined that the picture to be detected is non-face picture.
Alternatively, described device also includes training unit, for being trained to obtain the CNN graders according to following method:
According to the previously selected face samples pictures and non-face samples pictures, the CNN graders are instructed
Practice, obtain the first CNN graders, the first CNN graders include Softmax determinants;
The Softmax determinants in the first CNN graders are replaced with into support vector machines determinant;
Classified according in the previously selected face samples pictures and the non-face samples pictures by the cascade
The samples pictures that device is correctly classified, the first CNN graders are trained, obtain the 2nd CNN graders;
The CNN graders that the 2nd CNN graders are obtained as training.
Alternatively, the determinant in the CNN graders is following any:Softmax determinants, SVM determinants.
The embodiment of the present invention judges whether picture to be detected is face picture according to cascade graders, described
Cascade graders train to obtain according to previously selected face samples pictures and non-face samples pictures;If according to institute
State cascade graders and determine that the picture to be detected is face picture, then according to convolutional neural networks CNN graders, judge
Whether the picture to be detected is face picture, and the CNN graders are according to selected face samples pictures and described
The samples pictures correctly classified by the cascade graders in non-face samples pictures train what is obtained;If according to the CNN
Grader determines that the picture to be detected is face picture, it is determined that the picture to be detected is face picture.The present invention is implemented
Example filters out non-face picture by cascade graders first, therefore reduces the picture number of CNN detection of classifier, can
To effectively improve the bulk velocity of CNN graders;And detected before CNN detection of classifier by cascade graders,
The degree of accuracy of detection can thus be effectively improved.
Brief description of the drawings
Technical scheme in order to illustrate the embodiments of the present invention more clearly, make required in being described below to embodiment
Accompanying drawing is briefly introduced, it should be apparent that, drawings in the following description are only some embodiments of the present invention, for this
For the those of ordinary skill in field, without having to pay creative labor, it can also be obtained according to these accompanying drawings
His accompanying drawing.
Fig. 1 is a kind of face picture detection method flow chart provided in an embodiment of the present invention;
Fig. 2 is CNN graders configuration diagram provided in an embodiment of the present invention;
Fig. 3 is face picture detection framework schematic diagram provided in an embodiment of the present invention;
Fig. 4 is a kind of face picture detection method detail flowchart provided in an embodiment of the present invention;
Fig. 5 is a kind of face picture detection means schematic diagram provided in an embodiment of the present invention.
Embodiment
In order that the object, technical solutions and advantages of the present invention are clearer, the present invention is made below in conjunction with accompanying drawing into
One step it is described in detail, it is clear that described embodiment is only part of the embodiment of the present invention, rather than whole implementation
Example.Based on the embodiment in the present invention, what those of ordinary skill in the art were obtained under the premise of creative work is not made
All other embodiment, belongs to the scope of protection of the invention.
The embodiment of the present invention is described in further detail with reference to Figure of description.
As shown in figure 1, a kind of face picture detection method provided in an embodiment of the present invention, including:
Step 101, according to cascade graders, judge whether picture to be detected is face picture, the cascade points
Class device trains to obtain according to previously selected face samples pictures and non-face samples pictures;
If step 102, according to the cascade graders determine that the picture to be detected is face picture, according to volume
Product neutral net CNN graders, judge whether the picture to be detected is face picture, and the CNN graders are according to
The samples pictures correctly classified by the cascade graders in selected face samples pictures and the non-face samples pictures
What training obtained;
If step 103, according to the CNN graders determine that the picture to be detected is face picture, it is determined that described to treat
Detection picture is face picture.
Cascade graders are a kind of existing graders, can be previously according to needing to instruct cascade graders
Practice, the cascade graders after being trained, then sample data is divided using the cascade graders after training
Class.
CNN graders are also a kind of existing grader, as shown in Fig. 2 being CNN graders provided in an embodiment of the present invention
Configuration diagram, CNN graders are to be based on CNN (Convolutional Neural Network, convolutional neural networks) algorithm
A kind of obtained grader, and CNN graders are trained in advance (mainly the CNN shown in Fig. 2 classified as needed
Weight parameter in device is trained, the weight parameter after being trained), the CNN graders after being trained, wherein, in CNN
Included in grader
For example, for the structure of CNN graders, 5 layers of neural convolutional network as shown in Figure 2 be could be arranged to.Its
In, Image represents the face picture of input, and first layer convolutional layer C1 is 5x5 convolution kernel, and feature map (characteristic pattern) are
20 convolutional network layer (4 only having been drawn in figure, other ellipsis represent), then meets 2x2 max pooling (most
Bigization pond) layer S1, the second layer is similar with the composition of first layer with third layer, and difference is that second layer C2 feature map are 40,
Convolution kernel is 3x3, and third layer C3 feature map are 60, convolution kernel 3x3.Then output is sent into the output of third layer
For full connect (full connection) layer, i.e. FC levels of 128 dimensions.4th layer is the traditional two soft-max layers classified
(or referred to as soft-max determinants).Finally, the face picture of input final can be obtained by the CNN graders shown in Fig. 2
To classification results, you can be identified as face picture or be identified as non-face picture.
Face picture detection method provided in an embodiment of the present invention is illustrated with reference to Fig. 1 and Fig. 3.
In above-mentioned steps 101, first according to cascade graders, judge whether picture to be detected is face picture, its
In, the cascade graders train to obtain according to previously selected face samples pictures and non-face samples pictures.
It is face picture detection framework schematic diagram provided in an embodiment of the present invention, first by a mapping to be checked with reference to figure 3
Piece is input to cascade graders, and then cascade graders can be determined that the picture to be detected is face picture or right and wrong
Face picture.
If determine that the picture to be detected is face picture according to the cascade graders, by the picture to be detected
CNN graders are sent into be determined whether;If determine that the picture to be detected is non-face according to the cascade graders
Picture, it is determined that the picture to be detected is non-face picture.
In above-mentioned steps 102, when CNN graders receive the picture to be detected from cascade graders, then after
Continue according to CNN graders, judge whether the picture to be detected is face picture, if according to determining the CNN graders
Picture to be detected is face picture, then it is face picture that the picture to be detected is determined in above-mentioned steps 103;If according to described
CNN graders determine that the picture to be detected is non-face picture, it is determined that the picture to be detected is non-face picture.
I.e. in embodiments of the present invention, picture to be detected is differentiated according to cascade graders first, if
Cascade graders judge that picture to be detected is non-face picture, then it is non-face picture directly to determine the picture to be detected;If
Cascade graders judge that picture to be detected is face picture, then the picture to be detected is sent into CNN points by cascade graders
Class device is determined whether, if CNN graders judge that the picture to be detected is face picture, it is determined that the picture to be detected is behaved
Face picture, if CNN graders judge that the picture to be detected is non-face picture, it is determined that the picture to be detected is non-face figure
Piece.
The application scenarios for needing to judge substantial amounts of picture to be detected are can be seen that from above decision process
In, picture to be detected can once be filtered using cascade graders first, be clearly not non-face figure for some
The picture to be detected of piece, can directly be filtered out, and the characteristic due to cascade graders in itself, more difficult for some
The picture to be detected judged, cascade graders generally tend to picture to be detected being determined as face picture, that is to say, that
Cascade graders will not typically cause face picture being mistaken for non-face image filtering to fall, but may will be non-face
Picture is mistaken for face picture, and in other words, by the filtering of cascade graders, what final filtration fell is mostly non-face
Picture, and be determined as in all pictures of face picture, small portion (is accounted for by face picture (accounting for most of ratio) and non-face picture
Point) composition.
As an example it is assumed that one shares 10000 pictures to be detected, wherein there is 6000 face pictures, 4000 inhuman
Face picture, then eventually pass through the filtering of cascade graders, it may be possible to by 5990 correct identifications in 6000 face pictures
For face picture, non-face picture is mistaken for by remaining 10;It is and 3900 in 4000 non-face pictures are correct
Non-face picture is identified as, is face picture by remaining 100 wrong identification, therefore finally enter CNN graders is:
5990 face pictures correctly identified+100 are mistaken for the non-face picture of face picture, there it can be seen that cascade
Grader is that something lost miss out 10 face pictures, all pictures to be detected for being input to CNN graders, then can be passed through
CNN graders do further judgement.
Pass through the above method, on the one hand, most non-face picture can be filtered out by cascade graders, is carried
High overall recognition speed;On the other hand, picture to be detected is judged by binary classifier, is only all determined as face figure
Piece, just by picture recognition to be detected it is finally face picture, thus the degree of accuracy of identification can be improved.
The CNN graders be according in the selected face samples pictures and the non-face samples pictures by described
The samples pictures that cascade graders are correctly classified train what is obtained.
In CNN graders, determinant therein can be Softmax determinants, and Softmax determinants can also be entered
Row is replaced, such as could alternatively be SVM determinants.
Training method when below to Softmax determinants are replaced with into SVM determinants is described.
Step A, according to previously selected face samples pictures and non-face samples pictures, CNN graders are trained,
The first CNN graders are obtained, the first CNN graders include Softmax determinants;
Step B, the Softmax determinants in the first CNN graders are replaced with into support vector machines to judge
Device;
Step C, divided according in the previously selected face samples pictures and the non-face samples pictures by cascade
The samples pictures that class device is correctly classified, the first CNN graders are trained, obtain the 2nd CNN graders;
The CNN graders that the 2nd CNN graders are obtained as training.
In above-mentioned steps A, for example, assuming that previously selected face samples pictures are 5000, it is previously selected
Non-face samples pictures are 5000, then first according to this 5000 face samples pictures and 5000 non-face samples pictures pair
CNN graders are trained, and obtain the first CNN graders, and the first CNN graders include Softmax determinants.
In above-mentioned steps B, the Softmax determinants in the first CNN graders are replaced with into support vector machines
(Support Vector Machine, SVMs) determinant.
In above-mentioned steps C, it is assumed that 10000 pictures of above-mentioned selection, will by the classification of the cascade graders
5000 non-face picture recognitions therein are:4900 non-face pictures of face picture+100, then, will be upper in step C
The positive sample used in step A to 5000 face pictures as training CNN graders is stated, will be wherein by cascade graders
4900 non-face pictures of non-face picture are correctly identified as negative sample, then using selected positive sample and negative sample
CNN graders are trained, obtain the 2nd CNN graders, the institute for then obtaining the 2nd CNN graders as training
State CNN graders.
In the above-mentioned methods, due to being the output result for having used cascade graders when training CNN graders
As the training sample of CNN graders, thus the CNN graders that training obtains can be caused to have one with cascade graders
Fixed contact, so as to more accurately may finally correctly be classified to picture to be detected.
The embodiment of the present invention judges whether picture to be detected is face picture according to cascade graders, described
Cascade graders train to obtain according to previously selected face samples pictures and non-face samples pictures;If according to institute
State cascade graders and determine that the picture to be detected is face picture, then according to convolutional neural networks CNN graders, judge
Whether the picture to be detected is face picture, and the CNN graders are according to selected face samples pictures and described
The samples pictures correctly classified by the cascade graders in non-face samples pictures train what is obtained;If according to the CNN
Grader determines that the picture to be detected is face picture, it is determined that the picture to be detected is face picture.The present invention is implemented
Example filters out non-face picture by cascade graders first, therefore reduces the picture number of CNN detection of classifier, can
To effectively improve the bulk velocity of CNN graders;And detected before CNN detection of classifier by cascade graders,
The degree of accuracy of detection can thus be effectively improved.
A kind of face picture detection method provided in an embodiment of the present invention is described in detail below, as shown in figure 4, bag
Include:
Step 401, according to cascade graders, judge whether picture to be detected is face picture, if so, then going to step
Rapid 402, otherwise go to step 404;
Step 402, according to CNN graders, judge whether the picture to be detected is face picture, if so, then going to step
Rapid 403, otherwise go to step 404;
Step 403, determine that the picture to be detected is face picture;
Step 404, determine that the picture to be detected is non-face picture.
The embodiment of the present invention judges whether picture to be detected is face picture according to cascade graders, described
Cascade graders train to obtain according to previously selected face samples pictures and non-face samples pictures;If according to institute
State cascade graders and determine that the picture to be detected is face picture, then according to CNN graders, judge the mapping to be checked
Whether piece is face picture, and the CNN graders are according to the selected face samples pictures and the non-face sample graph
The samples pictures correctly classified by the cascade graders in piece train what is obtained;If institute is determined according to the CNN graders
It is face picture to state picture to be detected, it is determined that the picture to be detected is face picture.The embodiment of the present invention passes through first
Cascade graders filter out non-face picture, therefore reduce the picture number of CNN detection of classifier, can effectively improve
The bulk velocity of CNN graders;And detected, thus can be had by cascade graders before CNN detection of classifier
Effect improves the degree of accuracy of detection.
Based on identical technical concept, the embodiment of the present invention also provides a kind of face picture detection means, as shown in figure 5,
Including:
First facial image detection unit 501, for according to cascade graders, judging whether picture to be detected is people
Face picture, the cascade graders are to train to obtain according to previously selected face samples pictures and non-face samples pictures
's;
Second facial image detection unit 502, if for determining the picture to be detected according to the cascade graders
For face picture, then according to convolutional neural networks CNN graders, judge whether the picture to be detected is face picture, it is described
CNN graders are to be classified according in the selected face samples pictures and the non-face samples pictures by the cascade
The samples pictures that device is correctly classified train what is obtained;
Face picture determining unit 503, if for determining that the picture to be detected is face figure according to the CNN graders
Piece, it is determined that the picture to be detected is face picture.
Alternatively, the face picture determining unit 503, is additionally operable to:If according to determining the cascade graders
Picture to be detected is non-face picture, it is determined that the picture to be detected is non-face picture.
Alternatively, the face picture determining unit 503, is additionally operable to:If determined according to the CNN graders described to be checked
Mapping piece is non-face picture, it is determined that the picture to be detected is non-face picture.
Alternatively, described device also includes training unit 504, for being trained to obtain the CNN classification according to following method
Device:
According to the previously selected face samples pictures and non-face samples pictures, the CNN graders are instructed
Practice, obtain the first CNN graders, the first CNN graders include Softmax determinants;
The Softmax determinants in the first CNN graders are replaced with into support vector machines determinant;
Classified according in the previously selected face samples pictures and the non-face samples pictures by the cascade
The samples pictures that device is correctly classified, the first CNN graders are trained, obtain the 2nd CNN graders;
The CNN graders that the 2nd CNN graders are obtained as training.
Alternatively, the determinant in the CNN graders is following any:Softmax determinants, SVM determinants.
The embodiment of the present invention judges whether picture to be detected is face picture according to cascade graders, described
Cascade graders train to obtain according to previously selected face samples pictures and non-face samples pictures;If according to institute
State cascade graders and determine that the picture to be detected is face picture, then according to convolutional neural networks CNN graders, judge
Whether the picture to be detected is face picture, and the CNN graders are according to selected face samples pictures and described
The samples pictures correctly classified by the cascade graders in non-face samples pictures train what is obtained;If according to the CNN
Grader determines that the picture to be detected is face picture, it is determined that the picture to be detected is face picture.The present invention is implemented
Example filters out non-face picture by cascade graders first, therefore reduces the picture number of CNN detection of classifier, can
To effectively improve the bulk velocity of CNN graders;And detected before CNN detection of classifier by cascade graders,
The degree of accuracy of detection can thus be effectively improved.
The present invention is the flow with reference to method according to embodiments of the present invention, equipment (system) and computer program product
Figure and/or block diagram describe.It should be understood that can be by every first-class in computer program instructions implementation process figure and/or block diagram
Journey and/or the flow in square frame and flow chart and/or block diagram and/or the combination of square frame.These computer programs can be provided
The processors of all-purpose computer, special-purpose computer, Embedded Processor or other programmable data processing devices is instructed to produce
A raw machine so that produced by the instruction of computer or the computing device of other programmable data processing devices for real
The device for the function of being specified in present one flow of flow chart or one square frame of multiple flows and/or block diagram or multiple square frames.
These computer program instructions, which may be alternatively stored in, can guide computer or other programmable data processing devices with spy
Determine in the computer-readable memory that mode works so that the instruction being stored in the computer-readable memory, which produces, to be included referring to
Make the manufacture of device, the command device realize in one flow of flow chart or multiple flows and/or one square frame of block diagram or
The function of being specified in multiple square frames.
These computer program instructions can be also loaded into computer or other programmable data processing devices so that counted
Series of operation steps is performed on calculation machine or other programmable devices to produce computer implemented processing, so as in computer or
The instruction performed on other programmable devices is provided for realizing in one flow of flow chart or multiple flows and/or block diagram one
The step of function of being specified in individual square frame or multiple square frames.
Although preferred embodiments of the present invention have been described, but those skilled in the art once know basic creation
Property concept, then can make other change and modification to these embodiments.So appended claims be intended to be construed to include it is excellent
Select embodiment and fall into having altered and changing for the scope of the invention.
Obviously, those skilled in the art can carry out the essence of various changes and modification without departing from the present invention to the present invention
God and scope.So, if these modifications and variations of the present invention belong to the scope of the claims in the present invention and its equivalent technologies
Within, then the present invention is also intended to comprising including these changes and modification.
Claims (10)
- A kind of 1. face picture detection method, it is characterised in that including:According to cascade graders, judge whether picture to be detected is face picture, the cascade graders are according to pre- First selected face samples pictures and non-face samples pictures train what is obtained;If determine that the picture to be detected is face picture according to the cascade graders, according to convolutional neural networks CNN Grader, judge whether the picture to be detected is face picture, the CNN graders are according to the selected face sample The samples pictures correctly classified by the cascade graders in picture and the non-face samples pictures train what is obtained;If determine that the picture to be detected is face picture according to the CNN graders, it is determined that the picture to be detected is behaved Face picture.
- 2. the method as described in claim 1, it is characterised in that it is described judge whether picture to be detected is face picture after, Also include:If determine that the picture to be detected is non-face picture according to the cascade graders, it is determined that the mapping to be checked Piece is non-face picture.
- 3. the method as described in claim 1, it is characterised in that it is described according to CNN graders, judge that the picture to be detected is It is no be face picture after, in addition to:If determine that the picture to be detected is non-face picture according to the CNN graders, it is determined that the picture to be detected is Non-face picture.
- 4. such as the method any one of claim 1-3, it is characterised in that trained to obtain the CNN according to following method Grader:According to the previously selected face samples pictures and non-face samples pictures, the CNN graders are trained, obtained To the first CNN graders, the first CNN graders include Softmax determinants;The Softmax determinants in the first CNN graders are replaced with into support vector machines determinant;According in the previously selected face samples pictures and the non-face samples pictures by the cascade graders just The samples pictures really classified, the first CNN graders are trained, obtain the 2nd CNN graders;The CNN graders that the 2nd CNN graders are obtained as training.
- 5. such as the method any one of claim 1-3, it is characterised in that under the determinant in the CNN graders is Arrange any:Softmax determinants, SVM determinants.
- A kind of 6. face picture detection means, it is characterised in that including:First facial image detection unit, for according to cascade graders, judging whether picture to be detected is face picture, The cascade graders train to obtain according to previously selected face samples pictures and non-face samples pictures;Second facial image detection unit, if for determining that the picture to be detected is face according to the cascade graders Picture, then according to convolutional neural networks CNN graders, judge whether the picture to be detected is face picture, the CNN classification Device is correctly divided by the cascade graders according in the selected face samples pictures and the non-face samples pictures The samples pictures of class train what is obtained;Face picture determining unit, if for determining that the picture to be detected is face picture according to the CNN graders, really The fixed picture to be detected is face picture.
- 7. device as claimed in claim 6, it is characterised in that the face picture determining unit, be additionally operable to:If determine that the picture to be detected is non-face picture according to the cascade graders, it is determined that the mapping to be checked Piece is non-face picture.
- 8. device as claimed in claim 6, it is characterised in that the face picture determining unit, be additionally operable to:If determine that the picture to be detected is non-face picture according to the CNN graders, it is determined that the picture to be detected is Non-face picture.
- 9. such as the device any one of claim 6-8, it is characterised in that described device also includes training unit, is used for Trained to obtain the CNN graders according to following method:According to the previously selected face samples pictures and non-face samples pictures, the CNN graders are trained, obtained To the first CNN graders, the first CNN graders include Softmax determinants;The Softmax determinants in the first CNN graders are replaced with into support vector machines determinant;According in the previously selected face samples pictures and the non-face samples pictures by the cascade graders just The samples pictures really classified, the first CNN graders are trained, obtain the 2nd CNN graders;The CNN graders that the 2nd CNN graders are obtained as training.
- 10. such as the device any one of claim 6-8, it is characterised in that under the determinant in the CNN graders is Arrange any:Softmax determinants, SVM determinants.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610663327.6A CN107729796A (en) | 2016-08-11 | 2016-08-11 | A kind of face picture detection method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610663327.6A CN107729796A (en) | 2016-08-11 | 2016-08-11 | A kind of face picture detection method and device |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107729796A true CN107729796A (en) | 2018-02-23 |
Family
ID=61199937
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610663327.6A Pending CN107729796A (en) | 2016-08-11 | 2016-08-11 | A kind of face picture detection method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107729796A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109977835A (en) * | 2019-03-19 | 2019-07-05 | 中国联合网络通信集团有限公司 | Face-image recognition methods, device and equipment |
WO2021135933A1 (en) * | 2019-12-30 | 2021-07-08 | 中兴通讯股份有限公司 | Target recognition method and device, storage medium and electronic device |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102332094A (en) * | 2011-10-24 | 2012-01-25 | 西安电子科技大学 | Semi-supervised online study face detection method |
CN103824054A (en) * | 2014-02-17 | 2014-05-28 | 北京旷视科技有限公司 | Cascaded depth neural network-based face attribute recognition method |
CN105303193A (en) * | 2015-09-21 | 2016-02-03 | 重庆邮电大学 | People counting system for processing single-frame image |
CN105631519A (en) * | 2015-12-31 | 2016-06-01 | 北京工业大学 | Convolution nerve network acceleration method based on pre-deciding and system |
-
2016
- 2016-08-11 CN CN201610663327.6A patent/CN107729796A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102332094A (en) * | 2011-10-24 | 2012-01-25 | 西安电子科技大学 | Semi-supervised online study face detection method |
CN103824054A (en) * | 2014-02-17 | 2014-05-28 | 北京旷视科技有限公司 | Cascaded depth neural network-based face attribute recognition method |
CN105303193A (en) * | 2015-09-21 | 2016-02-03 | 重庆邮电大学 | People counting system for processing single-frame image |
CN105631519A (en) * | 2015-12-31 | 2016-06-01 | 北京工业大学 | Convolution nerve network acceleration method based on pre-deciding and system |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109977835A (en) * | 2019-03-19 | 2019-07-05 | 中国联合网络通信集团有限公司 | Face-image recognition methods, device and equipment |
CN109977835B (en) * | 2019-03-19 | 2022-01-25 | 中国联合网络通信集团有限公司 | Facial image recognition method, device and equipment |
WO2021135933A1 (en) * | 2019-12-30 | 2021-07-08 | 中兴通讯股份有限公司 | Target recognition method and device, storage medium and electronic device |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111179229B (en) | Industrial CT defect detection method based on deep learning | |
CN106485215B (en) | Face shielding detection method based on deep convolutional neural network | |
CN106845421B (en) | Face feature recognition method and system based on multi-region feature and metric learning | |
CN106776842B (en) | Multimedia data detection method and device | |
CN107657249A (en) | Method, apparatus, storage medium and the processor that Analysis On Multi-scale Features pedestrian identifies again | |
CN105303193B (en) | A kind of passenger number statistical system based on single-frame images processing | |
CN110110715A (en) | Text detection model training method, text filed, content determine method and apparatus | |
WO2017143919A1 (en) | Method and apparatus for establishing data identification model | |
CN106960195A (en) | A kind of people counting method and device based on deep learning | |
CN107563431A (en) | A kind of image abnormity detection method of combination CNN transfer learnings and SVDD | |
CN107358157A (en) | A kind of human face in-vivo detection method, device and electronic equipment | |
CN108765131A (en) | Credit authorization method, apparatus, terminal and readable storage medium storing program for executing based on micro- expression | |
CN107610113A (en) | The detection method and device of Small object based on deep learning in a kind of image | |
CN109359666A (en) | A kind of model recognizing method and processing terminal based on multiple features fusion neural network | |
CN106355188A (en) | Image detection method and device | |
CN108932523B (en) | Image classification and classified data creation system and method, and storage medium | |
CN110232404A (en) | A kind of recognition methods of industrial products surface blemish and device based on machine learning | |
CN111914665A (en) | Face shielding detection method, device, equipment and storage medium | |
CN112818871B (en) | Target detection method of full fusion neural network based on half-packet convolution | |
CN112017245B (en) | Certificate positioning method | |
CN110503099A (en) | Information identifying method and relevant device based on deep learning | |
CN114612472A (en) | SegNet improvement-based leather defect segmentation network algorithm | |
CN106778800A (en) | A kind of AdaBoost cascade classifiers method for quick | |
CN107729796A (en) | A kind of face picture detection method and device | |
CN112560584A (en) | Face detection method and device, storage medium and terminal |
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: 20180223 |
|
RJ01 | Rejection of invention patent application after publication |