CN107609493A - Method and device for optimizing human face image quality evaluation model - Google Patents
Method and device for optimizing human face image quality evaluation model Download PDFInfo
- Publication number
- CN107609493A CN107609493A CN201710743257.XA CN201710743257A CN107609493A CN 107609493 A CN107609493 A CN 107609493A CN 201710743257 A CN201710743257 A CN 201710743257A CN 107609493 A CN107609493 A CN 107609493A
- Authority
- CN
- China
- Prior art keywords
- face picture
- face
- measured
- similarity
- identity information
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 38
- 238000013441 quality evaluation Methods 0.000 title claims abstract description 13
- 238000012360 testing method Methods 0.000 claims abstract description 39
- 238000012549 training Methods 0.000 claims abstract description 20
- 238000013210 evaluation model Methods 0.000 claims description 49
- 230000007613 environmental effect Effects 0.000 claims description 48
- 238000005457 optimization Methods 0.000 claims description 30
- 230000007935 neutral effect Effects 0.000 claims description 8
- 238000003860 storage Methods 0.000 claims description 7
- 238000004590 computer program Methods 0.000 claims description 4
- 238000000605 extraction Methods 0.000 claims description 4
- SBNFWQZLDJGRLK-UHFFFAOYSA-N phenothrin Chemical compound CC1(C)C(C=C(C)C)C1C(=O)OCC1=CC=CC(OC=2C=CC=CC=2)=C1 SBNFWQZLDJGRLK-UHFFFAOYSA-N 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims 1
- 235000013399 edible fruits Nutrition 0.000 claims 1
- 210000005036 nerve Anatomy 0.000 claims 1
- 238000011156 evaluation Methods 0.000 abstract description 6
- 238000013528 artificial neural network Methods 0.000 abstract description 2
- 230000001537 neural effect Effects 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 3
- 241000208340 Araliaceae Species 0.000 description 2
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 2
- 235000003140 Panax quinquefolius Nutrition 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 235000008434 ginseng Nutrition 0.000 description 2
- 238000003062 neural network model Methods 0.000 description 2
- 238000002360 preparation method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 239000012141 concentrate Substances 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000010191 image analysis Methods 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
- Collating Specific Patterns (AREA)
Abstract
The invention relates to a method and a device for optimizing a human face image quality evaluation model. The method comprises the following steps: establishing a face picture test set; identifying the similarity between the face picture to be detected and a sample face picture in a preset face database, and obtaining the identification result of each face picture to be detected according to the similarity and the picture identity information; determining the quality score of each face picture to be detected according to the identification result; and (4) carrying out neural network training by taking the face picture to be tested and the corresponding quality score thereof as training data to obtain an optimized face picture quality evaluation model and parameters. By the human face picture quality evaluation model and the parameters, the evaluation on the human face picture quality is not influenced by artificial subjective factors.
Description
Technical field
The present invention relates to image analysis technology field, method more particularly to optimization face picture Environmental Evaluation Model,
Device, storage medium and computer equipment.
Background technology
With the development of deep learning and face recognition technology, recognition of face is gone soon applied to increasing scene
Identification a person's identity of speed.Standard of the existing face picture quality evaluating method in actual nonrestrictive application scenarios
True rate can be more much lower than the accuracy rate in experiment, traces it to its cause:In general face picture quality evaluating method is subjective selects
The different characteristic or attribute of face picture, e.g., the quantity of pixel evaluates the quality of picture in unit image.But in reality
In application scenarios, meeting recognition accuracy of the subjective good face picture of human eye in face recognition algorithms might not be high.Cause
This, existing face picture Environmental Evaluation Model is poor to the evaluation effect of face picture quality.
The content of the invention
Based on this, the invention provides the method and device of optimization face picture Environmental Evaluation Model, it is possible to increase face
The objectivity that picture quality evaluation model is evaluated picture quality, and then preferably evaluate the quality of face picture.
The present invention program includes:
In a first aspect, the embodiment of the present invention provides a kind of method for optimizing face picture Environmental Evaluation Model, including:
Establish face picture test set;Test set includes multiple face pictures to be measured, and each face picture to be measured is labeled with
Corresponding first identity information;
The similarity of face picture to be measured and the sample face picture in default face database is identified, according to described similar
Degree, the first identity information and the second identity information draw the recognition result of each face picture to be measured;Wrapped in the face database
Multiple sample face pictures are included, this face picture of various kinds is labeled with corresponding second identity information;
The mass fraction of each face picture to be measured is determined according to the recognition result;
Using each face picture to be measured and its corresponding mass fraction as training data, using regression neutral net to initial
Face picture Environmental Evaluation Model and parameter are trained, the face picture Environmental Evaluation Model and parameter optimized.
Second aspect, the embodiment of the present invention provide a kind of device for optimizing face picture Environmental Evaluation Model, including:
Test set establishes module, for establishing face picture test set;Test set includes multiple face pictures to be measured, respectively
Face picture to be measured is labeled with corresponding first identity information;
Initial identification module, for identifying the phase of face picture to be measured and the sample face picture in default face database
Like degree, the recognition result of each face picture to be measured is drawn according to the similarity, the first identity information and the second identity information;Institute
Stating face database includes multiple sample face pictures, and this face picture of various kinds is labeled with corresponding second identity information;
Mass fraction computing module, for determining the mass fraction of each face picture to be measured according to the recognition result;
And
Training module, for using each face picture to be measured and its corresponding mass fraction as training data, using regression
Neutral net is trained to Initial Face picture quality evaluation model and parameter, the face picture quality evaluation mould optimized
Type and parameter.
A kind of computer-readable recording medium, is stored thereon with computer program, and the program is realized when being executed by processor
The step of methods described.
A kind of computer equipment, including memory, processor and storage can be run on a memory and on a processor
Computer program, the step of realizing methods described during the computing device described program.
The beneficial effect that technical scheme provided in an embodiment of the present invention is brought:
By establishing face picture test set, test set includes multiple face pictures to be measured, inputs the people pre-established
Face database;The similarity of face picture to be measured and the sample face picture in default face database is identified, according to the phase
Show that the recognition result of each face picture to be measured draws the identification of each face picture to be measured like the identity information of degree and each picture
As a result;The mass fraction of each face picture to be measured is determined, with the face picture to be measured in test set and its corresponding quality point
Number carries out neural metwork training for training data, by training obtained face picture Environmental Evaluation Model and parameter, by this
The evaluation of face picture Environmental Evaluation Model and parameter to face picture quality is more objective and accurate, overcomes artificial subjective factor pair
The evaluation of picture quality influences.
Brief description of the drawings
Fig. 1 is the indicative flowchart of the method for the optimization face picture Environmental Evaluation Model of an embodiment;
Fig. 2 is the indicative flowchart of the method for the optimization face picture Environmental Evaluation Model under a concrete application scene;
Fig. 3 is the schematic diagram of the device of the optimization face picture Environmental Evaluation Model of an embodiment.
Embodiment
In order to make the purpose , technical scheme and advantage of the present invention be clearer, it is right below in conjunction with drawings and Examples
The present invention is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, and
It is not used in the restriction present invention.
Although the step in the present invention is arranged with label, it is not used to limit the precedence of step, unless
It specify that the order of step or based on the execution of certain step needs other steps, otherwise the relative rank of step is
It is adjustable.
Fig. 1 is the indicative flowchart of the method for the optimization face picture Environmental Evaluation Model of an embodiment;Such as Fig. 1 institutes
Show, the method for the optimization face picture Environmental Evaluation Model in the present embodiment includes step:
S11, establishes face picture test set, and the test set includes multiple face pictures to be measured, each face figure to be measured
Piece is provided with corresponding first identity information.
Face picture to be measured in face picture test set is the face picture gathered under concrete application scene, can be covered
Lid is more personal.Alternatively, everyone plurality of pictures, it is derived from different angles.The identity information of face picture refers to can be unique
The information of people's identity is identified, such as can be the name of personage, identification number at one's side.
S12, the similarity of face picture to be measured and the sample face picture in default face database is identified, according to described
Similarity, the first identity information and the second identity information draw the recognition result of each face picture to be measured.
Wherein, preset face database and include some personal face pictures, and be labeled with per pictures corresponding the
Two identity informations.Alternatively, everyone pictures.
Face picture to be measured is compared with the sample face picture in default face database, is to calculate two faces
Similarity, can determine whether face picture to be measured and sample face picture are same according to similarity and default similarity threshold
One people.
In one embodiment, identify that face picture to be measured is similar to the sample face picture in default face database
Degree, the tool of the recognition result of each face picture to be measured is drawn according to the similarity, the first identity information and the second identity information
Body mode includes:The similarity of face picture to be measured and this face picture of various kinds in face database is calculated respectively, is drawn and is respectively treated
Survey similarity maximum corresponding to face picture, and corresponding sample face picture during similarity maximum;Obtain people to be measured
First identity information of face picture, the second identity information of sample face picture corresponding to acquisition;It is maximum according to the similarity
Value, default similarity threshold, the first identity information and the second identity information obtain the recognition result of each face picture to be measured.
Alternatively, the recognition result includes a kind of result, two class results, three class results and four class results;Specifically for example:
A kind of result is that the face picture to be measured with common identity information and sample face picture correctly are identified as into same face;
Face picture to be measured with common identity information and sample face picture are identified as different faces by two class results for mistake;
Face picture to be measured with different identity information and sample face picture are identified as same face by three class results for mistake;
Four class results are that the face picture to be measured with different identity information and sample face picture correctly are identified as into different faces.
In an alternative embodiment, face recognition algorithms are preset, are calculated respectively in face picture to be measured and face database
The similarity of this face picture of various kinds, and similarity maximum corresponding to each face picture to be measured is drawn, and similarity is maximum
Corresponding sample face picture during value.By calculating each face picture to be measured and each sample face picture in face database
Similarity, a sample face picture maximum with the face picture similarity to be measured (also referred to as comparative sample can be obtained
This face picture) and corresponding similarity (i.e. similarity maximum).And then identify the face picture to be measured and the comparison
Whether sample face picture is same face.Optionally, according to the similarity maximum, default similarity threshold and right
The identity information answered, obtain the recognition result of each face picture to be measured.
S13, the mass fraction of each face picture to be measured is determined according to the recognition result.
Optionally, according to the recognition result, similarity maximum and similarity threshold, each face picture to be measured is determined
Mass fraction, in order to carry out neural metwork training.
S14, using the face picture to be measured in test set and its corresponding mass fraction as training data, using regression god
Face picture Environmental Evaluation Model and parameter are trained through network, the face picture Environmental Evaluation Model and ginseng optimized
Number.
It should be understood that when being trained using regression neutral net, when meeting the default condition of convergence described in acquisition
The model and network parameter of regression neutral net, it can thus be concluded that face picture Environmental Evaluation Model and parameter to optimization.
To sum up, the method for the optimization face picture Environmental Evaluation Model of above-described embodiment, first, face picture test is established
Collection;Then, the recognition result of each face picture to be measured is drawn according to face picture test set and default face database;Then,
The mass fraction of each face picture to be measured is determined according to the recognition result;Finally, with the face picture to be measured in test set and
Its corresponding face mass fraction is that training data carries out neural metwork training, the face picture Environmental Evaluation Model optimized
And parameter.Thus, it is possible to optimize face picture Environmental Evaluation Model and parameter, using the face picture quality evaluation mould after optimization
Type and parameter, the evaluation to face picture quality will not influenceed by artificial subjective factor, and evaluation result is more objective more accurate.
In an alternative embodiment, calculated respectively by face recognition algorithms each in face picture to be measured and face database
The similarity of sample face picture, including:Face characteristic, the face of sample face picture of each face picture to be measured are extracted respectively
Feature, the face characteristic based on extraction calculate similarity.
In an alternative embodiment, according to the similarity maximum, default similarity threshold and corresponding identity
Information, the recognition result of each face picture to be measured is obtained, including:The similarity maximum and default similarity threshold are entered
Row compares, and obtains the first comparison result;First identity information and the second identity information are compared, obtain the second comparison
As a result;The recognition result of each face picture to be measured is obtained according to the first comparison result, the second comparison result.Alternatively, the knowledge
Other result includes:A kind of result, two class results, three class results and four class results;Wherein, in a kind of result and three class results, phase
It is more than or equal to the similarity threshold like degree maximum;In two class results and four class results, similarity maximum is less than described
Similarity threshold.
Alternatively, the above-mentioned implementation for being identified result is specific for example:
If score>=threshold, and name_labeled is identical with test_name, then is identified as a kind of result;
If score<Threshold, but name_labeled is identical with test_name, then is identified as two class results;
If score>=threshold, but name_labeled is different from test_name, then is identified as three class results;
If score<Threshold, and name_labeled is different from test_name, then is identified as four class results;
Wherein, score represents the similarity maximum corresponding to face picture to be measured, and threshold represents similarity threshold
Value, name_labeled represent identity information corresponding to sample face picture, and test_name is represented corresponding to face picture to be measured
Identity information.
In an alternative embodiment, calculating the mode of mass fraction can be:
Quality_score=total+total × flag × | score-threshold |/delta;
Wherein, 2*total represents the full marks value of mass fraction;If score<Threshold, then delta=
Threshold, otherwise delta=1-threshold.For example, when total in formula is 50, then the full marks value of mass fraction
For 100.
Fig. 2 is the indicative flowchart of the method for the optimization face picture Environmental Evaluation Model under a specific embodiment, is had
Body is divided into four parts:Data set arranges S21, recognition of face S22, face quality annotation S23 and the study of face quality algorithm
S24。
It refer to such as Fig. 2, the method for the optimization face picture Environmental Evaluation Model in the present embodiment includes step:
S21, data preparation.
Data preparation part includes:Face recognition algorithms face_identify_algorithm setting, face database
Face_gallery_set default and face test set face_test_set foundation.
Similarity threshold threshold corresponding to being set to face recognition algorithms, the similarity threshold are used to judge two
Whether face picture belongs to a people;The default face database includes multiple sample face pictures, and (such as N, N is big
In 1), every sample face picture is provided with corresponding second identity information;Optionally, everyone only has a pictures.Establish
Face test set includes multiple face pictures to be measured, and every face picture to be measured is provided with corresponding first identity information;Can
Choosing, face picture to be measured is collected in the face picture of more people's different angles.It should be understood that face picture to be measured set the
The second identity information that one identity information is set with sample face picture is same type of information, for example, being name.Sample
The name name_labeled of face picture, the name test_name of face picture to be measured.
S22, recognition of face.
Recognition of face part is the recognition result that face picture to be measured is determined according to similarity threshold threshold.
In one embodiment, specific practice is:With face recognition algorithms extract every test face picture (assuming that it is current this
Test face picture name be labeled as test_name) characteristic information (characteristic information had both included the spy of subjective perception
Reference ceases, such as pixel, brightness;Also other characteristic informations artificially not perceived are included);Use the test face picture of extraction
Characteristic information characteristic information corresponding with the sample face pictures of the N in face database is contrasted respectively, obtain N number of phase
Like degree;This N number of similarity result is sorted, takes out the maximum similarity score score of similarity and corresponding sample face figure
The name name_labeled of piece, and then can obtain following recognition result identify_result:
TP (i.e. a kind of result):If score>=threshold, and name_labeled is identical with test_name;
FP (i.e. two class results):If score<Threshold, but name_labeled is identical with test_name;
FN (i.e. three class results):If score>=threshold, but name_labeled is different from test_name;
TN (i.e. four class results):If score<Threshold, and name_labeled is different from test_name.
A kind of result is correctly to be identified as the face picture to be measured with common identity information and sample face picture
Same face;Face picture to be measured with common identity information and sample face picture are identified as by two class results for mistake
Different faces;Face picture to be measured with different identity information and sample face picture are identified as by three class results for mistake
Same face;Four class results are correctly to be identified as the face picture to be measured with different identity information and sample face picture
Different faces.
S23, face quality annotation.
The mass fraction of face test pictures is obtained according to step S22 recognition result.Face quality annotation is to calculate
Face mass fraction out is added in face picture test set, obtains face quality test collection, the face quality test
Collect to include the test pictures collection of the face mass fraction calculated.
S24, the study of face quality algorithm.
The process uses regression neural metwork training data, obtains regression neural network parameter.Specifically, return
Type neural network model can be considered face picture Environmental Evaluation Model to be optimized, parameter in regression neutral net and described
The parameter of face picture Environmental Evaluation Model, neural metwork training are trained face picture quality evaluation model and parameter, that is, adopted
The parameter being related in face picture Environmental Evaluation Model is trained with regression neutral net, when the default condition of convergence of satisfaction
When, the regression neural network model and parameter are obtained, it can thus be concluded that face picture Environmental Evaluation Model and ginseng to optimization
Number.
In order to simply represent, face_identify_algorithm, face_ are used in above-mentioned technical characteristic respectively
Gallery_set, face_test_set, threshold represent face recognition algorithms, face database, face picture test set
And similarity threshold.Above-mentioned representation is not construed as to face recognition algorithms, face database, face picture test set
With the limitation of similarity threshold.
Based on the method identical thought with the optimization face picture Environmental Evaluation Model in above-described embodiment, the present invention is also
The device of optimization face picture Environmental Evaluation Model is provided, the device can be used for performing above-mentioned optimization face picture quality evaluation mould
The method of type.For convenience of description, in the structural representation of device embodiment for optimizing face picture Environmental Evaluation Model, only
The part related to the embodiment of the present invention is shown, it will be understood by those skilled in the art that schematic structure not structure twin installation
Restriction, can include than illustrating more or less parts, either combine some parts or different parts arrangement.
Fig. 3 is the schematic diagram of the device of the optimization face picture Environmental Evaluation Model of one embodiment of the invention, such as
Shown in Fig. 3, the device of the optimization face picture Environmental Evaluation Model of the present embodiment includes:Test set establishes module 310, initial knowledge
Other module 320, mass fraction computing module 330 and training module 340, details are as follows for each module:
The test set establishes module 310, for establishing face picture test set;Test set includes multiple faces to be measured
Picture, each face picture to be measured are provided with corresponding first identity information;
The initial identification module 320, for identifying face picture to be measured and the sample face in default face database
The similarity of picture, the knowledge of each face picture to be measured is drawn according to the similarity, the first identity information and the second identity information
Other result;The face database includes multiple sample face pictures, and this face picture of various kinds is labeled with corresponding second body
Part information;
The mass fraction computing module 330, for determining the matter of each face picture to be measured according to the recognition result
Measure fraction;
The training module 340, for using each face picture to be measured and its corresponding mass fraction as training data, using
Regression neutral net is trained to Initial Face picture quality evaluation model and parameter, the face picture quality optimized
Evaluation model and parameter.
In an alternative embodiment, in the device of optimization face picture Environmental Evaluation Model, the initial identification module
320, for calculating the similarity of face picture to be measured and this face picture of various kinds in face database respectively, draw each people to be measured
Similarity maximum corresponding to face picture, and corresponding sample face picture during similarity maximum;Obtain face figure to be measured
First identity information of piece, the second identity information of sample face picture corresponding to acquisition;According to the similarity maximum, in advance
If similarity threshold, the first identity information and the second identity information obtain the recognition result of each face picture to be measured.
In an alternative embodiment, in the device of the optimization face picture Environmental Evaluation Model, the initial identification
Module 320 may include similarity calculated and face identification unit.
Wherein, the similarity calculated, for calculating face picture to be measured and each sample in face database respectively
The similarity of face picture, similarity maximum corresponding to each face picture to be measured is drawn, and corresponded to during similarity maximum
Sample face picture.
Wherein, the face identification unit, for according to the similarity maximum, default similarity threshold and figure
Identity information corresponding to piece, obtain the recognition result of each face picture to be measured.
In an alternative embodiment, the similarity calculated, the face spy that each face picture to be measured can be extracted respectively
Sign, the face characteristic of sample face picture, the face characteristic based on extraction calculate similarity.
In an alternative embodiment, the face identification unit, for by the similarity maximum and default similar
Degree threshold value is compared, and obtains the first comparison result;First identity information and the second identity information are compared, obtained
Second comparison result;The recognition result of each face picture to be measured is obtained according to the first comparison result, the second comparison result.Specifically may be used
The recognition result of each face picture to be measured is obtained as follows, including:
If score>=threshold, and name_labeled is identical with test_name, then is identified as a kind of result;
If score<Threshold, but name_labeled is identical with test_name, then is identified as two class results;
If score>=threshold, but name_labeled is different from test_name, then is identified as three class results;
If score<Threshold, and name_labeled is different from test_name, then is identified as four class results;
Wherein, score represents the similarity maximum corresponding to face picture to be measured, and threshold represents similarity threshold
Value, name_labeled represent identity information corresponding to sample face picture, and test_name is represented corresponding to face picture to be measured
Identity information.
In an alternative embodiment, the mass fraction computing module 330, for extracting the phase corresponding to each recognition result
Like degree maximum and similarity threshold;According to the recognition result, similarity maximum and similarity threshold, each people to be measured is determined
The mass fraction of face picture.
Such as:The mass fraction computing module 330 can calculate the quality point of each face picture to be measured by equation below
Number:
Quality_score=total+total × flag × | score-threshold |/delta;
Wherein, 2*total represents the full marks value of mass fraction;If score<Threshold, then delta=
Threshold, otherwise delta=1-threshold.
In an alternative embodiment, in the device embodiment of above-mentioned optimization face picture Environmental Evaluation Model, the test
Concentrate, face picture to be measured corresponding to each identity information is at least two;In the face database, each identity information pair
The sample face picture answered is one.
It should be noted that in the embodiment of the device of the optimization face picture Environmental Evaluation Model of above-mentioned example, respectively
The contents such as information exchange, implementation procedure between module/unit, due to being based on same structure with preceding method embodiment of the present invention
Think, its technique effect brought is identical with preceding method embodiment of the present invention, and particular content can be found in the inventive method embodiment
In narration, here is omitted.
In addition, in the embodiment of the device of the optimization face picture Environmental Evaluation Model of above-mentioned example, each program module
Logical partitioning be merely illustrative of, can be as needed in practical application, for example, configuration requirement for corresponding hardware or
The convenient consideration of the realization of software, above-mentioned function distribution is completed by different program modules, will the optimization face picture
The internal structure of the device of Environmental Evaluation Model is divided into different program modules, described above all or part of to complete
Function.
It will appreciated by the skilled person that realizing all or part of flow in above-described embodiment method, being can
To instruct the hardware of correlation to complete by computer program, described program can be stored in a computer-readable storage and be situated between
In matter, as independent production marketing or use.Described program upon execution, can perform the complete of such as embodiment of above-mentioned each method
Portion or part steps.In addition, the storage medium may also be disposed in a kind of computer equipment, also wrapped in the computer equipment
Include processor, during program in storage medium described in the computing device, can realize above-mentioned each method embodiment it is complete
Portion or part steps.Wherein, described storage medium can be magnetic disc, CD, read-only memory (Read-Only
Memory, ROM) or random access memory (Random Access Memory, RAM) etc..
In the above-described embodiments, the description to each embodiment all emphasizes particularly on different fields, and does not have the portion being described in detail in some embodiment
Point, it may refer to the associated description of other embodiments.It is appreciated that wherein used term " first ", " second " etc. are at this
It is used to distinguish object in text, but these objects should not be limited by these terms.
Embodiment described above only expresses the several embodiments of the present invention, it is impossible to is interpreted as to the scope of the claims of the present invention
Limitation.It should be pointed out that for the person of ordinary skill of the art, without departing from the inventive concept of the premise,
Various modifications and improvements can be made, these belong to protection scope of the present invention.Therefore, the protection domain of patent of the present invention
It should be determined by the appended claims.
Claims (10)
- A kind of 1. method for optimizing face picture Environmental Evaluation Model, it is characterised in that including:Establish face picture test set;Test set includes multiple face pictures to be measured, and each face picture to be measured is labeled with correspondingly The first identity information;Identify the similarity of the sample face picture in face picture to be measured and default face database, according to the similarity, First identity information and the second identity information draw the recognition result of each face picture to be measured;The face database includes more Individual sample face picture, this face picture of various kinds are labeled with corresponding second identity information;The mass fraction of each face picture to be measured is determined according to the recognition result;Using each face picture to be measured and its corresponding mass fraction as training data, using regression neutral net to Initial Face Picture quality evaluation model and parameter are trained, the face picture Environmental Evaluation Model and parameter optimized.
- 2. the method for optimization face picture Environmental Evaluation Model according to claim 1, it is characterised in that identify people to be measured The similarity of sample face picture in face picture and default face database, according to the similarity, the first identity information and Second identity information draws the recognition result of each face picture to be measured, including:The similarity of face picture to be measured and this face picture of various kinds in face database is calculated respectively, draws each face figure to be measured Similarity maximum corresponding to piece, and corresponding sample face picture during similarity maximum;The first identity information of face picture to be measured is obtained, the second identity information of sample face picture corresponding to acquisition;According to The similarity maximum, default similarity threshold, the first identity information and the second identity information, obtain each face figure to be measured The recognition result of piece.
- 3. the method for optimization face picture Environmental Evaluation Model according to claim 2, it is characterised in that calculate treat respectively The similarity of face picture and this face picture of various kinds in face database is surveyed, including:Face characteristic, the face characteristic of sample face picture of each face picture to be measured are extracted respectively, and the face based on extraction is special Sign calculates similarity.
- 4. the method for the optimization face picture Environmental Evaluation Model according to Claims 2 or 3, it is characterised in that according to institute Similarity maximum, default similarity threshold, the first identity information and the second identity information are stated, obtains each face picture to be measured Recognition result, including:The similarity maximum and default similarity threshold are compared, obtain the first comparison result;First identity information and the second identity information are compared, obtain the second comparison result;The recognition result of each face picture to be measured is obtained according to the first comparison result, the second comparison result.
- 5. the method for optimization face picture Environmental Evaluation Model according to claim 4, it is characterised in that the identification knot Fruit includes:A kind of result, two class results, three class results and four class results;Wherein, in a kind of result and three class results, similarity Maximum is more than or equal to the similarity threshold;In two class results and four class results, similarity maximum is less than described similar Spend threshold value;The mass fraction of each face picture to be measured is determined according to the recognition result, including:Extract similarity maximum and the similarity threshold corresponding to each recognition result;According to the recognition result, similarity maximum and similarity threshold, the mass fraction of each face picture to be measured is determined.
- 6. the method for optimization face picture Environmental Evaluation Model according to claim 5, it is characterised in that by following public Formula calculates the mass fraction of each face picture to be measured:Quality_score=total+total × flag × | score-threshold |/delta;Wherein, 2*total represents the full marks value of mass fraction;If score<Threshold, then delta=threshold, Otherwise delta=1-threshold;Quality_score represents mass fraction, and score is represented corresponding to face picture to be measured Similarity maximum, threshold represent similarity threshold.
- 7. the method for the optimization face picture Environmental Evaluation Model according to claim 1,2,3,5 or 6, it is characterised in thatIn the test set, face picture to be measured corresponding to same identity information is at least two;In the face database, together Sample face picture corresponding to one identity information is one.
- A kind of 8. device for optimizing face picture Environmental Evaluation Model, it is characterised in that including:Test set establishes module, for establishing face picture test set;Test set includes multiple face pictures to be measured, each to be measured Face picture is labeled with corresponding first identity information;Initial identification module, for identifying that face picture to be measured is similar to the sample face picture in default face database Degree, the recognition result of each face picture to be measured is drawn according to the similarity, the first identity information and the second identity information;It is described Face database includes multiple sample face pictures, and this face picture of various kinds is labeled with corresponding second identity information;Mass fraction computing module, for determining the mass fraction of each face picture to be measured according to the recognition result;AndTraining module, for using each face picture to be measured and its corresponding mass fraction as training data, using regression nerve Network is trained to Initial Face picture quality evaluation model and parameter, the face picture Environmental Evaluation Model that is optimized and Parameter.
- 9. a kind of computer-readable recording medium, is stored thereon with computer program, it is characterised in that the program is held by processor The step of claim 1 to 7 any methods described is realized during row.
- 10. a kind of computer equipment, including memory, processor and storage are on a memory and the meter that can run on a processor Calculation machine program, it is characterised in that the step of any methods described of claim 1 to 7 is realized during the computing device described program Suddenly.
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710743257.XA CN107609493B (en) | 2017-08-25 | 2017-08-25 | Method and device for optimizing human face image quality evaluation model |
PCT/CN2017/116217 WO2019037346A1 (en) | 2017-08-25 | 2017-12-14 | Method and device for optimizing human face picture quality evaluation model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710743257.XA CN107609493B (en) | 2017-08-25 | 2017-08-25 | Method and device for optimizing human face image quality evaluation model |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107609493A true CN107609493A (en) | 2018-01-19 |
CN107609493B CN107609493B (en) | 2021-04-13 |
Family
ID=61055766
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710743257.XA Active CN107609493B (en) | 2017-08-25 | 2017-08-25 | Method and device for optimizing human face image quality evaluation model |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN107609493B (en) |
WO (1) | WO2019037346A1 (en) |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109544523A (en) * | 2018-11-14 | 2019-03-29 | 北京智芯原动科技有限公司 | Quality of human face image evaluation method and device based on more attribute face alignments |
CN109753917A (en) * | 2018-12-29 | 2019-05-14 | 中国科学院重庆绿色智能技术研究院 | Face quality optimization method, system, computer readable storage medium and equipment |
CN110427888A (en) * | 2019-08-05 | 2019-11-08 | 北京深醒科技有限公司 | A kind of face method for evaluating quality based on feature clustering |
WO2020001084A1 (en) * | 2018-06-30 | 2020-01-02 | 东南大学 | Online learning facial recognition method |
CN111126121A (en) * | 2018-11-01 | 2020-05-08 | 百度在线网络技术(北京)有限公司 | Method, device and equipment for adjusting face recognition model and storage medium |
CN111709906A (en) * | 2020-04-13 | 2020-09-25 | 北京深睿博联科技有限责任公司 | Medical image quality evaluation method and device |
CN112989869A (en) * | 2019-12-02 | 2021-06-18 | 深圳云天励飞技术有限公司 | Optimization method, device and equipment of face quality detection model and storage medium |
CN113095672A (en) * | 2021-04-09 | 2021-07-09 | 公安部物证鉴定中心 | Method and system for evaluating face image comparison algorithm |
CN117372405A (en) * | 2023-10-31 | 2024-01-09 | 神州通立电梯有限公司 | Face image quality evaluation method, device, storage medium and equipment |
Families Citing this family (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109977815B (en) * | 2019-03-13 | 2024-07-12 | 上海商汤智能科技有限公司 | Image quality evaluation method and device, electronic equipment and storage medium |
CN111797845A (en) * | 2019-03-19 | 2020-10-20 | 北京沃东天骏信息技术有限公司 | Picture processing method and device, storage medium and electronic equipment |
CN110276243A (en) * | 2019-05-07 | 2019-09-24 | 平安科技(深圳)有限公司 | Score mapping method, face comparison method, device, equipment and storage medium |
CN112488985A (en) * | 2019-09-11 | 2021-03-12 | 上海高德威智能交通***有限公司 | Image quality determination method, device and equipment |
CN110866437B (en) * | 2019-09-23 | 2024-06-28 | 平安科技(深圳)有限公司 | Face value judgment model optimization method and device, electronic equipment and storage medium |
CN111191584B (en) * | 2019-12-30 | 2024-02-09 | 电信科学技术第十研究所有限公司 | Face recognition method and device |
CN111311581A (en) * | 2020-02-20 | 2020-06-19 | 杭州涂鸦信息技术有限公司 | Image scoring method based on illumination and system and device thereof |
CN111382681B (en) * | 2020-02-28 | 2023-11-14 | 浙江大华技术股份有限公司 | Face registration method, device and storage medium |
CN111382693A (en) * | 2020-03-05 | 2020-07-07 | 北京迈格威科技有限公司 | Image quality determination method and device, electronic equipment and computer readable medium |
CN111814570B (en) * | 2020-06-12 | 2024-04-30 | 深圳禾思众成科技有限公司 | Face recognition method, system and storage medium based on dynamic threshold |
CN111966852B (en) * | 2020-06-28 | 2024-04-09 | 北京百度网讯科技有限公司 | Face-based virtual face-lifting method and device |
CN112148907A (en) * | 2020-10-23 | 2020-12-29 | 北京百度网讯科技有限公司 | Image database updating method and device, electronic equipment and medium |
CN112148908A (en) * | 2020-10-23 | 2020-12-29 | 北京百度网讯科技有限公司 | Image database updating method and device, electronic equipment and medium |
CN112686847B (en) * | 2020-12-23 | 2024-05-14 | 平安银行股份有限公司 | Identification card image shooting quality evaluation method and device, computer equipment and medium |
CN112633200A (en) * | 2020-12-29 | 2021-04-09 | 平安普惠企业管理有限公司 | Human face image comparison method, device, equipment and medium based on artificial intelligence |
CN113705496A (en) * | 2021-08-31 | 2021-11-26 | 深圳市酷开网络科技股份有限公司 | Poster selection method, device, equipment and storage medium |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2012109712A1 (en) * | 2011-02-18 | 2012-08-23 | National Ict Australia Limited | Image quality assessment |
US20130170545A1 (en) * | 2011-12-28 | 2013-07-04 | Canon Kabushiki Kaisha | Image encoding apparatus, image encoding method and program |
US8620092B2 (en) * | 2010-03-04 | 2013-12-31 | Hewlett-Packard Development Company, L.P. | Determining similarity of two images |
CN104636730A (en) * | 2015-02-10 | 2015-05-20 | 北京信息科技大学 | Method and device for face verification |
CN104866864A (en) * | 2015-05-07 | 2015-08-26 | 天津大学 | Extreme learning machine for three-dimensional image quality objective evaluation |
CN105608700A (en) * | 2015-12-24 | 2016-05-25 | 广州视源电子科技股份有限公司 | Photo screening method and system |
CN106503691A (en) * | 2016-11-10 | 2017-03-15 | 广州视源电子科技股份有限公司 | Identity labeling method and device for face picture |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7760917B2 (en) * | 2005-05-09 | 2010-07-20 | Like.Com | Computer-implemented method for performing similarity searches |
GB0512869D0 (en) * | 2005-06-24 | 2005-08-03 | Ibm | Method and system for facial recognition in groups |
CN106096538B (en) * | 2016-06-08 | 2019-08-23 | 中国科学院自动化研究所 | Face identification method and device based on sequencing neural network model |
CN106203333A (en) * | 2016-07-08 | 2016-12-07 | 乐视控股(北京)有限公司 | Face identification method and system |
CN106815566B (en) * | 2016-12-29 | 2021-04-16 | 天津中科智能识别产业技术研究院有限公司 | Face retrieval method based on multitask convolutional neural network |
-
2017
- 2017-08-25 CN CN201710743257.XA patent/CN107609493B/en active Active
- 2017-12-14 WO PCT/CN2017/116217 patent/WO2019037346A1/en active Application Filing
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8620092B2 (en) * | 2010-03-04 | 2013-12-31 | Hewlett-Packard Development Company, L.P. | Determining similarity of two images |
WO2012109712A1 (en) * | 2011-02-18 | 2012-08-23 | National Ict Australia Limited | Image quality assessment |
US20130170545A1 (en) * | 2011-12-28 | 2013-07-04 | Canon Kabushiki Kaisha | Image encoding apparatus, image encoding method and program |
CN104636730A (en) * | 2015-02-10 | 2015-05-20 | 北京信息科技大学 | Method and device for face verification |
CN104866864A (en) * | 2015-05-07 | 2015-08-26 | 天津大学 | Extreme learning machine for three-dimensional image quality objective evaluation |
CN105608700A (en) * | 2015-12-24 | 2016-05-25 | 广州视源电子科技股份有限公司 | Photo screening method and system |
CN106503691A (en) * | 2016-11-10 | 2017-03-15 | 广州视源电子科技股份有限公司 | Identity labeling method and device for face picture |
Non-Patent Citations (2)
Title |
---|
VIGNESH S ETAL.: "Face Image Quality Assessment for Face Selection", 《2015 IEEE GLOBAL CONFERENCE ON SIGNAL AND INFORMATION PROCESSING (GLOBALSIP)》 * |
高修峰: "人脸图像质量评估标准方法研究", 《中国博士学位论文全文数据库 信息科技辑》 * |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2020001084A1 (en) * | 2018-06-30 | 2020-01-02 | 东南大学 | Online learning facial recognition method |
CN111126121B (en) * | 2018-11-01 | 2023-04-04 | 百度在线网络技术(北京)有限公司 | Method, device and equipment for adjusting face recognition model and storage medium |
CN111126121A (en) * | 2018-11-01 | 2020-05-08 | 百度在线网络技术(北京)有限公司 | Method, device and equipment for adjusting face recognition model and storage medium |
CN109544523B (en) * | 2018-11-14 | 2021-01-01 | 北京智芯原动科技有限公司 | Method and device for evaluating quality of face image based on multi-attribute face comparison |
CN109544523A (en) * | 2018-11-14 | 2019-03-29 | 北京智芯原动科技有限公司 | Quality of human face image evaluation method and device based on more attribute face alignments |
CN109753917A (en) * | 2018-12-29 | 2019-05-14 | 中国科学院重庆绿色智能技术研究院 | Face quality optimization method, system, computer readable storage medium and equipment |
CN110427888A (en) * | 2019-08-05 | 2019-11-08 | 北京深醒科技有限公司 | A kind of face method for evaluating quality based on feature clustering |
CN112989869A (en) * | 2019-12-02 | 2021-06-18 | 深圳云天励飞技术有限公司 | Optimization method, device and equipment of face quality detection model and storage medium |
CN112989869B (en) * | 2019-12-02 | 2024-05-07 | 深圳云天励飞技术有限公司 | Optimization method, device, equipment and storage medium of face quality detection model |
CN111709906A (en) * | 2020-04-13 | 2020-09-25 | 北京深睿博联科技有限责任公司 | Medical image quality evaluation method and device |
CN113095672A (en) * | 2021-04-09 | 2021-07-09 | 公安部物证鉴定中心 | Method and system for evaluating face image comparison algorithm |
CN113095672B (en) * | 2021-04-09 | 2024-06-07 | 公安部物证鉴定中心 | Evaluation method and system for facial image comparison algorithm |
CN117372405A (en) * | 2023-10-31 | 2024-01-09 | 神州通立电梯有限公司 | Face image quality evaluation method, device, storage medium and equipment |
Also Published As
Publication number | Publication date |
---|---|
WO2019037346A1 (en) | 2019-02-28 |
CN107609493B (en) | 2021-04-13 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107609493A (en) | Method and device for optimizing human face image quality evaluation model | |
CN109522815B (en) | Concentration degree evaluation method and device and electronic equipment | |
WO2020151489A1 (en) | Living body detection method based on facial recognition, and electronic device and storage medium | |
CN104143079B (en) | The method and system of face character identification | |
CN106446754A (en) | Image identification method, metric learning method, image source identification method and devices | |
CN106557726A (en) | A kind of band is mourned in silence the system for face identity authentication and its method of formula In vivo detection | |
CN107423700A (en) | The method and device of testimony verification | |
CN109284733A (en) | A kind of shopping guide's act of omission monitoring method based on yolo and multitask convolutional neural networks | |
KR102284096B1 (en) | System and method for estimating subject image quality using visual saliency and a recording medium having computer readable program for executing the method | |
CN103824059A (en) | Facial expression recognition method based on video image sequence | |
CN105809178A (en) | Population analyzing method based on human face attribute and device | |
CN108960142B (en) | Pedestrian re-identification method based on global feature loss function | |
CN110400293B (en) | No-reference image quality evaluation method based on deep forest classification | |
CN107743225B (en) | A method of it is characterized using multilayer depth and carries out non-reference picture prediction of quality | |
CN110414350A (en) | The face false-proof detection method of two-way convolutional neural networks based on attention model | |
CN103996195A (en) | Image saliency detection method | |
CN107590460B (en) | Face classification method, apparatus and intelligent terminal | |
CN110135282A (en) | A kind of examinee based on depth convolutional neural networks model later plagiarizes cheat detection method | |
CN111709914B (en) | Non-reference image quality evaluation method based on HVS characteristics | |
CN109146873A (en) | A kind of display screen defect intelligent detecting method and device based on study | |
CN110309768A (en) | The staff's detection method and equipment of car test station | |
CN109510981A (en) | A kind of stereo-picture comfort level prediction technique based on multiple dimensioned dct transform | |
CN109685756A (en) | Image feature automatic identifier, system and method | |
CN117542121B (en) | Computer vision-based intelligent training and checking system and method | |
CN113705310A (en) | Feature learning method, target object identification method and corresponding device |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |