CN107609493B - 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
- CN107609493B CN107609493B CN201710743257.XA CN201710743257A CN107609493B CN 107609493 B CN107609493 B CN 107609493B CN 201710743257 A CN201710743257 A CN 201710743257A CN 107609493 B CN107609493 B CN 107609493B
- Authority
- CN
- China
- Prior art keywords
- face
- face picture
- similarity
- identity information
- 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.)
- Active
Links
- 238000013441 quality evaluation Methods 0.000 title claims abstract description 64
- 238000000034 method Methods 0.000 title claims abstract description 38
- 238000012360 testing method Methods 0.000 claims abstract description 60
- 238000012549 training Methods 0.000 claims abstract description 26
- 238000013528 artificial neural network Methods 0.000 claims abstract description 17
- 238000004364 calculation method Methods 0.000 claims description 5
- 238000004590 computer program Methods 0.000 claims description 5
- 238000002372 labelling Methods 0.000 claims description 3
- 238000011156 evaluation Methods 0.000 abstract description 4
- 238000004422 calculation algorithm Methods 0.000 description 12
- 238000010586 diagram Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 3
- 238000001303 quality assessment method Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000003062 neural network model Methods 0.000 description 2
- 230000000306 recurrent effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000007429 general method Methods 0.000 description 1
- 238000010191 image analysis Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
Images
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 invention relates to the technical field of image analysis, in particular to a method, a device, a storage medium and computer equipment for optimizing a human face image quality evaluation model.
Background
With the development of deep learning and face recognition technologies, face recognition has been applied to more and more scenes to rapidly recognize the identity of a person. The accuracy rate of the existing human face image quality evaluation method in the practical non-limiting application scene is much lower than that in the experiment, and the reason is as follows: a general method for evaluating the quality of a face picture is to subjectively select different features or attributes of the face picture, such as the number of pixels in a unit image, to evaluate the quality of the picture. However, in practical application scenarios, the recognition accuracy of a human face image which is good in accordance with human eyes subjectivity in a human face recognition algorithm is not necessarily high. Therefore, the existing human face image quality evaluation model has poor evaluation effect on the human face image quality.
Disclosure of Invention
Based on the method and the device, the objectivity of the human face image quality evaluation model on image quality evaluation can be improved, and the quality of the human face image can be better evaluated.
The scheme of the invention comprises the following steps:
in a first aspect, an embodiment of the present invention provides a method for optimizing a human face image quality evaluation model, including:
establishing a face picture test set; the test set comprises a plurality of human face pictures to be tested, and each human face picture to be tested is marked with corresponding first identity information;
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, the first identity information and the second identity information; the face database comprises a plurality of sample face pictures, and each sample face picture is marked with corresponding second identity information;
determining the quality score of each face picture to be detected according to the identification result;
and taking each face picture to be tested and the corresponding quality score thereof as training data, and training the initial face picture quality evaluation model and parameters by adopting a regression neural network to obtain an optimized face picture quality evaluation model and parameters.
In a second aspect, an embodiment of the present invention provides an apparatus for optimizing a human face image quality evaluation model, including:
the test set establishing module is used for establishing a face picture test set; the test set comprises a plurality of human face pictures to be tested, and each human face picture to be tested is marked with corresponding first identity information;
the initial identification module is used for 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, the first identity information and the second identity information; the face database comprises a plurality of sample face pictures, and each sample face picture is marked with corresponding second identity information;
the quality score calculation module is used for determining the quality score of each face picture to be detected according to the identification result; and the number of the first and second groups,
and the training module is used for training the initial human face picture quality evaluation model and parameters by using the human face pictures to be tested and the corresponding quality scores thereof as training data and adopting a regression neural network to obtain the optimized human face picture quality evaluation model and parameters.
A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the method.
A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the steps of the method when executing the program.
The technical scheme provided by the embodiment of the invention has the following beneficial effects:
establishing a face picture test set, wherein the test set comprises a plurality of face pictures to be tested, and inputting the face pictures into a pre-established face database; 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 identity information of each picture to obtain the identification result of each face picture to be detected; the quality scores of all the face pictures to be tested are determined, the face pictures to be tested in the test set and the corresponding quality scores of the face pictures to be tested are used as training data for neural network training, the quality evaluation model and parameters of the face pictures are obtained through training, the quality evaluation of the face pictures is more objective and accurate through the quality evaluation model and parameters of the face pictures, and the evaluation influence of human subjective factors on the picture quality is overcome.
Drawings
FIG. 1 is a schematic flow chart diagram of a method for optimizing a face picture quality assessment model according to an embodiment;
FIG. 2 is a schematic flow chart of a method for optimizing a face picture quality assessment model in a particular application scenario;
fig. 3 is a schematic structural diagram of an apparatus for optimizing a human face image quality evaluation model according to an embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
Although the steps in the present invention are arranged by using reference numbers, the order of the steps is not limited, and the relative order of the steps can be adjusted unless the order of the steps is explicitly stated or other steps are required for the execution of a certain step.
FIG. 1 is a schematic flow chart diagram of a method for optimizing a face picture quality assessment model according to an embodiment; as shown in fig. 1, the method for optimizing a face image quality evaluation model in this embodiment includes the steps of:
s11, establishing a face picture test set, wherein the test set comprises a plurality of face pictures to be tested, and each face picture to be tested is provided with corresponding first identity information.
The face pictures to be tested in the face picture test set are face pictures collected in a specific application scene, and can cover multiple persons. Optionally, multiple pictures per person, taken from different angles. The identity information of the face picture refers to information that can uniquely identify a person, and may be, for example, a name and a side identification number of a person.
And S12, identifying the similarity between the face picture to be detected and the sample face picture in the preset face database, and obtaining the identification result of each face picture to be detected according to the similarity, the first identity information and the second identity information.
The preset face database comprises a plurality of personal face pictures, and each picture is marked with corresponding second identity information. Optionally, one picture per person.
And comparing the face picture to be detected with the sample face picture in the preset face database, namely calculating the similarity of the two faces, and judging whether the face picture to be detected and the sample face picture are the same person or not according to the similarity and a preset similarity threshold.
In an embodiment, a specific manner of identifying similarity between a face picture to be detected and a sample face picture in a preset face database and obtaining an identification result of each face picture to be detected according to the similarity, the first identity information and the second identity information includes: respectively calculating the similarity between the face picture to be detected and each sample face picture in the face database to obtain the maximum similarity value corresponding to each face picture to be detected and the corresponding sample face picture when the similarity is maximum; acquiring first identity information of a face picture to be detected, and acquiring second identity information of a corresponding sample face picture; and obtaining the recognition result of each face picture to be detected according to the maximum similarity value, a preset similarity threshold value, the first identity information and the second identity information.
Optionally, the recognition result includes a first type result, a second type result, a third type result, and a fourth type result; specific examples thereof include: one type of result is that the face picture to be detected and the sample face picture with the same identity information are correctly identified as the same face; the second type of result is that the face picture to be detected and the sample face picture with the same identity information are identified as different faces in a wrong way; identifying the face picture to be detected and the sample face picture with different identity information as the same face if the three types of results are wrong; the four types of results are that the face picture to be detected and the sample face picture with different identity information are correctly identified as different faces.
In an optional embodiment, a face recognition algorithm is preset, and the similarity between the face picture to be detected and each sample face picture in the face database is respectively calculated, and the maximum value of the similarity corresponding to each face picture to be detected and the sample face picture corresponding to the maximum value of the similarity are obtained. By calculating the similarity between each face picture to be detected and each sample face picture in the face database, a sample face picture (also called a comparison sample face picture) with the largest similarity with the face picture to be detected and the corresponding similarity (namely the maximum similarity) can be obtained. And then identifying whether the face picture to be detected and the comparison sample face picture are the same face. Optionally, the recognition result of each face picture to be detected is obtained according to the maximum similarity value, a preset similarity threshold value and the corresponding identity information.
And S13, determining the quality score of each face picture to be detected according to the identification result.
Optionally, the quality score of each face picture to be detected is determined according to the recognition result, the maximum similarity value and the similarity threshold value, so as to facilitate neural network training.
And S14, training the face picture quality evaluation model and parameters by using the face pictures to be tested in the test set and the corresponding quality scores thereof as training data and adopting a regression neural network to obtain the optimized face picture quality evaluation model and parameters.
It can be understood that when the regression neural network is adopted for training, the model and the network parameters of the regression neural network are obtained when the preset convergence condition is met, so that the optimized human face image quality evaluation model and the optimized human face image quality evaluation parameters can be obtained.
In summary, in the method for optimizing the human face image quality evaluation model according to the embodiment, firstly, a human face image test set is established; then, obtaining the recognition result of each face picture to be detected according to the face picture test set and a preset face database; then, determining the quality score of each face picture to be detected according to the identification result; and finally, carrying out neural network training by taking the face picture to be tested in the test set and the face quality score corresponding to the face picture as training data to obtain an optimized face picture quality evaluation model and parameters. Therefore, the quality evaluation model and parameters of the face picture can be optimized, the quality evaluation of the face picture is not influenced by artificial subjective factors by adopting the optimized quality evaluation model and parameters of the face picture, and the evaluation result is more objective and accurate.
In an optional embodiment, the calculating the similarity between the face image to be detected and each sample face image in the face database by using a face recognition algorithm includes: and respectively extracting the face features of the face pictures to be detected and the face features of the sample face pictures, and calculating the similarity based on the extracted face features.
In an optional embodiment, obtaining the recognition result of each face picture to be detected according to the maximum similarity value, a preset similarity threshold value and corresponding identity information includes: comparing the maximum similarity with a preset similarity threshold to obtain a first comparison result; comparing the first identity information with the second identity information to obtain a second comparison result; and obtaining the identification result of each face picture to be detected according to the first comparison result and the second comparison result. Optionally, the recognition result includes: first class results, second class results, third class results, and fourth class results; wherein, in the first class result and the third class result, the maximum value of the similarity is greater than or equal to the similarity threshold; in the second category of results and the fourth category of results, the maximum value of the similarity is smaller than the similarity threshold.
Optionally, the implementation manner of obtaining the recognition result specifically includes:
if score > is threshold and name _ label is the same as test _ name, then it is identified as a type of result;
if score < threshold, but name _ label is the same as test _ name, then identify as a type two result;
if score > is threshold, but name _ label is different from test _ name, then three types of results are identified;
if score < threshold, and name _ label is different from test _ name, then four types of results are identified;
the score represents the maximum value of the similarity corresponding to the face picture to be detected, the threshold represents the threshold of the similarity, the name _ label represents the identity information corresponding to the sample face picture, and the test _ name represents the identity information corresponding to the face picture to be detected.
In an alternative embodiment, the way to calculate the quality score may be:
quality_score=total+total×flag×|score–threshold|/delta;
wherein 2 total represents the full score of the mass fraction; if score < threshold, then delta is threshold, otherwise delta is 1-threshold. For example, if total in the formula is 50, the full score of the quality score is 100.
Fig. 2 is a schematic flow chart of a method for optimizing a face image quality evaluation model in a specific embodiment, which is specifically divided into four parts: data set sorting S21, face recognition S22, face quality labeling S23 and face quality algorithm learning S24.
Referring to fig. 2, the method for optimizing a human face image quality evaluation model in this embodiment includes the steps of:
and S21, data arrangement.
The data arrangement part comprises: setting a face identification algorithm face _ identity _ algorithm, presetting face database face _ challenge _ set and establishing a face test set face _ test _ set.
Setting a corresponding similarity threshold for a face recognition algorithm, wherein the similarity threshold is used for judging whether two face pictures belong to a person; the preset face database comprises a plurality of sample face pictures (for example, N are larger than 1), and each sample face picture is provided with corresponding second identity information; alternatively, there is only one picture per person. The established face test set comprises a plurality of face pictures to be tested, and each face picture to be tested is provided with corresponding first identity information; optionally, the face picture to be detected is collected from face pictures of multiple people at different angles. It can be understood that the first identity information set by the face picture to be detected and the second identity information set by the sample face picture are the same type of information, for example, names. Name _ labeled of the sample face picture and name _ name of the face picture to be detected.
And S22, recognizing human faces.
The face recognition part determines the recognition result of the face picture to be detected according to the similarity threshold value threshold.
In one embodiment, the method comprises: extracting feature information of each tested face picture (assuming that the name of the current tested face picture is labeled as test _ name) by using a face recognition algorithm (the feature information comprises subjective perception feature information such as pixels, brightness and the like and other feature information which is not perceived by people); respectively comparing the extracted feature information of the tested face picture with feature information corresponding to N sample face pictures in a face database to obtain N similarity; the N similarity results are sorted, and the similarity score with the maximum similarity and the name _ labeled of the corresponding sample face picture are taken out, so that the following identification result identification _ result can be obtained:
TP (i.e. one type of outcome): if score > -threshold, and name _ label is the same as test _ name;
FP (i.e., two types of results): if score < threshold, but name _ labeled is the same as test _ name;
FN (i.e. three types of results): if score > threshold, but name _ label is different from test _ name;
TN (i.e. four types of results): if score < threshold, and name _ labeled is different from test _ name.
One type of result is that the face picture to be detected and the sample face picture with the same identity information are correctly identified as the same face; the second type of result is that the face picture to be detected and the sample face picture with the same identity information are identified as different faces in a wrong way; identifying the face picture to be detected and the sample face picture with different identity information as the same face if the three types of results are wrong; the four types of results are that the face picture to be detected and the sample face picture with different identity information are correctly identified as different faces.
And S23, marking the face quality.
And obtaining the quality score of the face test picture according to the recognition result of the step S22. And the face quality labeling is to add the calculated face quality scores to a face picture test set to obtain a face quality test set, wherein the face quality test set is a test picture set containing the calculated face quality scores.
And S24, learning a face quality algorithm.
The process uses the recurrent neural network training data to obtain recurrent neural network parameters. Specifically, the regression neural network model can be regarded as a face picture quality evaluation model to be optimized, parameters in the regression neural network and parameters of the face picture quality evaluation model, and neural network training is to train the face picture quality evaluation model and the parameters, that is, parameters related to the face picture quality evaluation model are trained by adopting the regression neural network, and when a preset convergence condition is met, the regression neural network model and the parameters are obtained, so that the optimized face picture quality evaluation model and the optimized face picture quality evaluation parameters can be obtained.
For simple representation, the above technical features respectively represent a face recognition algorithm, a face database, a face picture test set, and a similarity threshold with face _ identification _ algorithm, face _ dictionary _ set, face _ test _ set, and threshold. The above representation should not be understood as limiting the face recognition algorithm, the face database, the face picture test set, and the similarity threshold.
Based on the same idea as the method for optimizing the human face image quality evaluation model in the embodiment, the invention further provides a device for optimizing the human face image quality evaluation model, and the device can be used for executing the method for optimizing the human face image quality evaluation model. For convenience of explanation, in the schematic structural diagram of the embodiment of the apparatus for optimizing the human face image quality evaluation model, only the part related to the embodiment of the present invention is shown, and those skilled in the art will understand that the illustrated structure does not constitute a limitation to the apparatus, and may include more or less components than those illustrated, or combine some components, or arrange different components.
Fig. 3 is a schematic structural diagram of an apparatus for optimizing a human face image quality evaluation model according to an embodiment of the present invention, and as shown in fig. 3, the apparatus for optimizing a human face image quality evaluation model according to the embodiment includes: a test set establishing module 310, an initial identifying module 320, a quality score calculating module 330, and a training module 340, each of which is described in detail as follows:
the test set establishing module 310 is configured to establish a face image test set; the test set comprises a plurality of human face pictures to be tested, and each human face picture to be tested is provided with corresponding first identity information;
the initial recognition module 320 is configured to recognize similarity between a face picture to be detected and a sample face picture in a preset face database, and obtain a recognition result of each face picture to be detected according to the similarity, the first identity information, and the second identity information; the face database comprises a plurality of sample face pictures, and each sample face picture is marked with corresponding second identity information;
the quality score calculating module 330 is configured to determine a quality score of each to-be-detected face picture according to the recognition result;
the training module 340 is configured to train an initial face picture quality evaluation model and parameters by using each to-be-detected face picture and a corresponding quality score thereof as training data and using a regression neural network to obtain an optimized face picture quality evaluation model and parameters.
In an optional embodiment, in the apparatus for optimizing a face image quality evaluation model, the initial identification module 320 is configured to calculate similarities between a face image to be detected and sample face images in a face database, respectively, to obtain a maximum similarity value corresponding to each face image to be detected, and a sample face image corresponding to the maximum similarity value; acquiring first identity information of a face picture to be detected, and acquiring second identity information of a corresponding sample face picture; and obtaining the recognition result of each face picture to be detected according to the maximum similarity value, a preset similarity threshold value, the first identity information and the second identity information.
In an alternative embodiment, in the apparatus for optimizing a human face image quality evaluation model, the initial recognition module 320 may include a similarity calculation unit and a human face recognition unit.
The similarity calculation unit is used for calculating the similarity between the face picture to be detected and each sample face picture in the face database respectively, and obtaining the maximum similarity value corresponding to each face picture to be detected and the sample face picture corresponding to the maximum similarity value.
And the face recognition unit is used for obtaining the recognition result of each face picture to be detected according to the maximum similarity value, a preset similarity threshold value and the identity information corresponding to the picture.
In an optional embodiment, the similarity calculation unit may respectively extract the face features of each to-be-detected face picture and the face features of the sample face pictures, and calculate the similarity based on the extracted face features.
In an optional embodiment, the face recognition unit is configured to compare the maximum similarity with a preset similarity threshold to obtain a first comparison result; comparing the first identity information with the second identity information to obtain a second comparison result; and obtaining the identification result of each face picture to be detected according to the first comparison result and the second comparison result. Specifically, the recognition results of the face pictures to be detected can be obtained as follows:
if score > is threshold and name _ label is the same as test _ name, then it is identified as a type of result;
if score < threshold, but name _ label is the same as test _ name, then identify as a type two result;
if score > is threshold, but name _ label is different from test _ name, then three types of results are identified;
if score < threshold, and name _ label is different from test _ name, then four types of results are identified;
the score represents the maximum value of the similarity corresponding to the face picture to be detected, the threshold represents the threshold of the similarity, the name _ label represents the identity information corresponding to the sample face picture, and the test _ name represents the identity information corresponding to the face picture to be detected.
In an optional embodiment, the quality score calculating module 330 is configured to extract a maximum similarity value and a similarity threshold value corresponding to each recognition result; and determining the quality score of each face picture to be detected according to the identification result, the maximum similarity value and the similarity threshold value.
For example: the quality score calculating module 330 may calculate the quality score of each face picture to be detected according to the following formula:
quality_score=total+total×flag×|score–threshold|/delta;
wherein 2 total represents the full score of the mass fraction; if score < threshold, then delta is threshold, otherwise delta is 1-threshold.
In an optional embodiment, in the apparatus for optimizing a human face image quality evaluation model, in the test set, at least two human face images to be tested are corresponding to each identity information; in the face database, one sample face picture corresponding to each identity information is provided.
It should be noted that, in the embodiment of the apparatus for optimizing a human face image quality evaluation model in the foregoing example, because the contents of information interaction, execution process, and the like between the modules/units are based on the same concept as the foregoing method embodiment of the present invention, the technical effect brought by the contents is the same as the foregoing method embodiment of the present invention, and specific contents may refer to the description in the method embodiment of the present invention, and are not described herein again.
In addition, in the above-mentioned embodiment of the apparatus for optimizing a human face image quality evaluation model, the logical division of each program module is only an example, and in practical applications, the above-mentioned function allocation may be performed by different program modules according to needs, for example, due to the configuration requirements of corresponding hardware or the convenience of implementation of software, that is, the internal structure of the apparatus for optimizing a human face image quality evaluation model is divided into different program modules to perform all or part of the above-described functions.
It will be understood by those skilled in the art that all or part of the processes of the methods of the above embodiments may be implemented by a computer program, which is stored in a computer readable storage medium and sold or used as a stand-alone product. The program, when executed, may perform all or a portion of the steps of the embodiments of the methods described above. In addition, the storage medium may be provided in a computer device, and the computer device further includes a processor, and when the processor executes the program in the storage medium, all or part of the steps of the embodiments of the methods described above can be implemented. The storage medium may be a magnetic disk, an optical disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), or the like.
In the above embodiments, the descriptions of the respective embodiments have respective emphasis, and for parts that are not described in detail in a certain embodiment, reference may be made to related descriptions of other embodiments. It will be understood that the terms "first," "second," and the like as used herein are used herein to distinguish one object from another, but the objects are not limited by these terms.
The above-described examples merely represent several embodiments of the present invention and should not be construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the inventive concept, which falls within the scope of the present invention. Therefore, the protection scope of the present patent shall be subject to the appended claims.
Claims (10)
1. A method for optimizing a human face image quality evaluation model is characterized by comprising the following steps:
establishing a face picture test set; the test set comprises a plurality of human face pictures to be tested, and each human face picture to be tested is marked with corresponding first identity information;
respectively extracting the face features of each face picture to be detected and the face features of the sample face pictures, and calculating the similarity based on the extracted face features to obtain the maximum similarity value corresponding to each face picture to be detected and the sample face picture corresponding to the maximum similarity value;
acquiring first identity information of a face picture to be detected, and acquiring second identity information of a corresponding sample face picture; obtaining the recognition result of each human face picture to be detected according to the maximum similarity value, a preset similarity threshold value, the first identity information and the second identity information; the recognition result comprises: one type of result is that the face picture to be detected and the sample face picture with the same identity information are correctly identified as the same face; the second type of result is that the face picture to be detected and the sample face picture with the same identity information are identified as different faces in a wrong way; identifying the face picture to be detected and the sample face picture with different identity information as the same face if the three types of results are wrong; the four types of results are that the face picture to be detected and the sample face picture with different identity information are correctly identified as different faces; wherein, in the first class result and the third class result, the maximum value of the similarity is greater than or equal to the similarity threshold; in the second type of result and the fourth type of result, the maximum value of the similarity is smaller than the similarity threshold value;
extracting a similarity maximum value and a similarity threshold value corresponding to each recognition result;
determining the quality score of each face picture to be detected according to the identification result, the maximum similarity value and the similarity threshold value;
and taking each face picture to be tested and the corresponding quality score thereof as training data, and training the initial face picture quality evaluation model and parameters by adopting a regression neural network to obtain an optimized face picture quality evaluation model and parameters.
2. The method for optimizing the human face image quality evaluation model according to claim 1, wherein obtaining the recognition result of each human face image to be detected according to the maximum similarity value, a preset similarity threshold value, the first identity information and the second identity information comprises:
comparing the maximum similarity with a preset similarity threshold to obtain a first comparison result;
comparing the first identity information with the second identity information to obtain a second comparison result;
and obtaining the identification result of each face picture to be detected according to the first comparison result and the second comparison result.
3. The method for optimizing a human face image quality evaluation model according to claim 2, wherein the recognition result comprises:
one type of result: if score > -threshold, and name _ label is the same as test _ name;
the second type of results: if score < threshold, but name _ labeled is the same as test _ name;
three types of results: if score > threshold, but name _ label is different from test _ name;
four types of results: if score < threshold, and name _ labeled is different from test _ name;
wherein, score represents the maximum value of the similarity corresponding to the face picture to be detected, threshold represents the threshold of the similarity, name _ label is the name of the sample face picture, and test _ name is the name of the face picture to be detected.
4. The method for optimizing a human face image quality evaluation model according to claim 1, further comprising:
and (3) human face quality labeling: and adding the calculated face quality score into the face picture test set to obtain a face quality test set.
5. The method for optimizing the human face image quality evaluation model according to claim 3, wherein the quality score of each human face image to be tested is calculated by the following formula:
quality_score=total+total×flag×|score–threshold|/delta;
wherein 2 total represents the full score of the mass fraction; if score < threshold, then delta is threshold, otherwise delta is 1-threshold; quality _ score represents a quality score, score represents a maximum value of similarity corresponding to the face picture to be detected, and threshold represents a similarity threshold.
6. The method for optimizing a human face image quality evaluation model according to any one of claims 1 to 5,
in the test set, at least two face pictures to be tested corresponding to the same identity information are obtained; and in the face database, one sample face picture corresponding to the same identity information is used.
7. The method for optimizing a human face image quality evaluation model according to claim 6,
the identity information refers to information for uniquely identifying the identity of the face.
8. An apparatus for optimizing a human face image quality evaluation model, comprising:
the test set establishing module is used for establishing a face picture test set; the test set comprises a plurality of human face pictures to be tested, and each human face picture to be tested is marked with corresponding first identity information;
the initial identification module is used for respectively extracting the face features of each face picture to be detected and the face features of the sample face pictures, calculating the similarity based on the extracted face features, and obtaining the maximum value of the similarity corresponding to each face picture to be detected and the sample face picture corresponding to the maximum value of the similarity; acquiring first identity information of a face picture to be detected, and acquiring second identity information of a corresponding sample face picture; obtaining the recognition result of each human face picture to be detected according to the maximum similarity value, a preset similarity threshold value, the first identity information and the second identity information; the recognition result comprises: one type of result is that the face picture to be detected and the sample face picture with the same identity information are correctly identified as the same face; the second type of result is that the face picture to be detected and the sample face picture with the same identity information are identified as different faces in a wrong way; identifying the face picture to be detected and the sample face picture with different identity information as the same face if the three types of results are wrong; the four types of results are that the face picture to be detected and the sample face picture with different identity information are correctly identified as different faces; wherein, in the first class result and the third class result, the maximum value of the similarity is greater than or equal to the similarity threshold; in the second type of result and the fourth type of result, the maximum value of the similarity is smaller than the similarity threshold value;
the quality score calculation module is used for extracting the similarity maximum value and the similarity threshold value corresponding to each recognition result; determining the quality score of each face picture to be detected according to the identification result, the maximum similarity value and the similarity threshold value; and the number of the first and second groups,
and the training module is used for training the initial human face picture quality evaluation model and parameters by using the human face pictures to be tested and the corresponding quality scores thereof as training data and adopting a regression neural network to obtain the optimized human face picture quality evaluation model and parameters.
9. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 7.
10. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the steps of the method of any of claims 1 to 7 are implemented when the program is executed by the processor.
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 CN107609493A (en) | 2018-01-19 |
CN107609493B true 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) |
Families Citing this family (25)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109145717B (en) * | 2018-06-30 | 2021-05-11 | 东南大学 | Face recognition method for online learning |
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 |
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 |
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 |
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 |
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 |
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 |
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 |
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 |
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 |
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 |
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 |
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 |
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 |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2012109712A1 (en) * | 2011-02-18 | 2012-08-23 | National Ict Australia Limited | Image quality assessment |
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 (6)
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 |
JP5988577B2 (en) * | 2011-12-28 | 2016-09-07 | キヤノン株式会社 | Image coding apparatus, image coding method, and program |
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 (6)
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 |
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 (1)
Title |
---|
人脸图像质量评估标准方法研究;高修峰;《中国博士学位论文全文数据库 信息科技辑》;20090615;第2009年卷(第6期);全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN107609493A (en) | 2018-01-19 |
WO2019037346A1 (en) | 2019-02-28 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107609493B (en) | Method and device for optimizing human face image quality evaluation model | |
CN109522815B (en) | Concentration degree evaluation method and device and electronic equipment | |
Wu et al. | The machine knows what you are hiding: an automatic micro-expression recognition system | |
CN109740446A (en) | Classroom students ' behavior analysis method and device | |
KR101725651B1 (en) | Identification apparatus and method for controlling identification apparatus | |
Bekhouche et al. | Facial age estimation and gender classification using multi level local phase quantization | |
Mathe et al. | Action from still image dataset and inverse optimal control to learn task specific visual scanpaths | |
US11804071B2 (en) | Method for selecting images in video of faces in the wild | |
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 | |
CN111783505A (en) | Method and device for identifying forged faces and computer-readable storage medium | |
JP2004348674A (en) | Region detection method and its device | |
JP2019057815A (en) | Monitoring system | |
CN110418204B (en) | Video recommendation method, device, equipment and storage medium based on micro expression | |
CN110751069A (en) | Face living body detection method and device | |
Prudêncio et al. | Analysis of instance hardness in machine learning using item response theory | |
CN107578015B (en) | First impression recognition and feedback system and method based on deep learning | |
Boyd et al. | Cyborg: Blending human saliency into the loss improves deep learning | |
Jilani et al. | A machine learning approach for ethnic classification: the British Pakistani face | |
Neto et al. | Myope models-are face presentation attack detection models short-sighted? | |
CN113705310A (en) | Feature learning method, target object identification method and corresponding device | |
Algaraawi et al. | Study on aging effect on facial expression recognition | |
CN111222374A (en) | Lie detection data processing method and device, computer equipment and storage medium | |
CN113918471A (en) | Test case processing method and device and computer readable storage medium | |
CN111160263B (en) | Method and system for acquiring face recognition threshold | |
Tavakolian et al. | Face recognition under occlusion for user authentication and invigilation in remotely distributed online assessments |
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 |