CN109063656B - Method and device for carrying out face query by using multiple face engines - Google Patents
Method and device for carrying out face query by using multiple face engines Download PDFInfo
- Publication number
- CN109063656B CN109063656B CN201810895823.3A CN201810895823A CN109063656B CN 109063656 B CN109063656 B CN 109063656B CN 201810895823 A CN201810895823 A CN 201810895823A CN 109063656 B CN109063656 B CN 109063656B
- Authority
- CN
- China
- Prior art keywords
- face
- similarity
- engine
- faces
- ranking
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/30—Scenes; Scene-specific elements in albums, collections or shared content, e.g. social network photos or video
Abstract
The invention discloses a method and a device for inquiring human faces by utilizing a plurality of human face engines, wherein the human face inquiring method comprises the following steps: s1: respectively using m personal faces to inquire a certain face to obtain m groups of k faces with the most images; s2: carrying out first normalization processing on the similarity of the k most similar faces of each face engine by using a normalization function to obtain a set of m similarities; s3: and combining the m groups of the k faces which are most similar, configuring different ranking weights for each ranking position, and respectively calculating the total ranking weight and the maximum similarity of each face in the t faces. S4: and carrying out second normalization processing on the maximum similarity according to the ranking total weight of the t comprehensive similarities to obtain a set of t comprehensive similarities. S5: and (4) sequencing the set of the comprehensive similarity in a descending order again, and intercepting the front k faces to serve as a final query result. The invention can realize the face query by using a plurality of face engines.
Description
Technical Field
The invention relates to the field of face engines, in particular to a method and a device for carrying out face query by utilizing a plurality of face engines.
Background
With the development of artificial intelligence, the human face image comparison is more and more accurate, and the recognition rate is higher and higher, but due to the difference of model algorithms of various human face engines or the difference of training data, the query results of the human face engines of various models have certain difference. Different engines have respective advantages and disadvantages for different angles of faces and different quality pictures of faces, and the query accuracy is different.
The current method for judging the accuracy of the recognition result of the face engine generally adopts the following modes: firstly, a face image is uploaded by a face engine, and the similarity of k faces and the k faces which are the most similar in n faces in a bottom library is obtained through query. And then, a set is made for the query results obtained by the face engines, t faces exist in the set, t is more than or equal to k and less than or equal to m and k, and the final comprehensive similarity is obtained according to the weight coefficient of the face engine which is configured and evaluated in advance and the obtained similarity. The comprehensive similarity of t faces is calculated by the following formula
Wherein s isiThe similarity corresponding to the face is returned by the ith engine.
aiWeight coefficients for evaluating the face engine are configured in advance and
results stAnd finally obtaining t comprehensive similarities, and sequencing and intercepting the front k faces in a descending order to obtain the k faces which are the most similar in the comprehensive query results of the face engines.
The ability to have the contrast similarity between different faces of the same person as high as possible over the contrast similarity between faces of different persons is called the segmentation ability of the engine. The segmentation capability of each face engine is different, and if each face engine is aggregated according to a fixed weight, an engine with insufficient segmentation capability affects an engine with good segmentation capability. And the recognition weight coefficient a of each face engineiThe method is obtained by testing or observing the model capability, and the model capability is difficult to be fully expressed in limited test data, so that the comprehensive similarity results of the finally obtained query results of various human face engines are very different. Therefore, it is an urgent need to solve the problem of providing an evaluation standard and method for evaluating the accuracy of various face engine query results.
Disclosure of Invention
In view of the above, the present invention provides a method for performing face query by using multiple face engines, comprising the following steps:
s1: respectively using m personal face query engines to query a certain face in a face library, and respectively returning k faces with the most images, thereby obtaining m groups of k faces with the most images;
s2: carrying out first normalization processing on the similarity of the k most similar faces of each face engine by using a normalization function to obtain a set of m similarities; the normalization processing can eliminate the influence on the data analysis result caused by different segmentation capabilities in different face engines.
S3: combining the k faces of the m groups of the most images to obtain t faces, wherein t is more than or equal to k and less than or equal to m x k, recording the ranking and similarity of the t faces in each face engine respectively, configuring different ranking weights for each ranking position, and calculating the total ranking weight and the maximum similarity of each face in the t faces respectively; and further processing the ranking weight and the similarity of each face.
S4: aiming at each face in the t faces, carrying out second normalization processing on the maximum similarity according to the ranking total weight of the face to obtain t sets of comprehensive similarities; and normalization processing is carried out again, so that the influence of different ranking weights of each face in a face engine on the comprehensive similarity can be eliminated.
S5: and performing descending order sorting on the t human faces according to the obtained t comprehensive similarity sets, and intercepting the front k human faces in the t human faces to serve as a final query result.
In a further embodiment, the first normalization function takes the following function:
wherein X is input similarity, Y is normalized output similarity, w and b are parameters, e is a natural base number, each face engine corresponds to different parameters w and b, wherein b is the output similarity of each face engine representing the same person, w is the similarity of the same person after normalization, and the parameter w selects the same value when each face engine is initialized.
In a further embodiment, the ranking and similarity of t faces in each face engine, i.e. the ranking and similarity of t faces in step S3 are recorded
The 1 st face, the ranking of < face engine 1, similarity > the ranking of < face engine i, similarity > the ranking of < face engine m, similarity >;
...
the t-th face, < alignment of face engine 1, similarity > the alignment of face engine i, similarity >. the alignment of face engine m, similarity >.
Therefore, t faces have t × m groups of rows and similarity.
In a further embodiment, the ranking total weight rank of the t-th face in the face engine is calculated in step S3tMaximum value s of face similarity with t-th facetNamely:
st＝max{sti}1≤i≤mwherein s istiAnd (5) similarity of the t-th face in the ith engine.
The 1 st face of the person is displayed,<rank1>，<s1>；
...
the jth face of the person is presented,<rankj>，<sj>；
...
the t-th human face is displayed,<rankt>，<st>. Therefore, the maximum value of the ranking total weight and the face similarity of the t-th face in the face engine is obtained.
In a further embodiment, the second normalization function takes the following function:
wherein X is input similarity, Y is normalized output similarity, w and b are parameters, e is a natural base number, b is reference similarity, a value close to the initial value of w in the first-time normalization function can be selected, w represents the similarity value of the same person or not in final comprehensive judgment, and rank is the sum of ranking configuration weights of each face in each engine. And further improving all parameters by adopting quadratic normalization.
In a further embodiment, the parameters for determining the second normalization function include the total ranking weight of the jth face, the maximum value of the similarity of the jth face in the face engine, and fixed parameters, and the obtained result is a comprehensive similarity set of the t faces in the face engine.
In a further embodiment, the final query result samples are subjected to manual verification, and the first normalization function parameters of each face engine are adjusted according to the judgment whether the head face of the final query result and the query face are the same person. The first normalization function used by each face engine is more accurate.
In a further embodiment, the parameters w and b of the first normalization function are adjusted as follows: calculating the average value of the ranks of the head face in the t faces in the m personal face engines according to the ranks of the head face in the t faces in the m personal face engines; and adjusting parameters w and b of the normalization function of the face engine according to the size relation between the ranking of the head face in the t faces and the average value in a certain face engine so as to adjust the weight of the face engine in the method. The adjustment makes the result obtained by the normalization function more accurate.
In a further embodiment, the parameters w and b of the first normalization function are adjusted according to the judgment of the m personal face engines one by one, and if the ranking of the current face engine is smaller than the average ranking and the normalized output similarity of the current face engine is smaller, the parameter w of the normalization function of the current face engine is increased or the parameter b of the normalization function of the current face engine is decreased. The values of the parameters w and b are adjusted by means of this idea.
The invention also provides a device for inquiring human faces by using a plurality of human face engines, which comprises the following steps:
the resource set acquisition unit is configured to acquire a resource set of the face engine query result, and the resource set comprises a plurality of face engine data;
the first normalization processing unit is configured to perform first normalization processing on the resource set obtained by the resource set obtaining unit and the similarity of each face engine respectively;
the second normalization processing unit is configured to perform second normalization processing on the maximum similarity of each face according to the ranking total weight of each face in the face engines;
and the analysis processing unit is configured to calculate the ranking total weight and the maximum similarity of each face in the face engines, and rank each comprehensive similarity to obtain a final query result.
In a further embodiment, the first-time normalization processing unit includes a feedback adjustment unit configured to adjust parameters used by the first-time normalization processing unit according to a final query result obtained by the analysis processing unit.
The present invention also provides a computer device, including a memory, a processor, and a program stored in the memory and executable on the processor, wherein the processor implements any one of the methods described above when executing the program.
The invention also proposes a computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the method of any one of the above-mentioned.
The invention provides a comprehensive score calculation method for query results by utilizing various face engines, which is characterized in that the similarity obtained by the face engines is normalized, and then the ranking total weight in the face engines and the maximum value of the similarity of the face engines are further normalized. Therefore, the comprehensive score calculation method does not use each model to aggregate according to a certain weight, but adopts a result ranking voting mode to aggregate, so that the overall segmentation cannot be reduced due to the model with low segmentation capability, the problem that various face engines complement each other in the ranking capability and the segmentation capability is solved, and the comprehensive effect of the plurality of face engines is better than that of any face engine. In addition, feedback adjustment is carried out on the normalization function, and a feedback mechanism is added to dynamically adjust the normalization parameters of each engine. The scheme can be suitable for the comprehensive score calculation of a plurality of face comparison engines. The method is also suitable for the calculation of the comprehensive scores of other similar similarity query engines, such as the calculation of the comprehensive scores of a plurality of fingerprint identification engines and the calculation of the comprehensive scores of a plurality of voice identification engines, and has extremely high popularization and applicability.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
FIG. 1 is a schematic diagram of a method for performing face query using multiple face engines according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of an apparatus for performing face query using multiple face engines according to an embodiment of the present invention;
fig. 3 is a schematic diagram of a hardware structure of a computer device for performing face query by using multiple face engines according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention clearer, the present invention will be described in further detail with reference to the accompanying drawings, and it is apparent that the described embodiments are only a part of the embodiments of the present invention, not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention provides a method for inquiring human faces by using a plurality of human face engines, which comprises the following steps as shown in figure 1:
s1: and respectively querying a certain face in a face library by using m personal face query engines, and respectively returning k faces with the most images, thereby obtaining m groups of k faces with the most images.
S2: and carrying out first normalization processing on the similarity of the k most similar faces of each face engine by using a normalization function to obtain a set of m similarities.
In a preferred embodiment, the first normalization function takes the following function:
wherein X is input similarity, Y is normalized output similarity, w and b are parameters, e is a natural base number, each face engine corresponds to different parameters w and b, wherein b is that each face engine represents output similarity of the same person, w represents similarity of the same person after normalization, and the parameter w selects the same value when each face engine is initialized. Only one of the suitable normalization functions is provided in this implementation. In alternative embodiments, other realizable normalization functions may also be selected.
S3: and merging the k faces of the obtained m groups of the most images to obtain t faces, wherein t is more than or equal to k and less than or equal to m x k. Recording the ranking and the similarity of the t faces in each face engine respectively, configuring different ranking weights for each ranking position, and calculating the total ranking weight and the maximum similarity of each face in the t faces respectively. Recording the ranking and similarity of t faces in each face engine respectively, i.e.
The 1 st face, the ranking of < face engine 1, similarity > the ranking of < face engine i, similarity > the ranking of < face engine m, similarity >; k is not less than t not more than m
...
The t-th face, < alignment of face engine 1, similarity > the alignment of face engine i, similarity >. the alignment of face engine m, similarity >.
Calculating the ranking total weight rank of the t-th face in the face enginetMaximum value s of face similarity with t-th facetNamely:
st＝max{sti}1≤i≤mwherein s istiAnd (5) similarity of the t-th face in the ith engine.
The 1 st face of the person is displayed,<rank1>，<s1>；
...
the jth face of the person is presented,<rankj>，<sj>；
...
the t-th human face is displayed,<rankt>，<st>。
in a preferred embodiment, the weights of the rank configuration are represented by a piecewise linear function, i.e., assuming that the first rank bit weight is reused at 4.0, the second rank bit weight is used at 3.8, the third rank bit weight is used at 3.6, and so on until the weight is 0.
S4: and aiming at each face in the t faces, carrying out second normalization processing on the maximum similarity according to the ranking total weight of the face to obtain t sets of comprehensive similarities.
The second normalization function takes the following function:
wherein X is input similarity, Y is normalized output similarity, w and b are parameters, e is a natural base number, b is reference similarity, a value close to the initial value of w in the first-time normalization function can be selected, w represents the similarity value of the same person or not in final comprehensive judgment, and rank is the sum of ranking configuration weights of each face in each engine.
In a preferred embodiment, the second normalization function is chosen the same as the first normalization function. And determining parameters of a second normalization function, wherein the parameters comprise the total ranking weight of the jth face, the maximum value of the similarity of the jth face in the face engine and fixed parameters, and the obtained result is a comprehensive similarity set of the t faces in the face engine. w and b select an appropriate set of fixed parameters, and stAnd normalizing the input similarity X to obtain the comprehensive similarity Y. In an alternative embodiment, the second normalization function may also be selected to be a different normalization function than the first normalization function.
S5: and performing descending order sorting on the t human faces according to the obtained t comprehensive similarity sets, and intercepting the front k human faces in the t human faces to serve as a final query result.
And S6, performing manual check on the final query result samples, and adjusting the first normalization function parameters of each face engine according to the judgment whether the head face and the query face of the final query result are the same person. And a feedback mechanism is added to dynamically adjust the parameters of each model, so that the reliability of the final result is high, and the final result is more accurate.
The adjustment rules of the parameters w and b of the first normalization function are as follows: and (4) ranking the first face in the t faces in the m personal face engine. The average of the ranks of the top face of the t faces in the m personal face engine is calculated. And adjusting parameters w and b of the normalization function of the face engine according to the size relation between the ranking of the head face in the t faces and the average value in a certain face engine so as to adjust the weight of the face engine in the method. The values of the parameters w and b are adjusted according to different selected face engines. When the result of the adjustment and the result of the last adjustment are within a certain threshold range, the adjustment of the parameters w and b is close to reasonable, the final result after the two adjustments is not greatly different, and therefore the adjustment of the parameters w and b can be stopped.
And adjusting the parameters w and b according to the first normalization function, firstly judging m personal face engines one by one, and if the ranking of the current face engine is smaller than the average ranking and the normalized output similarity of the current face engine is smaller, increasing the parameter w of the normalization function of the current face engine or reducing the parameter b of the normalization function of the current face engine. In alternative embodiments, if other normalization functions are selected, their corresponding parameter values may be adjusted accordingly.
In a preferred implementation, the first and second normalization functions are selected, and an a engine and a B engine are selected as examples, where the magnitude of parameter B of the first normalization a engine is 0.55, and the magnitude of parameter w is 0.87; the size of the parameter B of the first normalization B engine is 0.65, and the size of the parameter w is 0.87; the fact that the normal similarity of the engine A exceeds 55% and is represented as the same person, the normal similarity of the engine B exceeds 65% and is represented as the same person is assumed, and the output similarity of the engine after the first normalization exceeds 87% and is represented as the same person, so that the output similarities of the engine A and the engine B after the first normalization are the same, and the influence of different similarity selected by different face engines on results can be reduced. The second normalization parameter b has a magnitude of 0.8 and the parameter w has a magnitude of 0.92. When the second normalization is performed, 80% is selected as the reference similarity, and the finally obtained result similarity is not less than 92%, namely, the result is the same person.
The present invention further provides an apparatus 200 for performing face query by using multiple face engines, as shown in fig. 2, including:
a resource set obtaining unit 201 configured to obtain a resource set 2011 of the face engine query result, where the resource set 2011 includes a plurality of face engine data;
a first normalization processing unit 202, configured to perform first normalization processing on the resource set 2011 obtained by the resource set obtaining unit 201 and the similarity of each face engine respectively;
a second normalization processing unit 203, configured to perform a second normalization processing on the maximum similarity of each face according to the ranking total weight of each face in the plurality of face engines;
and the analysis processing unit 204 is configured to calculate the ranking total weight and the maximum similarity of each face in the plurality of face engines, and rank each comprehensive similarity to obtain a final query result.
In a further embodiment, the first-time normalization processing unit 202 includes a feedback adjusting unit 2021, and the feedback adjusting unit 2021 is configured to adjust parameters used by the first-time normalization processing unit 202 according to a final query result obtained by the analysis processing unit 204.
The method and apparatus for performing face query using multiple face engines according to the embodiment of the present invention described in conjunction with fig. 1 to 2 can be implemented by a computer device 300. Fig. 3 is a schematic hardware structure diagram of a computer device 300 according to an embodiment of the present invention.
The computer device 300 comprises a memory 301, a processor 302 and a computer program stored on the memory 301 and executable on the processor 302.
In one example, the processor 302 described above may include a Central Processing Unit (CPU), or A Specific Integrated Circuit (ASIC), or one or more integrated circuits that may be configured to implement embodiments of the present invention.
Memory 301 may include mass storage for data or instructions. By way of example, and not limitation, memory 301 may include an HDD, floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or Universal Serial Bus (USB) drive, or a combination of two or more of these. Memory 301 may include removable or non-removable (or fixed) media, where appropriate. The memory 301 may be internal or external to the computer device 300, where appropriate. In a particular embodiment, the memory 301 is a non-volatile solid-state memory. In certain embodiments, memory 301 comprises Read Only Memory (ROM). Where appropriate, the ROM may be mask-programmed ROM, Programmable ROM (PROM), Erasable PROM (EPROM), Electrically Erasable PROM (EEPROM), electrically rewritable ROM (EAROM), or flash memory or a combination of two or more of these.
The processor 302 runs a program corresponding to the executable program code stored in the memory 301 by reading the executable program code for performing the synonym mining method in the above-described embodiments.
In one example, computer device 300 may also include a communication interface 303 and a bus 304. As shown in fig. 3, the memory 301, the processor 302, and the communication interface 303 are connected via a bus 304 to complete communication therebetween.
The communication interface 303 is mainly used for implementing communication between modules, apparatuses, units and/or devices in the embodiment of the present invention. The communication interface 303 may also access input devices and/or output devices.
Bus 304 comprises hardware, software, or both coupling the components of computer device 300 to one another. By way of example, and not limitation, the bus 304 may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an infiniband interconnect, a Low Pin Count (LPC) bus, a memory bus, a Micro Channel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a video electronics standards association local (VLB) bus, or other suitable bus, or a combination of two or more of these. Bus 304 may include one or more buses, where appropriate. Although specific buses have been described and shown in the embodiments of the invention, any suitable buses or interconnects are contemplated by the invention.
An embodiment of the present invention further provides a computer-readable storage medium, where a program is stored on the computer-readable storage medium, and when the program is executed by a processor, the program implements the method for performing face query by using multiple face engines in the foregoing embodiments.
It will be apparent to those skilled in the art that various modifications and variations can be made to the embodiments of the present invention without departing from the spirit and scope of the invention. In this way, if these modifications and changes are within the scope of the claims of the present invention and their equivalents, the present invention is also intended to cover these modifications and changes. The word "comprising" does not exclude the presence of other elements or steps than those listed in a claim. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage. Any reference signs in the claims shall not be construed as limiting the scope.
Claims (11)
1. A method for inquiring human faces by using a plurality of human face engines is characterized by comprising the following steps:
s1: respectively using m personal face query engines to query a certain face in a face library, and respectively returning k faces with the most images, thereby obtaining m groups of k faces with the most images;
s2: carrying out first normalization processing on the similarity of the k most-image faces of each face engine by using a normalization function to obtain a set of m similarities, wherein the parameter w and b adjustment rules in the first normalization function adopted in the first normalization processing are as follows: calculating the average value of the ranks of the head face in the t faces in the m personal face engines according to the ranks of the head face in the t faces in the m personal face engines; according to the size relation between the ranking of the head face in a certain face engine and the average value in the t faces, adjusting parameters w and b of a normalization function of the face engine to adjust the weight of the face engine in the method, wherein the first normalization function adopts the following function:
s3: combining the k faces of the m groups of the most images to obtain t faces, wherein t is more than or equal to k and less than or equal to m x k, recording the ranking and similarity of the t faces in each face engine respectively, configuring different ranking weights for each ranking position, and calculating the total ranking weight and the maximum similarity of each face in the t faces respectively;
s4: aiming at each face in the t faces, carrying out second normalization processing on the maximum similarity according to the ranking total weight of the face to obtain t sets of comprehensive similarities;
s5: and performing descending order sorting on the t human faces according to the obtained t comprehensive similarity sets, and intercepting the front k human faces in the t human faces to serve as a final query result.
2. The method of claim 1, wherein the ranking and similarity of t faces in each face engine, namely, ranking and similarity of t faces in each face engine, are recorded in step S3
The 1 st face, the ranking of < face engine 1, similarity > the ranking of < face engine i, similarity > the ranking of < face engine m, similarity >;
...
the t-th face, < alignment of face engine 1, similarity > the alignment of face engine i, similarity >. the alignment of face engine m, similarity >.
3. The method of claim 2, wherein the total ranking weight rank of the t-th face in the face engines is calculated in step S3tMaximum value s of face similarity with t-th facetNamely:
st＝max{sti}1≤i≤mwherein s istiSimilarity of the t-th face in the ith engine;
the 1 st face of the person is displayed,<rank1>，<s1>；
...
the jth personThe face of the user is provided with a face,<rankj>，<sj>；
...
the t-th human face is displayed,<rankt>，<st>。
4. a method for performing face query using multiple face engines according to any of claims 1-3, wherein the second normalization process uses the following function:
wherein X is input similarity, Y is normalized output similarity, w and b are parameters, e is a natural base number, b is reference similarity, a value close to the initial value of w in the first-time normalization function can be selected, w represents the similarity value of the same person or not in final comprehensive judgment, and rank is the sum of ranking configuration weights of each face in each engine.
5. The method of claim 4, wherein the parameters defining the second normalization function include a total ranking weight of the jth face, a maximum similarity of the jth face in the face engines, and the parameters, and the result is a comprehensive similarity set of the t faces in the face engines.
6. The method of claim 1, wherein the final query result samples are manually verified, and the first normalization function parameters of each face engine are adjusted according to whether the top face of the final query result and the query face are the same person.
7. The method of claim 1, wherein the parameters w and b of the first normalization function are adjusted according to the judgment of m individual face engines one by one, and if the ranking of the current face engine is smaller than the average ranking and the normalized output similarity of the current face engine is smaller, the parameter w of the normalization function of the current face engine is increased or the parameter b of the normalization function of the current face engine is decreased.
8. An apparatus for performing face queries using a plurality of face engines, comprising:
a resource set obtaining unit configured to obtain a resource set of face engine query results, the resource set including a plurality of face engine data;
the first normalization processing unit is configured to perform first normalization processing on the resource set obtained by the resource set obtaining unit and the similarity of each face engine respectively, and the adjustment rules of parameters w and b in a first normalization function adopted in the first normalization processing are as follows: calculating the average value of the ranks of the head face in the t faces in the m personal face engines according to the ranks of the head face in the t faces in the m personal face engines; according to the size relation between the ranking of the head face in a certain face engine and the average value, adjusting parameters w and b of a normalization function of the face engine to adjust the weight of the face engine in the device, wherein the first normalization function adopts the following function:
the second normalization processing unit is configured to perform second normalization processing on the maximum similarity of each face according to the ranking total weight of each face in the face engines;
and the analysis processing unit is configured to calculate the ranking total weight and the maximum similarity of each face in the face engines, and rank each comprehensive similarity to obtain a final query result.
9. The apparatus of claim 8, wherein the first normalization processing unit comprises a feedback adjustment unit configured to adjust parameters used by the first normalization processing unit according to a final query result obtained by the analysis processing unit.
10. A computer device comprising a memory, a processor and a program stored on the memory and executable on the processor, the processor implementing the method of any one of claims 1-7 when executing the program.
11. A computer-readable storage medium, on which a computer program is stored, which program, when being executed by a processor, is adapted to carry out the method of any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810895823.3A CN109063656B (en) | 2018-08-08 | 2018-08-08 | Method and device for carrying out face query by using multiple face engines |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810895823.3A CN109063656B (en) | 2018-08-08 | 2018-08-08 | Method and device for carrying out face query by using multiple face engines |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109063656A CN109063656A (en) | 2018-12-21 |
| CN109063656B true CN109063656B (en) | 2021-08-24 |
Family
ID=64678625
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810895823.3A Active CN109063656B (en) | 2018-08-08 | 2018-08-08 | Method and device for carrying out face query by using multiple face engines |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109063656B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110222823B (en) * | 2019-05-31 | 2022-12-30 | 甘肃省祁连山水源涵养林研究院 | Hydrological flow fluctuation situation identification method and system |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106815566A (en) * | 2016-12-29 | 2017-06-09 | 天津中科智能识别产业技术研究院有限公司 | A kind of face retrieval method based on multitask convolutional neural networks |
| CN108052864A (en) * | 2017-11-17 | 2018-05-18 | 平安科技（深圳）有限公司 | Face identification method, application server and computer readable storage medium |
| CN108197250A (en) * | 2017-12-29 | 2018-06-22 | 深圳云天励飞技术有限公司 | Picture retrieval method, electronic equipment and storage medium |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP5358083B2 (en) * | 2007-11-01 | 2013-12-04 | 株式会社日立製作所 | Person image search device and image search device |
| CN100464332C (en) * | 2007-03-20 | 2009-02-25 | 北京中星微电子有限公司 | Picture inquiry method and system |
| CN101398832A (en) * | 2007-09-30 | 2009-04-01 | 国际商业机器公司 | Image searching method and system by utilizing human face detection |
| CN102855245A (en) * | 2011-06-28 | 2013-01-02 | 北京百度网讯科技有限公司 | Image similarity determining method and image similarity determining equipment |
| CN106933861A (en) * | 2015-12-30 | 2017-07-07 | 北京大唐高鸿数据网络技术有限公司 | A kind of customized across camera lens target retrieval method of supported feature |
| CN107958073B (en) * | 2017-12-07 | 2020-07-17 | 电子科技大学 | Particle cluster algorithm optimization-based color image retrieval method |
-
2018
- 2018-08-08 CN CN201810895823.3A patent/CN109063656B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106815566A (en) * | 2016-12-29 | 2017-06-09 | 天津中科智能识别产业技术研究院有限公司 | A kind of face retrieval method based on multitask convolutional neural networks |
| CN108052864A (en) * | 2017-11-17 | 2018-05-18 | 平安科技（深圳）有限公司 | Face identification method, application server and computer readable storage medium |
| CN108197250A (en) * | 2017-12-29 | 2018-06-22 | 深圳云天励飞技术有限公司 | Picture retrieval method, electronic equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109063656A (en) | 2018-12-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2019120115A1 (en) | Facial recognition method, apparatus, and computer apparatus | |
| Sellahewa et al. | Image-quality-based adaptive face recognition | |
| KR101725651B1 (en) | Identification apparatus and method for controlling identification apparatus | |
| CN111052131B (en) | Authentication device, authentication system, authentication method, and storage medium | |
| CN111967392A (en) | Face recognition neural network training method, system, equipment and storage medium | |
| CN109711358B (en) | Neural network training method, face recognition system and storage medium | |
| EP2523149A2 (en) | A method and system for association and decision fusion of multimodal inputs | |
| JP5710748B2 (en) | Biometric authentication system | |
| CN109934300B (en) | Model compression method, device, computer equipment and storage medium | |
| US11126827B2 (en) | Method and system for image identification | |
| CN102216958A (en) | Object detection device and object detection method | |
| CN109063656B (en) | Method and device for carrying out face query by using multiple face engines | |
| WO2021051602A1 (en) | Lip password-based face recognition method and system, device, and storage medium | |
| Ferizal et al. | Gender recognition using PCA and LDA with improve preprocessing and classification technique | |
| CN109858328B (en) | Face recognition method and device based on video | |
| CN116958868A (en) | Method and device for determining similarity between text and video | |
| US11437044B2 (en) | Information processing apparatus, control method, and program | |
| CN112907541B (en) | Palm image quality evaluation model construction method and device | |
| CN107403199A (en) | Data processing method and device | |
| CN114519520A (en) | Model evaluation method, model evaluation device and storage medium | |
| CN113536947A (en) | Face attribute analysis method and device | |
| CN112784661B (en) | Real face recognition method and real face recognition device | |
| Prayogo et al. | A Novel Approach for Face Recognition: YOLO-Based Face Detection and Facenet | |
| CN113420689B (en) | Character recognition method, device, computer equipment and medium based on probability calibration | |
| CN112749625B (en) | Time sequence behavior detection method, time sequence behavior detection device and terminal equipment |
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 |