CN110276243A - Score mapping method, face comparison method, device, equipment and storage medium - Google Patents
Score mapping method, face comparison method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN110276243A CN110276243A CN201910377068.4A CN201910377068A CN110276243A CN 110276243 A CN110276243 A CN 110276243A CN 201910377068 A CN201910377068 A CN 201910377068A CN 110276243 A CN110276243 A CN 110276243A
- Authority
- CN
- China
- Prior art keywords
- similarity score
- sample
- threshold
- facial image
- score
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/172—Classification, e.g. identification
Abstract
The present invention provides a kind of similarity score mapping method, is compared by the multipair facial image sample to preset number and generates level threshold value, actual threshold, negative sample error rate/positive sample percent of pass mapping table.The present invention also provides a kind of face comparison methods, when the face comparison method carries out facial image comparison, the similarity score that human face recognition model compares out is mapped as criterion score according to the mapping table, then criterion score is obtained to final face alignment result compared with default similarity threshold.The present invention also provides a kind of similarity score mapping device, face comparison method, face alignment device, electronic equipment and computer readable storage mediums.The present invention does not need modification similarity threshold after human face recognition model replacement yet, so that it may guarantee that positive sample percent of pass/negative sample error rate is constant, reduce the complexity that facial image compares operation, improve user experience.
Description
Technical field
The present invention relates to field of face identification more particularly to similarity score mapping method, the faces of a kind of recognition of face
Comparison method, similarity score mapping device, face alignment device, electronic equipment and computer readable storage medium.
Background technique
In face identification system, same two photos, the similarity score compared using different models is past
It is past to be different;Using same a collection of photo, the similarity score distribution compared using different models is also different.In this way
It will result in the different models similarity threshold that identical negative sample error rate or when positive sample percent of pass are taken in order to obtain not
Together, if that business side will guarantee that negative sample error rate or positive sample percent of pass are constant, each model change can all need to repair
It uses instead in the similarity threshold for judging whether same people, which increase workloads when face alignment system upgrade.
Summary of the invention
In view of the foregoing, it is necessary to which a kind of similarity score mapping method, face comparison method, device, equipment are provided
And storage medium does not need modification similarity threshold it is also ensured that negative sample is wrong when using different human face recognition models
Accidentally rate or positive sample percent of pass are constant.
First aspect present invention provides a kind of similarity score mapping method, which comprises
When receiving the operational order for comparing facial image, the multipair people of preset number is inputted into human face recognition model
Face image sample, and obtain the similarity score for every a pair of of facial image sample that the human face recognition model compares out, wherein
Every a pair of of facial image sample standard deviation includes for carrying out the two of facial image comparison facial images;
The preset number is counted to people according to the similarity score of every a pair of of the facial image sample got
The similarity score of face image sample is distributed, and generates multiple actual thresholds and negative sample mistake according to similarity score distribution
Rate/positive sample percent of pass corresponding relationship;
It is instructed according to predetermined registration operation and generates level threshold value and the actual threshold, negative sample error rate/positive sample percent of pass
Mapping table.
Preferred embodiment according to the present invention, the multipair facial image sample of the preset number are positive sample, each pair of positive sample
Two facial images in this are identical.
Preferred embodiment according to the present invention, the multipair facial image sample of the preset number are negative sample, each pair of negative sample
Two facial images in this are different.
Preferred embodiment according to the present invention, the similarity point of every a pair of of the facial image sample got according to
Number counts the preset number and is distributed to the similarity score of facial image sample, is distributed and is generated according to the similarity score
Multiple actual thresholds include: with negative sample error rate/positive sample percent of pass corresponding relationship
By the similarity score of the multipair facial image sample of the preset number according to from small to large or from big to small
Sequence arranges, and obtains the similarity score distribution results;
Multiple actual thresholds and the corresponding negative sample of each actual threshold are determined according to the similarity score distribution results
Error rate/positive sample percent of pass;
It is generated according to the multiple actual threshold and the corresponding negative sample error rate of each actual threshold/positive sample percent of pass
The negative sample error rate/positive sample percent of pass and actual threshold corresponding relationship;
Wherein, the negative sample error rate is in the multipair negative sample of preset number, and human face recognition model is by negative sample
In two different faces image recognitions be identical face probability;
The positive sample percent of pass is in the multipair positive sample of preset number, and human face recognition model is by two in positive sample
Identical facial image is identified as the probability of identical face.
Preferred embodiment according to the present invention, the method also includes:
The level threshold value and the actual threshold, negative sample error rate/positive sample percent of pass mapping table are sent out
Preset memory locations are sent to be stored, for sharing the level threshold value and reality for the human face recognition model in other electronic equipments
Border threshold value, negative sample error rate/positive sample percent of pass mapping table carry out facial image comparison.
Second aspect of the present invention provides a kind of face comparison method, which comprises
When receiving the operational order for comparing facial image, at least a pair of of facial image is inputted into human face recognition model
Sample, and obtain the human face recognition model and compare the similarity score that the facial image sample obtains;
Judge that the similarity score is reflected in level threshold value and actual threshold, negative sample error rate/positive sample percent of pass
Penetrate threshold interval locating in the threshold interval that actual threshold is constituted in relation table;
According to threshold interval locating for the similarity score and the level threshold value and actual threshold, negative sample mistake
The similarity score is mapped as criterion score according to Linear Mapping by rate/positive sample percent of pass mapping table;
The criterion score is compared with a preset similarity threshold, determines facial image comparison result.
In a preferred embodiment of the invention, the threshold interval according to locating for the similarity score and the mapping are closed
It is table, the similarity score, which is mapped as criterion score, according to Linear Mapping includes:
The similarity score is mapped as criterion score according to the following formula:
Criterion score=(similarity score-threshold interval lower limit value)/(threshold interval upper limit value-threshold interval lower limit
Value) * (the corresponding level threshold value of the threshold interval upper limit value-corresponding level threshold value of threshold interval lower limit value)+threshold interval lower limit
It is worth corresponding level threshold value.
Third aspect present invention provides a kind of similarity score mapping device, and described device includes:
Comparison module, for being inputted into human face recognition model default when receiving comparison facial image operational order
The multipair facial image sample of number, and obtain the similar of every a pair of of facial image sample that the human face recognition model compares out
Spend score, wherein every a pair of of sample standard deviation includes two facial images for being compared;
Relationship determination module, the similarity score for the every a pair of sample got according to count described default
The similarity score of number sample pair is distributed, and generates multiple actual thresholds and negative sample mistake according to similarity score distribution
Rate/positive sample percent of pass corresponding relationship;
Mapping table generation module, for instructing generation level threshold value and the actual threshold according to predetermined registration operation, bearing
Sample error rate/positive sample percent of pass mapping table, wherein the multiple actual threshold constitutes multiple and different threshold zones
Between.
Fourth aspect present invention provides a kind of face alignment device, and described device includes:
Similarity score obtain module, for receive compare facial image operational order when, to recognition of face mould
At least a pair of of facial image sample is inputted in type, and is obtained the human face recognition model comparison facial image sample and obtained phase
Like degree score;
Section determining module, for determining that the similarity score obtains the similarity score of module acquisition in the standard
The threshold interval that actual threshold is constituted in threshold value and actual threshold, negative sample error rate/positive sample percent of pass mapping table
In locating section;
Mapping block, for according to the section determining module determine the similarity score locating for threshold interval,
The similarity score is mapped as criterion score according to Linear Mapping;
Determining module is come true for the criterion score after the mapping to be compared with a preset similarity threshold
Determine the comparison result to facial image sample.
Fifth aspect present invention provides a kind of electronic equipment, and the electronic equipment includes memory and processor, described to deposit
Reservoir is for storing at least one instruction, and the processor is for executing at least one described instruction to realize in any embodiment
Recognition of face face comparison method described in any one of any one similarity score mapping method and/or any embodiment.
Sixth aspect present invention provides a kind of computer readable storage medium, and the computer-readable recording medium storage has
At least one instruction, at least one described instruction realize similarity described in any one of any embodiment point when being executed by processor
Face comparison method described in number any one of mapping method and/or any embodiment.
From the above technical scheme, the present invention by the similarity score mapping method generate the level threshold value,
After actual threshold, negative sample error rate/positive sample percent of pass mapping table, pass through the facial image comparison method every time
When carrying out facial image comparison, the similarity score that human face recognition model compares out is mapped as marking according to the mapping table
Quasi- score, then criterion score is obtained to final face alignment result compared with default similarity threshold.Business side only needs basis
Level threshold value, actual threshold, the mapping table of positive sample percent of pass or level threshold value, actual threshold, negative sample error rate
Mapping table determines the level threshold value of oneself, modification similarity threshold is not needed after human face recognition model replacement yet, just
It can guarantee that positive sample percent of pass/negative sample error rate is constant, reduce the complexity that facial image compares operation, improve user
Experience.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
The embodiment of invention for those of ordinary skill in the art without creative efforts, can also basis
The attached drawing of offer obtains other attached drawings.
Fig. 1 is similarity score mapping method flow chart in one embodiment of the invention.
Fig. 2 is face comparison method flow chart in one embodiment of the invention.
Fig. 3 is the module map of similarity score mapping device in one embodiment of the invention.
Fig. 4 is the module map of face alignment device in one embodiment of the invention.
Fig. 5 is the structural schematic diagram of the preferred embodiment of electronic equipment at least one example of the invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
In order to make the foregoing objectives, features and advantages of the present invention clearer and more comprehensible, with reference to the accompanying drawing and specific real
Applying mode, the present invention is described in further detail.
In order to enable those skilled in the art to better understand the solution of the present invention, below in conjunction in the embodiment of the present invention
Attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is only
The embodiment of a part of the invention, instead of all the embodiments.Based on the embodiments of the present invention, ordinary skill people
The model that the present invention protects all should belong in member's every other embodiment obtained without making creative work
It encloses.
Description and claims of this specification and term " first " in above-mentioned attached drawing, " second " and " third " etc. are
For distinguishing different objects, not for description particular order.In addition, term " includes " and their any deformations, it is intended that
Non-exclusive include in covering.Such as the process, method, system, product or equipment for containing a series of steps or units do not have
It is defined in listed step or unit, but optionally further comprising the step of not listing or unit, or optionally further comprising
For the intrinsic other step or units of these process, methods, product or equipment.
As shown in Figure 1, being the flow chart of the first preferred embodiment of similarity score mapping method of the present invention.According to difference
Demand, the sequence of step can change in the flow chart, and certain steps can be omitted.
Step S11, when receiving the operational order for comparing facial image, preset number is inputted into human face recognition model
Multipair facial image sample, and obtain the similarity point for every a pair of of facial image sample that the human face recognition model compares out
Number, wherein every a pair of of facial image sample standard deviation includes for carrying out the two of face alignment facial images.
After each pair of sample is to the human face recognition model is inputted, one can be all exported by the comparison of the human face recognition model
A corresponding similarity score, the similarity score is for indicating the sample centering that the human face recognition model compares out
The similarity degree of two pictures.
In an embodiment of the present invention, the multipair sample of the preset number inputted into human face recognition model is multipair positive sample
This, the positive sample refers to that two facial images of the sample centering are identical.
In another embodiment of the present invention, the multipair sample of the preset number inputted into human face recognition model is multipair negative
Sample pair, the negative sample refer to that two facial images of the sample centering are different.
The number of the sample pair, which can according to need, is set as different numbers, such as can be 10,000 pairs of positive samples or one
Ten thousand pairs of negative samples, this is not limited by the present invention.
Step S12, the similarity score of every a pair of of the facial image sample got according to counts described default
The similarity score distribution of several pairs of facial image samples generates multiple actual thresholds and negative sample according to similarity score distribution
This error rate/positive sample percent of pass corresponding relationship.
The similarity score for each pair of sample pair that the human face recognition model compares out may be due to the difference of picture material
It is different.
Threshold value in recognition of face determines the judgment criteria of image comparison result, when the similarity point of two face pictures
Number is more than the threshold value, then is determined as that the face in two pictures belongs to same people, if the similarity score of two face pictures is low
In the threshold value, then it is determined as that the face in two pictures belongs to different people.
In one embodiment, the similarity score for the sample pair that the human face recognition model compares out can be distributed in
Between 0-1, for example, the similarity score of negative sample pair may be 0.1 or 0.3 etc., and the similarity score of positive sample pair may
It is 0.7 or 0.9 etc..
In an embodiment of the present invention, the step S12 is specifically included:
1) similarity score of the sample pair of the preset number is arranged according to sequence from small to large or from big to small,
Obtain the similarity score distribution results;
2) multiple actual thresholds and the corresponding negative sample of each actual threshold are determined according to the similarity score after the arrangement
This error rate/positive sample percent of pass;
The negative sample error rate refers in the multipair negative sample of the preset number, and human face recognition model is by negative sample
The probability for being identified as identical face of two different faces image mistakes of centering;
The positive sample percent of pass refers in the multipair positive sample of the preset number, and human face recognition model is by positive sample
The probability that identical facial image is identified as identical face is opened in centering two;
3) raw according to the multiple actual threshold and the corresponding negative sample error rate of each actual threshold/positive sample percent of pass
At the negative sample error rate/positive sample percent of pass and actual threshold corresponding relationship.
For example, when the input human face recognition model is 10,000 pairs of negative samples pair, the human face recognition model ratio
To having the similarity score of 9999 pairs of negative samples less than 0.8 out, there is the similarity score of 1 pair of negative sample to be greater than 0.8, if by threshold
Value is set as 0.8, then 9999 pair negative sample comparison results of the similarity score less than 0.8 are this to two face figures in negative sample
As being not belonging to the same person, a pair of of negative sample comparison result of the similarity score greater than 0.8 is this to two people in negative sample
Face image belongs to the same person, then the comparison result of this negative sample of a pair of of similarity score greater than 0.8 is mistake, threshold at this time
The corresponding negative sample error rate of value 0.8 is 0.01%, i.e., it is 0.8 that negative sample error rate, which is the corresponding threshold value of a ten thousandth, is passed through
Similar approach can determine the corresponding relationship of multiple groups negative sample error rate and actual threshold.
Step S13, it is instructed according to predetermined registration operation and generates level threshold value and the actual threshold, negative sample error rate/positive sample
The mapping table of this percent of pass.Wherein, the multiple actual threshold constitutes multiple and different threshold intervals.
The predetermined registration operation instruction is for defining multiple level threshold values and the multiple level threshold value and the actual threshold
Mapping table between negative sample error rate/positive sample percent of pass.
For example, in one embodiment, when the number of the negative sample pair is 10,000 pairs, by 10,000 pairs of negative samples
It is wrong according to the determining level threshold value of similarity score distribution, actual threshold, negative sample after inputting the human face recognition model
Accidentally the mapping table of rate can be as shown in table 1 below:
Level threshold value | Actual threshold | Negative sample error rate |
0 | t0 | 100% |
0.2 | t1 | 10% |
0.3 | t2 | 1% |
0.4 | t3 | 0.1% |
0.5 | t4 | 0.01% |
0.6 | t5 | 0.001% |
0.7 | t6 | 0.0001% |
0.8 | t7 | 0.00001% |
1 | t8 | 0% |
Table 1
Again for example, in one embodiment, when the number of the positive sample pair is 10,000 pairs, by 10,000 pairs of positive samples
After this input human face recognition model, according to the determining level threshold value of similarity score distribution, actual threshold, positive sample
The mapping table of percent of pass can be as shown in table 2 below:
Table 2
Wherein, the level threshold value is the customized setting of user, when the different human face recognition model of use is to identical
When facial image compares, if wanting negative sample error rate/positive sample percent of pass identical, it is only necessary to set the level threshold value as threshold
Value, without changing threshold value for each human face recognition model, that is to say, that identical mark in all human face recognition models
It is identical that quasi- threshold value corresponds to identical negative sample error rate/positive sample percent of pass.
In one embodiment, the method can also include the following steps: the level threshold value and the actual threshold, bear
Sample error rate/positive sample percent of pass mapping table sends preset memory locations and is stored, the preset memory locations
It can include but is not limited to the portable electronic devices such as cloud server, individual server, desktop computer, laptop
Deng sharing the level threshold value and actual threshold, negative sample error rate/just with the human face recognition model into other electronic equipments
The mapping table of sample percent of pass carries out facial image comparison for the face identification model in other people electronic equipments.
In the present invention, the step S11-S13 can be executed in off-line case, be also possible to online processing.It is described
Step S11-S13 only needs to execute once, after generating the mapping table, is carrying out face figure subsequently through human face recognition model
When as comparing, directly using the step S11-S13 level threshold value generated and the actual threshold, negative sample mistake
Rate/positive sample percent of pass mapping table carries out similarity score mapping.
As shown in Fig. 2, being the flow chart of the first preferred embodiment of face comparison method of the present invention.The facial image is known
Other method using the level threshold value that is generated in the similarity score mapping method and actual threshold, negative sample error rate/
Facial image sample is compared using human face recognition model for the mapping table of positive sample percent of pass.According to different need
It asks, the sequence of step can change in the flow chart, and certain steps can be omitted.
Step S21 is inputted at least a pair of when receiving the operational order for comparing facial image into human face recognition model
Facial image sample, and obtain the human face recognition model comparison facial image sample and obtain similarity score S.
At least a pair of of facial image sample can be a pair of of positive sample, be also possible to a pair of of negative sample.
In an embodiment of the present invention, further include in the facial image sample inputted into the human face recognition model
For identifying the identification information that facial image sample is positive sample or negative sample.It for example, can be by input interface
It is middle that provide positive sample/negative sample options for user to select the facial image sample of the input be positive sample or negative sample, and
It is generated according to the selection operation described for identifying the identification information of positive sample and negative sample.
Step S22 judges that the similarity score S is located at level threshold value and actual threshold, negative sample error rate/positive sample
Which threshold interval that actual threshold is constituted in the mapping table of percent of pass.
For example, for level threshold value that front is illustrated, actual threshold, the mapping table of negative sample error rate, step
The similarity score S that S21 is compared out is likely located at the section actual threshold t1-t2 in table 1, it is also possible to be located at the section t3-t4.
In one embodiment, before the step S22, the method also includes:
1) judge the comparison be facial image sample is positive sample or negative sample;
For example, judging that the facial image sample is positive sample or negative sample by the identification information;
If 2) positive sample, then level threshold value, actual threshold, positive sample percent of pass mapping table are called, and determine institute
State similarity score S is located at which threshold interval that actual threshold is constituted in the mapping table;
If 3) negative sample, then level threshold value, actual threshold, the mapping table of negative sample error rate are called, and determine
The similarity score S is located at which threshold interval that actual threshold is constituted in the mapping table.
Step S23, the threshold interval according to locating for the similarity score S and the mapping table, according to linearly reflecting
It penetrates and the similarity score is mapped as criterion score S '.
In mathematics, Linear Mapping (also referred to as linear transformation or linear operator) is the letter between two vector spaces
Number, it keeps the operation of vectorial addition and scalar multiplication.
Specifically, in an embodiment of the present invention, the similarity score S is mapped as by following Linear Mapping formula
Criterion score S ':
S '=(S- threshold interval lower limit value)/(threshold interval upper limit value-threshold interval lower limit value) * (threshold interval upper limit
It is worth corresponding level threshold value-corresponding level threshold value of threshold interval lower limit value)+corresponding the level threshold value of threshold interval lower limit value.
For example, for level threshold value that front is illustrated, actual threshold, the mapping table of negative sample error rate, if
The similarity score S is located at threshold interval t3-t4, then similarity score S is mapped as standard according to Linear Mapping formula
The formula of score S ' are as follows:
S'=(s-t3)/(t4-t3) * (0.5-0.4)+0.4.
In another example for level threshold value that front is illustrated, actual threshold, the mapping table of positive sample percent of pass, such as
Similarity score S described in fruit is located at threshold interval t3-t4, then being mapped as marking by similarity score S according to Linear Mapping formula
The formula of quasi- score S ' are as follows:
S'=(s-t3)/(t4-t3) * (0.4-0.3)+0.3.
Step S24, the criterion score S ' after the mapping is compared with a preset similarity threshold, to determine this
To the comparison result of facial image sample.
Criterion score S ' after the mapping is less than the preset similarity threshold, it is determined that two in the sample
It opens facial image and is not belonging to the same person, the criterion score S ' after the mapping is greater than the preset similarity threshold, then
Determine that two in the sample facial images belong to the same person.
Wherein, the preset similarity threshold can be what user was configured as needed, the similarity threshold
It can be applied in multiple and different human face recognition models, due to reflecting the similarity score obtained in different faces identification model
When penetrating for criterion score, therefore carrying out facial image comparison using different faces identification model, need to obtain identical positive sample
This percent of pass/negative sample error rate can use the same preset similarity threshold, without user in each face
Different similarity thresholds is all set in identification model.
In the present invention, it is wrong that the level threshold value, actual threshold, negative sample are generated by the similarity score mapping method
Accidentally after rate/positive sample percent of pass mapping table, facial image comparison is carried out by the facial image comparison method every time
When, similarity score that human face recognition model compares out is mapped as criterion score according to the mapping table, then by standard
Score obtains final face alignment result compared with level threshold value.Business side is only needed according to level threshold value, actual threshold, positive sample
The mapping table or level threshold value, actual threshold, the mapping table of negative sample error rate of this percent of pass determine the mark of oneself
Quasi- threshold value does not need modification threshold value, so that it may guarantee technology positive sample percent of pass/negative sample after human face recognition model replacement yet
This error rate is constant, reduces the complexity that facial image compares operation, improves user experience.
As shown in figure 3, the Program modual graph of the first preferred embodiment of similarity score mapping device of the present invention.The phase
Include, but are not limited to one or more following module: comparison module 30, relationship determination module 31 like degree score mapping device 3
And mapping table generation module 32.The so-called unit of the present invention refers to that one kind can be by the place of similarity score mapping device 3
Reason device is performed and can complete the series of computation machine program segment of fixed function, and storage is in memory.About each list
The function of member will be described in detail in subsequent embodiment.
The comparison module 30 is used to input when receiving comparison facial image operational order into human face recognition model
The multipair facial image sample of preset number, and obtain every a pair of of facial image sample that the human face recognition model compares out
Similarity score, wherein every a pair of of sample standard deviation includes two facial images for being compared.
After each pair of sample is to the human face recognition model is inputted, one can be all exported by the comparison of the human face recognition model
A corresponding similarity score, the similarity score is for indicating the sample centering that the human face recognition model compares out
The similarity degree of two pictures.
In an embodiment of the present invention, the multipair sample of the preset number inputted into human face recognition model is multipair positive sample
This, the positive sample refers to that two facial images of the sample centering are identical.
In another embodiment of the present invention, the multipair sample of the preset number inputted into human face recognition model is multipair negative
Sample pair, the negative sample refer to that two facial images of the sample centering are different.
The number of the sample pair, which can according to need, is set as different numbers, such as can be 10,000 pairs of positive samples or one
Ten thousand pairs of negative samples, this is not limited by the present invention.
The similarity score of every a pair of sample that the relationship determination module 31 is used to get according to counts institute
The similarity score distribution for stating preset number sample pair generates multiple actual thresholds and negative sample according to similarity score distribution
This error rate/positive sample percent of pass corresponding relationship.
The similarity score for each pair of sample pair that the human face recognition model compares out may be due to the difference of picture material
It is different.
Threshold value in recognition of face determines the judgment criteria of image comparison result, when the similarity point of two face pictures
Number is more than the threshold value, then is determined as that the face in two pictures belongs to same people, if the similarity score of two face pictures is low
In the threshold value, then it is determined as that the face in two pictures belongs to different people.
In one embodiment, the similarity score for the sample pair that the human face recognition model compares out can be distributed in
Between 0-1, for example, the similarity score of negative sample pair may be 0.1 or 0.3 etc., and the similarity score of positive sample pair may
It is 0.7 or 0.9 etc..
In an embodiment of the present invention, the relationship determination module 31 generates the corresponding relationship and specifically includes following behaviour
Make:
1) similarity score of the sample pair of the preset number is arranged according to sequence from small to large or from big to small,
Obtain the similarity score distribution results;
2) multiple actual thresholds and the corresponding negative sample of each actual threshold are determined according to the similarity score after the arrangement
This error rate/positive sample percent of pass;
The negative sample error rate refers in the multipair negative sample of the preset number, and human face recognition model is by negative sample
The probability for being identified as identical face of two different faces image mistakes of centering;
The positive sample percent of pass refers in the multipair positive sample of the preset number, and human face recognition model is by positive sample
The probability that identical facial image is identified as identical face is opened in centering two;
3) raw according to the multiple actual threshold and the corresponding negative sample error rate of each actual threshold/positive sample percent of pass
At the negative sample error rate/positive sample percent of pass and actual threshold corresponding relationship.
For example, when the input human face recognition model is 10,000 pairs of negative samples pair, the human face recognition model ratio
To having the similarity score of 9999 pairs of negative samples less than 0.8 out, there is the similarity score of 1 pair of negative sample to be greater than 0.8, if by threshold
Value is set as 0.8, then 9999 pair negative sample comparison results of the similarity score less than 0.8 are this to two face figures in negative sample
As being not belonging to the same person, a pair of of negative sample comparison result of the similarity score greater than 0.8 is this to two people in negative sample
Face image belongs to the same person, then the comparison result of this negative sample of a pair of of similarity score greater than 0.8 is mistake, threshold at this time
The corresponding negative sample error rate of value 0.8 is 0.01%, i.e., it is 0.8 that negative sample error rate, which is the corresponding threshold value of a ten thousandth, is passed through
Similar approach can determine the corresponding relationship of multiple groups negative sample error rate and actual threshold.
The mapping table generation module 32, which is used to be instructed according to predetermined registration operation, generates level threshold value and the practical threshold
Value, negative sample error rate/positive sample percent of pass mapping table, wherein the multiple actual threshold constitutes multiple and different
Threshold interval.For example, foregoing table 1, the mapping relations in table 2 can be generated in the mapping table generation module 32
Table.
Predetermined registration operation instruction be user's input the multiple level threshold values of definition and the multiple level threshold value with it is described
The instruction of mapping table between actual threshold and negative sample error rate/positive sample percent of pass.
Wherein, the level threshold value is the customized setting of user, when the different human face recognition model of use is to identical
When facial image compares, if wanting negative sample error rate/positive sample percent of pass identical, it is only necessary to set the level threshold value as threshold
Value, without changing threshold value for each human face recognition model, that is to say, that identical mark in all human face recognition models
It is identical that quasi- threshold value corresponds to identical negative sample error rate/positive sample percent of pass.
In one embodiment, the mapping table generation module 32 is also used to: by the level threshold value and the practical threshold
Value, negative sample error rate/positive sample percent of pass mapping table are deposited by the preset memory locations of communication device transmission
Storage, the preset memory locations can include but is not limited to cloud server, individual server, desktop computer, notebook electricity
Portable electronic devices such as brain etc. share the level threshold value and practical threshold with the human face recognition model into other electronic equipments
Value, negative sample error rate/positive sample percent of pass mapping table carry out people for the face identification model in other people electronic equipments
Face image compares.
As shown in figure 4, the Program modual graph of the first preferred embodiment for face comparison device of the present invention.The face ratio
One or more following module included, but are not limited to device 4: similarity score obtain module 40, section determining module 41,
Mapping block 42 and determining module 43.The so-called unit of the present invention refers to that one kind can be by the processor institute of face alignment device 4
The series of computation machine program segment of fixed function is executed and can complete, storage is in memory.Function about each unit
It can will be described in detail in subsequent embodiment.
The similarity score obtains module 40 and is used to know when receiving the operational order for comparing facial image to face
At least a pair of of facial image sample is inputted in other model, and is obtained the human face recognition model comparison facial image sample and obtained
To similarity score S.
At least a pair of of facial image sample can be a pair of of positive sample, be also possible to a pair of of negative sample.
In an embodiment of the present invention, further include in the facial image sample inputted into the human face recognition model
For identifying the identification information that facial image sample is positive sample or negative sample.It for example, can be by input interface
It is middle that provide positive sample/negative sample options for user to select the facial image sample of the input be positive sample or negative sample, and
It is generated according to the selection operation described for identifying the identification information of positive sample and negative sample.
The section determining module 41 is used to determine that the similarity score to obtain the similarity score S that module 40 obtains and exists
Actual threshold is constituted in the level threshold value and actual threshold, negative sample error rate/positive sample percent of pass mapping table
Locating section in threshold interval.
For example, for level threshold value that front is illustrated, actual threshold, the mapping table of negative sample error rate, it is described
Similarity score S is likely located at the section t1-t2 in above-mentioned table 1, it is also possible to be located at the section t3-t4.
In one embodiment, the section determining module 41 is also used before determining section locating for the similarity score S
In performing the following operations:
1) judge the comparison be facial image sample is positive sample or negative sample;
For example, judging that the facial image sample is positive sample or negative sample by the identification information;
If 2) positive sample, then level threshold value, actual threshold, positive sample percent of pass mapping table are called, and determine institute
State similarity score S is located at which threshold interval that actual threshold is constituted in the mapping table;
If 3) negative sample, then level threshold value, actual threshold, the mapping table of negative sample error rate are called, and determine
The similarity score S is located at which threshold interval that actual threshold is constituted in the mapping table.
Threshold locating for the similarity score S that the mapping block 42 is used to be determined according to the section determining module 41
It is worth section, the similarity score is mapped as criterion score S ' according to Linear Mapping.
Specifically, in an embodiment of the present invention, the similarity score S is mapped as by following Linear Mapping formula
Criterion score S ':
S '=(S- threshold interval lower limit value)/(threshold interval upper limit value-threshold interval lower limit value) * (threshold interval upper limit
It is worth corresponding level threshold value-corresponding level threshold value of threshold interval lower limit value)+corresponding the level threshold value of threshold interval lower limit value.
For example, for level threshold value that front is illustrated, actual threshold, the mapping table of negative sample error rate, if
The similarity score S is located at threshold interval t3-t4, then similarity score S is mapped as standard according to Linear Mapping formula
The formula of score S ' are as follows:
S'=(s-t3)/(t4-t3) * (0.5-0.4)+0.4.
In another example for level threshold value that front is illustrated, actual threshold, the mapping table of positive sample percent of pass, such as
Similarity score S described in fruit is located at threshold interval t3-t4, then being mapped as marking by similarity score S according to Linear Mapping formula
The formula of quasi- score S ' are as follows:
S'=(s-t3)/(t4-t3) * (0.4-0.3)+0.3.
The determining module 43 is for comparing the criterion score S ' after the mapping with a preset similarity threshold
It is right, to determine the comparison result to facial image sample.
Criterion score S ' after the mapping is less than the preset similarity threshold, it is determined that two in the sample
It opens facial image and is not belonging to the same person, the criterion score S ' after the mapping is greater than the preset similarity threshold, then
Determine that two in the sample facial images belong to the same person.
By above embodiments, the present invention provides a kind of similarity score mapping device for generating level threshold value, reality
Threshold value, negative sample error rate/positive sample percent of pass mapping table, also provide a kind of face alignment device, pass through institute every time
When stating the progress facial image comparison of facial image comparison method, the similarity score that human face recognition model is compared out is according to
Mapping table is mapped as criterion score, then criterion score is obtained to final face alignment result compared with level threshold value.Business
Side only needs according to level threshold value, actual threshold, the mapping table of positive sample percent of pass or level threshold value, actual threshold, bears
The mapping table of sample error rate determines the level threshold value of oneself, and modification threshold is not needed after human face recognition model replacement yet
Value, so that it may guarantee that technology positive sample percent of pass/negative sample error rate is constant, reduce the complexity that facial image compares operation
Degree improves user experience.
One takes in storage medium.Above-mentioned software program module is stored in a computer readable storage medium, described
Software program module includes that some instructions are used so that an electronic equipment (can be personal computer, server or network
Equipment etc.) or processor (processor) execute the part steps of each embodiment the method for the present invention.
As shown in figure 5, the electronic equipment 5 includes at least one sending device 51, at least one processor 52, at least one
A processor 53, at least one reception device 54.
The electronic equipment 5 be it is a kind of can according to the instruction for being previously set or store, automatic progress numerical value calculating and/or
The equipment of information processing, hardware include but is not limited to microprocessor, specific integrated circuit (Application Specific
Integrated Circuit, ASIC), programmable gate array (Field-Programmable Gate Array, FPGA), number
Word processing device (Digital Signal Processor, DSP), embedded device etc..The electronic equipment 5 may also include network
Equipment and/or user equipment.Wherein, the network equipment includes but is not limited to single network server, multiple network servers
The server group of composition or the cloud being made of a large amount of hosts or network server for being based on cloud computing (Cloud Computing),
Wherein, cloud computing is one kind of distributed computing, a super virtual computing consisting of a loosely coupled set of computers
Machine.
The electronic equipment 5, which may be, but not limited to, any one, to pass through keyboard, touch tablet or voice-operated device with user
Etc. modes carry out the electronic product of human-computer interaction, for example, tablet computer, smart phone, personal digital assistant (Personal
Digital Assistant, PDA), intellectual wearable device, picture pick-up device, the terminals such as monitoring device.
Network locating for the electronic equipment 5 includes, but are not limited to internet, wide area network, Metropolitan Area Network (MAN), local area network, virtual
Dedicated network (Virtual Private Network, VPN) etc..
Wherein, the reception device 54 and the sending device 51 can be wired sending port, or wirelessly set
It is standby, for example including antenna assembly, for carrying out data communication with other equipment.
The memory 52 is for storing program code.The memory 52, which can be, does not have physical form in integrated circuit
The circuit with store function, such as RAM (Random-Access Memory, random access memory), FIFO (First In
First Out) etc..Alternatively, the memory 52 is also possible to the memory with physical form, such as memory bar, TF card
(Trans-flash Card), smart media card (smart media card), safe digital card (secure digital
Card), storage facilities such as flash memory cards (flash card) etc..
The processor 53 may include one or more microprocessor, digital processing unit.The processor 53 is adjustable
With the program code stored in memory 52 to execute relevant function.For example, modules described in Fig. 3 are stored in institute
The program code in memory 52 is stated, and as performed by the processor 53, to realize a kind of similarity score mapping method;
And/or modules described in Fig. 4 are stored in the program code in the memory 52, and are held by the processor 53
Row, to realize a kind of face comparison method.The processor 53 is also known as central processing unit (CPU, Central Processing
Unit), it is one piece of ultra-large integrated circuit, is arithmetic core (Core) and control core (Control Unit).
The embodiment of the present invention also provides a kind of computer readable storage medium, is stored thereon with computer instruction, the finger
It enables when the electronic equipment for being included one or more processors executes, executes electronic equipment as described in embodiment of the method above
Similarity score mapping method and/or face comparison method.
The characteristic means of present invention mentioned above can be realized by integrated circuit, and control above-mentioned of realization
The function of similarity score mapping method described in embodiment of anticipating, face comparison method.The similarity described in any embodiment
Function achieved by score mapping method, face comparison method can be transferred through integrated circuit of the invention and be installed on the electronics
In equipment, the electronic equipment is enable to play similarity score mapping method, face comparison method institute described in any embodiment
The function of realization, this will not be detailed here.
The characteristic means of present invention mentioned above can be realized by integrated circuit, and control above-mentioned of realization
The function of similarity score mapping method described in embodiment of anticipating, face comparison method.
It should be noted that for the various method embodiments described above, for simple description, therefore, it is stated as a series of
Combination of actions, but those skilled in the art should understand that, the present invention is not limited by the sequence of acts described because
According to the present invention, some steps may be performed in other sequences or simultaneously.Secondly, those skilled in the art should also know
It knows, the embodiments described in the specification are all preferred embodiments, and related actions and modules is not necessarily of the invention
It is necessary.
In the above-described embodiments, it all emphasizes particularly on different fields to the description of each embodiment, there is no the portion being described in detail in some embodiment
Point, reference can be made to the related descriptions of other embodiments.
In several embodiments provided herein, it should be understood that disclosed device, it can be by another way
It realizes.For example, the apparatus embodiments described above are merely exemplary, such as the division of the unit, it is only a kind of
Logical function partition, there may be another division manner in actual implementation, such as multiple units or components can combine or can
To be integrated into another system, or some features can be ignored or not executed.Another point, shown or discussed is mutual
Coupling, direct-coupling or communication connection can be through some interfaces, the indirect coupling or communication connection of device or unit,
It can be electrical or other forms.
The unit as illustrated by the separation member may or may not be physically separated, aobvious as unit
The component shown may or may not be physical unit, it can and it is in one place, or may be distributed over multiple
In network unit.It can select some or all of unit therein according to the actual needs to realize the mesh of this embodiment scheme
's.
In addition, each functional unit in various embodiments of the present invention can integrate in one processing unit, it can also
To be that each unit physically exists alone, can also be integrated in one unit with two or more units.It is above-mentioned integrated
Unit both can take the form of hardware realization, can also realize in the form of software functional units.
If the integrated unit is realized in the form of SFU software functional unit and sells or use as independent product
When, it can store in a computer readable storage medium.Based on this understanding, technical solution of the present invention is substantially
The all or part of the part that contributes to existing technology or the technical solution can be in the form of software products in other words
It embodies, which is stored in a storage medium, including some instructions are used so that a computer
Equipment (can for personal computer, server or network equipment etc.) execute each embodiment the method for the present invention whole or
Part steps.And storage medium above-mentioned includes: that USB flash disk, read-only memory (ROM, Read-Only Memory), arbitrary access are deposited
Reservoir (RAM, Random Access Memory), mobile hard disk, magnetic or disk etc. be various to can store program code
Medium.
The above, the above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations;Although referring to before
Stating embodiment, invention is explained in detail, those skilled in the art should understand that: it still can be to preceding
Technical solution documented by each embodiment is stated to modify or equivalent replacement of some of the technical features;And these
It modifies or replaces, the range for technical solution of various embodiments of the present invention that it does not separate the essence of the corresponding technical solution.
Claims (10)
1. a kind of similarity score mapping method, which is characterized in that the described method includes:
When receiving the operational order for comparing facial image, the multipair face figure of preset number is inputted into human face recognition model
Decent, and obtain the similarity score for every a pair of of facial image sample that the human face recognition model compares out, wherein it is described
Every a pair of facial image sample standard deviation includes for carrying out the two of facial image comparison facial images;
The preset number is counted to face figure according to the similarity score of every a pair of of the facial image sample got
Decent similarity score distribution, according to similarity score distribution generate multiple actual thresholds and negative sample error rate/
The corresponding relationship of positive sample percent of pass;
Generation level threshold value is instructed to reflect with the actual threshold, negative sample error rate/positive sample percent of pass according to predetermined registration operation
Penetrate relation table.
2. similarity score mapping method as described in claim 1, which is characterized in that the multipair face figure of the preset number
Decent is positive sample, and two facial images in each pair of positive sample are identical.
3. similarity score mapping method as described in claim 1, which is characterized in that the multipair face figure of the preset number
Decent is negative sample, and two facial images in each pair of negative sample are different.
4. similarity score mapping method as claimed in claim 2 or claim 3, which is characterized in that described to be got according to
The similarity score of every a pair of facial image sample counts the preset number to the similarity score point of facial image sample
Cloth generates multiple actual thresholds pass corresponding with negative sample error rate/positive sample percent of pass according to similarity score distribution
System includes:
By the similarity score of the multipair facial image sample of the preset number according to sequence from small to large or from big to small
Arrangement, obtains the similarity score distribution results;
Multiple actual thresholds and the corresponding negative sample mistake of each actual threshold are determined according to the similarity score distribution results
Rate/positive sample percent of pass;
According to the multiple actual threshold and the corresponding negative sample error rate of each actual threshold/positive sample percent of pass generation
Negative sample error rate/positive sample percent of pass and actual threshold corresponding relationship;
Wherein, the negative sample error rate is in the multipair negative sample of preset number, and human face recognition model will be in negative sample
Two different faces image recognitions are the probability of identical face;
The positive sample percent of pass is in the multipair positive sample of preset number, and human face recognition model is identical by two in positive sample
Facial image is identified as the probability of identical face.
5. a kind of face comparison method, which is characterized in that the described method includes:
When receiving the operational order for comparing facial image, at least a pair of of facial image sample is inputted into human face recognition model
This, and obtain the human face recognition model and compare the similarity score that the facial image sample obtains;
Judge that the similarity score is closed in level threshold value and actual threshold, the mapping of negative sample error rate/positive sample percent of pass
It is threshold interval locating in the threshold interval of actual threshold composition in table;
According to threshold interval locating for the similarity score and the level threshold value and actual threshold, negative sample error rate/just
The similarity score is mapped as criterion score according to Linear Mapping by the mapping table of sample percent of pass;
The criterion score is compared with a preset similarity threshold, determines facial image comparison result.
6. face comparison method as claimed in claim 5, which is characterized in that the threshold according to locating for the similarity score
It is worth section and the mapping table, the similarity score, which is mapped as criterion score, according to Linear Mapping includes:
The similarity score is mapped as criterion score according to the following formula:
Criterion score=(similarity score-threshold interval lower limit value)/(threshold interval upper limit value-threshold interval lower limit value) * (threshold
It is worth the corresponding level threshold value-corresponding level threshold value of threshold interval lower limit value of section upper limit value)+threshold interval lower limit value is corresponding
Level threshold value.
7. a kind of similarity score mapping device, which is characterized in that described device includes:
Comparison module, for inputting preset number into human face recognition model when receiving comparison facial image operational order
Multipair facial image sample, and obtain the similarity point for every a pair of of facial image sample that the human face recognition model compares out
Number, wherein every a pair of of sample standard deviation includes two facial images for being compared;
Relationship determination module, the similarity score for the every a pair of sample got according to count the preset number
The similarity score of sample pair is distributed, according to similarity score distribution generate multiple actual thresholds and negative sample error rate/
The corresponding relationship of positive sample percent of pass;
Mapping table generation module generates level threshold value and the actual threshold, negative sample for instructing according to predetermined registration operation
Error rate/positive sample percent of pass mapping table, wherein the multiple actual threshold constitutes multiple and different threshold intervals.
8. a kind of face alignment device, which is characterized in that the face alignment device includes:
Similarity score obtain module, for receive compare facial image operational order when, into human face recognition model
At least a pair of of facial image sample of input, and obtain the human face recognition model comparison facial image sample and obtain similarity
Score;
Section determining module, for determining that the similarity score obtains the similarity score of module acquisition in the level threshold value
Institute in the threshold interval constituted with actual threshold in actual threshold, negative sample error rate/positive sample percent of pass mapping table
Locate section;
Mapping block, for according to the section determining module determine the similarity score locating for threshold interval, according to
The similarity score is mapped as criterion score by Linear Mapping;
Determining module, for the criterion score after the mapping to be compared with a preset similarity threshold, to determine this
To the comparison result of facial image sample.
9. a kind of electronic equipment, which is characterized in that the electronic equipment includes memory and processor, and the memory is for depositing
At least one instruction is stored up, the processor is for executing at least one described instruction to realize such as any one of claims 1 to 4
The similarity score mapping method, and/or the face comparison method as described in any one of claim 5 to 6.
10. a kind of computer readable storage medium, which is characterized in that the computer-readable recording medium storage has at least one
Instruction, at least one described instruction realize that the similarity score as described in any one of claims 1 to 4 is reflected when being executed by processor
Shooting method, and/or the face comparison method as described in any one of claim 5 to 6.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910377068.4A CN110276243A (en) | 2019-05-07 | 2019-05-07 | Score mapping method, face comparison method, device, equipment and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910377068.4A CN110276243A (en) | 2019-05-07 | 2019-05-07 | Score mapping method, face comparison method, device, equipment and storage medium |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110276243A true CN110276243A (en) | 2019-09-24 |
Family
ID=67959752
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910377068.4A Pending CN110276243A (en) | 2019-05-07 | 2019-05-07 | Score mapping method, face comparison method, device, equipment and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110276243A (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111191018A (en) * | 2019-12-30 | 2020-05-22 | 华为技术有限公司 | Response method and device of dialog system, electronic equipment and intelligent equipment |
CN111429638A (en) * | 2020-04-13 | 2020-07-17 | 汪辽宁 | Access control method based on voice recognition and face recognition |
CN111582305A (en) * | 2020-03-26 | 2020-08-25 | 平安科技(深圳)有限公司 | Biological feature recognition method and device, computer equipment and storage medium |
CN111898495A (en) * | 2020-07-16 | 2020-11-06 | 云从科技集团股份有限公司 | Dynamic threshold management method, system, device and medium |
CN113095672A (en) * | 2021-04-09 | 2021-07-09 | 公安部物证鉴定中心 | Method and system for evaluating face image comparison algorithm |
-
2019
- 2019-05-07 CN CN201910377068.4A patent/CN110276243A/en active Pending
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111191018A (en) * | 2019-12-30 | 2020-05-22 | 华为技术有限公司 | Response method and device of dialog system, electronic equipment and intelligent equipment |
CN111191018B (en) * | 2019-12-30 | 2023-10-20 | 华为技术有限公司 | Response method and device of dialogue system, electronic equipment and intelligent equipment |
CN111582305A (en) * | 2020-03-26 | 2020-08-25 | 平安科技(深圳)有限公司 | Biological feature recognition method and device, computer equipment and storage medium |
CN111582305B (en) * | 2020-03-26 | 2023-08-18 | 平安科技(深圳)有限公司 | Biological feature recognition method, apparatus, computer device and storage medium |
CN111429638A (en) * | 2020-04-13 | 2020-07-17 | 汪辽宁 | Access control method based on voice recognition and face recognition |
CN111429638B (en) * | 2020-04-13 | 2021-10-26 | 重庆匠技智能科技有限公司 | Access control method based on voice recognition and face recognition |
CN111898495A (en) * | 2020-07-16 | 2020-11-06 | 云从科技集团股份有限公司 | Dynamic threshold management method, system, device and medium |
CN113095672A (en) * | 2021-04-09 | 2021-07-09 | 公安部物证鉴定中心 | Method and system for evaluating face image comparison algorithm |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108197532B (en) | The method, apparatus and computer installation of recognition of face | |
CN110276243A (en) | Score mapping method, face comparison method, device, equipment and storage medium | |
CN108229555B (en) | Sample weights distribution method, model training method, electronic equipment and storage medium | |
CN109583332B (en) | Face recognition method, face recognition system, medium, and electronic device | |
CN108776787A (en) | Image processing method and device, electronic equipment, storage medium | |
CN107545241A (en) | Neural network model is trained and biopsy method, device and storage medium | |
WO2015165372A1 (en) | Method and apparatus for classifying object based on social networking service, and storage medium | |
CN107729935B (en) | The recognition methods of similar pictures and device, server, storage medium | |
CN112329826A (en) | Training method of image recognition model, image recognition method and device | |
CN108985954A (en) | A kind of method and relevant device of incidence relation that establishing each mark | |
CN113407850B (en) | Method and device for determining and acquiring virtual image and electronic equipment | |
CN106599194A (en) | Label determining method and device | |
CN109871762B (en) | Face recognition model evaluation method and device | |
CN109034199B (en) | Data processing method and device, storage medium and electronic equipment | |
CN110458644A (en) | A kind of information processing method and relevant device | |
CN115130711A (en) | Data processing method and device, computer and readable storage medium | |
CN113780098A (en) | Character recognition method, character recognition device, electronic equipment and storage medium | |
CN112580666A (en) | Image feature extraction method, training method, device, electronic equipment and medium | |
CN110597719B (en) | Image clustering method, device and medium for adaptation test | |
CN115222443A (en) | Client group division method, device, equipment and storage medium | |
CN113947701B (en) | Training method, object recognition method, device, electronic equipment and storage medium | |
CN108399370A (en) | The method and cloud system of Expression Recognition | |
CN107818323A (en) | Method and apparatus for handling image | |
CN113962401A (en) | Federal learning system, and feature selection method and device in federal learning system | |
Charlier et al. | Profiling smart contracts interactions with tensor decomposition and graph mining |
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 |