CN107729924B - Picture review probability interval generation method and picture review determination method - Google Patents
Picture review probability interval generation method and picture review determination method Download PDFInfo
- Publication number
- CN107729924B CN107729924B CN201710876917.1A CN201710876917A CN107729924B CN 107729924 B CN107729924 B CN 107729924B CN 201710876917 A CN201710876917 A CN 201710876917A CN 107729924 B CN107729924 B CN 107729924B
- Authority
- CN
- China
- Prior art keywords
- probability
- picture
- review
- original
- interval
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Probability & Statistics with Applications (AREA)
- Image Analysis (AREA)
Abstract
The present invention relates to a kind of picture review probability interval generation method, device, storage medium and computer equipments.The described method includes: obtaining the bad probability that each picture in pictures corresponding with original review probability interval is imperfect picture;For every group of picture with identical bad probability, statistics is determined as the first quantity of normal picture by review, and is determined as the second quantity of imperfect picture by review;Picture review probability interval is regenerated according to corresponding first quantity of every group of bad probability and the second quantity.Above-mentioned picture review probability interval generation method, device, storage medium and computer equipment passes through the result after combining artificial review, review probability interval is adjusted, so that newly-generated picture review probability interval range is more accurate, the workload for reducing picture review, to improve the efficiency of imperfect picture review.
Description
Technical field
The present invention relates to fields of communication technology, more particularly to a kind of picture review probability interval generation method, device, deposit
Storage media and computer equipment and a kind of picture review determination method, device, storage medium and computer equipment.
Background technique
With the development of communication technology, for the ease of the communication of user, more and more network platforms are such as social flat
Platform, financial platform etc. all provide the function that picture uploads downloading for user.This is but also the propagation of imperfect picture is increasingly held
Easily.The propagation of imperfect picture in order to prevent, many network platforms by the picture that will transmit access imperfect picture identification system into
Row identifies, and picture is accredited as single class a kind of in normal, sexy, pornographic three classes by can be for imperfect picture identification system feedback
Other probability, or when picture is accredited as normal, sexy, pornographic three classes corresponding three probability.
However, identifying system by imperfect picture since there are omission factor or false detection rates for current imperfect picture identification system
Still some picture needs to carry out artificial review to confirm in picture after system identification.In imperfect picture identification system discrimination
In limited situation, the picture review probability interval that imperfect picture identification system generates is bigger, therefore the work of artificial review
Amount is also corresponding larger, needs to expend a large amount of unnecessary times, lower so as to cause the review efficiency of imperfect picture.
Summary of the invention
The embodiment of the present invention provides a kind of picture review probability interval generation method, device, storage medium and computer and sets
It is standby, it can effectively solve the lower technical problem of imperfect picture review efficiency.
The embodiment of the present invention provides a kind of picture review determination method, device, storage medium and computer equipment, Ke Yiyou
Effect solves the lower technical problem of imperfect picture review efficiency.
A kind of picture review probability interval generation method, which comprises obtain opposite with original review probability interval
Each picture is the bad probability of imperfect picture in the pictures answered;For every group of picture with identical bad probability, statistics
It is determined as the first quantity of normal picture by review, and is determined as the second quantity of imperfect picture by review;It is bad according to every group
Corresponding first quantity of probability and the second quantity regenerate picture review probability interval.
Each figure in pictures corresponding with original review probability interval is obtained described in one of the embodiments,
Before piece is the bad probability of imperfect picture, comprising: obtain in default pictures, each picture is that imperfect picture is initially general
Rate;The corresponding probability of each picture is calculated, generating each picture in the default pictures is imperfect picture
Bad probability;Original review probability interval is set, and extracts the corresponding picture of bad probability for being located at original review probability interval,
Form pictures corresponding with the original review probability interval.
It is described in one of the embodiments, to be given birth to again according to corresponding first quantity of every group of bad probability and the second quantity
At picture review probability interval, comprising: obtain the original Upper Probability and original lower limit probability of the original review probability interval;
Go out two class probabilities according to the original Upper Probability and original lower limit probability calculation, can be obtained by two class probability
The sum of second quantity corresponding to original lower limit probability to two class probabilities, and corresponding to two class probabilities to original Upper Probability
The sum of the first quantity, the minimum value of the two summation;It is general to original Upper Probability and original lower limit according to two class probability
Rate is calculated, and picture review probability interval is regenerated.
In one of the embodiments, it is described according to two class probability to original Upper Probability and original lower limit probability
It is calculated, regenerates picture review probability interval, comprising: when two class probabilities are less than or equal to predetermined probabilities, according to public affairs
FormulaIt is calculated, regenerates Upper Probability s2', wherein s1For original lower limit probability, s2
For original Upper Probability, t is two class probabilities, x0For predetermined probabilities;When two class probabilities are greater than predetermined probabilities, according to formulaIt is calculated, regenerates lower limit probability s1', wherein s1For original lower limit probability, s2It is original
Upper Probability, t are two class probabilities, x0For predetermined probabilities;According to the new Upper Probability s2' or the new lower limit probability
s1', regenerate picture review probability interval.
It is described in one of the embodiments, to obtain each picture in pictures corresponding with original review probability interval
After the bad probability of imperfect picture, further includes: obtain the picture system power of the corresponding image transmission system of the pictures
Value;It is described that original Upper Probability and original lower limit probability are calculated according to two class probability, it is multiple to regenerate picture
Examine probability interval, comprising: according to two class probability and the picture system weight, to original Upper Probability and original lower limit
Probability is calculated, and picture review probability interval is regenerated.
A kind of picture review determination method, which comprises obtain the bad probability that picture to be identified is imperfect picture;
The bad probability is judged whether within preset picture review probability interval, and the picture review probability interval is according to above-mentioned
Each picture review probability interval generation method as described in the examples and generate;If so, determining that the picture to be identified needs
Carry out picture review.
A kind of picture review probability interval generating means, described device include: that bad probability obtains module, for obtain with
Each picture is the bad probability of imperfect picture in the corresponding pictures of original review probability interval;Picture determines quantity statistics
Module, for having the picture of identical bad probability for every group, statistics is determined as the first quantity of normal picture by review, and
It is determined as the second quantity of imperfect picture by review;Picture review probability interval generation module, for according to every group of bad probability
Corresponding first quantity and the second quantity regenerate picture review probability interval.
A kind of picture review decision maker, described device includes: that the bad probability of picture to be identified obtains module, for obtaining
Picture to be identified is the bad probability of imperfect picture;Picture review judgment module, for judging the bad probability whether pre-
If picture review probability interval within, picture review probability interval picture described in any one of -5 according to claim 1
Review probability interval generation method generates;If so, determining that the picture to be identified needs to carry out picture review.
A kind of computer readable storage medium, is stored thereon with computer program, realization when which is executed by processor
The step of above-mentioned each picture review probability interval generation method as described in the examples.
A kind of computer readable storage medium, is stored thereon with computer program, realization when which is executed by processor
The step of above-mentioned each picture review determination method as described in the examples.
A kind of computer equipment can be run on a memory and on a processor including memory, processor and storage
Computer program, the processor realize above-mentioned each picture review probability interval as described in the examples when executing described program
The step of generation method.
A kind of computer equipment can be run on a memory and on a processor including memory, processor and storage
Computer program, the processor realize above-mentioned each picture review determination method as described in the examples when executing described program
The step of.
Above-mentioned picture review probability interval generation method, device, storage medium and computer equipment and picture review are sentenced
Method, apparatus, storage medium and computer equipment are determined, by the artificial review for corresponding to pictures in conjunction with original review probability interval
As a result, statistics is determined as the first quantity of normal picture by review, and is answered for every group of picture with identical bad probability
Trial is set to the second quantity of imperfect picture, is calculated according to the first quantity and the second quantity original review probability interval,
Regenerate picture review probability interval.Review probability interval is adjusted according to the review result of statistics, so that newly-generated
Picture review probability interval range it is more accurate, reduce picture review workload, to improve the effect of imperfect picture review
Rate.
Detailed description of the invention
Fig. 1 is the applied environment figure of picture review probability interval generation method in one embodiment;
Fig. 2 is the flow chart of picture review probability interval generation method in one embodiment;
Fig. 3 is the flow chart of picture review probability interval generation method in another embodiment;
Fig. 4 is the flow chart of picture review probability interval generation method in another embodiment;
Fig. 5 is the flow chart of picture review determination method in one embodiment;
Fig. 6 is the structural block diagram of picture review probability interval generating means in one embodiment;
Fig. 7 is the structural block diagram of picture review probability interval generating means in another embodiment;
Fig. 8 is the structural block diagram of picture review decision maker in one embodiment;
Fig. 9 is the internal structure chart of the first computer equipment in one embodiment.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated.It should be appreciated that specific embodiment described herein is only used to explain the present invention, not
For limiting the present invention.
It is appreciated that term " first " used in the present invention, " second " etc. can be used to describe various elements herein,
But these elements are not limited by these terms.These terms are only used to distinguish the first element from the other element.Citing
For, without departing from the scope of the invention, the first quantity can be known as the second quantity, and similarly, it can be by
Two quantity are known as the first quantity.First quantity and the second quantity both quantity, but it is not same quantity.
Picture review probability interval generation method provided by the embodiment of the present invention, can be applied to the application environment such as Fig. 1
In.Referring to Fig.1, which includes the first computer equipment 102 and second computer equipment 104.First computer equipment
102 and second computer equipment 104 can be terminal be also server.Wherein, terminal includes but is not limited to mobile phone, tablet computer
Or personal digital assistant or wearable device etc., server can be independent physical server, be also possible to multiple physics
The server cluster that server is constituted.First computer equipment 102 and second computer equipment can be the computer of same type
Equipment can also be different types of computer equipment.First computer equipment 102 can be used for executing the embodiment of the present invention and be provided
Picture review probability interval generation method.Second computer equipment 104 can be the end for being stored with imperfect picture identification system
End or server, can be used for identifying imperfect picture and export the probability of picture.First computer equipment 102 can be with the second meter
Machine equipment 104 is calculated to be connected to the network, such as, the first computer equipment 102 can obtain second computer equipment 104 and carry out to picture
Identify the data such as obtained probability, network connection includes but is not limited to wireless network, cable network etc..
In one embodiment, the first computer equipment 102 and second computer equipment 104 can set for same computer
It is standby.
In one embodiment, as shown in Fig. 2, providing a kind of picture review probability interval generation method, this method can
Applied to the first computer equipment 102 in application environment as shown in Figure 1, this method comprises:
Step S202 obtains in corresponding with original review probability interval pictures each picture as imperfect picture not
Good probability.
In the present embodiment, original review probability interval can refer to the review probability interval of preset fixation, before being also possible to
The review probability interval once adjusted.Wherein, review probability interval refers in imperfect picture qualification process, needs to carry out artificial
Probability interval corresponding to the picture of review.Bad probability refers to that picture is the possibility degree of imperfect picture, and bad probability is in
In original review probability interval.The probability that can be exported according to imperfect picture identification systems, calculates according to preset algorithm
To bad probability.It further, can be not plan deliberately according to each picture in pictures corresponding with original review probability interval
The bad probability of piece, statistics form the bad distribution of picture on original review probability interval, wherein horizontal in the bad distribution of picture
Coordinate can be bad probability, and ordinate can be picture number corresponding to identical bad probability.
Step S204, for every group of picture with identical bad probability, statistics is determined as the of normal picture by review
One quantity, and it is determined as by review the second quantity of imperfect picture.
In the present embodiment, the picture counted in original review probability interval is determined as normal picture or imperfect picture by review
Quantity.Wherein, in the picture of identical bad probability, it is determined as that the picture number of normal picture is the first quantity, quilt by review
Review is determined as that the picture number of imperfect picture is the second quantity.
Step S206 regenerates picture review probability according to corresponding first quantity of every group of bad probability and the second quantity
Section.
In the present embodiment, new picture review probability interval refers to for determining whether picture will carry out the general of artificial review
Rate section, when the bad probability of new received picture is among the new picture review probability interval that the above method is calculated
When, which can be judged to needing to carry out artificial review.It further, can also be according to corresponding to the picture of each bad probability
The first quantity and the second quantity, be respectively formed after the first distributed number and the second distributed number, according to every group of bad probability
Corresponding first distributed number and the second distributed number regenerate picture review probability interval.
In above-described embodiment, by corresponding to the artificial review of pictures in conjunction with original review probability interval as a result, statistics quilt
Review is determined as the first quantity of normal picture, and is determined as the second quantity of imperfect picture by review, according to the first quantity and
Second quantity calculates original review probability interval, regenerates picture review probability interval.By combining artificial review
Afterwards as a result, be adjusted to review probability interval so that newly-generated picture review probability interval range is more accurate, reduce
The workload of picture review, to improve the efficiency of imperfect picture review.
In one embodiment, as shown in figure 3, before step S202, further includes:
Step S302 is obtained in default pictures, and each picture is the probability of imperfect picture.
In the present embodiment, default pictures refer to that picture system inputs imperfect picture identification system within a preset period of time
The pictures that picture is constituted.Each picture of probability imperfect picture identification system output identify after probability, not plan deliberately
Picture is accredited as single class probability a kind of in normal, sexy, pornographic three classes by can be for piece identification system output, either
Corresponding three multi-class probability when picture is accredited as normal, sexy, pornographic three classes.Wherein, single class probability and
Multi-class probability can be that picture is the corresponding other confidence level of picture category.The value of single class probability and multi-class probability can be 0
To 1.
Step S304 calculates the corresponding probability of each picture, generates each picture in default pictures and is
The bad probability of imperfect picture.
In the present embodiment, bad probability is that have system obtained from being calculated according to preset algorithm according to probability
The probability of one standard.Specifically, single class probability or multi-class probability are uniformly mapped as to the probability on same term of reference.
Further, it obtains in default pictures, each picture is picture category corresponding to the probability of imperfect picture
Other weight;According to picture classification weight corresponding to the probability of identical picture, to the corresponding probability of each picture into
Row calculates, and generates the bad probability that each picture in default pictures is imperfect picture.Wherein, picture classification weight refers to bad
Weight corresponding to the picture classification of picture identification system output.When probability is single class probability, each picture institute
Corresponding picture classification weight is to preset one of normal weight, default sexy weight, default pornographic weight three classes;When initial
When probability is multi-class probability, picture classification weight corresponding to each picture can be three, be to preset normal weight, pre- respectively
If sexy weight and default pornographic weight.
Step S306 sets original review probability interval, and extracts the bad probability pair for being located at original review probability interval
The picture answered forms pictures corresponding with original review probability interval.
It, can be according to original review probability after calculating the bad probability that each picture is imperfect picture in the present embodiment
Section extracts corresponding picture, forms pictures corresponding with original review probability interval, which is to need to carry out
The pictures of artificial review.Further, pictures corresponding with original review probability interval can be extracted store to
Database under preset path enables picture review personnel directly to carry out review to the picture extracted.
In above-described embodiment, uniformly it is mapped to by the probability for the picture for exporting imperfect picture identification systems bad
Probability is compared, and permits a determination that the original review probability interval with unified reference standard, and general according to original review
It is multiple to regenerate picture as a result, calculate original review probability interval for the artificial reviews of the corresponding pictures in rate section
Probability interval is examined, the method for determination of picture review probability interval has been standardized, has improved the accuracy of picture review probability interval.
In one embodiment, step S206, further includes: obtain the original Upper Probability and original of original review probability interval
Beginning lower limit probability;Go out two class probabilities according to original Upper Probability and original lower limit probability calculation, it can by two class probabilities
The sum of second quantity corresponding to original lower limit probability to two class probabilities is obtained, with two class probabilities to original Upper Probability institute
The sum of corresponding first quantity, the minimum value of the two summation;It is general to original Upper Probability and original lower limit according to two class probabilities
Rate is calculated, and picture review probability interval is regenerated.
In the present embodiment, original Upper Probability and original lower limit probability refer to the endpoint of original review probability interval, and former
The value of beginning Upper Probability is greater than the value of original lower limit probability.Two class probabilities, which refer to, is divided into two for original review probability interval
The value in section, two class probabilities are greater than original lower limit threshold value and are less than original upper limit threshold.Specifically, by original lower limit probability
To the sum of second quantity corresponding to two class probabilities, with the first quantity corresponding to two class probabilities to original Upper Probability it
With, when the summation of the two is minimized, available two class probabilities.It is represented by formulaWherein,
s1For original lower limit probability, s2For original Upper Probability, MiIndicate that probability is the second quantity of i, NiIndicate that probability is the first of i
Quantity, by takingWithThe minimum value of sum of the two can acquire two class probability t.
Specifically, the original Upper Probability and original lower limit probability of original review probability interval, can be according to bad with calculating
The identical preset algorithm of probability is calculated.Since there are omission factor or false detection rates for each imperfect picture identification system, often
A imperfect picture identification system has itself preset probability section for needing to carry out review, can extract the probability section
Initial upper limit probability and initial lower limit probability calculated, obtain original Upper Probability and original lower limit probability.Original review
The original Upper Probability and original lower limit probability of probability interval can also be straight according to the preceding picture review probability interval once adjusted
It connects determining.
It further, can also be after forming the first distributed number and the second distributed number, by the second distributed number in original
Beginning lower limit probability to two class probabilities integral, with the first distributed number in two class probabilities to the integral of original Upper Probability,
The sum of integral of the two, which is minimized, calculates to obtain two class probabilities.According to two class probabilities to original Upper Probability and original lower limit
Probability is calculated separately or together according to preset, regenerates picture review probability interval.
In one embodiment, each picture is bad in obtaining pictures corresponding with original review probability interval
After the bad probability of picture, further includes: obtain the picture system weight of the corresponding image transmission system of pictures.Step S206
It include: to calculate original Upper Probability and original lower limit probability, give birth to again according to two class probabilities and picture system weight
At picture review probability interval.
In the present embodiment, picture system weight is and one or more weights corresponding to picture system.Wherein, picture system
System includes but is not limited to recreational system, such as blind date forum, social platform etc., or can also be for financial field
System, such as banking system, securities system etc..Specifically, picture system weight can be carried out according to the attribute or parameter of picture system
Assignment.For example, assignment can be carried out according to picture type, wherein picture type includes but is not limited to invoice document class, client's document
Class etc. is linked up in class, credentials file class, work circulation class, customer communication, according to inside to external traffic flow degree by as low as
High assignment;Assignment can also be carried out according to picture access frequency, wherein access frequency can be divided into outside access frequency and lower carrier frequency
Rate etc., according to the value assignment from low to high of frequency;Assignment can also be carried out according to picture significance level, such as can be according to picture
Carry out assignment with the correlation degree of transaction, can according to picture and transaction certificate, promotional and the degree of correlation of customer account management by
Low to high assignment.It further, can be according to one or more system weights of picture system and two class probabilities, to the original upper limit
Probability and original lower limit probability are calculated, and picture review probability interval is regenerated.It specifically, can also be by multiple system weights
Integrated system weight of the sum as the picture system.When integrated system weight is at preset weight section, take corresponding
Separately or concurrently original Upper Probability and original lower limit probability are calculated.
In one embodiment, weight q can be determined according to picture type1, weight q can be determined according to picture access frequency2,
Weight q can be determined according to picture significance level3.It specifically, can be according to q1、q2、q3Weights sum determines integrated system weight Q=q1
+q2+q3, and q can be limited1、q2、q3Value range.For example, q1Can value 1 to 10, q2Value 1 to 20, q3Value 1 to 10.Into
One step, corresponding calculation formula can be taken to regenerate the original upper limit according to the size of integrated system weight Q and preset threshold
Probability and original lower limit probability.For example, when integrated system weight Q is less than or equal to 20, adjustmentWork as synthesis
System weight Q is greater than 20, adjustmentWherein, wherein s1For original lower limit probability, s2It is original
Upper Probability, t are two class probabilities, s1' it is new lower limit probability, s2' it is new Upper Probability.
In one embodiment, the class of picture can be determined according to the keyword in the URL (resource unifies finger URL) of picture
Type.Such as if the relevant keyword of the URL of a picture is cargo, payment for goods, payment etc., document can be stamped to picture
Label.Specifically, the mapping relations of keyword and type can be generated with the database of predetermined keyword.When extracting in URL
When the keyword for including, which is matched with the predetermined keyword in database, obtains the key with successful match
Word has the picture type of mapping relations, as picture type corresponding with the picture.
In the present embodiment, corresponding picture review probability interval is generated for the case where different picture systems, so that raw
At picture review probability interval be more in line with actual demand, improve the accuracy in picture review section.
In one embodiment, there can also be corresponding picture system weight for different users.The picture system weight
Diagram data can be sent out according to the history of user carry out assignment.Such as the picture sent out in being recorded according to user's history is bad
Degree carries out assignment;The picture access frequency that figure can be also sent out according to user's history carries out assignment;It can also be recorded according to user's history
The undesirable level of the other forms content such as text of middle hair carries out assignment.It for example, is above to be transmitted through for a historical custom
Porny, and picture access frequency is greater than the user of preset threshold, can assign relatively high weight.
In one embodiment, the undesirable level for the picture sent out in being recorded according to user's history determines weight c1, can root
Weight c is determined according to the picture access frequency of user's history hair figure2, can be recorded according to user's history in the other forms such as the text sent out
The undesirable level of content determines weight c3.It specifically, can be according to c1、c2、c3Weights sum determines integrated system weight C=c1+c2+
c3, and c can be limited1、c2、c3Value range.For example, c1Can value 1 to 10, c2Value 1 to 20, c3Value 1 to 10.Further
Ground can take corresponding calculation formula to regenerate original Upper Probability according to the size of integrated system weight C and preset threshold
With original lower limit probability.For example, when integrated system weight C is less than or equal to 20, adjustmentWork as integrated system
Weight C is greater than 20, adjustmentWherein, wherein s1For original lower limit probability, s2For the original upper limit
Probability, t are two class probabilities, s1' it is new lower limit probability, s2' it is new Upper Probability.
In the present embodiment, corresponding picture review probability interval is generated for the historical record of different user, so that generating
Picture review probability interval be more in line with actual demand, improve the accuracy in picture review section.
In one embodiment, as shown in figure 4, providing another picture review probability interval generation method, this method
Include:
Step S402 is obtained in default pictures, and each picture is the probability of imperfect picture, corresponding to each picture
Probability calculated, generate the bad probability that each picture in default pictures is imperfect picture.
In the present embodiment, each picture of probability imperfect picture identification system output identify after probability, not plan deliberately
Picture is accredited as single class probability a kind of in normal, sexy, pornographic three classes by can be for piece identification system output, either
Corresponding three multi-class probability when picture is accredited as normal, sexy, pornographic three classes.
In one embodiment, it when probability is single class probability, can preset respectively normal, sexy and three kinds pornographic
The mapping equation of single class probability calculates and generates bad probability.Such as be single class probability A of normal picture for identification,
Mapping equation can be bad probabilityIt is single class probability B of sexy picture for identification, mapping equation can be bad general
RateFor the single class probability C identified as porny, mapping equation can be bad probabilityIts
In, if the value of A, B, C are all 0 to 1, bad probability y value corresponding with probability A be 0 toIt is corresponding with probability B not
Good probability y value isExtremelyBad probability y value corresponding with probability C isTo 1, therefore the value range of bad probability y
It is 0 to 1.In the present embodiment, by the way that three kinds of single class probabilities are mapped as the substandard bad probability of same value, ratio is reduced
The difficulty of poorer picture identification system output probability, has standardized the standard of bad probability, improves determining review probability interval
Accuracy.
In one embodiment, it when probability is multi-class probability, can integrate normal general corresponding to every picture
Rate, sexy probability and pornographic probability calculate according to unified mapping equation and generate bad probability.Such as identical picture
Corresponding normal probability D, sexuality probability E and pornographic probability F, and the sum of three probability can be 1, can preset mapping equation is not
Good probabilityIn the present embodiment, bad probability is obtained by comprehensive three multi-class probability calculations,
The difficulty for comparing imperfect picture identification system output probability is reduced, the standard of bad probability has been standardized, improves determining review
The accuracy of probability interval.
Step S404 sets original review probability interval, and extracts the bad probability pair for being located at original review probability interval
The picture answered forms pictures corresponding with original review probability interval.
Specifically, original review probability interval can be determined according to the mapping equation of the bad probability of determination.Wherein, original multiple
It examines in the value interval of probability interval bad probability corresponding to default pictures.
Step S406 obtains in corresponding with original review probability interval pictures each picture as imperfect picture not
Good probability.
Step S408, for every group of picture with identical bad probability, statistics is determined as the of normal picture by review
One quantity, and it is determined as by review the second quantity of imperfect picture.
In the present embodiment, using the corresponding pictures of original review probability interval as the picture for needing to carry out artificial review
Collection, so that the picture in the pictures is determined as normal picture or imperfect picture by picture review personnel.
In one embodiment, for picture review, it is possible to provide corresponding picture review interface is shown former on the surface
Picture in the corresponding pictures of beginning review probability interval, picture review personnel can select normogram therein according to judgement
Piece or imperfect picture.Further, corresponding picture can be shown according to each bad probability, when detecting picture review personnel's
Clicking operation can then stamp corresponding label to the picture, count the number of labels of uniform type as the first quantity or
Two quantity.
For example, picture review personnel can be clicked to the picture selected as imperfect picture and labeled as " 1 ", will not had
The picture that selects is clicked as normal picture and labeled as " 0 ", when all having carried out review to the corresponding picture of the bad probability
Later, it can count and mark the picture number for 0 " as the first quantity, picture of the statistics labeled as " 0 " under the bad probability
Quantity is as the second quantity.
Step S410 obtains the original Upper Probability and original lower limit probability of original review probability interval, on original
Limit probability and original lower limit probability calculation go out two class probabilities.
In the present embodiment, it can be obtained second corresponding to original lower limit probability to two class probabilities by two class probabilities
The sum of quantity, with the sum of first quantity corresponding to two class probabilities to original Upper Probability, the minimum value of the two summation.Specifically
Ground obtains the corresponding original Upper Probability s of original review probability interval2With original lower limit probability s1, wherein original Upper Probability
s2With original lower limit probability s1It can be determined according to the omission factor and false detection rate of imperfect picture identification system.
It in one embodiment, can the first distributed number P of the generation of the first quantity according to corresponding to each bad probability1
It (x), correspondingly, can the second distributed number P of the generation of the second quantity according to corresponding to each bad probability2(x).Further,
It can be according to formulaIt calculates and generates two class probability t, wherein two class probability t make, by the
Two distributed number P2(x) in original lower limit probability s1To the integral of two class probability t, with the first distributed number P1(x) in two classification
Probability t to original Upper Probability s2Integral, the sum of integral of the two is minimized.
Step S412, judges whether two class probabilities are greater than predetermined probabilities.
In the present embodiment, when two class probabilities are less than or equal to predetermined probabilities, step S314 is executed;When two class probabilities are big
When predetermined probabilities, step S316 is executed.
Step S414 is calculated according to preset first formula, regenerates Upper Probability.
Specifically, the first formula can be preset asWherein, s1For original lower limit probability,
s2For original Upper Probability, t is two class probabilities, x0For predetermined probabilities, s2' it is new Upper Probability.
Step S416 is calculated according to preset second formula, regenerates lower limit probability.
Specifically, the second formula can be preset asWherein, s1For original lower limit probability, s2For original
Beginning Upper Probability, t are two class probabilities, x0For predetermined probabilities, s1' it is new lower limit probability.
Step S418 regenerates picture review according to the Upper Probability regenerated or the lower limit probability regenerated
Probability interval.
Specifically, according to new Upper Probability s2' or new lower limit probability s1', picture review probability interval is regenerated,
The picture review probability interval is s1' to s2’。
In above-described embodiment, uniformly it is mapped to by the probability for the picture for exporting imperfect picture identification systems bad
Probability is compared, according to picture corresponding to original review probability interval after artificial review as a result, statistics by review
It is determined as the first quantity of normal picture, and is determined as the second quantity of imperfect picture by review, according to the first quantity and second
Quantity calculates original review probability interval, regenerates picture review probability interval.After combining artificial review
Two class probabilities are generated as a result, calculating.According to the size of two class probabilities and predetermined probabilities by the adjustment to review probability interval
It is divided into two kinds of situations, so that newly-generated picture review probability interval range is more accurate, reduces the workload of picture review, from
And improve the efficiency of imperfect picture review.
In one embodiment, as shown in figure 5, providing a kind of picture review determination method, this method comprises:
Step S502 obtains the bad probability that picture to be identified is imperfect picture.
In the present embodiment, picture to be identified is the picture for needing to be made to determine whether image review to be carried out.Specifically, it obtains
Take the bad probability and the calculation for calculating each bad probability of picture in default pictures that picture to be identified is imperfect picture
Unanimously.
Whether step S504 judges bad probability within preset picture review probability interval.
In the present embodiment, the generation method of picture review probability interval includes: that acquisition is opposite with original review probability interval
Each picture is the bad probability of imperfect picture in the pictures answered;For every group of picture with identical bad probability, statistics
It is determined as the first quantity of normal picture by review, and is determined as the second quantity of imperfect picture by review;It is bad according to every group
Corresponding first quantity of probability and the second quantity regenerate picture review probability interval.If so, thening follow the steps S506;If
It is no, then follow the steps S508.
Step S506 determines that picture to be identified needs to carry out picture review.
Step S508 determines that picture to be identified does not need to carry out picture review.
In one embodiment, the generation method of picture review probability interval further include: obtaining and original review probability
In the corresponding pictures in section, before each picture is the bad probability of imperfect picture, obtain in default pictures, Mei Getu
Piece is the probability of imperfect picture;The corresponding probability of each picture is calculated, is generated each in default pictures
Picture is the bad probability of imperfect picture;Original review probability interval is set, and extracts and is located at original review probability interval not
The corresponding picture of good probability, forms pictures corresponding with original review probability interval.
In one embodiment, to regenerate picture according to corresponding first quantity of every group of bad probability and the second quantity multiple
Examine probability interval, comprising: obtain the original Upper Probability and original lower limit probability of original review probability interval;According to the original upper limit
Probability and original lower limit probability calculation go out two class probabilities, can obtain original lower limit probability to two classification by two class probabilities
The sum of second quantity corresponding to probability, with the sum of first quantity, the two corresponding to two class probabilities to original Upper Probability
The minimum value of summation;Original Upper Probability and original lower limit probability are calculated according to two class probabilities, regenerate picture
Review probability interval.
In one embodiment, original Upper Probability and original lower limit probability are calculated according to two class probabilities, weight
Newly-generated picture review probability interval, comprising: when two class probabilities are less than or equal to predetermined probabilities, according to formulaIt is calculated, regenerates Upper Probability s2', wherein s1For original lower limit probability, s2For
Original Upper Probability, t are two class probabilities, x0For predetermined probabilities;When two class probabilities are greater than predetermined probabilities, according to formulaIt is calculated, regenerates lower limit probability s1', wherein s1For original lower limit probability, s2It is original
Upper Probability, t are two class probabilities, x0For predetermined probabilities;According to new Upper Probability s2' or new lower limit probability s1', again
Generate picture review probability interval.
In one embodiment, obtaining each picture in pictures corresponding with original review probability interval is not plan deliberately
After the bad probability of piece, further includes: obtain the picture system weight of the corresponding image transmission system of pictures;According to two classification
Probability calculates original Upper Probability and original lower limit probability, regenerates picture review probability interval, comprising: according to two
Class probability and picture system weight, calculate original Upper Probability and original lower limit probability, regenerate picture review
Probability interval.
In above-described embodiment, uniformly it is mapped to by the probability for the picture for exporting imperfect picture identification systems bad
Probability is compared, according to picture corresponding to original review probability interval after artificial review as a result, statistics by review
It is determined as the first quantity of normal picture, and is determined as the second quantity of imperfect picture by review, according to the first quantity and second
Quantity calculates original review probability interval, regenerates picture review probability interval.After combining artificial review
As a result, being adjusted to review probability interval, so that newly-generated picture review probability interval range is more accurate, picture is reduced
The workload of review, to improve the efficiency of imperfect picture review.
In one embodiment, as shown in fig. 6, providing a kind of picture review probability interval generating means 600, the device
Include: that bad probability obtains module 602, is for obtaining each picture in pictures corresponding with original review probability interval
The bad probability of imperfect picture;Picture determines quantity statistical module 604, for for every group of figure with identical bad probability
Piece, statistics are determined as the first quantity of normal picture by review, and are determined as the second quantity of imperfect picture by review;Picture is multiple
Probability interval generation module 606 is examined, for regenerating figure according to corresponding first quantity of every group of bad probability and the second quantity
Piece review probability interval.
In one embodiment, as shown in fig. 7, providing another picture review probability interval generating means 700, the dress
Setting further includes that original review pictures form module 601, and for obtaining in default pictures, each picture is the first of imperfect picture
Beginning probability;The corresponding probability of each picture is calculated, generating each picture in default pictures is imperfect picture
Bad probability;Original review probability interval is set, and extracts the corresponding picture of bad probability for being located at original review probability interval,
Form pictures corresponding with original review probability interval.
In one embodiment, picture review probability interval generation module 606 is also used to obtain original review probability interval
Original Upper Probability and original lower limit probability;Go out two class probabilities according to original Upper Probability and original lower limit probability calculation,
The sum of second quantity corresponding to original lower limit probability to two class probabilities can be obtained by two class probabilities, it is general with two classification
The sum of first quantity corresponding to rate to original Upper Probability, the minimum value of the two summation;According to two class probabilities on original
Limit probability and original lower limit probability are calculated, and picture review probability interval is regenerated.
In one embodiment, picture review probability interval generation module 606 is also used to be less than or equal to when two class probabilities
When predetermined probabilities, according to formulaIt is calculated, regenerates Upper Probability s2', wherein s1
For original lower limit probability, s2For original Upper Probability, t is two class probabilities, x0For predetermined probabilities;When two class probabilities are greater than in advance
If when probability, according to formulaIt is calculated, regenerates lower limit probability s1', wherein s1It is original
Lower limit probability, s2For original Upper Probability, t is two class probabilities, x0For predetermined probabilities;According to new Upper Probability s2' or new
Lower limit probability s1', regenerate picture review probability interval.
In one embodiment, picture review probability interval generation module 606 is also used to obtain the corresponding picture of pictures
The picture system weight of Transmission system;According to two class probabilities and picture system weight, to original Upper Probability and original lower limit
Probability is calculated, and picture review probability interval is regenerated.
In one embodiment, as shown in figure 8, providing a kind of picture review decision maker 800, the device include: to
Identify that the bad probability of picture obtains module 802, for obtaining the bad probability that picture to be identified is imperfect picture;Picture review is sentenced
Disconnected module 804, for judge bad probability whether within preset picture review probability interval, picture review probability interval
Generation method include: obtain each picture in corresponding with original review probability interval pictures be imperfect picture it is bad generally
Rate;For in every group of picture with identical bad probability, statistics is determined as the first quantity of normal picture by review, and is answered
Trial is set to the second quantity of imperfect picture;Figure is regenerated according to corresponding first quantity of every group of bad probability and the second quantity
Piece review probability interval;If so, determining that picture to be identified needs to carry out picture review.
In one embodiment, each picture is bad in obtaining pictures corresponding with original review probability interval
Before the bad probability of picture, comprising: obtain in default pictures, each picture is the probability of imperfect picture;To each
The corresponding probability of picture is calculated, and the bad probability that each picture in default pictures is imperfect picture is generated;Setting
Original review probability interval, and extract be located at original review probability interval the corresponding picture of bad probability, formed with it is original multiple
Examine the corresponding pictures of probability interval.
In one embodiment, to regenerate picture according to corresponding first quantity of every group of bad probability and the second quantity multiple
Examine probability interval, comprising: obtain the original Upper Probability and original lower limit probability of original review probability interval;According to the original upper limit
Probability and original lower limit probability calculation go out two class probabilities, can obtain original lower limit probability to two classification by two class probabilities
The sum of second quantity corresponding to probability, with the sum of first quantity, the two corresponding to two class probabilities to original Upper Probability
The minimum value of summation;Original Upper Probability and original lower limit probability are calculated according to two class probabilities, regenerate picture
Review probability interval.
In one embodiment, original Upper Probability and original lower limit probability are calculated according to two class probabilities, weight
Newly-generated picture review probability interval, comprising: when two class probabilities are less than or equal to predetermined probabilities, according to formulaIt is calculated, regenerates Upper Probability s2', wherein s1For original lower limit probability, s2For
Original Upper Probability, t are two class probabilities, x0For predetermined probabilities;When two class probabilities are greater than predetermined probabilities, according to formulaIt is calculated, regenerates lower limit probability s1', wherein s1For original lower limit probability, s2It is original
Upper Probability, t are two class probabilities, x0For predetermined probabilities;According to new Upper Probability s2' or new lower limit probability s1', again
Generate picture review probability interval.
In one embodiment, obtaining each picture in pictures corresponding with original review probability interval is not plan deliberately
After the bad probability of piece, further includes: obtain the picture system weight of the corresponding image transmission system of pictures;According to two classification
Probability calculates original Upper Probability and original lower limit probability, regenerates picture review probability interval, comprising: according to two
Class probability and picture system weight, calculate original Upper Probability and original lower limit probability, regenerate picture review
Probability interval.
Above-mentioned picture review probability interval generating means and picture review decision maker can be implemented as a kind of computer
The form of program, computer program can be run in the first computer equipment of such as Fig. 9.It is non-volatile in first computer equipment
Property storage medium can store composition the picture review probability interval generating means and picture review decision maker each program mould
Bad probability in block, such as Fig. 6 obtains module 602, picture determines that quantity statistical module 604, picture review probability interval are raw
Module 802 and picture review judgment module 804 etc. are obtained at the bad probability of picture to be identified in module 606 and Fig. 8.It is each
It include computer program in program module, computer program is each for making server execute the application described in this specification
Step in the picture review probability interval generation method and picture review determination method of embodiment, in the first computer equipment
Processor can call the picture review probability interval stored in the non-volatile memory medium of the first computer equipment to generate dress
Each program module with picture review decision maker is set, corresponding computer program is run, realizes that picture is multiple in this specification
Examine the corresponding function of modules of probability interval generating means and picture review decision maker.
In one embodiment, a kind of computer readable storage medium is provided, computer program is stored thereon with, the journey
Each picture in acquisition pictures corresponding with original review probability interval is performed the steps of when sequence is executed by processor is
The bad probability of imperfect picture;For every group of picture with identical bad probability, statistics is determined as normal picture by review
First quantity, and it is determined as by review the second quantity of imperfect picture;According to corresponding first quantity of every group of bad probability and
Two quantity regenerate picture review probability interval.
In one embodiment, when which is executed by processor, acquisition and the original review probability interval phase realized
It is further comprising the steps of before each picture is the bad probability step of imperfect picture in corresponding pictures: to obtain default figure
Piece is concentrated, and each picture is the probability of imperfect picture;The corresponding probability of each picture is calculated, is generated default
Each picture is the bad probability of imperfect picture in pictures;Original review probability interval is set, and extracts and is located at original review
The corresponding picture of bad probability of probability interval forms pictures corresponding with original review probability interval.
In one embodiment, when which is executed by processor, realized according to every group of bad probability corresponding
One quantity and the second quantity regenerate the step of picture review probability interval, specifically includes the following steps: obtaining original review
The original Upper Probability and original lower limit probability of probability interval;Go out two points according to original Upper Probability and original lower limit probability calculation
Class probability can obtain the sum of second quantity corresponding to original lower limit probability to two class probabilities by two class probabilities, with
The sum of first quantity corresponding to two class probabilities to original Upper Probability, the minimum value of the two summation;According to two class probabilities
Original Upper Probability and original lower limit probability are calculated, picture review probability interval is regenerated.
In one embodiment, when which is executed by processor, realized according to two class probabilities to the original upper limit
The step of probability and original lower limit probability are calculated, and picture review probability interval is regenerated, specifically includes the following steps: working as
When two class probabilities are less than or equal to predetermined probabilities, according to formulaIt is calculated, is regenerated
Upper Probability s2', wherein s1For original lower limit probability, s2For original Upper Probability, t is two class probabilities, x0For predetermined probabilities;
When two class probabilities are greater than predetermined probabilities, according to formulaIt is calculated, it is general to regenerate lower limit
Rate s1', wherein s1For original lower limit probability, s2For original Upper Probability, t is two class probabilities, x0For predetermined probabilities;According to new
Upper Probability s2' or new lower limit probability s1', regenerate picture review probability interval.
In one embodiment, when which is executed by processor, acquisition and the original review probability interval phase realized
After the step of each picture is the bad probability of imperfect picture in corresponding pictures, further includes: it is corresponding to obtain pictures
The picture system weight of image transmission system;Realized according to two class probabilities to original Upper Probability and original lower limit probability
The step of being calculated, regenerating picture review probability interval, specifically includes: being weighed according to two class probabilities and picture system
Value, calculates original Upper Probability and original lower limit probability, regenerates picture review probability interval.
In one embodiment, a kind of computer readable storage medium is provided, computer program is stored thereon with, the journey
The bad probability for obtaining that picture to be identified is imperfect picture is performed the steps of when sequence is executed by processor;Judge bad probability
Whether within preset picture review probability interval, picture review probability interval is multiple according to the picture in above-mentioned each embodiment
It examines probability interval generation method and generates.
In one embodiment, a kind of computer equipment is provided, including memory, processor and storage are on a memory
And the computer program that can be run on a processor, processor are performed the steps of when executing program: being obtained and original review
Each picture is the bad probability of imperfect picture in the corresponding pictures of probability interval;There is identical bad probability for every group
Picture, statistics is determined as the first quantity of normal picture by review, and is determined as the second quantity of imperfect picture by review;Root
Picture review probability interval is regenerated according to corresponding first quantity of every group of bad probability and the second quantity.
In one embodiment, when above-mentioned processor executes program, acquisition and the original review probability interval phase realized
It is further comprising the steps of before the step of each picture is the bad probability of imperfect picture in corresponding pictures: to obtain default
In pictures, each picture is the probability of imperfect picture;The corresponding probability of each picture is calculated, is generated pre-
If each picture is the bad probability of imperfect picture in pictures;Original review probability interval is set, and is extracted positioned at original multiple
The corresponding picture of bad probability of probability interval is examined, pictures corresponding with original review probability interval are formed.
In one embodiment, when above-mentioned processor executes program, realized according to every group of bad probability corresponding the
One quantity and the second quantity regenerate the step of picture review probability interval, specifically includes the following steps: obtaining original review
The original Upper Probability and original lower limit probability of probability interval;Go out two points according to original Upper Probability and original lower limit probability calculation
Class probability can obtain the sum of second quantity corresponding to original lower limit probability to two class probabilities by two class probabilities, with
The sum of first quantity corresponding to two class probabilities to original Upper Probability, the minimum value of the two summation;According to two class probabilities
Original Upper Probability and original lower limit probability are calculated, picture review probability interval is regenerated.
In one embodiment, when above-mentioned processor executes program, realized according to two class probabilities to the original upper limit
The step of probability and original lower limit probability are calculated, and picture review probability interval is regenerated, specifically includes the following steps: working as
When two class probabilities are less than or equal to predetermined probabilities, according to formulaIt is calculated, is regenerated
Upper Probability s2', wherein s1For original lower limit probability, s2For original Upper Probability, t is two class probabilities, x0For predetermined probabilities;
When two class probabilities are greater than predetermined probabilities, according to formulaIt is calculated, it is general to regenerate lower limit
Rate s1', wherein s1For original lower limit probability, s2For original Upper Probability, t is two class probabilities, x0For predetermined probabilities;According to new
Upper Probability s2' or new lower limit probability s1', regenerate picture review probability interval.
In one embodiment, when above-mentioned processor executes program, acquisition and the original review probability interval phase realized
After the step of each picture is the bad probability of imperfect picture in corresponding pictures, further includes: it is corresponding to obtain pictures
The picture system weight of image transmission system;Realized according to two class probabilities to original Upper Probability and original lower limit probability
The step of being calculated, regenerating picture review probability interval, specifically includes: being weighed according to two class probabilities and picture system
Value, calculates original Upper Probability and original lower limit probability, regenerates picture review probability interval.
In one embodiment, a kind of computer equipment is provided, including memory, processor and storage are on a memory
And the computer program that can be run on a processor, processor is performed the steps of when executing program obtains picture to be identified
For the bad probability of imperfect picture;Bad probability is judged whether within preset picture review probability interval, and picture review is general
Rate section is generated according to the picture review probability interval generation method in above-mentioned each embodiment.
In one embodiment, above-mentioned computer equipment may include but be not limited to independent physical server, either
The server cluster that multiple physical servers are constituted.Such as Fig. 9, the internal structure for the first computer equipment in one embodiment is shown
It is intended to.First computer equipment can be applied to the first computer equipment 102 in the application environment of Fig. 1.First computer
Equipment includes processor, non-volatile memory medium, built-in storage and the network interface connected by system bus.Wherein, should
The processor of first computer equipment supports the operation of entire first computer equipment for providing calculating and control ability.The
The non-volatile memory medium of one computer equipment is stored with operating system, database and computer program.It is deposited in the database
It contains and determines for realizing a kind of picture review probability interval generation method and picture review provided by above each embodiment
The relevant data of method, for example probability corresponding to each pictures in preset pictures can be stored with.The computer
Program can be performed by processor, to generate for realizing a kind of picture review probability interval provided by above each embodiment
Method and a kind of picture review determination method.Built-in storage in first computer equipment is the behaviour in non-volatile memory medium
Make system, database and computer program and the running environment of cache is provided.Network interface can be Ethernet card or wireless
Network interface card etc., for being communicated with external terminal or server, such as with the second computer equipment in the application environment of Fig. 1
104 are communicated, and the probability that each picture in default pictures is imperfect picture is obtained.
It will be understood by those skilled in the art that the structure of the first computer equipment shown in Fig. 9, only and the application
The block diagram of the relevant part-structure of scheme does not constitute the limit for the first computer equipment being applied thereon to application scheme
Fixed, specific first computer equipment may include perhaps combining certain components or tool than components more or fewer in figure
There is different component layouts.For example, the first computer equipment in the figure may also include display screen etc..
Those of ordinary skill in the art will appreciate that realizing all or part of the process in above-described embodiment method, being can be with
Relevant hardware is instructed to complete by computer program, the program can be stored in a non-volatile computer and can be read
In storage medium, the program is when being executed, it may include such as the process of the embodiment of above-mentioned each method.Wherein, the storage is situated between
Matter can be magnetic disk, CD, read-only memory (Read-Only Memory, ROM) etc..
Each technical characteristic of embodiment described above can be combined arbitrarily, for simplicity of description, not to above-mentioned reality
It applies all possible combination of each technical characteristic in example to be all described, as long as however, the combination of these technical characteristics is not deposited
In contradiction, all should be considered as described in this specification.
The embodiments described above only express several embodiments of the present invention, and the description thereof is more specific and detailed, but simultaneously
It cannot therefore be construed as limiting the scope of the patent.It should be pointed out that coming for those of ordinary skill in the art
It says, without departing from the inventive concept of the premise, various modifications and improvements can be made, these belong to protection of the invention
Range.Therefore, the scope of protection of the patent of the invention shall be subject to the appended claims.
Claims (8)
1. a kind of picture review probability interval generation method, which comprises
Obtain the bad probability that each picture in pictures corresponding with original review probability interval is imperfect picture;
Using the picture with identical bad probability as one group;
For in every group of picture with identical bad probability, statistics is determined as the first quantity of normal picture by review, and by
Review is determined as the second quantity of imperfect picture;
Obtain the original Upper Probability and original lower limit probability of the original review probability interval;
Go out two class probabilities according to the original Upper Probability and original lower limit probability calculation, two class probability enables to
The sum of second quantity corresponding to original lower limit probability to two class probabilities, and corresponding to two class probabilities to original Upper Probability
The sum of the first quantity, the two summation is minimized;
Original Upper Probability and original lower limit probability are calculated according to two class probability, it is general to regenerate picture review
Rate section.
2. the method according to claim 1, wherein corresponding with original review probability interval in the acquisition
Before each picture is the bad probability of imperfect picture in pictures, comprising:
It obtains in default pictures, each picture is the probability of imperfect picture;
The corresponding probability of each picture is calculated, generating each picture in the default pictures is imperfect picture
Bad probability;
Original review probability interval is set, and extracts the corresponding picture of bad probability for being located at original review probability interval, is formed
Pictures corresponding with the original review probability interval.
3. the method according to claim 1, wherein it is described according to two class probability to original Upper Probability
It is calculated with original lower limit probability, regenerates picture review probability interval, comprising:
When two class probabilities are less than or equal to predetermined probabilities, according to formulaIt is calculated, weight
Newly-generated Upper Probability s2', wherein s1For original lower limit probability, s2For original Upper Probability, t is two class probabilities, x0It is pre-
If probability;
When two class probabilities are greater than predetermined probabilities, according to formulaIt is calculated, is regenerated down
Limit probability s1', wherein s1For original lower limit probability, s2For original Upper Probability, t is two class probabilities, x0For predetermined probabilities;
According to the new Upper Probability s2' or the new lower limit probability s1', regenerate picture review probability interval.
4. the method according to claim 1, wherein described obtain figure corresponding with original review probability interval
Piece is concentrated after the bad probability that each picture is imperfect picture, further includes:
Obtain the picture system weight of the corresponding image transmission system of the pictures;
It is described that original Upper Probability and original lower limit probability are calculated according to two class probability, it is multiple to regenerate picture
Examine probability interval, comprising:
According to two class probability and the picture system weight, original Upper Probability and original lower limit probability are counted
It calculates, regenerates picture review probability interval.
5. a kind of picture review determination method, which comprises
Obtain the bad probability that picture to be identified is imperfect picture;
Judge the bad probability whether within preset picture review probability interval, the picture review probability interval according to
Picture review probability interval generation method described in any one of claims 1 to 4 generates;
If so, determining that the picture to be identified needs to carry out picture review.
6. a kind of picture review probability interval generating means, which is characterized in that described device includes:
Bad probability obtains module, is bad for obtaining each picture in pictures corresponding with original review probability interval
The bad probability of picture;
Picture determines quantity statistical module, for that will have the picture of identical bad probability as one group;There is phase for every group
In picture with bad probability, statistics is determined as the first quantity of normal picture by review, and is determined as imperfect picture by review
The second quantity;
Picture review probability interval generation module, for obtaining the original Upper Probability of the original review probability interval and original
Lower limit probability;Go out two class probabilities according to the original Upper Probability and original lower limit probability calculation, two class probability makes
The sum of second quantity corresponding to original lower limit probability to two class probabilities is obtained, it is right to original Upper Probability institute with two class probabilities
The sum of first quantity answered, the two summation are minimized;According to two class probability to original Upper Probability and original lower limit
Probability is calculated, and picture review probability interval is regenerated.
7. a kind of picture review decision maker, which is characterized in that described device includes:
The bad probability of picture to be identified obtains module, for obtaining the bad probability that picture to be identified is imperfect picture;
Picture review judgment module, for judge the bad probability whether within preset picture review probability interval, institute
Picture review probability interval is stated to generate to picture review probability interval generation method described in any one of 4 according to claim 1;If
It is then to determine that the picture to be identified needs to carry out picture review.
8. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the program is held by processor
The step of any one of claim 1 to 5 the method is realized when row.
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710876917.1A CN107729924B (en) | 2017-09-25 | 2017-09-25 | Picture review probability interval generation method and picture review determination method |
PCT/CN2017/108480 WO2019056496A1 (en) | 2017-09-25 | 2017-10-31 | Method for generating picture review probability interval and method for picture review determination |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710876917.1A CN107729924B (en) | 2017-09-25 | 2017-09-25 | Picture review probability interval generation method and picture review determination method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107729924A CN107729924A (en) | 2018-02-23 |
CN107729924B true CN107729924B (en) | 2019-02-19 |
Family
ID=61206890
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710876917.1A Active CN107729924B (en) | 2017-09-25 | 2017-09-25 | Picture review probability interval generation method and picture review determination method |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN107729924B (en) |
WO (1) | WO2019056496A1 (en) |
Families Citing this family (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108491866B (en) * | 2018-03-06 | 2022-09-13 | 平安科技(深圳)有限公司 | Pornographic picture identification method, electronic device and readable storage medium |
CN109491970B (en) * | 2018-10-11 | 2024-05-10 | 平安科技(深圳)有限公司 | Bad picture detection method and device for cloud storage and storage medium |
CN109886865B (en) * | 2019-01-07 | 2024-01-23 | 平安科技(深圳)有限公司 | Method, device, computer equipment and storage medium for automatically shielding bad information |
CN112445410B (en) * | 2020-12-07 | 2023-04-18 | 北京小米移动软件有限公司 | Touch event identification method and device and computer readable storage medium |
CN113095563A (en) * | 2021-04-07 | 2021-07-09 | 全球能源互联网研究院有限公司 | Method and device for reviewing prediction result of artificial intelligence model |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101692639A (en) * | 2009-09-15 | 2010-04-07 | 西安交通大学 | Bad webpage recognition method based on URL |
CN103748998B (en) * | 2010-06-09 | 2012-02-08 | 北京理工大学 | The supervisory control system of flame and violation service in large-scale internal network |
CN102663093A (en) * | 2012-04-10 | 2012-09-12 | 中国科学院计算机网络信息中心 | Method and device for detecting bad website |
Family Cites Families (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101359372B (en) * | 2008-09-26 | 2011-05-11 | 腾讯科技(深圳)有限公司 | Training method and device of classifier, method and apparatus for recognising sensitization picture |
CN101996180A (en) * | 2009-08-12 | 2011-03-30 | 升东网络科技发展(上海)有限公司 | Picture examination and filter system and method |
CN102340424B (en) * | 2010-07-21 | 2013-12-04 | 中国移动通信集团山东有限公司 | Bad message detection method and bad message detection device |
JP5707291B2 (en) * | 2011-09-29 | 2015-04-30 | 株式会社日立ハイテクノロジーズ | Charged particle beam system that supports image classification |
CN103064858B (en) * | 2011-10-19 | 2016-03-30 | 北京千橡网景科技发展有限公司 | Method and apparatus for objectionable image detection in social networking websites |
CN106485528A (en) * | 2015-09-01 | 2017-03-08 | 阿里巴巴集团控股有限公司 | The method and apparatus of detection data |
CN105654057A (en) * | 2015-12-31 | 2016-06-08 | 中国建设银行股份有限公司 | Picture auditing system and picture auditing method based on picture contents |
CN106776842B (en) * | 2016-11-28 | 2021-03-19 | 腾讯科技(上海)有限公司 | Multimedia data detection method and device |
CN107145904A (en) * | 2017-04-28 | 2017-09-08 | 北京小米移动软件有限公司 | Determination method, device and the storage medium of image category |
-
2017
- 2017-09-25 CN CN201710876917.1A patent/CN107729924B/en active Active
- 2017-10-31 WO PCT/CN2017/108480 patent/WO2019056496A1/en active Application Filing
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101692639A (en) * | 2009-09-15 | 2010-04-07 | 西安交通大学 | Bad webpage recognition method based on URL |
CN103748998B (en) * | 2010-06-09 | 2012-02-08 | 北京理工大学 | The supervisory control system of flame and violation service in large-scale internal network |
CN102663093A (en) * | 2012-04-10 | 2012-09-12 | 中国科学院计算机网络信息中心 | Method and device for detecting bad website |
Also Published As
Publication number | Publication date |
---|---|
WO2019056496A1 (en) | 2019-03-28 |
CN107729924A (en) | 2018-02-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107729924B (en) | Picture review probability interval generation method and picture review determination method | |
WO2020082733A1 (en) | Method for message push request flow control based on service rules, device and medium | |
CN110414987A (en) | Recognition methods, device and the computer system of account aggregation | |
CN108229367A (en) | A kind of face identification method and device | |
CN109447156A (en) | Method and apparatus for generating model | |
WO2022116491A1 (en) | Dbscan clustering method based on horizontal federation, and related device therefor | |
CN110225104A (en) | Data capture method, device and terminal device | |
CN108776692A (en) | Method and apparatus for handling information | |
CN104077723A (en) | Social network recommending system and social network recommending method | |
CN109583228B (en) | Privacy information management method, device and system | |
CN104484600B (en) | Intrusion detection method and device based on improved density clustering | |
CN107766470A (en) | Intelligent statistical method, Intelligent statistical display methods and the device of a kind of data sharing | |
CN109492031A (en) | Display methods, server and the computer storage medium of list data | |
CN110245684A (en) | Data processing method, electronic equipment and medium | |
CN104573132B (en) | Song lookup method and device | |
CN112966756A (en) | Visual access rule generation method and device, machine readable medium and equipment | |
CN110851868A (en) | Position representative element generation method for track data release | |
CN108629506A (en) | Modeling method, device, computer equipment and the storage medium of air control model | |
CN110851655B (en) | Method and system for simplifying complex network | |
CN113971602A (en) | Financial data processing method and device, storage medium and server | |
CN109754295A (en) | Method and apparatus for output information | |
CN113487350A (en) | Business product determination method and related device | |
CN117237124B (en) | Digital asset management method and device based on multi-terminal interaction | |
CN113342998B (en) | Multimedia resource recommendation method and device, electronic equipment and storage medium | |
CN110019808A (en) | A kind of method and apparatus of predictive information attribute |
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 |