WO2019056496A1 - Method for generating picture review probability interval and method for picture review determination - Google Patents

Abstract

A method for generating a picture review probability interval, the method comprising: acquiring a probability of defectiveness by which each picture in a picture set corresponding to an original review probability interval is a defective picture; for each group of pictures having the same probability of defectiveness, counting a first number for pictures that are determined by means of review to be normal pictures and a second number for pictures that are determined by means of review to be defective pictures; newly generating a picture review probability interval according to a first number and second number corresponding to the probability of defectiveness of each group.

Description

Image review probability interval generation method and picture complex trial method

This application is required to be submitted to the Chinese Patent Office on September 25, 2017, the application number is 2017108769171, and the invention name is the priority of the Chinese patent application of the "picture review probability interval generation method and picture re-determination method", the entire contents of which are incorporated by reference. In this application.

Technical field

The present invention relates to the field of communications technologies, and in particular, to a method, an apparatus, a storage medium, and a computer device for generating a picture review probability interval, and a picture retry method, apparatus, storage medium, and computer device.

Background technique

With the development of communication technology, in order to facilitate user communication, more and more network platforms, such as social platforms and financial platforms, provide users with the function of uploading and downloading pictures. This also makes the spread of bad pictures easier and easier. In order to prevent the spread of bad pictures, many network platforms identify the transmitted pictures by accessing the bad picture authentication system. The bad picture authentication system can feedback the single-category probability of identifying the pictures as one of normal, sexy and pornographic. Or the three probabilities corresponding to the identification of the picture as normal, sexy, and erotic.

However, due to the current flaw detection rate or false detection rate of the bad picture authentication system, some pictures in the pictures identified by the bad picture authentication system still need to be manually reviewed to confirm. In the case that the recognition rate of the bad picture identification system is limited, the picture review probability interval generated by the bad picture authentication system is relatively large, so the workload of the manual review is also relatively large, and it takes a lot of unnecessary time, which leads to the review of the bad picture. Less efficient.

Summary of the invention

According to various embodiments of the present application, a picture review probability interval generation method, apparatus, storage medium, and computer device are provided.

In accordance with various embodiments of the present application, a picture review method, apparatus, storage medium, and computer apparatus are provided.

A method for generating a picture review probability interval, the method comprising: obtaining a bad probability that each picture in the picture set corresponding to the original review probability interval is a bad picture; and for each group of pictures having the same bad probability, the statistical review is determined as The first number of normal pictures, and the second number determined to be a bad picture; the picture review probability interval is regenerated according to the first quantity and the second quantity corresponding to each group of bad probabilities.

In one embodiment, before the obtaining a bad probability of each picture in the picture set corresponding to the original review probability interval, the method includes: acquiring a preset picture set, where each picture is a bad picture initial probability; Calculating an initial probability corresponding to each picture, generating a bad probability that each picture in the preset picture set is a bad picture; setting an original review probability interval, and extracting a picture corresponding to the bad probability in the original review probability interval, forming A set of pictures corresponding to the original review probability interval.

In one embodiment, the regenerating the picture review probability interval according to the first quantity and the second quantity corresponding to each set of the bad probability comprises: obtaining an original upper limit probability and an original lower limit probability of the original review probability interval; Calculating a two-category probability by using the original upper bound probability and the original lower bound probability, by which the sum of the original lower bound probability to the second quantity corresponding to the second classification probability can be obtained, and the second classification probability corresponds to the original upper limit probability a sum of quantities, the minimum of the sum of the two; calculating the original upper limit probability and the original lower limit probability according to the two classification probability, and regenerating the picture review probability interval.

In one embodiment, the calculating the original upper limit probability and the original lower limit probability according to the two classification probability, and regenerating the picture review probability interval, including: when the two classification probability is less than or equal to the preset probability, according to the formula

Perform calculation to regenerate the upper limit probability s ₂ ', where s ₁ is the original lower bound probability, s ₂ is the original upper bound probability, t is the second classification probability, x ₀ is the preset probability; when the second classification probability is greater than the preset probability, According to the formula

Calculate and regenerate the lower bound probability s ₁ ', where s ₁ is the original lower bound probability, s ₂ is the original upper bound probability, t is the second classification probability, x ₀ is the preset probability; according to the new upper bound probability s ₂ ' Or the new lower limit probability s ₁ ', regenerating the picture review probability interval.

In one embodiment, after obtaining the bad probability that each picture in the picture set corresponding to the original review probability interval is a bad picture, the method further includes: acquiring a picture system weight of the picture transmission system corresponding to the picture set; Calculating the original upper limit probability and the original lower limit probability according to the second classification probability, and regenerating the picture review probability interval, comprising: calculating the original upper limit probability and the original lower limit probability according to the second classification probability and the picture system weight value Perform calculations to regenerate the image review probability interval.

A picture re-determination method, the method comprising: obtaining a bad probability that a picture to be identified is a bad picture; determining whether the bad probability is within a preset picture review probability interval, and the picture review probability interval is according to the foregoing implementation The picture review probability interval generation method described in the example is generated; if yes, determining that the picture to be identified requires a picture review.

A picture review probability interval generating device, the device comprising: a bad probability obtaining module, configured to acquire a bad probability that each picture in the picture set corresponding to the original review probability interval is a bad picture; and a picture determination quantity statistics module, configured to: For each group of pictures with the same bad probability, the statistics are determined to be the first number of normal pictures, and the second number is determined as the bad picture; the picture review probability interval generation module is used to correspond to each group of bad probabilities. The first quantity and the second quantity regenerate the picture review probability interval.

A picture re-determination device, the device comprising: a picture defect probability acquisition module to be used for obtaining a bad probability that a picture to be identified is a bad picture; a picture review judging module, configured to determine whether the bad probability is preset Within the picture review probability interval, the picture review probability interval is generated according to the picture review probability interval generation method described in the above embodiments; if yes, determining that the picture to be identified requires image review.

One or more non-transitory readable storage mediums storing computer readable instructions, when executed by one or more processors, cause the one or more processors to execute The following steps: obtaining the bad probability that each picture in the picture set corresponding to the original review probability interval is a bad picture; for each picture having the same bad probability, the statistics are determined to be the first quantity of the normal picture, and the trial is determined. a second number of bad pictures; and regenerating a picture review probability interval according to the first quantity and the second quantity corresponding to each set of bad probabilities.

One or more non-transitory readable storage mediums storing computer readable instructions, when executed by one or more processors, cause the one or more processors to perform the steps of: acquiring The probability that the picture to be identified is a bad picture; determining whether the bad probability is within a preset picture review probability interval, the step of generating the picture review probability interval includes acquiring each picture set corresponding to the original review probability interval The picture is a bad probability of a bad picture; for each group of pictures with the same bad probability, the statistics are determined to be the first number of normal pictures, and the second number of bad pictures is determined as the bad picture; A quantity and a second quantity regenerate the picture review probability interval; and if so, determining that the picture to be identified requires a picture review.

A computer device comprising a memory and one or more processors, the memory storing computer readable instructions, the computer readable instructions being executed by the processor, causing the one or more processors to execute The following steps: obtaining a bad probability that each picture in the picture set corresponding to the original review probability interval is a bad picture; for each picture having the same bad probability, the statistics are determined to be the first number of normal pictures, and the trial is determined a second number of bad pictures; and regenerating a picture review probability interval according to the first quantity and the second quantity corresponding to each set of bad probabilities.

A computer device comprising a memory and one or more processors, the memory storing computer readable instructions, the computer readable instructions being executed by the processor, causing the one or more processors to execute The following steps: obtaining a bad probability that the picture to be identified is a bad picture; determining whether the bad probability is within a preset picture review probability interval, and the generating step of the picture review probability interval includes obtaining a corresponding interval corresponding to the original review probability interval The bad probability of each picture being a bad picture in the picture set; for each group of pictures with the same bad probability, the statistics are determined to be the first number of normal pictures, and the second number of bad pictures is determined as a bad picture; The first quantity and the second quantity corresponding to the probability regenerate the picture review probability interval; and if so, determining that the picture to be identified requires a picture review.

Details of one or more embodiments of the present application are set forth in the accompanying drawings and description below. Other features, objects, and advantages of the invention will be apparent from the description and appended claims.

DRAWINGS

In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings to be used in the embodiments will be briefly described below. Obviously, the drawings in the following description are only some embodiments of the present application, Those skilled in the art can also obtain other drawings based on these drawings without any creative work.

1 is an application environment diagram of a method for generating a picture review probability interval in an embodiment;

2 is a flowchart of a method for generating a picture review probability interval in an embodiment;

3 is a flowchart of a method for generating a picture review probability interval in another embodiment;

4 is a flowchart of a method for generating a picture review probability interval in still another embodiment;

Figure 5 is a flow chart of a method for resuming a picture in an embodiment;

6 is a structural block diagram of a picture review probability interval generating apparatus in an embodiment;

FIG. 7 is a structural block diagram of a picture review probability interval generating apparatus in another embodiment; FIG.

Figure 8 is a block diagram showing the structure of a picture re-determining device in an embodiment;

Figure 9 is a diagram showing the internal structure of a first computer device in one embodiment.

Detailed ways

In order to make the objects, technical solutions, and advantages of the present application more comprehensible, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It is understood that the specific embodiments described herein are merely illustrative of the application and are not intended to be limiting.

It will be understood that the terms "first", "second" and the like, as used herein, may be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish one element from another. For example, without departing from the scope of the application, The first quantity is referred to as a second quantity, and similarly, the second quantity may be referred to as a first quantity. Both the first quantity and the second quantity are quantities, but they are not the same quantity.

The image review probability interval generation method provided by the embodiment of the present application can be applied to the application environment as shown in FIG. 1 . Referring to FIG. 1, the application environment includes a first computer device 102 and a second computer device 104. The first computer device 102 and the second computer device 104 can be terminals or servers. The terminal includes but is not limited to a mobile phone, a tablet computer, or a personal digital assistant or a wearable device. The server may be an independent physical server or a server cluster composed of multiple physical servers. The first computer device 102 and the second computer device can be the same type of computer device, or can be different types of computer devices. The first computer device 102 can be used to execute the picture review probability interval generation method provided by the embodiment of the present application. The second computer device 104 may be a terminal or server that stores a bad picture authentication system that can be used to identify bad pictures and output an initial probability of the picture. The first computer device 102 can be connected to the second computer device 104. For example, the first computer device 102 can obtain data such as an initial probability that the second computer device 104 authenticates the picture, and the network connection includes, but is not limited to, a wireless network. Wired network, etc.

In one embodiment, the first computer device 102 and the second computer device 104 can be the same computer device.

In one embodiment, as shown in FIG. 2, a picture review probability interval generation method is provided, which is applicable to the first computer device 102 in the application environment as shown in FIG. 1, the method comprising:

Step S202: Obtain a bad probability that each picture in the picture set corresponding to the original review probability interval is a bad picture.

The original review probability interval may refer to a preset fixed review probability interval, or may be a previous adjusted review probability interval. Among them, the review probability interval refers to the probability interval corresponding to the picture that needs to be manually reviewed in the process of bad picture identification. The bad probability refers to the possible degree of the picture being a bad picture, and the bad probability is in the original review probability interval. According to the initial probability of the bad picture identification system output, the bad probability can be calculated according to a preset algorithm. Further, according to the original image of the original review probability interval, the probability of each picture being a bad picture may be statistically shaped. The image is poorly distributed in the original review probability interval. In the bad distribution of the image, the abscissa may be a bad probability, and the ordinate may be the number of pictures corresponding to the same bad probability.

Step S204, for each group of pictures having the same bad probability, the first number of the normal picture is determined as the normal picture, and the second number is determined as the bad picture.

The number of pictures in the original review probability interval is determined to be the number of normal pictures or bad pictures. Among them, in the picture with the same bad probability, the number of pictures that are determined to be normal pictures is the first number, and the number of pictures that are determined to be bad pictures is the second number.

Step S206: Regenerate the picture review probability interval according to the first quantity and the second quantity corresponding to each group of the bad probability.

The new picture review probability interval refers to the probability interval used to determine whether the picture is to be manually reviewed. When the bad probability of the newly received picture is in the new picture review probability interval calculated by the above method, the picture can be determined. For manual review. Further, after the first quantity distribution and the second quantity distribution are respectively formed according to the first quantity and the second quantity corresponding to the picture of each bad probability, the first quantity distribution and the second quantity corresponding to each group of the bad probability are further The quantity distribution regenerates the image review probability interval.

In the above embodiment, by combining the manual review results of the original review probability interval corresponding picture set, the first number of the normal picture is determined as the normal picture, and the second number is determined as the bad picture, according to the first quantity and the second number. The quantity is calculated for the original review probability interval, and the image review probability interval is regenerated. By combining the results of the manual review, the review probability interval is adjusted, so that the newly generated image review probability interval range is more accurate, reducing the workload of image review, thereby improving the efficiency of poor image review.

In an embodiment, as shown in FIG. 3, before step S202, the method further includes:

Step S302: Acquire a preset picture set, and each picture is an initial probability of a bad picture.

The preset picture set refers to a picture set formed by the picture system inputting a picture of the bad picture authentication system within a preset time period. The initial probability is that the probability of each picture outputted by the image authentication system is different. The output of the bad picture authentication system may be a single-category probability that the picture is identified as one of normal, sexy, and erotic, or the picture is identified as normal. Three pairs of sexy, erotic, and erotic Multiple category probabilities. Among them, the single category probability and the multi-category probability may be the confidence that the picture is the corresponding picture category. The single category probability and the multi-class probability may range from 0 to 1.

Step S304, calculating an initial probability corresponding to each picture, and generating a bad probability that each picture in the preset picture set is a bad picture.

The probability of badness is the probability of having a uniform standard obtained by calculating according to the initial probability according to a preset algorithm. Uniform mapping of single category probabilities or multi-category probabilities to probabilities over the same reference range.

In an embodiment, the preset picture set is obtained, and each picture is a picture category weight corresponding to an initial probability of the bad picture; and an initial corresponding to each picture according to the picture category weight corresponding to the initial probability of the same picture The probability is calculated to generate a bad probability that each picture in the preset picture set is a bad picture. The picture category weight refers to the weight corresponding to the picture category output by the bad picture authentication system. When the initial probability is a single category probability, the picture category weight corresponding to each picture is one of three categories: a preset normal weight, a preset sexy weight, and a preset porn weight; when the initial probability is a multi-category In the case of probability, the picture category weight value corresponding to each picture may be three, which are a preset normal weight, a preset sexy weight, and a preset porn weight.

Step S306, setting an original review probability interval, and extracting a picture corresponding to the bad probability in the original review probability interval, and forming a picture set corresponding to the original review probability interval.

After calculating the bad probability that each picture is a bad picture, the corresponding picture may be extracted according to the original review probability interval to form a picture set corresponding to the original review probability interval, and the picture set is a picture set that needs manual review. Further, the picture set corresponding to the original review probability interval may be extracted and stored in a database under the preset path, so that the picture reviewer can directly review the extracted picture.

In the above embodiment, the initial probability of the picture outputted by the bad picture identification system is uniformly mapped to the bad probability for comparison, so that the original review probability interval having the unified reference standard can be determined, and the picture set corresponding to the original review probability interval is determined. The results of the manual review are used to calculate the original review probability interval, regenerate the image review probability interval, standardize the determination method of the image review probability interval, and improve the accuracy of the image review probability interval.

In an embodiment, step S206, further comprising: acquiring an original review probability interval on the original The probability of the limit and the probability of the original lower limit; the probability of the second class is calculated according to the original upper limit probability and the original lower limit probability, and the sum of the second lower limit probability corresponding to the second classification probability and the second classification probability to the original upper limit can be obtained by the two classification probability The sum of the first quantity corresponding to the probability, the minimum of the sum of the two; the original upper limit probability and the original lower limit probability are calculated according to the second classification probability, and the picture review probability interval is regenerated.

The original upper bound probability and the original lower bound probability refer to the endpoints of the original review probability interval, and the value of the original upper bound probability is greater than the original lower bound probability. The second classification probability refers to dividing the original review probability interval into two interval values, and the second classification probability is greater than the original lower limit threshold and less than the original upper limit threshold. When the sum of the original lower bound probability to the second quantity corresponding to the second classification probability and the first quantity corresponding to the second upper limit probability to the original upper limit probability, the sum of the two takes the minimum value, the two classification probability is obtained. Can be expressed as a formula

Where s ₁ is the original lower bound probability, s ₂ is the original upper bound probability, M _i represents the second quantity with probability i, and N _i represents the first quantity with probability i, by taking

versus

The minimum of the sum of the two can be obtained as the second classification probability t.

In one embodiment, the original upper bound probability and the original lower bound probability of the original review probability interval may be calculated according to a preset algorithm that is the same as the calculated bad probability. Since each bad picture identification system has a missed detection rate or a false detection rate, each bad picture authentication system has its own initial probability interval that needs to be reviewed, and the initial upper limit probability and the initial lower limit probability of the initial probability interval can be extracted. The calculation is performed to obtain the original upper limit probability and the original lower limit probability. The original upper bound probability and the original lower bound probability of the original review probability interval may also be directly determined according to the previously adjusted picture review probability interval.

Further, after the first quantity distribution and the second quantity distribution are formed, the second quantity is distributed from the original lower limit probability to the integral of the second classification probability, and the first quantity is distributed to the integral of the second classification probability to the original upper limit probability. The sum of the integrals of the two takes the minimum value to calculate the two-class probability. According to the second classification probability, the original upper limit probability and the original lower limit probability are calculated according to presets or simultaneously, and the picture review probability interval is regenerated.

In an embodiment, after obtaining the bad probability that each picture is a bad picture in the picture set corresponding to the original review probability interval, the method further includes: acquiring the picture transmission system corresponding to the picture set Image system weight. Step S206 includes: calculating the original upper limit probability and the original lower limit probability according to the second classification probability and the picture system weight, and regenerating the picture review probability interval.

The picture system weight is one or more weights corresponding to the picture system. The picture system includes, but is not limited to, an entertainment system, such as a blind date forum, a social platform, or the like, or a system in the financial field, such as a banking system, a securities system, and the like. Specifically, the picture system weight can be assigned according to the attributes or parameters of the picture system. For example, the value can be assigned according to the type of the picture, wherein the picture type includes but is not limited to the invoice document class, the customer document class, the voucher file class, the work circulation class, the customer communication class, etc., according to the degree of internal to external process business from low to High assignment; can also be assigned according to the frequency of image access, wherein the access frequency can be divided into external access frequency and download frequency, etc., according to the value of the frequency from low to high assignment; can also be assigned according to the importance of the image, for example, according to The degree of relevance of the image to the transaction is assigned, and can be assigned from low to high according to the degree of relevance of the image to the transaction voucher, promotion and customer management. Further, the original upper limit probability and the original lower limit probability may be calculated according to one or more system weights and two classification probabilities of the picture system, and the picture review probability interval is regenerated. The sum of multiple system weights can also be used as the overall system weight for the picture system. When the weight of the integrated system is in the preset weight interval, the corresponding upper limit probability and the original lower limit probability are calculated separately or simultaneously.

In one embodiment, the weights may be determined q _1, q ₂ weights may be determined according to the picture access frequency, may be determined according to weight q ₃ importance image based on the image type. According to q _1, q _{2, 3} to determine the weighted sum of q and integrated system of weights _{Q = q 1 + q 2 +} q 3, and may define q _1, q _2, q ₃ range. For example, q ₁ can take values from 1 to 10, q ₂ takes values from 1 to 20, and q ₃ takes values from 1 to 10. Further, according to the weight of the integrated system weight Q and the preset threshold, a corresponding calculation formula is used to regenerate the original upper limit probability and the original lower limit probability. For example, when the integrated system weight Q is less than or equal to 20, adjust

When the integrated system weight Q is greater than 20, adjust

Where s ₁ is the original lower bound probability, s ₂ is the original upper bound probability, t is the second classification probability, s ₁ ' is the new lower bound probability, and s ₂ ' is the new upper bound probability.

In one embodiment, the type of picture may be determined based on keywords in the URL (resource uniform locator) of the picture. For example, if the keyword related to the URL of a picture is goods, payment, payment, etc., the picture can be tagged with a document. A database of keywords can be preset to generate a mapping relationship between keywords and types. When the keyword included in the URL is extracted, the keyword is matched with the preset keyword in the database, and the image type having a mapping relationship with the keyword that matches the matching is obtained as the image type corresponding to the image. Corresponding image review probability intervals are generated for different picture system situations, so that the generated picture review probability interval is more in line with actual needs, and the accuracy of the picture review interval is improved.

In one embodiment, corresponding picture system weights may also be available for different users. The image system weight can be assigned according to the user's historical map data. For example, the value may be assigned according to the degree of the badness of the image sent in the user history record; the value may be assigned according to the image access frequency of the user history map; or may be performed according to the degree of badness of other forms of content such as text sent in the user history record. Assignment. For example, a user who has uploaded a pornographic picture and whose picture access frequency is greater than a preset threshold may be given a relatively high weight.

In one embodiment, the poor according to the degree of user history send images to determine the weight c _1, may determine the weight value c based on the image user historical development of FIG access frequency _2, according to the user history issued text, etc. The degree of badness of other forms of content determines the weight c ₃ . May be c _1, c _{2, 3} c, and the weighted sum is determined integrated system weight _{C = c 1 + c 2 +} c 3 according to, and may define a c _1, c _2, c ₃ range. For example, c ₁ may take values from 1 to 10, c _{2 may} take values from 1 to 20, and c _{3 may} take values from 1 to 10. Further, according to the weight of the integrated system weight C and the preset threshold, a corresponding calculation formula is used to regenerate the original upper limit probability and the original lower limit probability. For example, when the integrated system weight C is less than or equal to 20, adjust

When the integrated system weight C is greater than 20, adjust

Where s ₁ is the original lower bound probability, s ₂ is the original upper bound probability, t is the second classification probability, s ₁ ' is the new lower bound probability, and s ₂ ' is the new upper bound probability. Corresponding image review probability intervals are generated for the history records of different users, so that the generated image review probability interval is more in line with actual needs, and the accuracy of the image review interval is improved.

In an embodiment, as shown in FIG. 4, another method for generating a picture review probability interval is provided, and the method includes:

Step S402: Acquire an initial probability that each picture is a bad picture, and calculate an initial probability corresponding to each picture to generate a bad probability that each picture in the preset picture set is a bad picture.

The initial probability is that the probability of each picture outputted by the image authentication system is different. The output of the bad picture authentication system may be a single-category probability that the picture is identified as one of normal, sexy, and erotic, or the picture is identified as normal. Three multi-category probabilities corresponding to the three categories of sexy, erotic, and erotic.

In one embodiment, when the initial probability is a single-category probability, a mapping formula of three single-category probabilities of normal, sexy, and erotic may be preset, respectively, to calculate a probability of generating a bad. For example, for a single-category probability A that is identified as a normal picture, the mapping formula can be a bad probability.

For a single-category probability B identified as a sexy picture, the mapping formula can be a bad probability

For a single-category probability C that is identified as an erotic image, the mapping formula can be a bad probability

Wherein, if the values of A, B, and C are both 0 to 1, the probability y corresponding to the probability A is 0.

The probability of bad y corresponding to probability B is

to

The probability of bad y corresponding to probability C is

Up to 1, so the probability of failure y ranges from 0 to 1. By mapping the three single-category probabilities to the bad probability under the same value criterion, the difficulty of comparing the output probability of the bad picture identification system is reduced, the standard of the bad probability is standardized, and the accuracy of determining the review probability interval is improved.

In one embodiment, when the initial probability is a multi-category probability, the normal probability, the sexy probability, and the erotic probability corresponding to each picture may be integrated, and the probability of generating the defect is calculated according to a unified mapping formula. For example, for the normal probability D, the sexy probability E, and the erotic probability F corresponding to the same picture, and the sum of the three probabilities may be 1, the mapping formula may be preset as a bad probability.

By combining three multi-category probability calculations to obtain the bad probability, the difficulty of comparing the output probability of the bad picture identification system is reduced, the standard of the bad probability is standardized, and the accuracy of determining the review probability interval is improved.

Step S404, setting an original review probability interval, and extracting a picture corresponding to the bad probability in the original review probability interval, and forming a picture set corresponding to the original review probability interval.

The original review probability interval can be determined according to the mapping formula for determining the probability of failure. The original review probability interval is in the value interval of the bad probability corresponding to the preset picture set.

Step S406, obtaining a bad probability that each picture in the picture set corresponding to the original review probability interval is a bad picture.

Step S408, for each group of pictures having the same bad probability, the first number of the normal picture is determined as the normal picture, and the second number is determined as the bad picture.

The picture set corresponding to the original review probability interval is used as a picture set that needs to be manually reviewed, so that the picture reviewer determines the picture in the picture set as a normal picture or a bad picture.

In an embodiment, for the image review, a corresponding image review interface may be provided, and the image in the image set corresponding to the original review probability interval is displayed on the interface, and the image reviewer may select a normal image or a bad image according to the judgment. . Further, the corresponding picture may be displayed according to each bad probability. When the click operation of the picture reviewing personnel is detected, the picture may be marked with a corresponding label, and the number of labels of the unified type is counted as the first quantity or the second quantity.

For example, the picture reviewer may click on the selected picture as a bad picture and mark it as "1", and the picture that has not been clicked as the normal picture is marked as "0", and the picture corresponding to the bad probability is performed. After the review, the number of pictures marked as “0” under the probability of failure may be counted as the first quantity, and the number of pictures marked as “0” as the second quantity.

Step S410: Acquire an original upper bound probability and an original lower bound probability of the original review probability interval, and calculate a second classification probability according to the original upper bound probability and the original lower bound probability.

The sum of the second lower quantity probability to the second quantity corresponding to the second classification probability, the sum of the first quantity corresponding to the second classification probability to the original upper limit probability, and the sum of the sums of the two can be obtained by the two classification probability. The original upper limit probability s ₂ and the original lower limit probability s ₁ corresponding to the original review probability interval are obtained, wherein the original upper limit probability s ₂ and the original lower limit probability s ₁ may be determined according to the missed detection rate and the false detection rate of the bad picture authentication system.

In an embodiment, the first quantity distribution P ₁ (x) may be generated according to the first quantity corresponding to each bad probability. Correspondingly, the second quantity distribution P may be generated according to the second quantity corresponding to each bad probability. ₂ (x). Further, according to the formula

Calculating generates a two-class probability t, wherein the two-class probability t is such that the integral of the second quantity distribution P ₂ (x) between the original lower-order probability s _{1 and the} second-class probability t, and the first quantity distribution P ₁ (x) The integral of the classification probability t to the original upper limit probability s ₂ , and the sum of the integrals of the two takes the minimum value.

In step S412, it is determined whether the two classification probability is greater than a preset probability.

When the two classification probability is less than or equal to the preset probability, step S314 is performed; when the second classification probability is greater than the preset probability, step S316 is performed.

In step S414, the calculation is performed according to the preset first formula, and the upper limit probability is regenerated.

For example, the first formula can be preset to

Where s ₁ is the original lower bound probability, s ₂ is the original upper bound probability, t is the second classification probability, x ₀ is the preset probability, and s ₂ ' is the new upper bound probability.

In step S416, the calculation is performed according to the preset second formula, and the lower limit probability is regenerated.

For example, the second formula can be preset to

Where s ₁ is the original lower bound probability, s ₂ is the original upper bound probability, t is the second classification probability, x ₀ is the preset probability, and s ₁ ' is the new lower bound probability.

Step S418: Regenerate the picture review probability interval according to the regenerated upper limit probability or the regenerated lower limit probability.

According to the new upper limit probability s ₂ ' or the new lower limit probability s ₁ ', the picture review probability interval is regenerated, and the picture review probability interval is s ₁ ' to s ₂ '.

In the above embodiment, the initial probability of the picture outputted by the bad picture identification system is uniformly mapped to the bad probability for comparison, and the result of the manual review after the picture corresponding to the original review probability interval is counted as the normal picture. A quantity, and a second quantity determined as a bad picture by the re-trial, calculates the original review probability interval according to the first quantity and the second quantity, and regenerates the picture review probability interval. The combined two-class probability is calculated by combining the results of the manual review. According to the size of the second classification probability and the preset probability, the adjustment of the review probability interval is divided into two cases, which makes the newly generated picture review probability interval range more accurate, reduces the workload of the picture review, and improves the efficiency of the bad picture review.

In one embodiment, as shown in FIG. 5, a picture retrieving method is provided, the method comprising:

Step S502, obtaining a bad probability that the picture to be identified is a bad picture.

The picture to be identified is a picture that needs to be judged whether or not an image review is to be performed. The bad probability of obtaining the picture to be identified as a bad picture is consistent with the calculation method of calculating the probability of each picture in the preset picture set.

Step S504, determining whether the probability of failure is within a preset picture review probability interval.

The method for generating the image review probability interval includes: obtaining a bad probability that each picture in the picture set corresponding to the original review probability interval is a bad picture; for each group of pictures having the same bad probability, the statistically retried is determined as the first picture of the normal picture. The quantity, and the second number determined to be a bad picture; the picture review probability interval is regenerated according to the first quantity and the second quantity corresponding to each group of bad probabilities. If yes, go to step S506; if no, go to step S508.

In step S506, it is determined that the picture to be identified needs to be reviewed.

In step S508, it is determined that the picture to be identified does not need to be reviewed.

In an embodiment, the method for generating a picture review probability interval further includes: acquiring a preset picture set in each of the picture sets corresponding to the original review probability interval, each picture being a bad picture, and each picture is The initial probability of a bad picture; the initial probability corresponding to each picture is calculated, and the bad probability that each picture in the preset picture set is a bad picture is generated; The probability interval is reviewed, and the picture corresponding to the bad probability in the original review probability interval is extracted to form a picture set corresponding to the original review probability interval.

In an embodiment, the image review probability interval is regenerated according to the first quantity and the second quantity corresponding to each set of the bad probability, including: obtaining the original upper limit probability and the original lower limit probability of the original review probability interval; according to the original upper limit probability and the original lower limit The probability of calculating the two-category probability, the sum of the second quantity corresponding to the second-class probability, and the sum of the first quantity corresponding to the two-class probability to the original upper-level probability, and the sum of the two, can be obtained by the two-class probability Minimum value; the original upper limit probability and the original lower limit probability are calculated according to the two-class probability, and the picture review probability interval is regenerated.

In an embodiment, the original upper bound probability and the original lower bound probability are calculated according to the second classification probability, and the image review probability interval is regenerated, including: when the second classification probability is less than or equal to the preset probability, according to the formula

Perform calculation to regenerate the upper limit probability s ₂ ', where s ₁ is the original lower bound probability, s ₂ is the original upper bound probability, t is the second classification probability, x ₀ is the preset probability; when the second classification probability is greater than the preset probability, According to the formula

Calculate and regenerate the lower bound probability s ₁ ', where s ₁ is the original lower bound probability, s ₂ is the original upper bound probability, t is the second classification probability, x ₀ is the preset probability; according to the new upper limit probability s ₂ ' or new The lower bound probability s ₁ ', regenerates the picture review probability interval.

In an embodiment, after obtaining the bad probability that each picture in the picture set corresponding to the original review probability interval is a bad picture, the method further includes: acquiring a picture system weight of the picture transmission system corresponding to the picture set; The original upper limit probability and the original lower limit probability are calculated, and the image review probability interval is regenerated, including: calculating the original upper limit probability and the original lower limit probability according to the second classification probability and the picture system weight, and regenerating the picture review probability interval.

In the above embodiment, the initial probability of the picture outputted by the bad picture identification system is uniformly mapped to the bad probability for comparison, and the picture corresponding to the original review probability interval is manually reviewed. After the result, the statistics are determined to be the first number of normal pictures, and the second number determined to be a bad picture, and the original review probability interval is calculated according to the first quantity and the second quantity, and the picture review probability interval is regenerated. . By combining the results of the manual review, the review probability interval is adjusted, so that the newly generated image review probability interval range is more accurate, reducing the workload of image review, thereby improving the efficiency of poor image review.

In an embodiment, as shown in FIG. 6, a picture review probability interval generation device 600 is provided. The device includes: a bad probability acquisition module 602, configured to acquire each picture in the picture set corresponding to the original review probability interval. The bad probability of the bad picture; the picture determination quantity statistics module 604 is configured to, for each group of pictures having the same bad probability, count the first number of the normal picture to be determined as the normal picture, and the second quantity determined as the bad picture by the repeated trial; The picture review probability interval generation module 606 is configured to regenerate the picture review probability interval according to the first quantity and the second quantity corresponding to each set of bad probability.

In an embodiment, as shown in FIG. 7, another picture review probability interval generating apparatus 700 is provided. The apparatus further includes an original review picture set forming module 601, configured to acquire a preset picture set, and each picture is bad. The initial probability of the picture; the initial probability corresponding to each picture is calculated, and the bad probability of each picture in the preset picture set as a bad picture is generated; the original review probability interval is set, and the bad probability corresponding to the original review probability interval is extracted. The picture forms a set of pictures corresponding to the original review probability interval.

In one embodiment, the picture review probability interval generation module 606 is further configured to obtain an original upper limit probability and an original lower limit probability of the original review probability interval; calculate a second classification probability according to the original upper limit probability and the original lower limit probability, and obtain the second classification probability by using the second classification probability The sum of the original lower bound probability to the second quantity corresponding to the second classification probability, the sum of the first quantity corresponding to the second classification probability to the original upper limit probability, and the sum of the sums; the original upper bound probability and the original according to the second classification probability The lower bound probability is calculated and the image review probability interval is regenerated.

In an embodiment, the picture review probability interval generation module 606 is further configured to: when the two classification probability is less than or equal to the preset probability, according to the formula

Perform calculation to regenerate the upper limit probability s ₂ ', where s ₁ is the original lower bound probability, s ₂ is the original upper bound probability, t is the second classification probability, x ₀ is the preset probability; when the second classification probability is greater than the preset probability, According to the formula

Calculate and regenerate the lower bound probability s ₁ ', where s ₁ is the original lower bound probability, s ₂ is the original upper bound probability, t is the second classification probability, x ₀ is the preset probability; according to the new upper limit probability s ₂ ' or new The lower bound probability s ₁ ', regenerates the picture review probability interval.

In an embodiment, the picture review probability interval generation module 606 is further configured to acquire a picture system weight of the picture transmission system corresponding to the picture set; and calculate the original upper limit probability and the original lower limit probability according to the second classification probability and the picture system weight value. , regenerate the image review probability interval. In an embodiment, as shown in FIG. 8, a picture re-determination device 800 is provided. The device includes: a picture defect probability acquisition module 802 to be identified, which is used to obtain a bad probability that a picture to be identified is a bad picture; The determining module 804 is configured to determine whether the bad probability is within a preset image review probability interval, and the method for generating the image review probability interval includes: obtaining a bad probability that each image in the image set corresponding to the original review probability interval is a bad image For each group of pictures with the same bad probability, the statistics are determined to be the first number of normal pictures, and the second number determined to be a bad picture; the first quantity and the second quantity corresponding to each group of bad probabilities The image review probability interval is regenerated; if it is, then the image to be identified needs to be reviewed.

In an embodiment, before acquiring the bad probability that each picture is a bad picture in the picture set corresponding to the original review probability interval, the method includes: acquiring a preset picture set, each picture being an initial probability of a bad picture; The initial probability corresponding to the picture is calculated, and the bad probability of each picture in the preset picture set as a bad picture is generated; the original review probability interval is set, and the picture corresponding to the bad probability in the original review probability interval is extracted, and the original review probability interval is formed. Corresponding picture set.

In an embodiment, the image review probability interval is regenerated according to the first quantity and the second quantity corresponding to each set of the bad probability, including: obtaining the original upper limit probability and the original lower limit probability of the original review probability interval; according to the original upper limit probability and the original lower limit Probability calculates the probability of two classifications, passing the two points The class probability can obtain the sum of the original lower bound probability to the second quantity corresponding to the second classification probability, the sum of the first quantity corresponding to the second classification probability to the original upper limit probability, and the sum of the sums; the original according to the second classification probability The upper limit probability and the original lower limit probability are calculated, and the picture review probability interval is regenerated.

In an embodiment, the original upper bound probability and the original lower bound probability are calculated according to the second classification probability, and the image review probability interval is regenerated, including: when the second classification probability is less than or equal to the preset probability, according to the formula

Perform calculation to regenerate the upper limit probability s ₂ ', where s ₁ is the original lower bound probability, s ₂ is the original upper bound probability, t is the second classification probability, x ₀ is the preset probability; when the second classification probability is greater than the preset probability, According to the formula

Calculate and regenerate the lower bound probability s ₁ ', where s ₁ is the original lower bound probability, s ₂ is the original upper bound probability, t is the second classification probability, x ₀ is the preset probability; according to the new upper limit probability s ₂ ' or new The lower bound probability s ₁ ', regenerates the picture review probability interval.

In an embodiment, after obtaining the bad probability that each picture in the picture set corresponding to the original review probability interval is a bad picture, the method further includes: acquiring a picture system weight of the picture transmission system corresponding to the picture set; The original upper limit probability and the original lower limit probability are calculated, and the image review probability interval is regenerated, including: calculating the original upper limit probability and the original lower limit probability according to the second classification probability and the picture system weight, and regenerating the picture review probability interval.

The above picture review probability interval generating means and picture retrieving means may be implemented in the form of a computer readable instruction which may be run on the first computer device as in FIG.

Each of the modules in the picture review probability interval generation device and the picture review device may be implemented in whole or in part by software, hardware, and combinations thereof. Each of the above modules may be embedded in or independent of the memory of the computer device in hardware, or may be stored in the memory of the computer device in software form, so that the processor invokes the operations corresponding to the above modules. The processor can It is a central processing unit (CPU), a microprocessor, a single chip microcomputer, and the like.

In one embodiment, there is provided one or more non-transitory readable storage mediums storing computer readable instructions that, when executed by one or more processors, cause one or more processors to execute The steps of the picture review probability interval generation method in each of the above embodiments.

In one embodiment, there is provided one or more non-transitory readable storage media having computer readable instructions that, when executed by one or more processors, cause one or more processors to execute The steps of the picture review method in the various embodiments described above.

In one embodiment, a computer apparatus is provided comprising a memory and one or more processors having stored therein computer readable instructions that, when executed by a processor, cause one or more processors to execute The steps of the picture review probability interval generation method in each of the above embodiments.

In one embodiment, a computer apparatus is provided comprising a memory and one or more processors having stored therein computer readable instructions that, when executed by a processor, cause one or more processors to execute The steps of the picture review method in the various embodiments described above.

In one embodiment, the computer device described above may include, but is not limited to, a stand-alone physical server, or a server cluster of multiple physical servers. FIG. 9 is a schematic diagram showing the internal structure of a computer device in an embodiment. The computer device is applicable to the first computer device 102 in the application environment of FIG. The computer device includes a processor coupled through a system bus, a non-volatile storage medium, an internal memory, and a network interface. The processor of the computer device is used to provide computing and control capabilities to support the operation of the entire computer device. A non-volatile storage medium of a computer device stores an operating system, a database, and computer readable instructions. The database stores data related to a picture review probability interval generation method and a picture re-determination method provided by the foregoing various embodiments, for example, an initial probability corresponding to each picture set in a preset picture set may be stored. The computer readable instructions are executable by the processor for implementing a picture review probability interval generation method and a picture retry method provided by the above various embodiments. The internal memory in the computer device provides a cached operating environment for operating systems, databases, and computer readable instructions in a non-volatile storage medium. The network interface can be an Ethernet card or a wireless network card, etc., for external use. The terminal or server communicates, such as communicating with the second computer device 104 in the application environment of FIG. 1, to obtain an initial probability that each picture in the preset picture set is a bad picture.

It will be understood by those skilled in the art that the structure of the computer device shown in FIG. 9 is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation of the computer device to which the solution of the present application is applied. The computer device may include more or fewer components than those in the figures, or some components may be combined, or have different component arrangements. For example, the computer device in the figure may also include a display screen or the like.

One of ordinary skill in the art can understand that all or part of the process of implementing the above embodiments can be completed by computer readable instructions, which can be stored in a non-volatile computer. The readable storage medium, which when executed, may include the flow of an embodiment of the methods as described above. The storage medium may be a magnetic disk, an optical disk, a read-only memory (ROM), or the like.

The technical features of the above-described embodiments may be arbitrarily combined. For the sake of brevity of description, all possible combinations of the technical features in the above embodiments are not described. However, as long as there is no contradiction between the combinations of these technical features, All should be considered as the scope of this manual.

The above-mentioned embodiments are merely illustrative of several embodiments of the present application, and the description thereof is more specific and detailed, but is not to be construed as limiting the scope of the invention. It should be noted that a number of variations and modifications may be made by those skilled in the art without departing from the spirit and scope of the present application. Therefore, the scope of the invention should be determined by the appended claims.

Claims (20)

1. A method for generating a picture review probability interval, comprising:

   Obtaining a bad probability that each picture in the picture set corresponding to the original review probability interval is a bad picture;

   For each group of pictures with the same probability of failure, the statistics are determined to be the first number of normal pictures, and the second number determined to be a bad picture; and

   The picture review probability interval is regenerated according to the first quantity and the second quantity corresponding to each group of bad probabilities.
2. The method according to claim 1, wherein before the obtaining, in the set of pictures corresponding to the original review probability interval, each picture is a bad probability of a bad picture, the method further comprises:

   Obtain the initial probability that each picture is a bad picture in the preset picture set;

   Calculating an initial probability corresponding to each picture, and generating a bad probability that each picture in the preset picture set is a bad picture; and

   The original review probability interval is set, and the picture corresponding to the bad probability in the original review probability interval is extracted, and a picture set corresponding to the original review probability interval is formed.
3. The method according to claim 1, wherein the regenerating the picture review probability interval according to the first quantity and the second quantity corresponding to each set of the bad probability comprises:

   Obtaining an original upper limit probability and an original lower limit probability of the original review probability interval;

   Calculating a binary classification probability according to the original upper bound probability and the original lower bound probability, by which the sum of the original lower bound probability to the second quantity corresponding to the second classification probability may be obtained, corresponding to the second classification probability to the original upper bound probability The sum of the first quantities, the sum of the sum of the two; and

   Calculating the original upper bound probability and the original lower bound probability according to the binary classification probability, and regenerating the picture review probability interval.
4. The method according to claim 3, wherein the calculating the original upper limit probability and the original lower limit probability according to the two classification probability, and regenerating the picture review probability interval, comprising:

   When the two classification probability is less than or equal to the preset probability, according to the formula

   

   Calculate and regenerate the upper limit probability s ₂ ', where s ₁ is the original lower bound probability, s ₂ is the original upper bound probability, t is the second classification probability, and x ₀ is the preset probability;

   When the two classification probability is greater than the preset probability, according to the formula

   

   Calculate and regenerate the lower bound probability s ₁ ', where s ₁ is the original lower bound probability, s ₂ is the original upper bound probability, t is the second classification probability, and x ₀ is the preset probability;

   The picture review probability interval is regenerated according to the new upper limit probability s ₂ ' or the new lower limit probability s ₁ '.
5. The method according to claim 3, further comprising: after the obtaining the bad probability that each picture is a bad picture in the picture set corresponding to the original review probability interval, further comprising:

   Obtaining a picture system weight of the picture transmission system corresponding to the picture set; and

   The calculating the original upper limit probability and the original lower limit probability according to the two classification probability, and regenerating the picture review probability interval, including:

   Calculating the original upper bound probability and the original lower bound probability according to the two classification probability and the picture system weight, and regenerating the picture review probability interval.
6. A picture re-determination method, the method comprising:

   Obtaining a bad probability that the picture to be identified is a bad picture;

   Determining whether the bad probability is within a preset picture review probability interval, and the picture review probability interval is generated according to the picture review probability interval generation method according to any one of claims 1 to 5;

   If yes, determining that the picture to be identified requires a picture review.
7. A picture review probability interval generating device, wherein the device comprises:

   a bad probability acquisition module, configured to obtain a bad probability that each picture in the picture set corresponding to the original review probability interval is a bad picture;

   a picture determination quantity statistics module, configured to determine, for each group of pictures having the same bad probability, a first number of normal pictures to be determined as a normal picture, and a second quantity to be determined as a bad picture by the repeated trial; and

   The picture review probability interval generating module is configured to regenerate the picture review probability interval according to the first quantity and the second quantity corresponding to each set of bad probability.
8. A picture re-determination device, characterized in that the device comprises:

   A bad probability acquisition module for identifying a picture, which is used to obtain a bad probability that the picture to be identified is a bad picture;

   a picture review judging module, configured to determine whether the bad probability is within a preset picture review probability interval, and the picture review probability interval is generated according to the picture review probability interval generation method according to any one of claims 1 to 5; If yes, determining that the picture to be identified requires a picture review.
9. One or more non-transitory readable storage mediums storing computer readable instructions, when executed by one or more processors, cause the one or more processors to perform the following steps:

   Obtaining a bad probability that each picture in the picture set corresponding to the original review probability interval is a bad picture;

   For each group of pictures with the same probability of failure, the statistics are determined to be the first number of normal pictures, and the second number determined to be a bad picture; and

   The picture review probability interval is regenerated according to the first quantity and the second quantity corresponding to each group of bad probabilities.
10. The storage medium according to claim 9, wherein before the step of acquiring, by the processor, the probability that each picture in the picture set corresponding to the original review probability interval is a bad picture, the method further comprises:

    Obtain the initial probability that each picture is a bad picture in the preset picture set;

    Calculating an initial probability corresponding to each picture, and generating a bad probability that each picture in the preset picture set is a bad picture; and

    The original review probability interval is set, and the picture corresponding to the bad probability in the original review probability interval is extracted, and a picture set corresponding to the original review probability interval is formed.
11. A storage medium according to claim 9, wherein said processor executes The steps of regenerating the image review probability interval according to the first quantity and the second quantity corresponding to each group of bad probabilities, including:

    Obtaining an original upper limit probability and an original lower limit probability of the original review probability interval;

    Calculating a binary classification probability according to the original upper bound probability and the original lower bound probability, by which the sum of the original lower bound probability to the second quantity corresponding to the second classification probability may be obtained, corresponding to the second classification probability to the original upper bound probability The sum of the first quantities, the sum of the sum of the two; and

    Calculating the original upper bound probability and the original lower bound probability according to the binary classification probability, and regenerating the picture review probability interval.
12. The storage medium according to claim 11, wherein the step of calculating, by the processor, the original upper limit probability and the original lower limit probability according to the two classification probability, and regenerating the picture review probability interval comprises:

    When the two classification probability is less than or equal to the preset probability, according to the formula

    

    Perform calculation to regenerate the upper limit probability s ₂ ', where s ₁ is the original lower bound probability, s ₂ is the original upper bound probability, t is the second classification probability, and x ₀ is the preset probability;

    When the two classification probability is greater than the preset probability, according to the formula

    

    Calculate and regenerate the lower bound probability s ₁ ', where s ₁ is the original lower bound probability, s ₂ is the original upper bound probability, t is the second classification probability, and x ₀ is the preset probability;

    The picture review probability interval is regenerated according to the new upper limit probability s ₂ ' or the new lower limit probability s ₁ '.
13. The storage medium according to claim 11, wherein after the step of obtaining, by the processor, the probability that each picture in the picture set corresponding to the original review probability interval is a bad picture, the method further includes:

    Obtaining a picture system weight of the picture transmission system corresponding to the picture set; and

    Calculating the original upper bound probability and the original lower bound probability according to the two classification probability, The newly generated image review probability interval includes:

    Calculating the original upper bound probability and the original lower bound probability according to the two classification probability and the picture system weight, and regenerating the picture review probability interval.
14. One or more non-transitory readable storage mediums storing computer readable instructions, when executed by one or more processors, cause the one or more processors to perform the following steps:

    Obtaining a bad probability that the picture to be identified is a bad picture;

    Determining whether the bad probability is within a preset picture review probability interval, and the step of generating the picture review probability interval includes obtaining a bad probability that each picture in the picture set corresponding to the original review probability interval is a bad picture; The group has the same bad probability picture, the statistics are determined to be the first number of normal pictures, and the second number is determined as the bad picture; the first quantity and the second quantity corresponding to each group of bad probability are regenerated. Probability interval; and

    If yes, determining that the picture to be identified requires a picture review.
15. A computer device comprising a memory and one or more processors, the memory storing computer readable instructions, the computer readable instructions being executed by the processor, causing the one or more processors to execute The following steps:

    Obtaining an original upper limit probability and an original lower limit probability of the original review probability interval;

    Calculating a binary classification probability according to the original upper bound probability and the original lower bound probability, by which the sum of the original lower bound probability to the second quantity corresponding to the second classification probability may be obtained, corresponding to the second classification probability to the original upper bound probability The sum of the first quantities, the sum of the sum of the two; and

    Calculating the original upper bound probability and the original lower bound probability according to the binary classification probability, and regenerating the picture review probability interval.
16. The computer device according to claim 15, wherein the step of calculating, by the processor, the original upper limit probability and the original lower limit probability according to the two classification probability, and regenerating the picture review probability interval comprises:

    When the two classification probability is less than or equal to the preset probability, according to the formula

    

    Calculate and regenerate the upper limit probability s ₂ ', where s ₁ is the original lower bound probability, s ₂ is the original upper bound probability, t is the second classification probability, and x ₀ is the preset probability;

    When the two classification probability is greater than the preset probability, according to the formula

    

    Calculate and regenerate the lower bound probability s ₁ ', where s ₁ is the original lower bound probability, s ₂ is the original upper bound probability, t is the second classification probability, and x ₀ is the preset probability;

    The picture review probability interval is regenerated according to the new upper limit probability s ₂ ' or the new lower limit probability s ₁ '.
17. The computer device according to claim 15, wherein after the step of acquiring, by the processor, the probability that each picture in the picture set corresponding to the original review probability interval is a bad picture, the method further comprises:

    Obtaining a picture system weight of the picture transmission system corresponding to the picture set; and

    The calculating the original upper limit probability and the original lower limit probability according to the two classification probability, and regenerating the picture review probability interval, including:

    Calculating the original upper bound probability and the original lower bound probability according to the two classification probability and the picture system weight, and regenerating the picture review probability interval.
18. The computer device according to claim 17, wherein the step of calculating, by the processor, the original upper limit probability and the original lower limit probability according to the two classification probability, and regenerating the picture review probability interval comprises:

    When the two classification probability is less than or equal to the preset probability, according to the formula

    

    Perform calculation to regenerate the upper limit probability s ₂ ', where s ₁ is the original lower bound probability, s ₂ is the original upper bound probability, t is the second classification probability, and x ₀ is the preset probability;

    When the two classification probability is greater than the preset probability, according to the formula

    

    Calculate and regenerate the lower bound probability s ₁ ', where s ₁ is the original lower bound probability, s ₂ is the original upper bound probability, t is the second classification probability, and x ₀ is the preset probability;

    The picture review probability interval is regenerated according to the new upper limit probability s ₂ ' or the new lower limit probability s ₁ '.
19. The computer device according to claim 17, wherein after the step of acquiring, by the processor, the probability that each picture in the picture set corresponding to the original review probability interval is a bad picture, the method further comprises:

    Obtaining a picture system weight of the picture transmission system corresponding to the picture set; and

    The calculating the original upper limit probability and the original lower limit probability according to the two classification probability, and regenerating the picture review probability interval, including:

    Calculating the original upper bound probability and the original lower bound probability according to the two classification probability and the picture system weight, and regenerating the picture review probability interval.
20. A computer device comprising a memory and one or more processors, the memory storing computer readable instructions, the computer readable instructions being executed by the processor, causing the one or more processors to execute The following steps:

    Obtaining a bad probability that the picture to be identified is a bad picture;

    Determining whether the bad probability is within a preset picture review probability interval, and the step of generating the picture review probability interval includes obtaining a bad probability that each picture in the picture set corresponding to the original review probability interval is a bad picture; The group has the same bad probability picture, the statistics are determined to be the first number of normal pictures, and the second number is determined as the bad picture; the first quantity and the second quantity corresponding to each group of bad probability are regenerated. Probability interval; and

    If yes, determining that the picture to be identified requires a picture review.

Images