CN110705402A - Face recognition confidence value mapping algorithm - Google Patents
Face recognition confidence value mapping algorithm Download PDFInfo
- Publication number
- CN110705402A CN110705402A CN201910887638.4A CN201910887638A CN110705402A CN 110705402 A CN110705402 A CN 110705402A CN 201910887638 A CN201910887638 A CN 201910887638A CN 110705402 A CN110705402 A CN 110705402A
- Authority
- CN
- China
- Prior art keywords
- score
- recognition rate
- face recognition
- recognition
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000013507 mapping Methods 0.000 title claims abstract description 32
- 238000012360 testing method Methods 0.000 claims abstract description 12
- 230000011218 segmentation Effects 0.000 claims abstract description 8
- 238000000034 method Methods 0.000 claims description 3
- 230000008569 process Effects 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 4
- 230000009286 beneficial effect Effects 0.000 abstract 1
- 230000003631 expected effect Effects 0.000 abstract 1
- 241000282414 Homo sapiens Species 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/172—Classification, e.g. identification
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a face recognition confidence value mapping algorithm, which comprises the following steps: evaluating the test sample by the face recognition model to obtain recognition rate and error recognition rate data; performing data fitting on the recognition rate and the false recognition rate; selecting data segmentation mapping; and mapping the output value according to the fitted formula. The invention has the beneficial effects that: after the product is replaced by different algorithms, external parameters are not required to be modified, and only the recognition effect of the product is required to be mapped by using the algorithm, so that the expected effect of the algorithm can be achieved, and the accuracy of face recognition is improved.
Description
Technical Field
The invention belongs to the field of face recognition, and particularly relates to a face recognition confidence value mapping algorithm.
Background
With the development of deep learning technology, the recognition rate of the face recognition algorithm exceeds that of human beings in many occasions. Face recognition algorithms are widely used in various products. Due to different algorithm manufacturers or due to model iteration, default confidence values input by the recognition algorithm are changed continuously, so that an upper-layer system can be debugged continuously to meet different algorithms to achieve the best effect. This leads to a cumbersome adaptation and high professional requirements.
Disclosure of Invention
In view of the above, the present invention is directed to a face recognition confidence value mapping algorithm to solve the above-mentioned problems.
In order to achieve the purpose, the technical scheme of the invention is realized as follows:
the face recognition confidence value mapping algorithm comprises the following steps:
A. evaluating the test sample by the face recognition model to obtain recognition rate and error recognition rate data;
B. b, performing data fitting on the recognition rate and the false recognition rate obtained in the step A;
C. selecting data segmentation mapping;
D. and mapping the output value according to the fitted formula.
Further, in the step a, the face recognition model is used to evaluate the recognition rate and the false recognition rate of the test sample, so as to obtain the recognition rate and the false recognition rate corresponding to each score in the test sample, wherein the score represents the probability of recognizing one sample.
Further, in the step B, the data obtained in the step A are respectively fitted by taking the score as a horizontal coordinate and the recognition rate and the false recognition rate as a vertical coordinate to obtain a data curve.
Further, the step C process is as follows:
C1. taking the minimum score point with the false recognition rate of 0 in the step B as a reference segmentation point, and recording the corresponding score as a;
C2. mapping the score of the interval a-b to the interval b-1.0, and considering the score which is larger than b before mapping as 1.0, wherein b is the minimum score when the identification accuracy tends to be stable;
C3. carrying out polynomial fitting on data of the false recognition rate corresponding to the interval a-b values;
C4. and equally dividing the score of the interval b-1.0, and performing polynomial fitting on the false recognition rate.
Further, in the step D, when the output score of the face recognition model is within the interval a-b, the output score is substituted into the polynomial obtained in the step C3, then the false recognition rate parameters corresponding to the polynomial obtained by fitting in the steps C3 and C4 are equal to each other, a set of solution x of the score parameters is obtained, the score solution larger than 1.0 is removed, and the other score solution is smaller than 1.0 and is determined according to MAX (x, b).
Compared with the prior art, the face recognition confidence value mapping algorithm has the following advantages:
the face recognition confidence value mapping algorithm does not need to modify external parameters after the product is replaced by different algorithms, and only needs to map the recognition effect of the product by using the algorithm, so that the output of the algorithm is mapped to the score in the score interval with high recognition rate, the accuracy of the face recognition model is improved, and the optimal effect of the model is achieved.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate an embodiment of the invention and, together with the description, serve to explain the invention and not to limit the invention. In the drawings:
FIG. 1 is a flowchart illustrating the steps of a face recognition confidence value mapping algorithm according to an embodiment of the present invention;
FIG. 2 is a recognition rate polynomial fit curve in an embodiment of the present invention;
FIG. 3 is a fitting curve of a polynomial of the false positive rate according to an embodiment of the present invention;
FIG. 4 is a 0.6-0.8 segmented data fitting curve according to an embodiment of the present invention;
FIG. 5 is a curve fitted to 0.8-1.0 segmented data in an embodiment of the present invention.
Detailed Description
It should be noted that the embodiments and features of the embodiments may be combined with each other without conflict.
The present invention will be described in detail below with reference to the embodiments with reference to the attached drawings.
As shown in fig. 1, the face recognition confidence value mapping algorithm includes the following steps:
A. evaluating the test sample by the face recognition model to obtain recognition rate and error recognition rate data;
B. b, performing data fitting on the recognition rate and the false recognition rate obtained in the step A;
C. selecting data segmentation mapping;
D. and mapping the output value according to the fitted formula.
And in the step A, the face recognition model is used for evaluating the recognition rate and the false recognition rate of the test sample to obtain the recognition rate and the false recognition rate corresponding to each score in the test sample, wherein the score represents the probability of recognizing one sample.
And in the step B, fitting the data obtained in the step A respectively by taking the score as a horizontal coordinate and the recognition rate and the error recognition rate as a vertical coordinate to obtain a data curve.
The step C process is as follows:
C1. taking the minimum score point with the false recognition rate of 0 in the step B as a reference segmentation point, and recording the corresponding score as a;
C2. mapping the scores of the intervals a-b to the intervals b-1.0, and considering the scores which are larger than b before mapping as 1.0, because the recognition rates are basically consistent when the scores are larger than 0.8, the score mapping has no meaning, wherein b is the minimum score when the recognition accuracy tends to be stable; the recognition accuracy is higher when the score is higher and lower, and the recognition accuracy is higher when the score is lower, but the recognition accuracy is lower.
C3. Carrying out polynomial fitting on data of the false recognition rate corresponding to the interval a-b values;
C4. and equally dividing the score of the interval b-1.0, and performing polynomial fitting on the false recognition rate.
And D, when the output score of the face recognition model is in an interval a-b, substituting the output score into the polynomial obtained in the step C3, and then enabling the corresponding false recognition rate parameters in the polynomial obtained by fitting in the step C3 and the step C4 to be equal to each other, obtaining a solution x of a group of score parameters, removing the score solution larger than 1.0, and determining the solution x according to MAX (x, b) if the other score solution is smaller than 1.0.
In this embodiment, a group of face recognition models is given to evaluate a test sample to obtain recognition rate and error recognition rate data, and with 0.01 score as an interval, see table 1:
TABLE 1 identification and error Rate data enumeration
As shown in fig. 2 and 3, the data are respectively subjected to polynomial fitting by using the score as an abscissa and the recognition rate and the false recognition rate as an ordinate to obtain a data curve, wherein the fitting polynomial of the recognition rate is as follows:
y=-159.44x6+470.04x5-503.86x4+238.05x3-49.631x2+3.875x+0.9258
in fig. 3, a point with a false recognition rate of 0 is found as a reference segmentation point, and the test model in this embodiment is selected to be 0.6;
as shown in FIG. 2, when the score of the model is greater than 0.8, the recognition rate is basically consistent, and the score mapping has no significance, so that the scores in the interval of 0.6-0.8 are mapped to the interval of 0.8-1.0, wherein the scores greater than 0.8 before mapping are all regarded as 1.0;
as shown in fig. 4, the score data of 0.6-0.8 is taken to perform polynomial fitting, and the fitting polynomial is obtained as:
y=14.806x2-24.047x+9.7656;
as shown in fig. 5, the interval 0.8-1.0 is divided equally, and then polynomial fitting is performed to obtain a fitting polynomial as:
y=16.337x12-32.735x1+16.4;
when the model output score is 0.6-0.8, here 0.6 for example, it can be obtained from the two fitting curves:
16.337*x12-32.735*x1+16.4=14.806*0.62-24.047*0.6+9.7656,
by solving the equations, x is 0.75131 and 1.252424, and since the range of data to be mapped is between 0.8-1.0, values greater than 1.0 are removed and a smaller solution 0.7513 is selected. Since 0.75131 does not completely meet the expected 0.8 requirement, the value can be taken to be a proper value according to MAX (x,0.8), which also avoids the complex operation of too complex fitting function.
Finally, the algorithm is used for outputting scores in the score interval mapped to high recognition rate, and the accuracy of the face recognition model is improved.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.
Claims (5)
1. The face recognition confidence value mapping algorithm is characterized by comprising the following steps of:
A. evaluating the test sample by the face recognition model to obtain recognition rate and error recognition rate data;
B. b, performing data fitting on the recognition rate and the false recognition rate obtained in the step A;
C. selecting data segmentation mapping;
D. and mapping the output value according to the fitted formula.
2. The face recognition confidence value mapping algorithm of claim 1, wherein: and in the step A, the face recognition model is used for evaluating the recognition rate and the false recognition rate of the test sample to obtain the recognition rate and the false recognition rate corresponding to each score in the test sample, wherein the score represents the probability of recognizing one sample.
3. The face recognition confidence value mapping algorithm of claim 2, wherein: and in the step B, fitting the data obtained in the step A respectively by taking the score as a horizontal coordinate and the recognition rate and the error recognition rate as a vertical coordinate to obtain a data curve.
4. The face recognition confidence value mapping algorithm of claim 3, wherein the step C process is as follows:
C1. taking the minimum score point with the false recognition rate of 0 in the step B as a reference segmentation point, and recording the corresponding score as a;
C2. mapping the score of the interval a-b to the interval b-1.0, and considering the score which is larger than b before mapping as 1.0, wherein b is the minimum score when the identification accuracy tends to be stable;
C3. carrying out polynomial fitting on data of the false recognition rate corresponding to the interval a-b values;
C4. and equally dividing the score of the interval b-1.0, and performing polynomial fitting on the false recognition rate.
5. The face recognition confidence value mapping algorithm of claim 4, wherein: and D, when the output score of the face recognition model is in an interval a-b, substituting the output score into the polynomial obtained in the step C3, and then enabling the corresponding false recognition rate parameters in the polynomial obtained by fitting in the step C3 and the step C4 to be equal to each other, obtaining a solution x of a group of score parameters, removing the score solution larger than 1.0, and determining the solution x according to MAX (x, b) if the other score solution is smaller than 1.0.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910887638.4A CN110705402A (en) | 2019-09-19 | 2019-09-19 | Face recognition confidence value mapping algorithm |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910887638.4A CN110705402A (en) | 2019-09-19 | 2019-09-19 | Face recognition confidence value mapping algorithm |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110705402A true CN110705402A (en) | 2020-01-17 |
Family
ID=69195810
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910887638.4A Pending CN110705402A (en) | 2019-09-19 | 2019-09-19 | Face recognition confidence value mapping algorithm |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110705402A (en) |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106096547A (en) * | 2016-06-11 | 2016-11-09 | 北京工业大学 | A kind of towards the low-resolution face image feature super resolution ratio reconstruction method identified |
CN108648239A (en) * | 2018-05-04 | 2018-10-12 | 苏州富强科技有限公司 | The scaling method of phase height mapping system based on segmented fitting of a polynomial |
CN108921123A (en) * | 2018-07-17 | 2018-11-30 | 重庆科技学院 | A kind of face identification method based on double data enhancing |
CN109494716A (en) * | 2018-11-15 | 2019-03-19 | 沈阳工业大学 | Wind power output power confidence interval prediction technique based on Bootstrap |
US10346693B1 (en) * | 2019-01-22 | 2019-07-09 | StradVision, Inc. | Method and device for attention-based lane detection without post-processing by using lane mask and testing method and testing device using the same |
CN110083517A (en) * | 2019-04-29 | 2019-08-02 | 秒针信息技术有限公司 | A kind of optimization method and device of user's portrait confidence level |
-
2019
- 2019-09-19 CN CN201910887638.4A patent/CN110705402A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106096547A (en) * | 2016-06-11 | 2016-11-09 | 北京工业大学 | A kind of towards the low-resolution face image feature super resolution ratio reconstruction method identified |
CN108648239A (en) * | 2018-05-04 | 2018-10-12 | 苏州富强科技有限公司 | The scaling method of phase height mapping system based on segmented fitting of a polynomial |
CN108921123A (en) * | 2018-07-17 | 2018-11-30 | 重庆科技学院 | A kind of face identification method based on double data enhancing |
CN109494716A (en) * | 2018-11-15 | 2019-03-19 | 沈阳工业大学 | Wind power output power confidence interval prediction technique based on Bootstrap |
US10346693B1 (en) * | 2019-01-22 | 2019-07-09 | StradVision, Inc. | Method and device for attention-based lane detection without post-processing by using lane mask and testing method and testing device using the same |
CN110083517A (en) * | 2019-04-29 | 2019-08-02 | 秒针信息技术有限公司 | A kind of optimization method and device of user's portrait confidence level |
Non-Patent Citations (3)
Title |
---|
吴蔚杰: "图像质量与人脸识别关系量化模型的研究与实现", 《中国优秀硕士学位论文全文数据库 (信息科技辑)》 * |
周丽芳等: "一种基于分治策略的Huffman-LBP多姿态人脸识别", 《小型微型计算机系统》 * |
陈晓方等: "基于主元导数特征聚类的加氢裂化动态调整区间识别", 《清华大学学报(自然科学版)》 * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2018028546A1 (en) | Key point positioning method, terminal, and computer storage medium | |
CN106683680B (en) | Speaker recognition method and device, computer equipment and computer readable medium | |
WO2018108129A1 (en) | Method and apparatus for use in identifying object type, and electronic device | |
TWI752455B (en) | Image classification model training method, image processing method, data classification model training method, data processing method, computer device, and storage medium | |
WO2021120834A1 (en) | Biometrics-based gesture recognition method and apparatus, computer device, and medium | |
CN107423773B (en) | Automatic registration method and device for three-dimensional skull | |
CN109101481A (en) | A kind of name entity recognition method, device and electronic equipment | |
CN110147732A (en) | Refer to vein identification method, device, computer equipment and storage medium | |
CN107945791B (en) | Voice recognition method based on deep learning target detection | |
CN111832462B (en) | Frequency hopping signal detection and parameter estimation method based on deep neural network | |
WO2020082734A1 (en) | Text emotion recognition method and apparatus, electronic device, and computer non-volatile readable storage medium | |
JP2018500645A (en) | System and method for tracking objects | |
CN109034095A (en) | A kind of face alignment detection method, apparatus and storage medium | |
CN109656366B (en) | Emotional state identification method and device, computer equipment and storage medium | |
WO2022027913A1 (en) | Target detection model generating method and apparatus, device and storage medium | |
CN111274955A (en) | Emotion recognition method and system based on audio-visual feature correlation fusion | |
CN113539501A (en) | Breathing machine man-machine asynchronous classification method, system, terminal and storage medium | |
CN112131322B (en) | Time sequence classification method and device | |
CN114343577B (en) | Cognitive function evaluation method, terminal device, and computer-readable storage medium | |
CN107729947A (en) | A kind of Face datection model training method, device and medium | |
CN110705402A (en) | Face recognition confidence value mapping algorithm | |
CN108009157B (en) | Statement classification method and device | |
CN111126566A (en) | Abnormal furniture layout data detection method based on GAN model | |
CN109976555A (en) | A kind of frequency calibrating method and device | |
CN114996466A (en) | Method and system for establishing medical standard mapping model and using method |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200117 |
|
RJ01 | Rejection of invention patent application after publication |