CN108073914A - A kind of animal face key point mask method - Google Patents
A kind of animal face key point mask method Download PDFInfo
- Publication number
- CN108073914A CN108073914A CN201810023304.8A CN201810023304A CN108073914A CN 108073914 A CN108073914 A CN 108073914A CN 201810023304 A CN201810023304 A CN 201810023304A CN 108073914 A CN108073914 A CN 108073914A
- Authority
- CN
- China
- Prior art keywords
- key point
- sample image
- prediction result
- animal face
- mark
- 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.)
- Granted
Links
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
- G06V40/161—Detection; Localisation; Normalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
- G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
-
- 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
- G06V40/168—Feature extraction; Face representation
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30196—Human being; Person
- G06T2207/30201—Face
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- General Health & Medical Sciences (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
The present invention discloses a kind of animal face key point mask method, including：Animal face in the sample image of input is detected, obtains animal surface frame and animal face key point position；The sample image of the input is handled according to the animal surface frame and animal face key point position, by treated, sample image is sent into crucial point prediction network trained in advance, obtains the first key point prediction result；The first key point prediction result is adjusted, obtains the mark point of animal face key point；Quality testing is carried out to the mark point according to pre-defined standard picture, obtains the quality score of the mark point.Technical solution provided by the invention can accurately mark animal face key point, and mark point can be audited automatically, so as to greatly improve work efficiency while improving and marking precision.
Description
Technical field
The present invention relates to technical field of image processing more particularly to a kind of animal face key point mask methods.
Background technology
In recent years, self-timer makeups receive more and more attention, and also demonstrate the brilliance to the demand of cute pet makeups.With face
Makeups depend on being accurately positioned equally for facial key point, and cute pet makeups also have animal face key point very strong dependence.
At present, the algorithm on animal face key point location is all fewer in science circle and industrial quarters, and reason may be compared to people
For face key point, the mark sample of animal face key point is fewer, lacks disclosed evaluation and test database.And to animal face
The research of portion's key point location algorithm is largely dependent upon the mark to animal face key point, by animal face
Key point is labeled, and forms a training sample of animal face key point location algorithm after treatment.As it can be seen that key point
Mark can directly affect the precision of key point location algorithm.
Face key point labeling system is continued to use for the mark of animal face key point at present, which includes following several
A module：Face detection module, key point prediction module and key point labeling module, wherein, key point prediction module and key
Point labeling module is two completely self-contained modules, which is simply simply led into samples pictures, for the mark of key point
It fully relies on engineer to be operated manually by rule of thumb, therefore, the facial key point mark carried out for samples pictures will not be very
Accurately, so as to generate many invalid mark samples, and these invalid mark samples are to promoting key point location algorithm
Performance has no to help.In addition, manually also needing to carry out manual examination and verification to all mark samples after the completion of mark, this is also one non-
The process often taken time and effort is unfavorable for the raising of work efficiency.
The content of the invention
The present invention is intended to provide a kind of animal face key point mask method, can carry out animal face key point accurate
Ground marks, and mark point can be audited automatically, so as to greatly improve work efficiency while improving and marking precision.
In order to achieve the above objectives, the technical solution adopted by the present invention is as follows：
A kind of animal face key point mask method, including：
Animal face in the sample image of input is detected, obtains animal surface frame and animal face key point
It puts；The sample image of the input is handled according to the animal surface frame and animal face key point position, after processing
Sample image be sent into advance training crucial point prediction network, obtain the first key point prediction result；It is crucial to described first
Point prediction result is adjusted, and obtains the mark point of animal face key point；According to pre-defined standard picture to the mark
Note point carries out quality testing, obtains the quality score of the mark point.
Preferably, the standard picture that the basis pre-defines carries out quality testing to the mark point, obtains the mark
Noting the method for the quality score of point includes：Obtain the crucial point coordinates of the pre-defined standard picture；Calculate the mark
The similitude transformation matrix of the coordinate of point and the crucial point coordinates of the standard picture；According to the similitude transformation matrix to described defeated
The sample image entered does affine transformation, obtains normalized image；Extract the characteristics of image of the normalized image；By described image
Feature is sent into grader trained in advance, obtains the quality score of the mark point.
Preferably, described image feature includes：HOG features and/or depth characteristic；The selection mode bag of the grader
It includes：Decision tree or logistic regression or depth network.
Further, before the animal face in the sample image of the input at described pair is detected, further include：To being intended to mark
The importance of the sample image of note is assessed.
Preferably, the method that the importance of the described pair of sample image to be marked is assessed includes：Obtain described to be marked
The key point prediction result of the sample image of note is the second key point prediction result；The sample image to be marked is carried out
Left and right overturning obtains overturning sample image；The key point prediction result of the overturning sample image is obtained, is that the 3rd key point is pre-
Survey result；The coordinate of the 3rd key point prediction result is subjected to left and right overturning, obtains the 4th key point prediction result；It calculates
Error between the 4th key point prediction result and the second key point prediction result；By the error and default the
Two threshold values are compared, and obtain the importance of the sample image to be marked.
Preferably, the mistake calculated between the 4th key point prediction result and the second key point prediction result
Difference method be：
Wherein, error be the error, S1For the second key point prediction result, S3For the described 4th crucial point prediction
As a result, (x10,y10), (x11,y11) for the coordinate of two of which key point in the second key point prediction result.
Further, further include：Supplement the sample image of particular category.
Preferably, the method for the sample image of the supplement particular category includes：Count the category of the sample image marked
Property distribution, lacked sample image is known according to the property distribution；According to default keyword, described lack is downloaded from network
Sample image, and duplicate removal processing is carried out to the scarce sample image of institute of download.
Preferably, the attribute of the sample image marked includes：It closes one's eyes, opens eyes, shut up, open one's mouth, positive face, side face.
Preferably, the method for institute's scarce sample image progress duplicate removal processing of described pair of download is：It is gone based on pixel similarity
Again or based on semantic Hash duplicate removal.
Animal face key point mask method provided in an embodiment of the present invention carries out face by the sample image to input
Detection obtains the key point prediction result of input sample image, is adjusted on the basis of the key point prediction result, you can
The mark point of animal face key point is obtained, manual operations is carried out by rule of thumb completely without engineer, so as to greatly drop
The low difficulty of artificial mark, and improve the precision of mark point.Meanwhile quality testing is carried out to mark point, it can mark
Operator is reminded in real time in the process, so as to improve mark quality and work efficiency.In addition, technical solution provided by the invention is also
The importance for the sample image to be marked is assessed, and then the importance for the sample image to be marked is ranked up, very
The possibility of mark invalid sample is reduced in big degree so that the sample image of input is effective sample, is closed so as to improve
The performance of key point location Algorithm for Training model；By the sample image for supplementing particular category so that the property distribution of sample image
It meets the requirements, further improves the performance of key point location algorithm training pattern.
Description of the drawings
Fig. 1 is the method flow diagram of the embodiment of the present invention.
Specific embodiment
In order to make the purpose , technical scheme and advantage of the present invention be clearer, below in conjunction with attached drawing, to the present invention into
Row is further described.
Step 101, the animal face in the sample image of input is detected, obtains animal surface frame and animal face closes
Key point position；
Step 102, the sample image of the input is carried out according to the animal surface frame and animal face key point position
Processing, the processing include the sample image of input being cut out, scaled and rotate etc. conversion, and treated that sample image is sent by general
Enter the crucial point prediction network of training in advance, obtain the first key point prediction result；
In the present embodiment, special face detection and key point prediction module are set.Above-mentioned " animal face key point
Put " when referring to the do face detection while position of 5 key points of prediction：Left eye center, right eye center, nose, the left corners of the mouth are right
The corners of the mouth；Above-mentioned " key point prediction result " refers to the key point predicted with existing key point location algorithm, is typically more than above-mentioned 5
A point.
Step 103, the first key point prediction result is adjusted, obtains the mark point of animal face key point；
In the present embodiment, special key point is set to adjust module, which is the major part manually marked, basic
Function is exactly mark, and person can be provided by dragging the actions such as mouse to adjust above-mentioned face detection and key point prediction module
First key point prediction result, makes it more accurate.In addition, the module further includes picture zoom function, specified region can be scaled,
More accurately to mark key point.
Step 104, quality testing is carried out to the mark point according to pre-defined standard picture, obtains the mark point
Quality score, in the specific implementation, special Automatic quality inspection module is set, which detects the matter of mark sample in real time
Amount is fed back in time to mark person.Specific method is as follows：
Obtain the crucial point coordinates of the pre-defined standard picture；Calculate the coordinate of the mark point and the standard
The similitude transformation matrix of the crucial point coordinates of image；The sample image of the input is done according to the similitude transformation matrix affine
Conversion, obtains normalized image；Extract the characteristics of image of the normalized image；Described image feature is sent into training in advance
Grader obtains the quality score of the mark point；When the quality score is less than default first threshold, return " serious
Mistake " alerts, while the person that reminds mark pays attention to changing annotation results.
Above-mentioned characteristics of image includes：HOG features and/or depth characteristic；The selection mode of above-mentioned grader includes：Decision-making
Tree or logistic regression or depth network.Grader needs training in advance, and training sample includes the positive sample that correctly marks and right
The negative sample that the annotation results random perturbation of positive sample generates, the floating number of the output result of grader between one (0,1),
Indicate the putting property degree that mark sample is positive sample.
In practical operation, face detection and key point prediction module pass data to key point adjustment module.Key point
Module is adjusted by the data transfer marked to Automatic quality inspection module, the matter of Automatic quality inspection module estimation data mark
Then assessment result is fed back to key point adjustment module by amount.
In the present embodiment, before the animal face in the sample image to input is detected, further include：To being intended to mark
The importance of sample image assessed." importance for the sample to be marked " refers to add the sample image to training pattern
The influence of performance if being with the addition of the sample image, can improve the accuracy of model prediction key point, then it is assumed that the sample is weight
It wants, otherwise, then it is assumed that the sample image is less important.In the specific implementation, special mark sample importance is set to comment
Estimate module, the importance of the module estimation sample image to be marked sorts to the importance for the sample image to be marked.Tool
Body method is as follows：
(1) the key point prediction result of the sample image to be marked is obtained, is the second key point prediction result S1；The
Two key point prediction result S1Acquisition performed by the face detection in the present embodiment and key point prediction module, method and the
The acquisition methods of one key point prediction result are identical；
(2) sample image to be marked is subjected to left and right overturning, obtains overturning sample image；Obtain the overturning sample
The key point prediction result of this image is the 3rd key point prediction result S2；3rd key point prediction result S2Acquisition by this reality
The face detection in example and key point prediction module are applied to perform, the acquisition methods phase of method and the first key point prediction result
Together；
(3) by the 3rd key point prediction result S2Coordinate carry out left and right overturning, obtain the 4th crucial point prediction knot
Fruit S3；
For convenience of description, it is assumed herein that the number of the key point of prediction is 5, it is respectively：Left eye center, right eye center,
Nose, the left corners of the mouth, the right corners of the mouth.By the 3rd key point prediction result S2It is denoted as (x20,y20,x21,y21,x22,y22,x23,y23,x24,
y24), by the 4th key point prediction result S3It is denoted as (x30,y30,x31,y31,x32,y32,x33,y33,x34,y34), the width of picture is
W, then the 4th key point prediction result S3It can be obtained with equation below：
(x30,y30)=(w-x21,y21)
(x31,y31)=(w-x20,y20)
(x32,y32)=(x22,y22)
(x33,y33)=(w-x24,y24)
(x34,y34)=(w-x23,y23)
(4) the 4th key point prediction result S is calculated3With the second key point prediction result S1Between error
Error, formula are as follows：
Wherein, error be the error, S1For the second key point prediction result, S3For the described 4th crucial point prediction
As a result, (x10,y10), (x11,y11) it is the second key point prediction result S1The coordinate of middle two of which key point, this implementation
It is S in example1In first, second key point coordinate.
(5) by the error e rror compared with default second threshold, the sample image to be marked is obtained
Importance.When error is more than default second threshold, it is believed that the sample image is effective sample, and the error is bigger, this sample
The importance of image is higher.
Practice have shown that in key point location algorithm, there are problems that so-called mirror image prejudice, i.e., for same pre- measuring and calculating
Method, the prediction result Y1 of input figure I1 are in the main true, and the prediction result Y2 of the image I2 after mirror image switch then may be completely wrong
By mistake.And in general, when Y2 is correct, Y1 is also correct, and when Y2 is completely wrong, the error of Y1 is also very big；In addition, if by Y2
It does mirror image switch and obtains Y3, then the error of Y1 and Y3 can then be used for weighing performance of the algorithm on image I1, if by mistake
Difference is smaller, illustrates that prediction result is pretty good, image I1 is simple sample, conversely, image I1 is difficult sample, adds the sample and helps
In lift scheme performance.Therefore, the importance of judgement sample image can by the value of calculation error error, be carried out.Certainly, may be used also
, to calculate this error, for example to be normalized using other manner with the coordinate of the 4th key point prediction result S3, with animal face
The length or width of frame normalize, and can select different computational methods according to actual needs.
Supplement particular category sample module is also specially provided in the present embodiment, can easily supplement spy as needed
Determine the sample of classification.Due to, there are the unbalanced problem of sample distribution, lacking some specific in the training sample of early application
The sample of classification, for example, the sample closed one's eyes, the sample turned one's head, header planes rotate very big sample, sample that mouth magnifies etc. outside
Deng the shortages of, these samples, to directly result in trained model prediction effect on these classification pictures bad, so needing to mend
The sample of particular category is filled, specific method includes：
(1) property distribution for the sample image that statistics has marked, knows lacked sample image according to the property distribution；It should
Attribute includes：It closes one's eyes, eye opening shuts up, opens one's mouth, positive face, the features such as side face.
(2) according to default keyword, the lacked sample image is downloaded from network.The main body of this module is a net
Network reptile inputs special key words, you can downloads network picture to locally.
(3) duplicate removal processing is carried out to the scarce sample image of institute of download.Since the picture multiplicity that network crawls is higher, because
This needs the module of a duplicate removal.Here can be selected there are many method, for example, the simplest duplicate removal based on pixel similarity；
More complicated duplicate removal based on semantic Hash etc..
The attribute of sample has been marked herein, can be simply calculated by crucial point coordinates, for example, being closed by eyes
Key point coordinates can easily be closed one's eyes or eye opening attribute；It can easily be obtained by the crucial point coordinates of face part
To shutting up or attribute of opening one's mouth；By eyes, nose, lower jaw part crucial point coordinates, head drift angle can be calculated, obtained just
Face or side face attribute.Specific computational methods are exemplified below：
The crucial point coordinates that can be gone out according to existing model prediction seeks the ratio of width to height of eyes, is then believed that more than some threshold value
It is to close one's eyes.For example, left eye includes 6 key point { x0,y0,x1,y1,x2,y2,x3,y3,x4,y4,x5,y5, then can be approximate
Width to left eye is：W=max (x0,x1,x2,x3,x4,x5)-min(x0,x1,x2,x3,x4,x5), approximate altitude is：H=max
(y0,y1,y2,y3,y4,y5)-min(y0,y1,y2,y3,y4,y5), the ratio of width to height is：Ratio=w/h.Similarly, it can calculate and open one's mouth to belong to
Property.
The computational methods of head drift angle are as follows：6 key points of the eyes of prediction, nose, lower jaw part are mapped to and dote on
The correspondence key point of object plane ministerial standard threedimensional model, acquiring rotating vector and translation vector (can use the solvePnP letters of opencv
Number is realized)；Then it is (available tri- yaw angle yaw, roll angle roll, pitch angle pitch angles to be obtained by the two vectors
The decomposeProjectionMatrix functions of opencv are realized).
Supplement particular category sample module needs to call face detection and key point prediction module, is required supplementation in selection
It is called during sample.
Animal face key point mask method provided in an embodiment of the present invention carries out face by the sample image to input
Detection obtains the key point prediction result of input sample image, is adjusted on the basis of the key point prediction result, you can
The mark point of animal face key point is obtained, manual operations is carried out by rule of thumb completely without engineer, so as to greatly drop
The low difficulty of artificial mark, and improve the precision of mark point.Meanwhile quality testing is carried out to mark point, it can mark
Operator is reminded in real time in the process, so as to improve mark quality and work efficiency.In addition, technical solution provided by the invention is also
The importance for the sample image to be marked is assessed, and then the importance for the sample image to be marked is ranked up, very
The possibility of mark invalid sample is reduced in big degree so that the sample image of input is effective sample, is closed so as to improve
The performance of key point location Algorithm for Training model；By the sample image for supplementing particular category so that the property distribution of sample image
It meets the requirements, further improves the performance of key point location algorithm training pattern.As it can be seen that animal face provided by the present invention closes
Key point mask method can fast and efficiently mark effective sample, and then promote the essence of animal face key point location algorithm
Degree.Animal face key point anchor point algorithm research, it can also be used to animal face Expression Recognition, pain identification etc..
The above description is merely a specific embodiment, but protection scope of the present invention is not limited thereto, any
Those familiar with the art in the technical scope disclosed by the present invention, can readily occur in change or replacement, should all contain
Lid is within protection scope of the present invention.
Claims (10)
1. a kind of animal face key point mask method, which is characterized in that including：
Animal face in the sample image of input is detected, obtains animal surface frame and animal face key point position；
The sample image of the input is handled according to the animal surface frame and animal face key point position, after processing
Sample image be sent into advance training crucial point prediction network, obtain the first key point prediction result；
The first key point prediction result is adjusted, obtains the mark point of animal face key point；
Quality testing is carried out to the mark point according to pre-defined standard picture, obtains the quality score of the mark point.
2. animal face key point mask method according to claim 1, which is characterized in that the basis pre-defined
Standard picture carries out quality testing to the mark point, obtains the method for the quality score of the mark point and includes：
Obtain the crucial point coordinates of the pre-defined standard picture；
Calculate the similitude transformation matrix of the coordinate and the crucial point coordinates of the standard picture of the mark point；
Affine transformation is done to the sample image of the input according to the similitude transformation matrix, obtains normalized image；
Extract the characteristics of image of the normalized image；
Described image feature is sent into grader trained in advance, obtains the quality score of the mark point.
3. animal face key point mask method according to claim 2, which is characterized in that described image feature includes：
HOG features and/or depth characteristic；The selection mode of the grader includes：Decision tree or logistic regression or depth network.
4. animal face key point mask method according to claim 1, which is characterized in that the sample of the input at described pair
Before animal face in image is detected, further include：The importance for the sample image to be marked is assessed.
5. animal face key point mask method according to claim 4, which is characterized in that the described pair of sample to be marked
The method that the importance of image is assessed includes：
The key point prediction result of the sample image to be marked is obtained, is the second key point prediction result；
The sample image to be marked is subjected to left and right overturning, obtains overturning sample image；Obtain the overturning sample image
Key point prediction result, be the 3rd key point prediction result；
The coordinate of the 3rd key point prediction result is subjected to left and right overturning, obtains the 4th key point prediction result；
Calculate the error between the 4th key point prediction result and the second key point prediction result；
By the error compared with default second threshold, the importance of the sample image to be marked is obtained.
6. animal face key point mask method according to claim 5, which is characterized in that described to calculate the 4th pass
The method of error between key point prediction result and the second key point prediction result is：
<mrow>
<mi>e</mi>
<mi>r</mi>
<mi>r</mi>
<mi>o</mi>
<mi>r</mi>
<mo>=</mo>
<mfrac>
<mrow>
<mo>|</mo>
<mo>|</mo>
<msub>
<mi>S</mi>
<mn>1</mn>
</msub>
<mo>-</mo>
<msub>
<mi>S</mi>
<mn>3</mn>
</msub>
<mo>|</mo>
<msubsup>
<mo>|</mo>
<mn>2</mn>
<mn>2</mn>
</msubsup>
</mrow>
<mrow>
<mn>5</mn>
<mo>*</mo>
<msqrt>
<mrow>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<mn>10</mn>
</msub>
<mo>-</mo>
<msub>
<mi>x</mi>
<mn>11</mn>
</msub>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>+</mo>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>y</mi>
<mn>10</mn>
</msub>
<mo>-</mo>
<msub>
<mi>y</mi>
<mn>11</mn>
</msub>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
</mrow>
</msqrt>
</mrow>
</mfrac>
</mrow>
Wherein, error be the error, S1For the second key point prediction result, S3For the described 4th crucial point prediction knot
Fruit, (x10,y10), (x11,y11) for the coordinate of two of which key point in the second key point prediction result.
7. animal face key point mask method according to claim 1, which is characterized in that further include：Supplement certain kinds
Other sample image.
8. animal face key point mask method according to claim 7, which is characterized in that the supplement particular category
The method of sample image includes：
The property distribution of the sample image marked is counted, lacked sample image is known according to the property distribution；
According to default keyword, the lacked sample image is downloaded from network, and the scarce sample image of institute of download is carried out
Duplicate removal processing.
9. animal face key point mask method according to claim 8, which is characterized in that the sample graph marked
The attribute of picture includes：It closes one's eyes, opens eyes, shut up, open one's mouth, positive face, side face.
10. animal face key point mask method according to claim 8, which is characterized in that described pair download lack
Sample image carry out duplicate removal processing method be：Based on pixel similarity duplicate removal or based on semantic Hash duplicate removal.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810023304.8A CN108073914B (en) | 2018-01-10 | 2018-01-10 | Animal face key point marking method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810023304.8A CN108073914B (en) | 2018-01-10 | 2018-01-10 | Animal face key point marking method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108073914A true CN108073914A (en) | 2018-05-25 |
| CN108073914B CN108073914B (en) | 2022-02-18 |
Family
ID=62156750
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810023304.8A Active CN108073914B (en) | 2018-01-10 | 2018-01-10 | Animal face key point marking method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108073914B (en) |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109214343A (en) * | 2018-09-14 | 2019-01-15 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating face critical point detection model |
| CN109241910A (en) * | 2018-09-07 | 2019-01-18 | 高新兴科技集团股份有限公司 | A kind of face key independent positioning method returned based on the cascade of depth multiple features fusion |
| CN110110811A (en) * | 2019-05-17 | 2019-08-09 | 北京字节跳动网络技术有限公司 | Method and apparatus for training pattern, the method and apparatus for predictive information |
| CN110210526A (en) * | 2019-05-14 | 2019-09-06 | 广州虎牙信息科技有限公司 | Predict method, apparatus, equipment and the storage medium of the key point of measurand |
| CN110288512A (en) * | 2019-05-16 | 2019-09-27 | 成都品果科技有限公司 | Illumination for image synthesis remaps method, apparatus, storage medium and processor |
| CN110414369A (en) * | 2019-07-05 | 2019-11-05 | 安徽省农业科学院畜牧兽医研究所 | A kind of training method and device of ox face |
| CN111259822A (en) * | 2020-01-19 | 2020-06-09 | 杭州微洱网络科技有限公司 | Method for detecting key point of special neck in E-commerce image |
| CN111382612A (en) * | 2018-12-28 | 2020-07-07 | 北京市商汤科技开发有限公司 | Animal face detection method and device |
| CN112990335A (en) * | 2021-03-31 | 2021-06-18 | 江苏方天电力技术有限公司 | Intelligent recognition self-learning training method and system for power grid unmanned aerial vehicle inspection image defects |
| WO2021164251A1 (en) * | 2020-02-21 | 2021-08-26 | 平安科技（深圳）有限公司 | Image annotation task pre-verification method and apparatus, device, and storage medium |
| CN113470077A (en) * | 2021-07-15 | 2021-10-01 | 郑州布恩科技有限公司 | Mouse open field experiment movement behavior analysis method based on key point detection |
| CN114550207A (en) * | 2022-01-17 | 2022-05-27 | 北京新氧科技有限公司 | Method and device for detecting key points of neck and method and device for training detection model |
| CN115546845A (en) * | 2022-11-24 | 2022-12-30 | 中国平安财产保险股份有限公司 | Multi-view cow face identification method and device, computer equipment and storage medium |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104504376A (en) * | 2014-12-22 | 2015-04-08 | 厦门美图之家科技有限公司 | Age classification method and system for face images |
| US9008366B1 (en) * | 2012-01-23 | 2015-04-14 | Hrl Laboratories, Llc | Bio-inspired method of ground object cueing in airborne motion imagery |
| CN105426870A (en) * | 2015-12-15 | 2016-03-23 | 北京文安科技发展有限公司 | Face key point positioning method and device |
| CN105844582A (en) * | 2015-01-15 | 2016-08-10 | 北京三星通信技术研究有限公司 | 3D image data registration method and device |
| CN105868769A (en) * | 2015-01-23 | 2016-08-17 | 阿里巴巴集团控股有限公司 | Method and device for positioning face key points in image |
| CN106203376A (en) * | 2016-07-19 | 2016-12-07 | 北京旷视科技有限公司 | Face key point localization method and device |
| CN106204429A (en) * | 2016-07-18 | 2016-12-07 | 合肥赑歌数据科技有限公司 | A kind of method for registering images based on SIFT feature |
| CN106570459A (en) * | 2016-10-11 | 2017-04-19 | 付昕军 | Face image processing method |
| CN107403145A (en) * | 2017-07-14 | 2017-11-28 | 北京小米移动软件有限公司 | Image characteristic points positioning method and device |
-
2018
- 2018-01-10 CN CN201810023304.8A patent/CN108073914B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9008366B1 (en) * | 2012-01-23 | 2015-04-14 | Hrl Laboratories, Llc | Bio-inspired method of ground object cueing in airborne motion imagery |
| CN104504376A (en) * | 2014-12-22 | 2015-04-08 | 厦门美图之家科技有限公司 | Age classification method and system for face images |
| CN105844582A (en) * | 2015-01-15 | 2016-08-10 | 北京三星通信技术研究有限公司 | 3D image data registration method and device |
| CN105868769A (en) * | 2015-01-23 | 2016-08-17 | 阿里巴巴集团控股有限公司 | Method and device for positioning face key points in image |
| CN105426870A (en) * | 2015-12-15 | 2016-03-23 | 北京文安科技发展有限公司 | Face key point positioning method and device |
| CN106204429A (en) * | 2016-07-18 | 2016-12-07 | 合肥赑歌数据科技有限公司 | A kind of method for registering images based on SIFT feature |
| CN106203376A (en) * | 2016-07-19 | 2016-12-07 | 北京旷视科技有限公司 | Face key point localization method and device |
| CN106570459A (en) * | 2016-10-11 | 2017-04-19 | 付昕军 | Face image processing method |
| CN107403145A (en) * | 2017-07-14 | 2017-11-28 | 北京小米移动软件有限公司 | Image characteristic points positioning method and device |
Non-Patent Citations (2)
| Title |
|---|
| MAHEEN RASHID等: "Interspecies Knowledge Transfer for Facial Keypoint Detection", 《2017 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR)》 * |
| 张雄美 等: "基于改进SIFT的SAR图像配准方法", 《计算机工程》 * |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109241910A (en) * | 2018-09-07 | 2019-01-18 | 高新兴科技集团股份有限公司 | A kind of face key independent positioning method returned based on the cascade of depth multiple features fusion |
| CN109241910B (en) * | 2018-09-07 | 2021-01-01 | 高新兴科技集团股份有限公司 | Face key point positioning method based on deep multi-feature fusion cascade regression |
| CN109214343A (en) * | 2018-09-14 | 2019-01-15 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating face critical point detection model |
| CN111382612A (en) * | 2018-12-28 | 2020-07-07 | 北京市商汤科技开发有限公司 | Animal face detection method and device |
| CN110210526A (en) * | 2019-05-14 | 2019-09-06 | 广州虎牙信息科技有限公司 | Predict method, apparatus, equipment and the storage medium of the key point of measurand |
| CN110288512A (en) * | 2019-05-16 | 2019-09-27 | 成都品果科技有限公司 | Illumination for image synthesis remaps method, apparatus, storage medium and processor |
| WO2020233200A1 (en) * | 2019-05-17 | 2020-11-26 | 北京字节跳动网络技术有限公司 | Model training method and device and information prediction method and device |
| CN110110811A (en) * | 2019-05-17 | 2019-08-09 | 北京字节跳动网络技术有限公司 | Method and apparatus for training pattern, the method and apparatus for predictive information |
| CN110414369A (en) * | 2019-07-05 | 2019-11-05 | 安徽省农业科学院畜牧兽医研究所 | A kind of training method and device of ox face |
| CN110414369B (en) * | 2019-07-05 | 2023-04-18 | 安徽省农业科学院畜牧兽医研究所 | Cow face training method and device |
| CN111259822A (en) * | 2020-01-19 | 2020-06-09 | 杭州微洱网络科技有限公司 | Method for detecting key point of special neck in E-commerce image |
| WO2021164251A1 (en) * | 2020-02-21 | 2021-08-26 | 平安科技（深圳）有限公司 | Image annotation task pre-verification method and apparatus, device, and storage medium |
| CN112990335A (en) * | 2021-03-31 | 2021-06-18 | 江苏方天电力技术有限公司 | Intelligent recognition self-learning training method and system for power grid unmanned aerial vehicle inspection image defects |
| CN112990335B (en) * | 2021-03-31 | 2021-10-15 | 江苏方天电力技术有限公司 | Intelligent recognition self-learning training method and system for power grid unmanned aerial vehicle inspection image defects |
| CN113470077A (en) * | 2021-07-15 | 2021-10-01 | 郑州布恩科技有限公司 | Mouse open field experiment movement behavior analysis method based on key point detection |
| CN114550207A (en) * | 2022-01-17 | 2022-05-27 | 北京新氧科技有限公司 | Method and device for detecting key points of neck and method and device for training detection model |
| CN114550207B (en) * | 2022-01-17 | 2023-01-17 | 北京新氧科技有限公司 | Method and device for detecting key points of neck and method and device for training detection model |
| CN115546845A (en) * | 2022-11-24 | 2022-12-30 | 中国平安财产保险股份有限公司 | Multi-view cow face identification method and device, computer equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108073914B (en) | 2022-02-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108073914A (en) | A kind of animal face key point mask method | |
| CN106709909B (en) | A kind of flexible robot's visual identity and positioning system based on deep learning | |
| US20200210702A1 (en) | Apparatus and method for image processing to calculate likelihood of image of target object detected from input image | |
| CN107506693B (en) | Distort face image correcting method, device, computer equipment and storage medium | |
| Dehshibi et al. | A new algorithm for age recognition from facial images | |
| CN109635875A (en) | A kind of end-to-end network interface detection method based on deep learning | |
| CN108647588A (en) | Goods categories recognition methods, device, computer equipment and storage medium | |
| CN109977780A (en) | A kind of detection and recognition methods of the diatom based on deep learning algorithm | |
| CN107895160A (en) | Human face detection and tracing device and method | |
| CN108830188A (en) | Vehicle checking method based on deep learning | |
| CN106960202A (en) | A kind of smiling face's recognition methods merged based on visible ray with infrared image | |
| CN107358149A (en) | A kind of human body attitude detection method and device | |
| CN108710829A (en) | A method of the expression classification based on deep learning and the detection of micro- expression | |
| CN107463865A (en) | Face datection model training method, method for detecting human face and device | |
| CN103810491B (en) | Head posture estimation interest point detection method fusing depth and gray scale image characteristic points | |
| CN107590460B (en) | Face classification method, apparatus and intelligent terminal | |
| CN111723657B (en) | River foreign matter detection method and device based on YOLOv3 and self-optimization | |
| CN110889446A (en) | Face image recognition model training and face image recognition method and device | |
| CN106991408A (en) | The generation method and method for detecting human face of a kind of candidate frame generation network | |
| CN108805016A (en) | A kind of head and shoulder method for detecting area and device | |
| CN110634116A (en) | Facial image scoring method and camera | |
| Amalia et al. | Detection and description generation of diabetic retinopathy using convolutional neural network and long short-term memory | |
| CN109359577A (en) | A kind of Complex Background number detection system based on machine learning | |
| CN106127733A (en) | The method and apparatus of human body target identification | |
| CN105426882A (en) | Method for rapidly positioning human eyes in human face image |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |