CN107273871A - The training method and device of a kind of face characteristic model - Google Patents
The training method and device of a kind of face characteristic model Download PDFInfo
- Publication number
- CN107273871A CN107273871A CN201710562077.1A CN201710562077A CN107273871A CN 107273871 A CN107273871 A CN 107273871A CN 201710562077 A CN201710562077 A CN 201710562077A CN 107273871 A CN107273871 A CN 107273871A
- Authority
- CN
- China
- Prior art keywords
- face
- image
- face characteristic
- standard
- characteristic model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Abstract
The present invention relates to the training method and device of technical field of face recognition, more particularly to a kind of face characteristic model, training method includes:There is provided face sample image to be identified by step S1;Face sample image, is combined by step S2 with each standard decorative image respectively based on a preset strategy, forms corresponding composograph;Step S3, is pre-processed to each composograph;Step S4, extracts the face characteristic information in pretreated composograph;Step S5, the face characteristic model in face characteristic model library is updated according to the face characteristic information of extraction;Trainer includes memory module, input module, image synthesis unit, pretreatment module, characteristic extracting module, training module;The influence that the decorative features of recognition of face object are extracted to face characteristic can be effectively prevented from, to ensure the accurate extraction of face characteristic, the degree of accuracy of recognition of face is high, and reliability is high.
Description
Technical field
The present invention relates to the training method and device of technical field of face recognition, more particularly to a kind of face characteristic model.
Background technology
Recognition of face is mainly used in public safety field in early days as a kind of biometrics identification technology.Traditional people
Face identification technology is mainly based upon the recognition of face of visible images, and this is also familiar identification method.But this mode
There is the defect for being difficult to overcome, especially when ambient lighting changes, recognition effect can drastically decline, it is impossible to meet actual system
The need for system.Solving the scheme of lighting issues has 3-D view recognition of face, and thermal imaging recognition of face.But both technologies are also
Remote immature, recognition effect is unsatisfactory.
Nearly 2 years, with advances in technology, recognition of face was more and more applied in every field.But, wherein
The extraction of human face characteristic point (eye, mouth, nose etc.) influenceed greatly by external factor, such as recognition of face subject wears glasses, or shaved
Beard, or eyes are covered with hair, human face characteristic point can be caused to extract, and error is big, and accuracy rate is low, it is impossible to apply in practice.
The content of the invention
In view of the above-mentioned problems, the present invention proposes a kind of training method of face characteristic model, applied to a face characteristic
In the trainer of model, the trainer of the face characteristic model includes a face characteristic model library;
Wherein, including provide a standard decorative image storehouse, the standard decorative image storehouse includes the mark of the first predetermined number
Quasi- decorative image;Also include:
There is provided face sample image to be identified by step S1;
Step S2, is mutually tied the face sample image based on a preset strategy with each standard decorative image respectively
Close, form corresponding composograph;
Step S3, is pre-processed to each composograph;
Step S4, extracts the face characteristic information in the pretreated composograph;
Step S5, the face characteristic mould in the face characteristic model library is updated according to the face characteristic information of extraction
Type.
Above-mentioned training method, wherein, in the step S5, based on convolutional neural networks to the face characteristic model library
In the face characteristic model be updated.
Above-mentioned training method, wherein, in the step S5, based on the convolutional neural networks to the face characteristic mould
The specific method that the face characteristic model in type storehouse is updated is:
Extracted using the limitation graceful machine of bohr from the probability sample of the face characteristic information described in the second predetermined number
Probability sample, to train the convolutional neural networks to carry out more the face characteristic model in the face characteristic model library
Newly;
Wherein, known to the probability for the probability sample not being extracted.
Above-mentioned training method, wherein, in the step S2, the face sample image leads to the standard decorative image
Cross and the mode of the standard decorative image is superimposed on the basis of the face sample image is combined, form corresponding described close
Into image.
Above-mentioned training method, wherein, in the step S2, the preset strategy be by the face sample image once
It is combined with a standard decorative image, forms the composograph of first predetermined number.
Above-mentioned training method, wherein, in the step S2, the preset strategy is also included the face sample image
Once it is combined with multiple standard decorative images.
Above-mentioned training method, wherein, the standard decorative image is standard sunglasses image, standard beard image, standard
One in hair style image, standard mouth mask image, standard cap image, standard cosmetic image and standard faces plastic deformation image
Plant or a variety of.
Above-mentioned training method, wherein, in the step S1, the face sample image is stored in a face database
In.
A kind of trainer of face characteristic model, including a memory module, the memory module face that is stored with are special
Levy model library;
Wherein, be also stored with a standard decorative image storehouse in the memory module, and the standard decorative image storehouse includes the
The standard decorative image of one predetermined number;
The trainer also includes:
Input module, the face sample image to be identified for inputting;
Image synthesis unit, is connected with the input module and the memory module, for based on a preset strategy respectively
The face sample image is combined with each standard decorative image respectively, corresponding composograph is formed;
Pretreatment module, is connected with described image synthesis module, for being pre-processed to each composograph;
Characteristic extracting module, is connected with the pretreatment module, for extracting in the pretreated composograph
Face characteristic information;
Training module, is connected with the memory module and the characteristic extracting module, for according to extraction respectively
Face characteristic information updates the face characteristic model in the face characteristic model library.
Above-mentioned training method, wherein, the training module is based on convolutional neural networks to the face characteristic model library
In the face characteristic model be updated.
Beneficial effect:The training method and device of face characteristic model proposed by the present invention, can be effectively prevented from face
The influence that the decorative features of identification object are extracted to face characteristic, to ensure the accurate extraction of face characteristic, the standard of recognition of face
Exactness is high, and reliability is high.
Brief description of the drawings
Fig. 1 is the step flow chart of the training method of face characteristic model in one embodiment of the invention;
Fig. 2 be one embodiment of the invention in image synthesis before face sample image schematic diagram;
Fig. 3 is the schematic diagram of composograph in one embodiment of the invention;
Fig. 4 is the structure principle chart of the trainer of face characteristic model in one embodiment of the invention.
Embodiment
The present invention is further described with reference to the accompanying drawings and examples, it will be appreciated that described herein preferred
Embodiment is merely to illustrate and explain the present invention, and is not intended to limit the present invention.
Embodiment one
, can be with as shown in figure 1, in a preferred embodiment, it is proposed that a kind of training method of face characteristic model
In trainer applied to a face characteristic model, the trainer of face characteristic model can include a face characteristic model
Storehouse;
Wherein it is possible to which including providing a standard decorative image storehouse, standard decorative image storehouse includes the mark of the first predetermined number
Quasi- decorative image;It can also include:
There is provided face sample image to be identified by step S1;
Face sample image, is combined by step S2 with each standard decorative image respectively based on a preset strategy, is formed
Corresponding composograph;
Step S3, is pre-processed to each composograph;
Step S4, extracts the face characteristic information in pretreated composograph;
Step S5, the face characteristic model in face characteristic model library is updated according to the face characteristic information of extraction.
In above-mentioned technical proposal, the face characteristic in face sample image to be identified should be without modification, still
In most of application scenarios, recognition of face object often hides the recognition of face of computer by the modification of profile, for example
By wearing cap, the either identification to face characteristic by pasting usurped beard or by wearing wig interference calculation machine;So
And the training method of the face characteristic model in the present invention can accurately carry out recognition of face, even if recognition of face object is to certainly
Oneself is modified, and can also be come out the recognition of face Object identifying, the face characteristic the reliability of the adjustment model trained is higher, is known
The other degree of accuracy is high;The pretreatment to face sample image can also be included in step S3, so that the training method takes into account existing
To the extraction and the training of face characteristic model of face characteristic in the processing of somebody's face sample image and subsequent step.
, can be based on convolutional neural networks in face characteristic model library in step S5 in a preferred embodiment
Face characteristic model be updated.
In above-mentioned technical proposal, the convolutional layer of convolutional neural networks applies the discrete representation of convolution.If examined or check layer
K-th of characteristic plane is denoted as hk, the connection weight matrix with preceding layer is Wk, deviation is bk, for nonlinear tan,
Feature Mapping can be obtained as follows:
In formula,Exported for the mapping of convolutional layer.
As preferred embodiment, in step S5, based on convolutional neural networks to the face in face characteristic model library
The specific method that characteristic model is updated can be:
The probability sample of the second predetermined number is extracted from the probability sample of face characteristic information using the limitation graceful machine of bohr,
The face characteristic model in face characteristic model library is updated with training convolutional neural networks;
Wherein, known to the probability for the probability sample not being extracted.
In above-mentioned technical proposal, effective random sampling technology for being used in the graceful machine of each limitation bohr of training, it is necessary to from
Unknown joint probability distribution f (x1..., xk) n probability sample X of middle extraction(1), X(2)..., X(n), by
So as to when extracting probability sample, it is assumed that its dependent variable is, it is known that under using its dependent variable as the probability distribution of condition
Extracted, until extracting all samples, i.e., for sample X(i)J-th of variable, be from distributionMiddle extraction, until extracting n sample.
As shown in Figures 2 and 3, in a preferred embodiment, in step S2, face sample image is modified with standard schemes
As being combined by way of being superimposed standard decorative image on the basis of face sample image, corresponding composograph is formed;
Wherein, composograph can include overlap-add region and non-superimposed region;Image in non-superimposed region should be with face
Image in sample image in corresponding region is consistent, and this uniformity can show as the completely the same of all pixels point,
Position consistency of human face characteristic point etc. in image can be shown as (such as coordinate position of the human face characteristic point on image is consistent);
Image in the overlap-add region of composograph can or standard decorative image consistent with standard decorative image preset by one
Transparency is combined with face sample image, and the default transparency can be 30%, 40%, 50%, 60%, or 70% etc..
, can be with as shown in Fig. 2 composograph with reference to after can be with reference to preceding face sample image in above-mentioned technical proposal
As shown in Figure 3.
In a preferred embodiment, in step S2, preset strategy be by face sample image once with a standard
Decorative image is combined, and forms the composograph of the first predetermined number.
As preferred embodiment, in step S2, preset strategy also include by face sample image once with multiple marks
Quasi- decorative image is combined.
, can be with as shown in Fig. 2 composograph with reference to after can be with reference to preceding face sample image in above-mentioned technical proposal
As shown in figure 3, the standard decorative image combined in Fig. 3 is standard cap image and standard sunglasses image.
In a preferred embodiment, standard decorative image is standard sunglasses image, standard beard image, standard hair style
Image, standard mouth mask image, standard cap image, standard cosmetic image and standard faces plastic deformation image in one kind or
It is a variety of, to ensure to be identified when recognition of face object carries out a kind of standard modification or multiple standards modification.
In above-mentioned technical proposal, a kind of standard decorative image can also have multiple modification species, for example, standard is sent out
Type image can have multiple species, such as standard long hair types of image, standard short hair styles image etc.;Standard faces are plastically deformed
Image can also include multiple species, such as include standard faces smile's image, standard faces rotation image etc.;Other standards are modified
Image also can be similar, will not be repeated here.
In a preferred embodiment, in step S1, face sample image is stored in a face database, in order to
Storage and extraction.
In above-mentioned technical proposal, face sample image can have one or more, when there is multiple face sample images,
The composograph formed should correspond to corresponding face sample image so that each face sample image correspondence one is combined into figure
Picture.
Embodiment two
As shown in figure 4, in a preferred embodiment, except the training method of above-mentioned face characteristic model, also carrying
A kind of trainer of face characteristic model is gone out, a memory module 1 can be included, memory module is stored with a face characteristic mould
Type storehouse;
Wherein, can also be stored with a standard decorative image storehouse in memory module 1, and it is pre- that standard decorative image storehouse includes first
If the standard decorative image of quantity;
The trainer can also include:
Input module 2, the face sample image to be identified for inputting;
Image synthesis unit 3, is connected with input module 2 and memory module 1 respectively, for based on a preset strategy by face
Sample image is combined with each standard decorative image respectively, forms corresponding composograph;
Pretreatment module 4, is connected with image synthesis unit 3, for being pre-processed to each composograph;
Characteristic extracting module 5, is connected with pretreatment module 4, special for extracting the face in pretreated composograph
Reference ceases;
Training module 6, is connected with memory module 1 and characteristic extracting module 5 respectively, for being believed according to the face characteristic of extraction
Breath updates the face characteristic model in face characteristic model library.
In above-mentioned technical proposal, composograph can include overlap-add region and non-superimposed region;Figure in non-superimposed region
As image that should be in region corresponding with face sample image is consistent, this uniformity can show as the complete of all pixels point
It is complete consistent, (such as coordinate of the human face characteristic point on image of position consistency of human face characteristic point etc. in image can also be shown as
Position consistency);Image in the overlap-add region of composograph can or standard consistent with standard decorative image modification figure
As being combined by a default transparency with face sample image, the default transparency can be 30%, 40%, 50%, 60%, or
70% etc..
In a preferred embodiment, training module 6 can be based on convolutional neural networks in face characteristic model library
Face characteristic model be updated.
It should be noted that when the face identification device that above-described embodiment is provided states function in realization, only with above-mentioned work(
The division progress of energy module is for example, in practical application, as needed can distribute above-mentioned functions by different functions
Module is completed, i.e., the internal structure of equipment is divided into different functional modules, described above all or part of to complete
Function.In addition, the face identification device that above-described embodiment is provided belongs to same design with face identification method embodiment, its is specific
Implementation process refers to embodiment of the method, repeats no more here.
It should be understood by those skilled in the art that, embodiments of the invention can be provided as method, system or computer program
Product.Therefore, the present invention can be using the reality in terms of complete hardware embodiment, complete software embodiment or combination software and hardware
Apply the form of example.
The present invention is the flow with reference to method according to embodiments of the present invention, equipment (system) and computer program product
Figure and/or block diagram are described.It should be understood that can be by every first-class in computer program instructions implementation process figure and/or block diagram
Journey and/or the flow in square frame and flow chart and/or block diagram and/or the combination of square frame.These computer programs can be provided
The processor of all-purpose computer, special-purpose computer, Embedded Processor or other programmable data processing devices is instructed to produce
A raw machine so that produced by the instruction of computer or the computing device of other programmable data processing devices for real
The device for the function of being specified in present one flow of flow chart or one square frame of multiple flows and/or block diagram or multiple square frames.
These computer program instructions, which may be alternatively stored in, can guide computer or other programmable data processing devices with spy
Determine in the computer-readable memory that mode works so that the instruction being stored in the computer-readable memory, which is produced, to be included referring to
Make the manufacture of device, the command device realize in one flow of flow chart or multiple flows and/or one square frame of block diagram or
The function of being specified in multiple square frames.
By explanation and accompanying drawing, the exemplary embodiments of the specific structure of embodiment are given, based on essence of the invention
God, can also make other conversions.Although foregoing invention proposes existing preferred embodiment, however, these contents are not intended as
Limitation.
For a person skilled in the art, read after described above, various changes and modifications undoubtedly will be evident.
Therefore, appended claims should regard whole variations and modifications of the true intention and scope that cover the present invention as.In power
Any and all scope and content of equal value, are all considered as still belonging to the intent and scope of the invention in the range of sharp claim.
Claims (10)
1. in a kind of training method of face characteristic model, the trainer applied to a face characteristic model, the face is special
Levying the trainer of model includes a face characteristic model library;
It is characterised in that it includes providing a standard decorative image storehouse, the standard decorative image storehouse includes the first predetermined number
Standard decorative image;Also include:
There is provided face sample image to be identified by step S1;
The face sample image, is combined by step S2 with each standard decorative image respectively based on a preset strategy,
Form corresponding composograph;
Step S3, is pre-processed to each composograph;
Step S4, extracts the face characteristic information in the pretreated composograph;
Step S5, the face characteristic model in the face characteristic model library is updated according to the face characteristic information of extraction.
2. training method according to claim 1, it is characterised in that in the step S5, based on convolutional neural networks pair
The face characteristic model in the face characteristic model library is updated.
3. training method according to claim 2, it is characterised in that in the step S5, based on the convolutional Neural net
The specific method that network is updated to the face characteristic model in the face characteristic model library is:
The probability of the second predetermined number is extracted from the probability sample of the face characteristic information using the limitation graceful machine of bohr
Sample, to train the convolutional neural networks to be updated the face characteristic model in the face characteristic model library;
Wherein, known to the probability for the probability sample not being extracted.
4. training method according to claim 1, it is characterised in that in the step S2, the face sample image with
The standard decorative image is mutually tied by way of being superimposed the standard decorative image on the basis of the face sample image
Close, form the corresponding composograph.
5. training method according to claim 1, it is characterised in that in the step S2, the preset strategy is by institute
State face sample image to be once combined with a standard decorative image, form the synthesis of first predetermined number
Image.
6. training method according to claim 5, it is characterised in that in the step S2, the preset strategy also includes
The face sample image is once combined with multiple standard decorative images.
7. training method according to claim 1, it is characterised in that the standard decorative image is standard sunglasses image,
Standard beard image, standard hair style image, standard mouth mask image, standard cap image, standard cosmetic image and standard faces
It is plastically deformed the one or more in image.
8. training method according to claim 1, it is characterised in that in the step S1, the face sample image is deposited
It is stored in a face database.
9. a kind of trainer of face characteristic model, including a memory module, the memory module is stored with a face characteristic
Model library;
Characterized in that, the standard decorative image storehouse that is also stored with the memory module, the standard decorative image storehouse includes
The standard decorative image of first predetermined number;
The trainer also includes:
Input module, the face sample image to be identified for inputting;
Image synthesis unit, is connected with the input module and the memory module respectively, for based on a preset strategy by institute
State face sample image to be combined with each standard decorative image respectively, form corresponding composograph;
Pretreatment module, is connected with described image synthesis module, for being pre-processed to each composograph;
Characteristic extracting module, is connected with the pretreatment module, for extracting the face in the pretreated composograph
Characteristic information;
Training module, is connected with the characteristic extracting module, for updating the people according to the face characteristic information of extraction
Face characteristic model in face feature model library.
10. training method according to claim 9, it is characterised in that the training module is based on convolutional neural networks pair
The face characteristic model in the face characteristic model library is updated.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710562077.1A CN107273871A (en) | 2017-07-11 | 2017-07-11 | The training method and device of a kind of face characteristic model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710562077.1A CN107273871A (en) | 2017-07-11 | 2017-07-11 | The training method and device of a kind of face characteristic model |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107273871A true CN107273871A (en) | 2017-10-20 |
Family
ID=60071877
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710562077.1A Pending CN107273871A (en) | 2017-07-11 | 2017-07-11 | The training method and device of a kind of face characteristic model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107273871A (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107808373A (en) * | 2017-11-15 | 2018-03-16 | 北京奇虎科技有限公司 | Sample image synthetic method, device and computing device based on posture |
CN107862266A (en) * | 2017-10-30 | 2018-03-30 | 广东欧珀移动通信有限公司 | Image processing method and related product |
CN108764149A (en) * | 2018-05-29 | 2018-11-06 | 北京中庆现代技术股份有限公司 | A kind of training method for class student faceform |
CN109858524A (en) * | 2019-01-04 | 2019-06-07 | 北京达佳互联信息技术有限公司 | Gesture identification method, device, electronic equipment and storage medium |
CN110909654A (en) * | 2019-11-18 | 2020-03-24 | 深圳市商汤科技有限公司 | Training image generation method and device, electronic equipment and storage medium |
CN111914630A (en) * | 2020-06-19 | 2020-11-10 | 北京百度网讯科技有限公司 | Method, apparatus, device and storage medium for generating training data for face recognition |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20080273766A1 (en) * | 2007-05-03 | 2008-11-06 | Samsung Electronics Co., Ltd. | Face recognition system and method based on adaptive learning |
CN102034079A (en) * | 2009-09-24 | 2011-04-27 | 汉王科技股份有限公司 | Method and system for identifying faces shaded by eyeglasses |
CN104598935A (en) * | 2014-12-29 | 2015-05-06 | 天津瑞为拓新科技发展有限公司 | Incremental subspace learning algorithm-based face recognition method |
CN105184253A (en) * | 2015-09-01 | 2015-12-23 | 北京旷视科技有限公司 | Face identification method and face identification system |
CN106339702A (en) * | 2016-11-03 | 2017-01-18 | 北京星宇联合投资管理有限公司 | Multi-feature fusion based face identification method |
-
2017
- 2017-07-11 CN CN201710562077.1A patent/CN107273871A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20080273766A1 (en) * | 2007-05-03 | 2008-11-06 | Samsung Electronics Co., Ltd. | Face recognition system and method based on adaptive learning |
CN102034079A (en) * | 2009-09-24 | 2011-04-27 | 汉王科技股份有限公司 | Method and system for identifying faces shaded by eyeglasses |
CN104598935A (en) * | 2014-12-29 | 2015-05-06 | 天津瑞为拓新科技发展有限公司 | Incremental subspace learning algorithm-based face recognition method |
CN105184253A (en) * | 2015-09-01 | 2015-12-23 | 北京旷视科技有限公司 | Face identification method and face identification system |
CN106339702A (en) * | 2016-11-03 | 2017-01-18 | 北京星宇联合投资管理有限公司 | Multi-feature fusion based face identification method |
Non-Patent Citations (4)
Title |
---|
李凡长等: "《李群机器学习》", 30 April 2013, 中国科学技术大学出版社 * |
柴瑞敏等: "基于Gabor小波与深度信念网络的人脸识别方法", 《计算机应用》 * |
蒋政: "人脸识别中特征提取算法的研究与实现", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
闫新宝: "深度学习及其在人脸识别中的应用进展", 《无线互联科技》 * |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107862266A (en) * | 2017-10-30 | 2018-03-30 | 广东欧珀移动通信有限公司 | Image processing method and related product |
CN107808373A (en) * | 2017-11-15 | 2018-03-16 | 北京奇虎科技有限公司 | Sample image synthetic method, device and computing device based on posture |
CN108764149A (en) * | 2018-05-29 | 2018-11-06 | 北京中庆现代技术股份有限公司 | A kind of training method for class student faceform |
CN109858524A (en) * | 2019-01-04 | 2019-06-07 | 北京达佳互联信息技术有限公司 | Gesture identification method, device, electronic equipment and storage medium |
CN109858524B (en) * | 2019-01-04 | 2020-10-16 | 北京达佳互联信息技术有限公司 | Gesture recognition method and device, electronic equipment and storage medium |
US11163373B2 (en) | 2019-01-04 | 2021-11-02 | Beijing Dajia Internet Information Technology Co., Ltd. | Method and electronic device of gesture recognition |
CN110909654A (en) * | 2019-11-18 | 2020-03-24 | 深圳市商汤科技有限公司 | Training image generation method and device, electronic equipment and storage medium |
CN111914630A (en) * | 2020-06-19 | 2020-11-10 | 北京百度网讯科技有限公司 | Method, apparatus, device and storage medium for generating training data for face recognition |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107273871A (en) | The training method and device of a kind of face characteristic model | |
CN107358648B (en) | Real-time full-automatic high quality three-dimensional facial reconstruction method based on individual facial image | |
CN108701216B (en) | Face recognition method and device and intelligent terminal | |
KR102433971B1 (en) | Systems and Methods for Facial Expression Recognition and Annotation | |
CN105678232B (en) | A kind of face picture feature extraction and control methods based on deep learning | |
CN108229330A (en) | Face fusion recognition methods and device, electronic equipment and storage medium | |
CN107610209A (en) | Human face countenance synthesis method, device, storage medium and computer equipment | |
CN106778584A (en) | A kind of face age estimation method based on further feature Yu shallow-layer Fusion Features | |
US10911695B2 (en) | Information processing apparatus, information processing method, and computer program product | |
CN108829900A (en) | A kind of Research on face image retrieval based on deep learning, device and terminal | |
WO2022143645A1 (en) | Three-dimensional face reconstruction method and apparatus, device, and storage medium | |
CN105426882B (en) | The method of human eye is quickly positioned in a kind of facial image | |
CN111178128B (en) | Image recognition method, device, computer equipment and storage medium | |
CN112257503A (en) | Sex age identification method, device and storage medium | |
US9877791B2 (en) | System and method for virtual treatments based on aesthetic procedures | |
CN113570684A (en) | Image processing method, image processing device, computer equipment and storage medium | |
CN106803084A (en) | A kind of facial characteristics independent positioning method based on end-to-end recirculating network | |
CN110288513A (en) | For changing the method, apparatus, equipment and storage medium of face character | |
CN105718898B (en) | Face age estimation method, system based on sparse undirected probability graph model | |
CN107194995A (en) | A kind of method of rapid build true three-dimension person model | |
CN112200074A (en) | Attitude comparison method and terminal | |
US9514354B2 (en) | Facial analysis by synthesis and biometric matching | |
CN112115790A (en) | Face recognition method and device, readable storage medium and electronic equipment | |
CN115147261A (en) | Image processing method, device, storage medium, equipment and product | |
Jandová et al. | Sexual dimorphism in human facial expressions by 3D surface processing |
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 | ||
TA01 | Transfer of patent application right | ||
TA01 | Transfer of patent application right |
Effective date of registration: 20190805 Address after: Room 318, Building 2, 588 Zixing Road, Minhang District, Shanghai, 200241 Applicant after: Shanghai, like me, Mdt InfoTech Ltd Address before: Room 2304, 1399 Pudong Road, Pudong New Area, Shanghai, 200122 Applicant before: Xia Li |
|
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20171020 |