CN105138951B - Human face portrait-photo array the method represented based on graph model - Google Patents
Human face portrait-photo array the method represented based on graph model Download PDFInfo
- Publication number
- CN105138951B CN105138951B CN201510397326.7A CN201510397326A CN105138951B CN 105138951 B CN105138951 B CN 105138951B CN 201510397326 A CN201510397326 A CN 201510397326A CN 105138951 B CN105138951 B CN 105138951B
- Authority
- CN
- China
- Prior art keywords
- portrait
- test
- graph model
- photo
- block
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
Abstract
The invention discloses a kind of human face portrait photo array methods represented based on graph model, mainly solve the problems, such as that existing method ignores facial image spatial structural form when carrying out human face portrait photo array.Implementation step is：(1) division training portrait sample set, training photo sample set and test sample collection；(2) test portrait graph model is formed according to division result and represents that collection and test photo graph model represent collection；(3) represent that collection and test photo graph model represent that collection calculates similarity collection according to test portrait graph model；(4) human face portrait photo array rate is calculated according to similarity collection.The present invention compared with the conventional method, the spatial structural form of facial image is used during calculating graph model and representing, improves human face portrait photo array rate, the identification available for suspect in criminal investigation and case detection.
Description
Technical field
The invention belongs to technical field of image processing, more particularly to a kind of human face portrait-photo array method, available for punishment
The identification of suspect in detection case.
Background technology
During criminal investigation and case detection, it is difficult what is obtained that the photo of suspect, which is usually, at this time according to eye witness or
The portrait for the suspect that the description of victim is drawn out is to determine the important clue of suspect's identity.Since face is drawn
The difference of picture and photo on generation mechanism, portrait the texture of photo between exist very big difference, by traditional human face photo-
The discrimination that photo array method is applied directly to acquirement among human face portrait-photo array is very low, and solving a case for the police, it is tired to bring
It is difficult.Human face portrait-photo array technology is to reduce the difference drawn a portrait between two kinds of images of photo by signal processing technology, is carried
The discrimination of high human face portrait-photo array, therefore received significant attention in image processing field.
At present, largely it is suggested on human face portrait-photo array method, is broadly divided into three classes：Based on synthetic method,
The method of method and feature based based on subspace projection.
First, it is to synthesize pseudo- photo by that will draw a portrait based on synthetic method, then utilizes traditional human face photo-photograph
Piece recognition methods is identified between pseudo- photo and photo.N.Wang et al. document " N.Wang, D.Tao, X.Gao,
X.Li,and J.Li.Transductive face sketch-photo synthesis.IEEE Transactions on
Neural Networks and Learning System,24(9):One kind is proposed in 1364-1376,2013 " to be based on directly pushing away
Human face portrait-picture synthesis method of formula realizes the synthesis of human face portrait-photo, Ran Hou using the thought of transductive learning
It is upper between the pseudo- photo and photo of synthesis to carry out human face photo-photo array.Shortcoming is existing for this method, recognition effect
The quality of the pseudo- photo of synthesis is depended primarily upon, due to can cause to identify there are the deformation and distortion of image in the synthesis process
Rate is low.
2nd, the method based on subspace projection, be by will draw a portrait and photo project to simultaneously in a sub-spaces, then
Portrait and photo are compared in this sub-spaces, realize human face portrait-photo array.A.Sharma et al. is in document
“A.Sharma and D.Jacobs.Bypass synthesis:PLS for face recognition with pose,
low-resolution and sketch.In Proc.IEEE Int.Conference on Computer Vision and
A kind of human face portrait-photo based on offset minimum binary is proposed in Pattern Recognition, pp.593-600,2011 "
Recognition methods will draw a portrait by using Partial Least Squares and photographic projection is to same linear subspaces, then in this line
The enterprising pedestrian's face sketch-photo identification of subspace.Shortcoming is existing for this method, and letter is usually there will be in projection process
The loss of breath reduces recognition effect.
3rd, the method for feature based respectively encodes human face portrait and photo first with feature, then passes through meter
The distance relation of the feature after coding is calculated to realize human face portrait-photo array.A.Alex et al. document " A.Alex,
V.Asari,and A.Mathew.Local difference of Gaussian binary pattern:robust
features for face sketch recognition.In Proc.IEEE Int.Conference on Systems,
A kind of people based on difference of Gaussian binaryzation feature is proposed in Man, and Cybernetics, pp.1211-1216,2013 "
Face sketch-photo recognition methods respectively encodes human face portrait and photo using difference of Gaussian binaryzation feature, Ran Houli
Human face portrait-photo array is carried out with coding result.Shortcoming is existing for this method, does not have when being encoded using feature
The favourable spatial structural form for using facial image, causes recognition result poor.
The content of the invention
It is an object of the invention to be directed to the deficiency of above-mentioned existing method, propose that a kind of face represented based on graph model is drawn
Picture-photo array method, with by using the spatial structural form of facial image, the discrimination of raising human face portrait-photo.
Realize that the technical solution of the object of the invention includes the following steps：
(1) M portrait composition training portrait sample sets are taken out to concentrating from sketch-photo, and taken out and training portrait sample
One-to-one M photo composition training photo sample sets of portrait of this concentration, by sketch-photo to concentrating remaining N to drawing
Picture-photo forms test sample collection；
(2) composition test portrait graph model represents collection WSCollection W is represented with test photo graph modelP：
Every test portrait that test sample is concentrated is divided into same size and overlapped test portrait block, and will
Every test portrait is respectively with training portrait sample set combination learning, and the graph model for obtaining every test portrait represents, composition survey
Examination portrait graph model represents collectionWhereinFor
I-th of test portrait graph model expression,The graph model expression for testing portrait block for b-th, i=1,2 ..., N, b=1,
2 ..., B, B are the total number of test portrait block；
Every test photo that test sample is concentrated is divided into same size and overlapped test photo block, and will
For every test photo respectively with training photo sample set combination learning, the graph model for obtaining every test photo represents that composition is surveyed
It tries photo graph model and represents collectionWherein
It is represented for j-th of test photo graph model,For the graph model expression of b-th of test photo block, j=1,2 ..., N；
(3) statistical parameter u=0 is initialized；
(4) i-th of test portrait graph model is represented into Wi SCollection W is represented with test photo graph modelPIn each test photo
Graph model represents progress similarity calculation, obtains the similarity collection T={ T that i-th of test portrait graph model representsi,1,Ti,2,…,
Ti,j,…,Ti,N, wherein Ti,jW is represented for i-th of test portrait graph modeli SIt is represented with j-th of test photo graph model's
Similarity；
(5) similarity in similarity collection T is sorted from big to small, finds out maximum similarity Ti,hIf h is equal to i,
Then statistical parameter u increases by 1；
(6) step (4)-(5) are repeated, until having handled test portrait graph model represents collection WSIn all tests portrait figure
Model represents, the discrimination r of human face portrait-photo is calculated further according to following formula：
R=u/N.
The present invention realizes human face portrait-photo array due to being represented using graph model, and represents process calculating graph model
The middle spatial structural form using facial image overcomes the space structure that existing method ignores facial image in identification process
The problem of recognition effect that information band comes is poor improves the discrimination of human face portrait-photo.
To be described in further detail the step of below in conjunction with the accompanying drawings, realization to the present invention.
Description of the drawings
Fig. 1 is that the present invention is based on human face portrait-photo array flow charts that graph model represents；
Specific embodiment
With reference to Fig. 1, the step of realizing of the invention, is as follows：
Step 1, division training portrait sample set, training photo sample set and test sample collection.
M portrait composition training portrait sample sets are taken out to concentrating from sketch-photo, and are taken out and training portrait sample set
In portrait one-to-one M photos composition training photo sample sets, by sketch-photo to concentrating remaining N to portrait-photograph
Piece forms test sample collection.
Step 2, composition test portrait graph model represents collection WSCollection W is represented with test photo graph modelP。
Every test portrait that (2a) concentrates test sample is divided into same size and overlapped test portrait block,
And by every test portrait respectively with training portrait sample set combination learning, the graph model for obtaining every test portrait represents, group
Collection W is represented into test portrait graph modelS：
It is described test portrait with training portrait sample set combination learning, be using bibliography " H.Zhou, Z.Kuang,
and K.Wong.Markov Weight Fields for Face Sketch Synthesis.In Proc.IEEE
Method disclosed in Int.Conference on Computer Vision, pp.1091-1097,2012 " carries out, and step is such as
Under：
I-th test portrait S that (2a1) concentrates test sampleiIt is divided into same size and overlapped test picture
As block, using the pixel value of each test portrait block as feature vector, test portrait S is obtainediSet of eigenvectors f (Si)={ f
(Si,1),f(Si,2),…,f(Si,b),…,f(Si,B), wherein Si,bFor b-th of test portrait block, f (Si,b) tested for b-th
Draw a portrait block feature vector, i=1,2 ..., N, b=1,2 ..., B, B be test portrait block total number, described eigenvector bag
It includes pixel value, scale invariant feature, histogram of gradients feature and accelerates robust features etc., the present invention selects but is not limited to pixel
Value is as feature vector；
(2a2) is according to test portrait S in step (2a1)iDivision result, will training portrait sample set in training portrait
It is divided into same size and overlapped training portrait block, and using the pixel value of each training portrait block as feature vector；
(2a3) is to b-th of test portrait block Si,b, each identical bits for training portrait are taken out in drawing a portrait sample set from training
The training portrait block put obtains common M training portrait block and forms to wait to select the block collection D of drawing a portraiti,b, and will treat selection portrait block collection Di,b
In it is all training portrait blocks feature vectors form set of eigenvectors F to be selectedi,b；By test portrait block Si,bIt is adjacent with v-th
The overlapping region of test portrait block is denoted asIt will treat selection portrait block collection Di,bIn overlapping regionInterior pixel value composition weight
Folded provincial characteristics vector setV=1,2 ..., 4；
(2a4) calculates b-th of test portrait block S according to the following formulai,bGraph model represent
Wherein It represents v-th
Test portrait block Si,vTreat selection portrait block collection Di,vIn overlapping regionInterior pixel value,TIt represents to carry out transposition to matrix
Operation；
(2a5) repeats step (2a3)-(2a4), until the graph model for obtaining B test portrait block represents, composition test picture
As graph model represents
(2a6) repeats step (2a1)-(2a5), is represented until obtaining N number of test portrait graph model, composition test portrait figure
Model represents collectionWherein i=1,2 ..., N；
Every test photo that (2b) concentrates test sample is divided into same size and overlapped test photo block,
And by every test photo respectively with training photo sample set combination learning, the graph model for obtaining every test photo represents, group
Collection is represented into test photo graph modelWherein
It is represented for j-th of test photo graph model,For the graph model expression of b-th of test photo block, j=1,2 ..., N；
The specific implementation of this step is carried out with reference to step (2a).
Step 3, statistical parameter u=0 is initialized.
Step 4, calculate i-th of test portrait graph model and represent Wi SSimilarity collection.
It calculates i-th of test portrait graph model and represents Wi SSimilarity collection can be used the method estimated based on Euclidean distance and
Method based on COS distance etc., the present invention use but are not limited to following method：
(4a) represents test portrait graph modelIt is represented with test photo graph modelThe graph model for calculating b-th of test portrait block according to the following formula representsWith b-th
The graph model for testing photo block representsSimilarity tb：
WhereinRepresent that graph model representsZ dimension element value,Represent that graph model represents
Z ties up element value, nzParameter in order to control works as element valueAnd element valueIt is all higher than n when 0.0001z1 is taken, otherwise nz
0, b=1 is taken, 2 ..., B, B are the total number of test portrait block；
(4b) repeats step (4a), until obtaining B similarity, calculates test portrait graph model according to the following formula and represents Wi SWith
Photo graph model is tested to representSimilarity Ti,j：
(4c) repeats step (4a)-(4b), until having handled test photo graph model represents collection WPIn all test photos
Graph model represents, obtains the similarity collection T={ T that i-th of test portrait graph model representsi,1,Ti,2,…,Ti,j,…,Ti,N}。
Step 5, the similarity in similarity collection T is sorted from big to small, finds out maximum similarity Ti,hIf h is equal to
I, then statistical parameter u increases by 1.
Step 6, human face portrait-photo array rate is calculated.
Step 4-5 is repeated, until having handled test portrait graph model represents collection WSIn all tests portrait graph model table
Show, the discrimination r of human face portrait-photo is calculated further according to following formula：
R=u/N.
The effect of the present invention can be described further by following emulation experiment.
1. simulated conditions
The present invention is to be grasped in central processing unit for Intel (R) Core i3-530 2.93GHZ, memory 4G, WINDOWS 7
Make in system, the emulation carried out with MATLAB softwares.
The method compared in experiment includes following 2 kinds：
First, the method based on direct-push, TFSPS is denoted as in experiment；Bibliography is N.Wang, D.Tao, X.Gao,
X.Li,and J.Li.Transductive face sketch-photo synthesis.IEEE Transactions on
Neural Networks and Learning System,24(9):1364-1376,2013；
Second is that the method based on difference of Gaussian binaryzation feature, LDoGBP is denoted as in experiment；Bibliography is A.Alex,
V.Asari,and A.Mathew.Local difference of Gaussian binary pattern:robust
features for face sketch recognition.In Proc.IEEE Int.Conference on Systems,
Man,and Cybernetics,pp.1211-1216,2013。
The database used in experiment is CUHK Face Sketch FERET Database disclosed in Hong Kong Chinese University
Representation data storehouse.
2. emulation content
According to the present invention described in specific embodiment, calculate human face portrait-photo array rate, and with TFSPS methods and
The discrimination of LDoGBP methods is compared, and the results are shown in Table 1.
1 human face portrait of table-photo array rate
As seen from Table 1, discrimination of the invention is higher than two kinds of control methods, illustrates that the method for the present invention is represented using graph model
It realizes human face portrait-photo array, and the spatial structural form of facial image is used during calculating graph model and representing, it can be with
Preferable recognition effect is obtained, demonstrates the advance of the present invention.
Claims (2)
1. a kind of human face portrait-photo array method represented based on graph model, is included the following steps：
(1) M portrait composition training portrait sample sets are taken out to concentrating from sketch-photo, and taken out and training portrait sample set
In portrait one-to-one M photos composition training photo sample sets, by sketch-photo to concentrating remaining N to portrait-photograph
Piece forms test sample collection；
(2) composition test portrait graph model represents collection WSCollection W is represented with test photo graph modelP：
Every test portrait that test sample is concentrated is divided into same size and overlapped test portrait block, and by every
For test portrait respectively with training portrait sample set combination learning, the graph model for obtaining every test portrait represents that composition tests picture
As graph model represents collectionWhereinFor i-th
Test portrait graph model expression,For the graph model expression of b-th of test portrait block, i=1,2 ..., N, b=1,2 ..., B, B
For the total number of test portrait block；
Every test photo that test sample is concentrated is divided into same size and overlapped test photo block, and by every
Photo is tested respectively with training photo sample set combination learning, and the graph model for obtaining every test photo represents that composition test is shone
Piece graph model represents collectionWhereinFor jth
A test photo graph model expression,For the graph model expression of b-th of test photo block, j=1,2 ..., N；
The sample set combination learning that every test portrait is drawn a portrait respectively with training, carries out as follows：
I-th test portrait S that (2a) concentrates test sampleiSame size and overlapped test portrait block are divided into, it will
The pixel value of each test portrait block obtains test portrait S as feature vectoriSet of eigenvectors f (Si)={ f (Si,1),f
(Si,2),…,f(Si,b),…,f(Si,B), wherein Si,bFor b-th of test portrait block, f (Si,b) it is b-th of test portrait block
Feature vector, b=1,2 ..., B, B are the total number of test portrait block；
(2b) is according to test portrait S in step (2a)iDivision result, will training portrait sample set in training portrait be divided into phase
With size and overlapped training portrait block, and using the pixel value of each training portrait block as feature vector；
(2c) is to b-th of test portrait block Si,b, the instruction of the same position of each training portrait of taking-up from training portrait sample set
Practice portrait block, obtain common M training portrait block and form to wait to select the block collection D of drawing a portraiti,b, and will treat selection portrait block collection Di,bIn own
The feature vector of training portrait block forms set of eigenvectors F to be selectedi,b；By test portrait block Si,bTest picture adjacent with v-th
As the overlapping region of block is denoted asIt will treat selection portrait block collection Di,bIn overlapping regionInterior pixel value composition overlapping region
Set of eigenvectors
(2d) calculates b-th of test portrait block S according to the following formulai,bGraph model represent
Wherein Represent v-th of test
Draw a portrait block Si,vTreat selection portrait block collection Di,vIn overlapping regionInterior pixel value,TIt represents to carry out transposition behaviour to matrix
Make；
(2e) repeats step (2c)-(2d), until the graph model for obtaining B test portrait block represents, composition test portrait artwork
Type represents
(3) statistical parameter u=0 is initialized；
(4) i-th of test portrait graph model is represented into Wi SCollection W is represented with test photo graph modelPIn each test photo artwork
Type represents progress similarity calculation, obtains the similarity collection T={ T that i-th of test portrait graph model representsi,1,Ti,2,…,
Ti,j,…,Ti,N, wherein Ti,jW is represented for i-th of test portrait graph modeli SIt is represented with j-th of test photo graph model's
Similarity；
(5) similarity in similarity collection T is sorted from big to small, finds out maximum similarity Ti,hIf h is equal to i, unite
Counting parameter u increases by 1；
(6) step (4)-(5) are repeated, until having handled test portrait graph model represents collection WSIn all tests portrait graph model
It represents, the discrimination r of human face portrait-photo is calculated further according to following formula：
R=u/N.
2. according in claim 1 based on graph model represent human face portrait-photo array method, wherein described in step (4)
I-th of test portrait graph model is represented into Wi SCollection W is represented with test photo graph modelPIn each test photo graph model represent into
Row similarity calculation carries out as follows：
(3a) represents test portrait graph modelIt is represented with test photo graph modelThe graph model for calculating b-th of test portrait block according to the following formula representsWith b-th
The graph model for testing photo block representsSimilarity tb：
WhereinRepresent that graph model representsZ dimension element value,Represent that graph model representsZ dimension member
Element value, nzParameter in order to control works as element valueAnd element valueIt is all higher than n when 0.0001z1 is taken, otherwise nzTake 0, b
=1,2 ..., B, B are the total number of test portrait block；
(3b) repeats step (3a), until obtaining B similarity, calculates test portrait graph model according to the following formula and represents Wi SAnd test
Photo graph model representsSimilarity Ti,j：
(3c) repeats step (3a)-(3b), until having handled test photo graph model represents collection WPIn it is all test photo graph models
It represents, obtains the similarity collection T={ T that i-th of test portrait graph model representsi,1,Ti,2,…,Ti,j,…,Ti,N}。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510397326.7A CN105138951B (en) | 2015-07-08 | 2015-07-08 | Human face portrait-photo array the method represented based on graph model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510397326.7A CN105138951B (en) | 2015-07-08 | 2015-07-08 | Human face portrait-photo array the method represented based on graph model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105138951A CN105138951A (en) | 2015-12-09 |
| CN105138951B true CN105138951B (en) | 2018-05-25 |
Family
ID=54724295
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510397326.7A Active CN105138951B (en) | 2015-07-08 | 2015-07-08 | Human face portrait-photo array the method represented based on graph model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105138951B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105718889B (en) * | 2016-01-21 | 2019-07-16 | 江南大学 | Based on GB (2D)2The face personal identification method of PCANet depth convolution model |
| CN105608451B (en) * | 2016-03-14 | 2019-11-26 | 西安电子科技大学 | Human face portrait generation method based on subspace ridge regression |
| CN106412590B (en) * | 2016-11-21 | 2019-05-14 | 西安电子科技大学 | A kind of image processing method and device |
| CN108154133B (en) * | 2018-01-10 | 2020-04-14 | 西安电子科技大学 | Face portrait-photo recognition method based on asymmetric joint learning |
| CN109145704B (en) * | 2018-06-14 | 2022-02-22 | 西安电子科技大学 | Face portrait recognition method based on face attributes |
| CN110619777B (en) * | 2019-09-26 | 2021-08-27 | 重庆三原色数码科技有限公司 | Criminal investigation and experiment intelligent training and assessment system creation method based on VR technology |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101169830A (en) * | 2007-11-30 | 2008-04-30 | 西安电子科技大学 | Human face portrait automatic generation method based on embedded type hidden markov model and selective integration |
| EP2113865A1 (en) * | 2008-04-30 | 2009-11-04 | Siemens AG Österreich | Method for inspecting portrait photos |
-
2015
- 2015-07-08 CN CN201510397326.7A patent/CN105138951B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101169830A (en) * | 2007-11-30 | 2008-04-30 | 西安电子科技大学 | Human face portrait automatic generation method based on embedded type hidden markov model and selective integration |
| EP2113865A1 (en) * | 2008-04-30 | 2009-11-04 | Siemens AG Österreich | Method for inspecting portrait photos |
Non-Patent Citations (2)
| Title |
|---|
| Face photo-sketch synthesis and recognition;X Wang等;《IEEE》;20080912;第31卷(第11期);全文 * |
| 人脸画像-照片的合成与识别方法研究;肖冰;《中国博士学位论文全文数据库 信息科技辑》;20101015;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105138951A (en) | 2015-12-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105138951B (en) | Human face portrait-photo array the method represented based on graph model | |
| CN109344736B (en) | Static image crowd counting method based on joint learning | |
| CN107194341B (en) | Face recognition method and system based on fusion of Maxout multi-convolution neural network | |
| CN106897675B (en) | Face living body detection method combining binocular vision depth characteristic and apparent characteristic | |
| CN110443143B (en) | Multi-branch convolutional neural network fused remote sensing image scene classification method | |
| TWI753039B (en) | Image recognition method and device | |
| Bazzani et al. | Self-taught object localization with deep networks | |
| Rekha et al. | An efficient automated attendance management system based on Eigen Face recognition | |
| CN105095856B (en) | Face identification method is blocked based on mask | |
| WO2015101080A1 (en) | Face authentication method and device | |
| CN107085716A (en) | Across the visual angle gait recognition method of confrontation network is generated based on multitask | |
| CN104915673B (en) | A kind of objective classification method and system of view-based access control model bag of words | |
| CN110516616A (en) | A kind of double authentication face method for anti-counterfeit based on extensive RGB and near-infrared data set | |
| CN106228129A (en) | A kind of human face in-vivo detection method based on MATV feature | |
| CN104933414A (en) | Living body face detection method based on WLD-TOP (Weber Local Descriptor-Three Orthogonal Planes) | |
| CN105760833A (en) | Face feature recognition method | |
| Zeng et al. | Towards resolution invariant face recognition in uncontrolled scenarios | |
| CN101739555A (en) | Method and system for detecting false face, and method and system for training false face model | |
| Wyzykowski et al. | Level three synthetic fingerprint generation | |
| CN108108760A (en) | A kind of fast human face recognition | |
| CN109344709A (en) | A kind of face generates the detection method of forgery image | |
| CN107145841A (en) | A kind of low-rank sparse face identification method and its system based on matrix | |
| CN108154133A (en) | Human face portrait based on asymmetric combination learning-photo array method | |
| CN109614866A (en) | Method for detecting human face based on cascade deep convolutional neural networks | |
| Suvarnam et al. | Combination of CNN-GRU model to recognize characters of a license plate number without segmentation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |