CN108171255A - Picture association intensity ratings method and device based on image identification - Google Patents
Picture association intensity ratings method and device based on image identification Download PDFInfo
- Publication number
- CN108171255A CN108171255A CN201711178165.8A CN201711178165A CN108171255A CN 108171255 A CN108171255 A CN 108171255A CN 201711178165 A CN201711178165 A CN 201711178165A CN 108171255 A CN108171255 A CN 108171255A
- Authority
- CN
- China
- Prior art keywords
- picture
- main body
- evaluated
- body object
- title
- 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
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- 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
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30168—Image quality inspection
Abstract
The invention discloses the picture association intensity ratings methods identified based on image, include the following steps：Establish article identification model；The picture of object all included in picture to be evaluated, and all main body objects according to included in the picture size of each object determines picture to be evaluated are obtained, and feature extraction is carried out to the picture of each main body object and obtains character pair vector；The title of each main body object is obtained according to the feature vector of each main body object and article identification model；It is obtained according to the title of each main body object and knowledge mapping there are the number of incidence relation between main body object, and then the scoring of picture to be evaluated is obtained according to the number there are incidence relation between the number of main body object and main body object.The present invention also provides a kind of electronic equipment and storage medium.The present invention avoids the differentiation artificially to score picture by the way that the main body object in picture and incidence relation are identified and are scored, and then obtain the appraisal result of picture.
Description
Technical field
The present invention relates to picture points-scoring system more particularly to a kind of picture association intensity ratings sides based on image identification
Method, electronic equipment and storage medium.
Background technology
At present, the design of intention picture, photography creation works are more and more, are also emerged on network large quantities of with intention subject matter
The picture share-type website that picture is the theme for material picture or various camera shooting subject matters.Then how these intention pictures are carried out
Judge, become one it is important the problem of, and existing evaluation method be usually artificially subjectively observe picture, and to picture
Several aspects such as form, performance, effect, idea carry out a comprehensive marking, grade to picture.And these are evaluated
Based on standard is all based on people to the correlation theory of above-mentioned various aspects or to the limited understanding of picture, picture is carried out subjective
Evaluation, the theory and the difference of social experiences grasped due to everyone, evaluation result also tend to it is not comprehensive enough, can because umpire's
Difference causes evaluation result also different.
Invention content
For overcome the deficiencies in the prior art, one of the objects of the present invention is to provide the picture associations identified based on image
Intensity ratings method, can solve in the prior art for the evaluation of picture because standard difference due to so that evaluation result exist compared with
The problem of big difference.
The second object of the present invention is to provide a kind of electronic equipment, can solve commenting for picture in the prior art
Valency is due to standard difference so that evaluation result has larger difference.
The third object of the present invention is to provide a kind of computer readable storage medium, and it is right in the prior art to solve
So that evaluation result has larger difference due to the evaluation of picture is because standard difference.
An object of the present invention adopts the following technical scheme that realization：
Based on the picture association intensity ratings method of image identification, include the following steps：
Model foundation step：Article identification model is established, the title of each object is stored in the article identification model
And the set of feature vector；
Main body object extraction step：The picture of object all included in picture to be evaluated is obtained, and according to each
The picture size of object determines main body object all included in picture to be evaluated；
Characteristic vector pickup step：The spy that feature extraction obtains each main body object is carried out to the picture of each main body object
Sign vector；
Title identification step：Each main body object is obtained according to the feature vector of each main body object and article identification model
The title of part；
Relation recognition step：Master is obtained according to pre-stored knowledge mapping in the title and system of each main body object
There are the numbers of incidence relation between body object；Wherein, knowledge mapping stores the incidence relation between various main body objects；
Score step：Exist between the number of all main body objects and main body object in picture to be evaluated and close
The number of connection relationship obtains the scoring of picture to be evaluated.
Further, pre-treatment step is further included：Preprocessing process is carried out to the picture of each main body object, wherein pre- place
Reason process includes one or more combinations of following methods：Image binaryzation, removal noise spot, barycenter alignment schemes and line
Property interpolation amplification method.
Further, the model foundation step specifically includes：The plurality of pictures of each object is obtained first and to every
Picture carries out the feature vector that feature extraction obtains every pictures；Then pass through the feature vector of the plurality of pictures to each object
Training is identified, and then obtains the set of multiple feature vectors of each object, that is to say article identification model.
Further, the picture size of each object of the basis determines main body object all included in picture to be evaluated
During part, be specially the size of picture of object in the picture to be evaluated and the ratio of entire picture size to be evaluated be more than or
During equal to preset value, then the object is a main body object included in picture to be evaluated.
Further, the preset value is 15%.
The second object of the present invention adopts the following technical scheme that realization：
A kind of electronic equipment can be run on a memory and on a processor including memory, processor and storage
Computer program realizes that the picture association intensity as previously described based on image identification is commented when the processor performs described program
The step of dividing method.
The third object of the present invention adopts the following technical scheme that realization：
A kind of computer readable storage medium, is stored thereon with computer program, and the computer program is held by processor
The step of picture association intensity ratings method as previously described based on image identification is realized during row.
Compared with prior art, the beneficial effects of the present invention are：
The present invention is by pre-establishing the article identification model corresponding to each object, then when evaluating picture
The technology identification identified first by image obtains the object in the picture of needs assessment, and obtains figure with reference to object identification model
Then the title of object included in piece obtains between object exist according to knowledge mapping pre-stored in system to match
The number of incidence relation finally obtains the final scoring of picture according to incidence relation existing between the object of picture and object
As a result.During due to scoring in the present invention for picture, scored according to the main body object in the presence of picture first,
Therefore there is no as in the prior art by people grasps the difference of knowledge and the difference of standards of grading and to the scoring of picture
As a result there is very big otherness.
Description of the drawings
Fig. 1 is picture schematic diagram provided by the invention；
Fig. 2 is the flow chart of the picture association intensity ratings method provided by the invention based on image identification；
Fig. 3 is the module map of the picture association intensity ratings device provided by the invention based on image identification.
Specific embodiment
In the following, with reference to attached drawing and specific embodiment, the present invention is described further, it should be noted that not
Under the premise of conflicting, new implementation can be formed between various embodiments described below or between each technical characteristic in any combination
Example.
Embodiment
The present invention is to apply to image recognition technology, crawler technology in identification and evaluation to picture, beforehand through
The identification model of various objects is established using machine learning and object and object are captured from internet by crawler technology
Between existing incidence relation and form knowledge mapping, while scoring is formulated according to object.
When scoring picture, processing identification is carried out to picture to be evaluated first and obtains what picture to be evaluated was included
Object obtains the title of each object included in picture to be evaluated then in conjunction with article identification Model Identification, subsequently root
The number to obtain existing incidence relation between object is matched in knowledge mapping according to the title of each object, and then
According to included in picture to be evaluated between the number of object and object the number of existing incidence relation obtain it is to be evaluated
The appraisal result of picture, solve the problems, such as it is existing think subjective marking not enough it is comprehensive comprehensively and there are subjectivities, have certain
Commercial application value.
When in use, picture to be evaluated can be uploaded in system by user, and system automatically carries out picture to be evaluated corresponding
Processing and obtain corresponding appraisal result.It is by collecting in advance corresponding to each object firstly, for article identification model
Picture training be identified obtain.It is as follows that it specifically establishes process：
The plurality of pictures of same object is obtained by approach such as internets first.The picture can be various angles, side
Picture under, scene.
Then every pictures are pre-processed and feature extraction is carried out to every pictures and obtain corresponding feature vector.By
It is different in assemblage characteristics such as the shapes, color or material of each object, therefore every pictures of each object can be carried out
Then the feature vector extracted is identified the trained feature vector obtained corresponding to the object in the extraction of feature vector
Set.In addition, in recognition training, it should be pointed out that the title of each object.Such as the automobile for a certain brand and model, by
It is different in the picture corresponding to different angles, when extracting feature vector, can obtain it is multiple, by largely identifying instruction
Practice, eventually obtain the set of the feature vector corresponding to the automobile of the brand and model.The automobile is pointed out in recognition training
Brand and model.
By the above-mentioned method for establishing article identification model, corresponding article identification can be established for a variety of different objects
Model.When carrying out match cognization to the object in picture, it will be able to obtain the title of object included in picture.
In addition, due to each object, such as its shape such as people, plant, animal, color, texture, shape, structure or material
Deng equal difference, then its body feature having is also different, therefore when establishing article identification model, needs for different masters
Body establishes corresponding characteristic model respectively.
For example corresponding characteristic model is established respectively for different animals, different famous persons establishes corresponding feature respectively
Model establishes corresponding characteristic model for different objects (automobile, the trade marks of such as various models).
It, specifically can be by acquiring the picture under the different angle of each object, light, by object when establishing identification model
Every pictures be divided into 25 grid spaces of 5*5, calculate the points of each grid spaces and article always the ratio between points, and then
25 dimensional feature vectors are obtained, and then obtain the set of the feature vector corresponding to each object, is i.e. identification corresponding to the object
Model.
And for the knowledge mapping in this system, Baidupedia of the existing crawler technology to internet, dimension can be passed through
The class label of the encyclopaedia of base encyclopaedia and other encyclopaedic knowledge contents extractions each entry therein and according between entry
Incidence relation obtain incidence relation between various objects.Wherein, entry that is to say the object in corresponding the present embodiment.Than
Such as, it can be famous person, vehicle, animal, plant, building etc..
For incidence relation, for example, between two entries have hyponymy, coordination, inclusion relation and
Other relationships that are mutually related etc..
It that is to say：Knowledge mapping includes the label relational network figure between object and object with multiple hierarchical relationships, than
It such as pre-processed, analyzed by the entry content of each object to acquisition, keyword label extraction, structure is comprising multiple
The label relational network figure of hierarchical relationship.For example, for famous person：Angstrom Long Masike is the CEO of Tesla (CS) Koncern, Podebradska 186, Praha 9, Czechoslovakia, then recognizes
For angstrom there is incidence relation between Long Masike and Tesla (CS) Koncern, Podebradska 186, Praha 9, Czechoslovakia；And Tesla (CS) Koncern, Podebradska 186, Praha 9, Czechoslovakia is production electric vehicle, then it is assumed that
Between Tesla (CS) Koncern, Podebradska 186, Praha 9, Czechoslovakia and electric vehicle have incidence relation, by the description of above-mentioned relation can also obtain angstrom Long Masike with
Also there is incidence relation between tesla's electric vehicle.
In addition, when scoring picture, be the object obtained according to above-mentioned identification number and object between deposit
Incidence relation give a mark, and then obtain the appraisal result of picture.For marking when, based on standard can be according to difference
Demand formulate its rule, for example picture intensity association evaluation mechanism can be used in the present embodiment, assume that object
The number of incidence relation between number and object has formulated corresponding scoring.Such as there are an objects in picture
Part then remembers 1 point；When there are during an incidence relation between the object in picture, it is believed that the incidence relation is denoted as 2 for a connection
Point；When the incidence relation between object in picture is there are at two, it is denoted as 4 points；And so on.
In scoring, standard pre-establishes based on it, and artificial subjective limitation is not present, because
This, larger difference will not be caused for the evaluation of same pictures, have certain reference value.
As shown in Fig. 2, a kind of image associative strength methods of marking based on image identification, specifically includes following steps：
S1, picture to be evaluated is obtained.
The picture of S2, object all according to included in picture to be evaluated obtains picture to be evaluated, and according to each
The picture size of object determines main body object all included in picture to be evaluated.
In addition, when object is identified to picture to be evaluated, due to its position for occupying of some images in a picture
Very little is put, it is little to the whole influence of picture, therefore do not need to account for these images when evaluating picture.Cause
This, distinguishes object included in picture to be evaluated, for example, according to the image size of object distinguished based on object
Part and secondary object.And main body object refers to occupy main leading or position object in picture to be evaluated, such as such as Fig. 1 institutes
Show, contain more subgraph in the picture, but wherein vehicle, people, crowd which occupies picture most of position, to figure
Piece plays absolute leading role, therefore when carrying out article identification to picture to be evaluated, it is only necessary to in picture people, vehicle,
Crowd is identified.
Refer to the image size when an object with entirely treating the main body object in picture in addition, the present invention also indicates that
When evaluating the ratio of the image size of picture more than certain preset value, it is believed that the object is the main body object of picture to be evaluated.
Preset value can be set as 15%.For example, it is identified in Fig. 1 there are three the main body objects obtained.
In the main body object in identifying picture to be evaluated, can be known by the image to the object in picture to be evaluated
It does not divide, for example the image size corresponding to each object is obtained by color cluster, boundary demarcation etc., then according to object
Image size judge whether the object is main body object in picture to be evaluated.
In addition, when including multiple main body objects in picture to be evaluated, several connections are obtained using seed fill algorithm
Line, the main body object of adhesion non-in this way, that is, divisible；And for adhesion main body object when, can be used and paddy is looked in vertical projection diagram
Main body object is split by the method for point.
S3, feature extraction is carried out to the picture of each main body object and obtains the feature vector of each main body object.
In addition, before characteristic vector pickup is carried out to the picture of each main body object, picture is pre-processed first
Process by being pre-processed to picture, can improve processing recognition performance of the server to picture.
For example rational threshold value is set according to specific picture analyzing, by image binaryzation, removal noise spot.Pass through barycenter pair
Neat method and the method for linear interpolation amplification normalize picture, are set as unified specification.
S4, the title that each main body object is obtained according to the feature vector and article identification model of each main body object.
The set of the title and feature vector corresponding to each object is stored in article identification model.In matching, by that will lead
The feature vector of body object is identified and then matches with each feature vector of each object in article identification model and obtains
Title corresponding to the main body object.
S5, it is obtained according to pre-stored knowledge mapping in the title and system of each main body object between main body object
There are the numbers of incidence relation.
S6, the number of all main body objects in picture to be evaluated and there are incidence relations between main body object
Number obtain the scoring of picture to be evaluated.
For example, for Fig. 1, show that main body object has three included in picture by picture being identified processing
It is a, and show that main body object is behaved according to article identification Model Matching：Angstrom Long Masike, vehicle：Tesla's electric vehicle, crowd,
Then obtained between people and vehicle there are incidence relation according to the matching of knowledge mapping, then according to picture association intensity ratings mechanism,
Existing incidence relation carries out marking and show that the appraisal result of the Fig. 1 is 5 between main body object and main body object.
Since the main body object to each picture is automatic identification, while the same standards of grading of score basis, it is right in this way
It is just comparable in the evaluation result of picture, there is opposite objectivity, it is not comprehensive enough to solve existing artificial subjective marking
Comprehensively and there are problems that subjectivity.
The present invention also provides a kind of electronic equipment, including memory, processor and storage on a memory and can
The computer program run in processing, realization is as described herein when the processor performs described program is identified based on image
Picture association intensity ratings method the step of.
The present invention also provides a kind of computer readable storage mediums, are stored thereon with computer program, computer program
The step of picture association intensity ratings method based on image identification as described herein is realized when being executed by processor.
As shown in figure 3, the present invention also provides the picture association intensity ratings device identified based on image, including：
For establishing article identification model, each object is stored in the article identification model for model building module
The set of title and feature vector；
Main body object extraction module, for obtaining the picture of object all included in picture to be evaluated, and according to
The picture size of each object determines main body object all included in picture to be evaluated；
Characteristic vector pickup module obtains each main body object for carrying out feature extraction to the picture of each main body object
Feature vector；
Title identification module obtains each master for the feature vector according to each main body object and article identification model
The title of body object；
Relation recognition module obtains for pre-stored knowledge mapping in the title and system according to each main body object
Go out between main body object that there are the numbers of incidence relation；Wherein, the association that knowledge mapping is stored between various main body objects is closed
System；
Grading module, for being deposited between the number of all main body objects in picture to be evaluated and main body object
The scoring of picture to be evaluated is obtained in the number of incidence relation.
Further, preprocessing module is further included, for carrying out preprocessing process to the picture of each main body object, wherein
Preprocessing process includes one or more combinations of following methods：Image binaryzation, removal noise spot, barycenter alignment schemes with
And linear interpolation amplification method.
Present embodiment is only the preferred embodiment of the present invention, it is impossible to the scope of protection of the invention is limited with this, this
The variation and replacement for any unsubstantiality that the technical staff in field is done on the basis of the present invention belong to the present invention and want
Seek the range of protection.
Claims (7)
1. the picture association intensity ratings method based on image identification, it is characterised in that include the following steps：
Model foundation step：Establish article identification model, stored in the article identification model each object title and
The set of feature vector；
Main body object extraction step：The picture of object all included in picture to be evaluated is obtained, and according to each object
Picture size determine main body object all included in picture to be evaluated；
Characteristic vector pickup step：The picture of each main body object is carried out feature extraction obtain the feature of each main body object to
Amount；
Title identification step：Each main body object is obtained according to the feature vector of each main body object and article identification model
Title；
Relation recognition step：Main body object is obtained according to pre-stored knowledge mapping in the title and system of each main body object
There are the numbers of incidence relation between part；Wherein, knowledge mapping stores the incidence relation between various main body objects；
Score step：There is association between the number of all main body objects and main body object in picture to be evaluated to close
The number of system obtains the scoring of picture to be evaluated.
2. the method as described in claim 1, it is characterised in that：Further include pre-treatment step：To the picture of each main body object
Preprocessing process is carried out, wherein preprocessing process includes one or more combinations of following methods：Image binaryzation, removal are dry
Disturb point, barycenter alignment schemes and linear interpolation amplification method.
3. the method as described in claim 1, it is characterised in that：The model foundation step specifically includes：Obtain each first
The plurality of pictures of object simultaneously carries out every pictures the feature vector that feature extraction obtains every pictures；Then by each object
Training is identified in the feature vector of the plurality of pictures of part, so obtain multiple feature vectors of each object set namely
It is article identification model.
4. the method as described in claim 1, it is characterised in that：The picture size of each object of basis determines picture to be evaluated
Included in all main body objects when, be specially when the picture of the object in picture to be evaluated size with it is entirely to be evaluated
When the ratio of valency picture size is greater than or equal to preset value, then the object is a main body object included in picture to be evaluated
Part.
5. method as claimed in claim 4, it is characterised in that：The preset value is 15%.
6. a kind of electronic equipment including memory, processor and stores the meter that can be run on a memory and on a processor
Calculation machine program, it is characterised in that：The processor realizes the base as described in any one of claim 1-5 when performing described program
In image identification picture association intensity ratings method the step of.
7. a kind of computer readable storage medium, is stored thereon with computer program, it is characterised in that：The computer program quilt
The picture association intensity ratings method based on image identification as described in any one of claim 1-5 is realized when processor performs
The step of.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711178165.8A CN108171255A (en) | 2017-11-22 | 2017-11-22 | Picture association intensity ratings method and device based on image identification |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711178165.8A CN108171255A (en) | 2017-11-22 | 2017-11-22 | Picture association intensity ratings method and device based on image identification |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108171255A true CN108171255A (en) | 2018-06-15 |
Family
ID=62527423
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711178165.8A Pending CN108171255A (en) | 2017-11-22 | 2017-11-22 | Picture association intensity ratings method and device based on image identification |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108171255A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109597851A (en) * | 2018-09-26 | 2019-04-09 | 阿里巴巴集团控股有限公司 | Feature extracting method and device based on incidence relation |
| CN110210406A (en) * | 2019-06-04 | 2019-09-06 | 北京字节跳动网络技术有限公司 | Method and apparatus for shooting image |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1701343A (en) * | 2002-09-20 | 2005-11-23 | 德克萨斯大学董事会 | Computer program products, systems and methods for information discovery and relational analyses |
| WO2008134588A1 (en) * | 2007-04-25 | 2008-11-06 | Counsyl, Inc. | Methods and systems of automatic ontology population |
| CN105574468A (en) * | 2014-10-08 | 2016-05-11 | 深圳中兴力维技术有限公司 | Video flame detection method, device and system |
| CN106663221A (en) * | 2014-08-19 | 2017-05-10 | 高通股份有限公司 | Knowledge-graph biased classification for data |
| CN106855944A (en) * | 2016-12-22 | 2017-06-16 | 浙江宇视科技有限公司 | Pedestrian's Marker Identity method and device |
| CN106991673A (en) * | 2017-05-18 | 2017-07-28 | 深思考人工智能机器人科技（北京）有限公司 | A kind of cervical cell image rapid classification recognition methods of interpretation and system |
| CN107054937A (en) * | 2017-03-23 | 2017-08-18 | 广东数相智能科技有限公司 | A kind of refuse classification suggestion device and system based on image recognition |
| CN107273106A (en) * | 2016-04-08 | 2017-10-20 | 北京三星通信技术研究有限公司 | Object information is translated and derivation information acquisition methods and device |
| CN107290342A (en) * | 2017-05-09 | 2017-10-24 | 广东数相智能科技有限公司 | A kind of timber varieties of trees classification discrimination method and system based on cell analysis |
| CN107358596A (en) * | 2017-04-11 | 2017-11-17 | 阿里巴巴集团控股有限公司 | A kind of car damage identification method based on image, device, electronic equipment and system |
-
2017
- 2017-11-22 CN CN201711178165.8A patent/CN108171255A/en active Pending
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1701343A (en) * | 2002-09-20 | 2005-11-23 | 德克萨斯大学董事会 | Computer program products, systems and methods for information discovery and relational analyses |
| WO2008134588A1 (en) * | 2007-04-25 | 2008-11-06 | Counsyl, Inc. | Methods and systems of automatic ontology population |
| CN106663221A (en) * | 2014-08-19 | 2017-05-10 | 高通股份有限公司 | Knowledge-graph biased classification for data |
| CN105574468A (en) * | 2014-10-08 | 2016-05-11 | 深圳中兴力维技术有限公司 | Video flame detection method, device and system |
| CN107273106A (en) * | 2016-04-08 | 2017-10-20 | 北京三星通信技术研究有限公司 | Object information is translated and derivation information acquisition methods and device |
| CN106855944A (en) * | 2016-12-22 | 2017-06-16 | 浙江宇视科技有限公司 | Pedestrian's Marker Identity method and device |
| CN107054937A (en) * | 2017-03-23 | 2017-08-18 | 广东数相智能科技有限公司 | A kind of refuse classification suggestion device and system based on image recognition |
| CN107358596A (en) * | 2017-04-11 | 2017-11-17 | 阿里巴巴集团控股有限公司 | A kind of car damage identification method based on image, device, electronic equipment and system |
| CN107290342A (en) * | 2017-05-09 | 2017-10-24 | 广东数相智能科技有限公司 | A kind of timber varieties of trees classification discrimination method and system based on cell analysis |
| CN106991673A (en) * | 2017-05-18 | 2017-07-28 | 深思考人工智能机器人科技（北京）有限公司 | A kind of cervical cell image rapid classification recognition methods of interpretation and system |
Non-Patent Citations (2)
| Title |
|---|
| ZHILING LUO 等: "Deep Learning of Graphs with Ngram Convolutional Neural Networks", 《IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING》 * |
| 白亚龙: "基于深度神经网络的图像识别系统的研究与改进", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109597851A (en) * | 2018-09-26 | 2019-04-09 | 阿里巴巴集团控股有限公司 | Feature extracting method and device based on incidence relation |
| CN110210406A (en) * | 2019-06-04 | 2019-09-06 | 北京字节跳动网络技术有限公司 | Method and apparatus for shooting image |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104572804B (en) | A kind of method and its system of video object retrieval | |
| CN105518744B (en) | Pedestrian recognition methods and equipment again | |
| CN104933416B (en) | Micro- expression sequence characteristic extracting method based on optical flow field | |
| CN110135282B (en) | Examinee return plagiarism cheating detection method based on deep convolutional neural network model | |
| CN106033435B (en) | Item identification method and device, indoor map generation method and device | |
| CN106462771A (en) | 3D image significance detection method | |
| CN108492294B (en) | Method and device for evaluating harmony degree of image colors | |
| CN106610969A (en) | Multimodal information-based video content auditing system and method | |
| CN103218832B (en) | Based on the vision significance algorithm of global color contrast and spatial distribution in image | |
| CN110232379A (en) | A kind of vehicle attitude detection method and system | |
| CN110133443B (en) | Power transmission line component detection method, system and device based on parallel vision | |
| CN110188835A (en) | Data based on production confrontation network model enhance pedestrian's recognition methods again | |
| CN109785400B (en) | Silhouette image manufacturing method and device, electronic equipment and storage medium | |
| CN108470178B (en) | Depth map significance detection method combined with depth credibility evaluation factor | |
| CN107944459A (en) | A kind of RGB D object identification methods | |
| CN110428412A (en) | The evaluation of picture quality and model generating method, device, equipment and storage medium | |
| CN109191460A (en) | A kind of quality evaluating method for tone mapping image | |
| CN111414948B (en) | Target object detection method and related device | |
| CN109741268A (en) | A kind of breakage image complementing method for mural painting | |
| CN104484347B (en) | A kind of stratification Visual Feature Retrieval Process method based on geography information | |
| CN107909072A (en) | A kind of vegetation type recognition methods, electronic equipment, storage medium and device | |
| CN109583498A (en) | A kind of fashion compatibility prediction technique based on low-rank regularization feature enhancing characterization | |
| CN107066979A (en) | A kind of human motion recognition method based on depth information and various dimensions convolutional neural networks | |
| CN108875505A (en) | Pedestrian neural network based recognition methods and device again | |
| CN108171255A (en) | Picture association intensity ratings method and device based on image identification |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180615 |
|
| RJ01 | Rejection of invention patent application after publication |