CN110309727A - A kind of foundation of Building recognition model, Building recognition method and apparatus - Google Patents
A kind of foundation of Building recognition model, Building recognition method and apparatus Download PDFInfo
- Publication number
- CN110309727A CN110309727A CN201910500682.5A CN201910500682A CN110309727A CN 110309727 A CN110309727 A CN 110309727A CN 201910500682 A CN201910500682 A CN 201910500682A CN 110309727 A CN110309727 A CN 110309727A
- Authority
- CN
- China
- Prior art keywords
- building
- training
- sample
- construction characteristic
- convolutional neural
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention belongs to technical field of information processing, are related to foundation, the Building recognition method and apparatus of a kind of Building recognition model, in particular a kind of Building recognition method based on convolutional neural networks.The present invention chooses suitable APN training sample and is trained to convolutional neural networks, converts construction characteristic for deep layer or/and shallow-layer network characterization, forms Building recognition model;User inputs building query formulation into model, can obtain the construction characteristic vector of building query formulation and carry out judgement identification with distance between this construction characteristic vector with image in inquiry database, and result is fed back to user.Different from existing picture recognition technology, the present invention has carried out the feature extraction of specific area for building, reduces the difficulty of relevant parameter adjustment, and form perfect network based on convolutional Neural training and carry out Building recognition.And construction characteristic vector is optimized using HNSW algorithm, reduce computational complexity and calculates the time, so that Building recognition judgement is simple rapid.
Description
Technical field
The invention belongs to technical field of information processing, be related to the foundation of Building recognition model a kind of, Building recognition method and
Device, in particular a kind of Building recognition method based on convolutional neural networks.
Background technique
Building is as important one of man-made features, the extraction and identification of information closely bound up with public life
There is very big promotion to make the development such as numerical map, building information database, digital city modeling, virtual city, tourism
With.
The extraction of building information at present passes through pixel by comparing pixel distribution situation with identification traditional approach
Variance approximation find it is that may be present certain similar, this method accuracy is lower.Deep approach of learning is applied to count at present
According to identification field, such as recognition of face etc., in the case where data volume is bigger, deep learning rule has preferable accurate
Rate, but it is slow compared to traditional algorithm speed, meanwhile, also belong to blank in Building recognition field at present.So how to reduce correlation
The difficulty of parameter adjustment, reduces the difficulty of deep learning algorithm, trains more perfect network and carries out Building recognition, is then urgently
Problem to be solved.
Summary of the invention
It is an object of the invention to overcome above-mentioned the deficiencies in the prior art, a kind of foundation of Building recognition model is provided, is built
Recognition methods and device are built, Building recognition technology is intended to provide, reduce computational complexity and calculates the time, so that Building recognition
Judgement is simple rapidly, result is reliable.
First aspect of the present invention provides a kind of method for building up of Building recognition model, comprising:
Convolutional neural networks are trained using APN data sample;
The convolutional neural networks that training is completed are as Building recognition model:
It wherein, include training building query formulation, positive sample building picture and other architectural drawings in the APN data sample
Piece, the convolutional neural networks are;
It is described that convolutional neural networks are trained including being distinguished using Siamese network using APN data sample
Training building query formulation+positive sample building picture, training building query formulation+other building pictures information are received, model is carried out
Training.
Preferably, described be trained using APN data sample to convolutional neural networks includes:
Choose the APN training sample of setting quantity;
The APN training sample is input to convolutional neural networks and calculates generic features;
Construction characteristic is converted by generic features using full link neural computing;
The minimum error that training obtains APN training sample determines model building feature.
Preferably, the generic features include the feature in network shallow-layer and/or network deep layer, in the network shallow-layer
Feature includes image lines tendency, image inflection point, image fixed point feature, the feature in the network deep layer include image category,
Style.
Preferably, the vector algorithm of the determining model building feature be L (A, P, N)=max (| | f (A)-f (P) | | ^2-
||f(A)-f(N)||^2+α,0);
Wherein, L (A, P, N) is error brought by APN training sample;F (A), f (P), f (N) respectively indicate trained building
The training construction characteristic vector that query formulation, positive sample building picture, other building pictures obtain after training;α, 0 is positive number,
For increasing discrimination;
The feature vector of the f (A) and f (P) distance are less than the feature vector distance of f (A) and f (N), then training terminates,
Alternatively, the feature vector of the f (A) and f (P) distance are more than or equal to the feature vector distance of f (A) and f (N), then increase each figure
The quantity of the corresponding samples pictures of piece classification, to re-start training.
The second aspect of the present invention provides a kind of Building recognition method, comprising:
Obtain the building query formulation of user's input;
Image in the building query formulation and inquiry database is input in above-mentioned Building recognition model, is converted into respectively
Construction characteristic vector;
The construction characteristic vector of image in the construction characteristic vector and inquiry database of query formulation is built in Building recognition judgement
Between distance be less than threshold value when be same building, otherwise be different buildings.
Preferably, the Building recognition judgement optimizes construction characteristic vector using HNSW algorithm.
Third aspect of the present invention provides a kind of building knowledge that the method for building up using above-mentioned Building recognition model is established
The training device of other model, comprising:
Sample training library, the sample training library include APN data sample, including training building query formulation, positive sample are built
Build picture and other building pictures;
Training unit, for using sample training library training convolutional neural networks;
Model generation unit, the convolutional neural networks after completing training generate Building recognition model.
Preferably, the APN data sample is aided with handmarking by crawlers or handmarking or crawlers and obtains
?.
Preferably, the training unit includes:
Extraction unit of classifying is mentioned for using the Building recognition model extraction generic features using network shallow-layer feature
It takes;
Computing unit, for converting construction characteristic vector, the construction characteristic vector for the vector of the generic features
Algorithm be L (A, P, N)=max (| | f (A)-f (P) | | ^2- | | f (A)-f (N) | | ^2+ α, 0),
Wherein, L (A, P, N) is error brought by APN training sample;F (A), f (P), f (N) respectively indicate trained building
The training construction characteristic vector that query formulation, positive sample building picture, other building pictures obtain after training;α, 0 is positive number,
For increasing discrimination, i.e., if α >=0, max (α, 0)=α;If α < 0, max (α, 0)=0;
Judging unit, for judging whether convolutional neural networks reach sets requirement.
The 4th aspect of the present invention, provides a kind of Building recognition device, comprising:
Query formulation acquiring unit is built, for obtaining the building query formulation of user's input;
Computing unit is established for image in the building query formulation and inquiry database to be input to the training device
Building recognition model in, by calculate be converted into respective construction characteristic vector;
Database is inquired, for storing national famous structure information and its image;
Judging unit builds query formulation for carrying out identification judgement to building according to the construction characteristic vector being calculated
Construction characteristic vector with inquiry database in image construction characteristic vector between distance be less than threshold value when be identical building, it is on the contrary
For different buildings;
As a result output unit, for the result of the identification judgement to be supplied to the user.
The present invention chooses suitable APN training sample first and is trained to convolutional neural networks, by shallow-layer network characterization
It is converted into construction characteristic, forms Building recognition model;By image input in the building query formulation of user's input and inquiry database
Into model, the construction characteristic vector of building query formulation can be obtained and with the construction characteristic vector of this and image in inquiry database
Between distance carry out judgement identification, and result is fed back into user.Different from existing picture recognition technology, the present invention is directed to building
It has carried out the feature extraction of specific area, has reduced the difficulty of relevant parameter adjustment, and formed based on convolutional Neural training perfect
Network carries out Building recognition.And construction characteristic vector is optimized using HNSW algorithm, when reducing computational complexity and calculating
Between, so that Building recognition judgement is simple rapid.
Detailed description of the invention
Attached drawing is used to provide further understanding of the present invention, and constitutes part of specification, with reality of the invention
It applies example to be used to explain the present invention together, not be construed as limiting the invention.In the accompanying drawings:
Fig. 1 is an example flow of the training process of convolutional neural networks in the embodiment of the present invention;
Fig. 2 is an example flow for carrying out Building recognition in the embodiment of the present invention by building query formulation;
Fig. 3 is a kind of training device that Building recognition model is applied in the embodiment of the present invention;
Fig. 4 is a kind of Building recognition device in the embodiment of the present invention.
Specific embodiment
To make the objectives, technical solutions, and advantages of the present invention clearer, below in conjunction in the embodiment of the present invention
Attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is only
A part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art
Every other embodiment obtained without making creative work, shall fall within the protection scope of the present invention.
It should be mentioned that some exemplary embodiments are described as before exemplary embodiment is discussed in more detail further
The processing or method that flow chart is described.Although operations or step are described as the processing of sequence by flow chart, therein
Many operations can by concurrently, concurrently or simultaneously implement.In addition, the sequence of operations can be rearranged.When
When it operates completion, affiliated processing or step can be terminated, it is also possible to have the additional step being not included in attached drawing.
Affiliated processing or step can correspond to method, function, regulation, subprogram etc..
Embodiment gives Building recognition model convolutional neural networks model foundation, the training to convolutional neural networks, with
And the method based on convolutional neural networks progress Object identifying:
Embodiment 1
A kind of method for building up of Building recognition model, comprising:
Convolutional neural networks are trained using APN data sample;
The convolutional neural networks that training is completed are as Building recognition model:
Fig. 1 shows APN data sample to an example flow of the training process of convolutional neural networks.In the exemplary flow
In the various realizations of journey, each step can be deleted, combine or be divided into sub-step.The example flow may include preparation stage and instruction
Practice the stage.
In the preparation stage, need to prepare trained sample data, the training sample data includes a large amount of multichannel numbers
According to, such as the multichannel image sample of the thousands of orders of magnitude, and to mark the corresponding correct recognition result of each sample.This reality
It applies in example, includes training building query formulation, positive sample building picture and other building pictures in the APN data sample;
In the training stage, it is trained using known convolutional neural networks by the sample in input APN data sample.
It is described that convolutional neural networks are trained including receiving training respectively using Siamese network using APN data sample
Query formulation+positive sample building picture, training building query formulation+other building pictures information are built, model training is carried out.
Specifically, described be trained using APN data sample to convolutional neural networks includes the following steps 110~150:
In step 110, the APN training sample of setting quantity is chosen;
It should be noted that needing to calculate training one using deep learning to analyze the building in two different pictures
Whether the model of same building, this model of training needs a large amount of picture, and which good picture is marked to belong to same building,
Data volume more big data marks more accurate, and the precision for training the model come is higher.Usable crawlers are aided with manually,
Arrangement handles out good authority data, forms APN database.APN data sample in APN database includes that training building is looked into
Inquiry formula, positive sample building picture and other building pictures.
In the step 120, the APN training sample is input to convolutional neural networks and calculates generic features;Specifically, institute
Stating generic features includes the feature in network shallow-layer, the feature in the network shallow-layer include image lines tendency, image inflection point,
Image pinpoints feature, and the feature in the network deep layer includes image category, style.
In step 130, construction characteristic is converted by generic features using full link neural computing;
In step 140, the minimum error that training obtains APN training sample determines model building feature.
Specifically, the vector algorithm of the determining model building feature be L (A, P, N)=max (| | f (A)-f (P) | | ^2-
||f(A)-f(N)||^2+α,0);
Wherein, L (A, P, N) is error brought by APN training sample;F (A), f (P), f (N) respectively indicate trained building
The training construction characteristic vector that query formulation, positive sample building picture, other building pictures obtain after training;α, 0 is positive number,
For increasing discrimination;
The feature vector of the f (A) and f (P) distance are less than the feature vector distance of f (A) and f (N), then training terminates,
Alternatively, the feature vector of the f (A) and f (P) distance are more than or equal to the feature vector distance of f (A) and f (N), then increase each figure
The quantity of the corresponding samples pictures of piece classification, to re-start training.
Finally, when the loss calculated is sufficiently small or loss does not change for a long time in step 150,
Above-mentioned convolutional neural networks training process terminates with regard to this.Under normal conditions, terminate to have had read entire data when training
Collect up to a hundred times, Shang Qianci, or more.
It should be noted that when generic features vector pass through full Connection Neural Network when can be converted to construction characteristic to
Amount, this point is determined by loss function.Generic features vector by full Connection Neural Network can be converted into arbitrarily to
Amount, but loss function can allow the vector distance between identical building picture smaller, it means that full Connection Neural Network is necessary
Certain vectors that can represent the building are extracted from generic features, so that identical universal building vector is by the full connection mind
Through similar vector can be obtained when network.The vector is the numerical value of one group of regular length, the error, i.e. loss letter
Number determines fully-connected network, this step was completed in the training stage, and the parameter in full Connection Neural Network can be constantly updated, directly
Loss function can be minimized to it.Wherein, the parameter in full Connection Neural Network is determined how from generic features to meter
Calculate construction characteristic vector.
Embodiment 2
A kind of Building recognition method, comprising:
Obtain the building query formulation of user's input;
Construction characteristic vector obtains;
Building recognition judgement.
Fig. 2 shows input building query formulations to pass through an example of trained convolutional neural networks progress Building recognition
Process.In the various realizations of the example flow, each step can be deleted, combine or be divided into sub-step.
Specifically, the Building recognition method includes the following steps 210~250:
In step 210, the building query formulation of user's input is obtained;Specifically, the building query formulation is building picture,
It is also possible to the description of picture or certain a part of monolithic architecture picture.
In step 220~240, the building query formulation is input in training pattern, is mentioned by Building recognition model
Take, be converted into the construction characteristic vector of building query formulation;
In step 250, the construction characteristic vector of image in the construction characteristic vector and inquiry database of query formulation is built
Between distance be less than threshold value when be same building, otherwise be different buildings.Specifically, the Building recognition judgement uses HNSW algorithm
Construction characteristic vector is optimized.
It should be noted that the threshold value can be set at random, and carry out calculating judgement by calculating, optimal threshold is obtained.
Such as, different threshold values is randomly choosed, precision of its threshold value in test set is calculated, choosing that can allow test set precision highest
Threshold value.Wherein, test set can be APN data set, precision are as follows: the correct quantity/total quantity for distinguishing same building.
Embodiment 3
As shown in figure 3, a kind of training device using Building recognition model, comprising:
Sample training library, the sample training library include APN data sample, including training building query formulation, positive sample are built
Build picture and other building pictures;
Training unit, for using sample training library training convolutional neural networks;
Model generation unit, the convolutional neural networks after completing training generate Building recognition model.
Specifically, the APN data sample is aided with handmarking by crawlers or handmarking or crawlers and obtains
?.
Specifically, the training unit includes:
Extraction unit of classifying is mentioned for using the Building recognition model extraction generic features using network shallow-layer feature
It takes;
Computing unit, for converting construction characteristic vector, the construction characteristic vector for the vector of the generic features
Algorithm be L (A, P, N)=max (| | f (A)-f (P) | | ^2- | | f (A)-f (N) | | ^2+ α, 0),
Wherein, L (A, P, N) is error brought by APN training sample;F (A), f (P), f (N) respectively indicate trained building
The training construction characteristic vector that query formulation, positive sample building picture, other building pictures obtain after training;α, 0 is positive number,
For increasing discrimination;
Judging unit, for judging whether convolutional neural networks reach sets requirement.
Embodiment 4
As shown in figure 4, a kind of Building recognition device, comprising:
Query formulation acquiring unit is built, for obtaining the building query formulation of user's input;
Computing unit, for image in the building query formulation and inquiry database to be input to building for training device foundation
It builds in identification model, is converted into respective construction characteristic vector by calculating;
Database is inquired, for storing national famous structure information and its image;
Judging unit builds query formulation for carrying out identification judgement to building according to the construction characteristic vector being calculated
Construction characteristic vector with inquiry database in image construction characteristic vector between distance be less than threshold value when be identical building, it is on the contrary
For different buildings;
As a result output unit, for the result of the identification judgement to be supplied to the user.
Finally, it should be noted that the foregoing is only a preferred embodiment of the present invention, it is not intended to restrict the invention,
Although the present invention is described in detail referring to the foregoing embodiments, for those skilled in the art, still may be used
To modify the technical solutions described in the foregoing embodiments or equivalent replacement of some of the technical features.
All within the spirits and principles of the present invention, any modification, equivalent replacement, improvement and so on should be included in of the invention
Within protection scope.
Claims (10)
1. a kind of method for building up of Building recognition model characterized by comprising
Convolutional neural networks are trained using APN data sample;
The convolutional neural networks that training is completed are as Building recognition model:
It wherein, include training building query formulation, positive sample building picture and other building pictures in the APN data sample;
It is described that convolutional neural networks are trained including being received respectively using Siamese network using APN data sample
The information of training building query formulation and positive sample building picture, training building query formulation and other building pictures, carries out model instruction
Practice.
2. the method according to claim 1, wherein it is described using APN data sample to convolutional neural networks into
Row training include:
Choose the APN training sample of setting quantity;
The APN training sample is input to the convolutional neural networks and calculates generic features;
Construction characteristic is converted by generic features using full link neural computing;
The minimum error that training obtains APN training sample determines model building feature.
3. according to the method described in claim 2, it is characterized in that, the generic features include that network shallow-layer and/or network are deep
Feature in layer, the feature in the network shallow-layer include image lines tendency, image inflection point, image fixed point feature, the net
Feature in network deep layer includes image category, style.
4. according to the method described in claim 2, it is characterized in that, the vector algorithm of the determining model building feature be L (A,
P, N)=max (| | f (A)-f (P) | | ^2- | | f (A)-f (N) | | ^2+ α, 0);
Wherein, L (A, P, N) is error brought by APN training sample;F (A), f (P), f (N) respectively indicate training building inquiry
The training construction characteristic vector that formula, positive sample building picture, other building pictures obtain after training;α, 0 is positive number, is used for
Increase discrimination;
The feature vector of the f (A) and f (P) distance are less than the feature vector distance of f (A) and f (N), then training terminates, alternatively,
The feature vector of the f (A) and f (P) distance are more than or equal to the feature vector distance of f (A) and f (N), then increase each picture classification
The quantity of corresponding samples pictures, to re-start training.
5. a kind of Building recognition method characterized by comprising
Obtain the building query formulation of user's input;
Image in the building query formulation and inquiry database is input to and is built by method either described in Claims 1 to 4
In vertical Building recognition model, it is converted into respective construction characteristic vector;
The construction characteristic vector spacing of image in the construction characteristic vector and inquiry database of query formulation is built in Building recognition judgement
From being same building when being less than threshold value, on the contrary is different buildings.
6. according to the method described in claim 5, it is characterized in that, Building recognition judgement is special to building using HNSW algorithm
Sign vector optimizes.
7. a kind of training device for the Building recognition model established using method either described in Claims 1 to 4, feature exist
In, comprising:
Sample training library, the sample training library include APN data sample, including training building query formulation, positive sample architectural drawing
Piece and other building pictures;
Training unit, for using sample training library training convolutional neural networks;
Model generation unit, the convolutional neural networks after completing training generate Building recognition model.
8. training device according to claim 7, which is characterized in that the APN data sample passes through crawlers or people
Work label or crawlers are aided with handmarking's acquisition.
9. training device according to claim 7, which is characterized in that the training unit includes:
Classification extraction unit, for using the Building recognition model extraction generic features, using network shallow-layer feature extraction;
Computing unit, for converting construction characteristic vector, the calculation of the construction characteristic vector for the vector of the generic features
Method be L (A, P, N)=max (| | f (A)-f (P) | | ^2- | | f (A)-f (N) | | ^2+ α, 0),
Wherein, L (A, P, N) is error brought by APN training sample;F (A), f (P), f (N) respectively indicate training building inquiry
The training construction characteristic vector that formula, positive sample building picture, other building pictures obtain after training;α, 0 is positive number, is used for
Increase discrimination;
Judging unit, for judging whether convolutional neural networks reach sets requirement.
10. a kind of Building recognition device characterized by comprising
Query formulation acquiring unit is built, for obtaining the building query formulation of user's input;
Computing unit, for image in the building query formulation and inquiry database to be input to instruction described in claim 8 or 9
Practice in the Building recognition model that device is established, is converted into respective construction characteristic vector by calculating;
Database is inquired, for storing national famous structure information and its image;
Judging unit builds building for query formulation for carrying out identification judgement to building according to the construction characteristic vector being calculated
Building when distance is less than threshold value between the construction characteristic vector of image in feature vector and inquiry database is identical building, otherwise for not
With building;
As a result output unit, for the result of the identification judgement to be supplied to the user.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910500682.5A CN110309727B (en) | 2019-06-11 | 2019-06-11 | Building identification model establishing method, building identification method and building identification device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910500682.5A CN110309727B (en) | 2019-06-11 | 2019-06-11 | Building identification model establishing method, building identification method and building identification device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110309727A true CN110309727A (en) | 2019-10-08 |
CN110309727B CN110309727B (en) | 2022-02-18 |
Family
ID=68077097
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910500682.5A Active CN110309727B (en) | 2019-06-11 | 2019-06-11 | Building identification model establishing method, building identification method and building identification device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110309727B (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113269223A (en) * | 2021-03-16 | 2021-08-17 | 重庆市地理信息和遥感应用中心 | City style classification method based on spatial culture modular factorial analysis |
CN115394036A (en) * | 2022-08-24 | 2022-11-25 | 爱瑞克(大连)安全技术集团有限公司 | Monitoring and early warning method and system for building fire |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104899562A (en) * | 2015-05-29 | 2015-09-09 | 河南理工大学 | Texture segmentation and fusion based radar remote-sensing image artificial building recognition algorithm |
US20170140248A1 (en) * | 2015-11-13 | 2017-05-18 | Adobe Systems Incorporated | Learning image representation by distilling from multi-task networks |
US20180060699A1 (en) * | 2016-08-30 | 2018-03-01 | Imagry (Israel) Ltd. | System and method for image classification |
CN109002784A (en) * | 2018-06-29 | 2018-12-14 | 国信优易数据有限公司 | The training method and system of streetscape identification model, streetscape recognition methods and system |
CN109753928A (en) * | 2019-01-03 | 2019-05-14 | 北京百度网讯科技有限公司 | The recognition methods of architecture against regulations object and device |
-
2019
- 2019-06-11 CN CN201910500682.5A patent/CN110309727B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104899562A (en) * | 2015-05-29 | 2015-09-09 | 河南理工大学 | Texture segmentation and fusion based radar remote-sensing image artificial building recognition algorithm |
US20170140248A1 (en) * | 2015-11-13 | 2017-05-18 | Adobe Systems Incorporated | Learning image representation by distilling from multi-task networks |
US20180060699A1 (en) * | 2016-08-30 | 2018-03-01 | Imagry (Israel) Ltd. | System and method for image classification |
CN109002784A (en) * | 2018-06-29 | 2018-12-14 | 国信优易数据有限公司 | The training method and system of streetscape identification model, streetscape recognition methods and system |
CN109753928A (en) * | 2019-01-03 | 2019-05-14 | 北京百度网讯科技有限公司 | The recognition methods of architecture against regulations object and device |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113269223A (en) * | 2021-03-16 | 2021-08-17 | 重庆市地理信息和遥感应用中心 | City style classification method based on spatial culture modular factorial analysis |
CN113269223B (en) * | 2021-03-16 | 2022-04-22 | 重庆市地理信息和遥感应用中心 | City style classification method based on spatial culture modular factorial analysis |
CN115394036A (en) * | 2022-08-24 | 2022-11-25 | 爱瑞克(大连)安全技术集团有限公司 | Monitoring and early warning method and system for building fire |
Also Published As
Publication number | Publication date |
---|---|
CN110309727B (en) | 2022-02-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2020151528A1 (en) | Urban land automatic identification system integrating industrial big data and building forms | |
CN111553236B (en) | Road foreground image-based pavement disease target detection and example segmentation method | |
CN109697233A (en) | A kind of knowledge mapping system building method | |
CN108090499B (en) | Data active labeling method and system based on maximum information triple screening network | |
CN108171209A (en) | A kind of face age estimation method that metric learning is carried out based on convolutional neural networks | |
CN108960409A (en) | Labeled data generation method, equipment and computer readable storage medium | |
CN110135459A (en) | A kind of zero sample classification method based on double triple depth measure learning networks | |
CN108256431A (en) | A kind of hand position identification method and device | |
CN109214280A (en) | Shop recognition methods, device, electronic equipment and storage medium based on streetscape | |
CN109948522A (en) | A kind of X-ray hand osseous maturation degree interpretation method based on deep neural network | |
CN109284779A (en) | Object detecting method based on the full convolutional network of depth | |
CN101980211A (en) | Machine learning model and establishing method thereof | |
CN112016605A (en) | Target detection method based on corner alignment and boundary matching of bounding box | |
CN108805196A (en) | Auto-increment learning method for image recognition | |
CN112528934A (en) | Improved YOLOv3 traffic sign detection method based on multi-scale feature layer | |
CN112949517B (en) | Plant stomata density and opening degree identification method and system based on deep migration learning | |
CN110415815A (en) | The hereditary disease assistant diagnosis system of deep learning and face biological information | |
CN109086803A (en) | A kind of haze visibility detection system and method based on deep learning and the personalized factor | |
CN110309727A (en) | A kind of foundation of Building recognition model, Building recognition method and apparatus | |
CN111144462A (en) | Unknown individual identification method and device for radar signals | |
CN113269195A (en) | Reading table image character recognition method and device and readable storage medium | |
CN113792726B (en) | Method and system for rapidly generating POI (Point of interest) based on visual image | |
CN112700356B (en) | Method and system for managing information of training personnel for real-time online education and training | |
CN112348922B (en) | Method, system and device for automatically drawing logging curve and storage medium | |
CN114898215A (en) | Automatic arrangement method of sound barrier |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |