CN108257115A - Image enhancement detection method and system based on orientation consistency convolutional neural networks - Google Patents
Image enhancement detection method and system based on orientation consistency convolutional neural networks Download PDFInfo
- Publication number
- CN108257115A CN108257115A CN201711484389.1A CN201711484389A CN108257115A CN 108257115 A CN108257115 A CN 108257115A CN 201711484389 A CN201711484389 A CN 201711484389A CN 108257115 A CN108257115 A CN 108257115A
- Authority
- CN
- China
- Prior art keywords
- image
- neural networks
- convolutional neural
- image enhancement
- orientation consistency
- 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
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 57
- 238000001514 detection method Methods 0.000 title claims abstract description 28
- 238000012360 testing method Methods 0.000 claims abstract description 44
- 238000000034 method Methods 0.000 claims abstract description 18
- 238000012549 training Methods 0.000 claims abstract description 13
- 238000013135 deep learning Methods 0.000 claims description 23
- 239000011248 coating agent Substances 0.000 claims description 8
- 238000000576 coating method Methods 0.000 claims description 8
- 238000011478 gradient descent method Methods 0.000 claims description 6
- 230000002708 enhancing effect Effects 0.000 claims description 5
- 230000006870 function Effects 0.000 claims description 5
- 238000012544 monitoring process Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 230000008859 change Effects 0.000 claims description 2
- 230000007547 defect Effects 0.000 abstract description 2
- 238000010586 diagram Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 230000006872 improvement Effects 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- 230000001537 neural effect Effects 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
Classifications
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
Landscapes
- Engineering & Computer Science (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses the image enhancement detection methods based on orientation consistency convolutional neural networks, include the following steps:A width testing image is selected, and is cut to fixed dimension, selection center is cut and cuts;Image after cutting is input to trained based on orientation consistency convolutional neural networks model in advance, calculates testing image and pass through image enhancement operation and the probability without image enhancement operation;Compare testing image by image enhancement operation and the probability size without image enhancement operation, it is final to judge testing image whether by image enhancement operation.The invention also discloses the image enhancement detecting system based on orientation consistency convolutional neural networks, including acquisition module, computing module, judge templet.The present invention collects evidence for specific image enhancement operation, realizes higher image detection rate, and it is time-consuming and laborious and the defects of easily cause over-fitting to solve existing training method.
Description
Technical field
The present invention relates to image forensics field, more particularly, to the image based on orientation consistency convolutional neural networks
Enhance detection method and system.
Background technology
With the arrival in multimedia messages epoch, the high speed of digital device and image processing tool is popularized, and is on the one hand accelerated
The progress of Digital image technology, brings great convenience for people’s lives;On the other hand so that digital picture is easier
It is tampered, reduces the safety and reliability of image.Accordingly, it is determined that the primary source of image and checking image content
Authenticity and integrity is particularly important.In recent years, image forensics rapid technological improvement is based particularly on convolutional neural networks
Distorted image detection technique, greatly improve the accuracys rate of image forensics.It is however existing based on convolutional neural networks
Distorted image detection technique is required for training quantity of parameters greatly, and this requires we must pass through a huge image data base
It trains, this method is time-consuming and laborious, also easily cause over-fitting.The present invention proposes a kind of convolution god based on orientation consistency
It through network, collects evidence for specific image enhancement operation, realizes higher image detection rate.
Invention content
Present invention aim to address said one or multiple defects, propose a kind of based on orientation consistency convolutional Neural net
The image enhancement detection method and system of network
For realization more than goal of the invention, the technical solution adopted is that:
Image enhancement detection method based on orientation consistency convolutional neural networks, includes the following steps:
S1:A width testing image is selected, and is cut to fixed dimension, selection center is cut and cuts;
S2:Image after cutting is input to trained in advance based on orientation consistency convolutional neural networks model, meter
It calculates testing image and passes through image enhancement operation and the probability without image enhancement operation;
S3:Compare testing image by image enhancement operation and the probability size without image enhancement operation, finally sentence
Whether disconnected testing image passes through image enhancement operation.
Preferably, fixed dimension described in step S1 is 256*256.
It is trained based on orientation consistency convolutional neural networks model in advance that wherein step S2 further includes acquisition, including with
Lower step:
S2.1:Based on the supervised learning method of label label, the original image in image data base is labeled as label 0,
1 is labeled as by the correspondence image of image enhancement;
S2.2:Deep learning training is carried out to the label image, obtains corresponding to the label original image and process
The characteristic information of picture after image enhancement operation;
S2.3:The characteristic information is trained using stochastic gradient descent method, obtain be corresponding to monitoring image
The no deep learning network model by enhancing operation.
Wherein described deep learning network model is 5 layers of structure convolutional neural networks, including convolutional layer, normalization layer, is swashed
Work layer, pond layer and full articulamentum;
Wherein convolutional layer includes general convolutional layer, restrictive convolutional layer and orientation consistency convolutional layer, restrictive convolution
Each position weights pass through the following formula hard constraints in layer:
W (0.0)=- 1, ∑L, m ≠ 0W (l, m)=1;
Active coating is the active coating with parameter;
Pond layer includes maximum pond layer and average pond layer.
Wherein described convolutional neural networks further include loss function, and wherein loss function is the classification damage of overall training sample
It loses.
Image enhancement detecting system based on orientation consistency convolutional neural networks, including acquisition module, computing module and
Judgment module;
Wherein acquisition module is used to testing image being cut into the image of particular size;
Computing module is used to, using the advance trained convolutional neural networks model based on orientation consistency, calculate to be measured
Image is the probability of original image and the probability after image enhancement operation;
For judgment module for comparing probability, it is original image or after image enhancement operation to judge the testing image
Image.
The acquisition module includes acquiring unit;Wherein acquiring unit is cut by center, obtained for inputting testing image
To the image of fixed size.
The computing module network establishes unit and computing unit;
Wherein network establishes unit for establishing deep learning network model;Based on label label supervised learning method,
By the original image in image data base labeled as label 0, by the correspondence picture of image enhancement labeled as label 1, and utilize
Deep learning is carried out to the label image based on the convolutional neural networks of orientation consistency, obtains corresponding original image and warp
The feature of image after image enhancement operation is crossed, the characteristic information is trained using stochastic gradient descent method, is obtained pair
The deep learning network model of target is detected described in Ying Yu;
Computing unit is used to utilize trained orientation consistency convolutional neural networks model, and it is original to obtain testing image
The probability of image and be probability after image enhancement operation.
Compared with prior art, the beneficial effects of the invention are as follows:
The present invention provides image enhancement detection method and system based on orientation consistency convolutional neural networks, the sides of passing through
Deep learning is carried out to the convolutional neural networks of consistency, it is more to overcome conventional depth learning network parameter, easy over-fitting
Problem has higher Detection accuracy.
Description of the drawings
Fig. 1 is the flow chart of the present invention;
Fig. 2 is flow chart of the training of the present invention based on orientation consistency convolutional neural networks method;
Fig. 3 is based on orientation consistency convolutional neural networks schematic diagram to be of the present invention;
Fig. 4 is constraint convolutional neural networks layer schematic diagram of the present invention;
Fig. 5 is based on orientation consistency convolutional neural networks layer schematic diagram to be of the present invention;
Fig. 6 is that the structure of the image enhancement detection device of the present invention based on orientation consistency convolutional neural networks is shown
It is intended to.
Specific embodiment
The attached figures are only used for illustrative purposes and cannot be understood as limitating the patent;
Below in conjunction with drawings and examples, the present invention is further elaborated.
Embodiment 1
Image enhancement inspection provided in an embodiment of the present invention based on orientation consistency convolutional neural networks shown in Figure 1
Method flow diagram is surveyed, described method specifically comprises the following steps:
Step S1:A width testing image is selected, is cut into 256*256 sizes, selection center is cut and cuts out;
Specifically, in view of picture to be measured there may be different sizes, thus selection cut out with center be cut into it is fixed
256*256 sizes.The mode that selection center is cut out is the interference brought in order to avoid some edge effects, improves inspection to greatest extent
Survey accuracy rate.
The target detection image of particular size got by the operation of above-mentioned steps S1, S2 is calculated as follows
Image passes through probability whether image enhancement operation.
Step S2:Input an image into trained based on orientation consistency convolutional neural networks model in advance, calculating is treated
Altimetric image passes through image enhancement operation and the probability without image enhancement operation.
Specifically, with the progress of artificial intelligence field technology, deep learning it is more and more extensive be used in each neck
Domain, but data volume is big, parameter is more, easy over-fitting, is always deep learning method urgent problem to be solved.Only excellent depth
Learning network model, articulate model, could fully excavate the information in data in other words.So in present example
Be the improvement in original conventional depth learning network based on orientation consistency convolutional neural networks, by using direction one
The feature not influenced in cause property convolutional layer extraction picture by picture direction, and then more effective feature can be extracted, reduce parameter
Amount avoids the over-fitting in training process, improves classification accuracy.
By above-mentioned steps S2 operation obtain testing image whether the probability Jing Guo image enhancement operation, by as follows
Whether step S3 finally to judge image by image enhancement operation.
Step S3:Compare testing image by image enhancement operation and the probability without image enhancement operation, finally sentence
Whether disconnected testing image passes through image enhancement operation.
Specifically, after probability whether testing image is obtained by image enhancement operation, compare the size of probability, if
It is judged as that the probability of original image is more than the probability Jing Guo image enhancement operation, testing image is judged as original image by us,
Without image enhancement operation.Conversely, testing image is judged as the image by image enhancement operation by we, it is concluded that.
Image enhancement detection method provided in an embodiment of the present invention based on orientation consistency convolutional neural networks, it is and existing
Compared using the image enhancement detection method of deep learning, select a width testing image first, it is big to be cut into 256*256
It is small, then the image of particular size is input to trained in advance based on orientation consistency convolutional neural networks model, calculating
Testing image passes through image enhancement operation and the probability without image enhancement operation.Increase by comparing testing image by image
Strong operation and the probability without image enhancement operation, it is final to judge that testing image whether by image enhancement operation, uses
Orientation consistency convolutional neural networks structure, effectively reduces training parameter and data volume, avoids over-fitting, relatively significantly improves
Detection accuracy.
During whether testing image is obtained by image enhancement operation probability, it is necessary first to which acquisition trains in advance
Based on orientation consistency convolutional neural networks model, referring to Fig. 2, the acquisition process of above-mentioned neural network model specifically includes:
Step S2.1:Based on the supervised learning method of label label, by the original image in image data base labeled as mark
Label 0 are labeled as 1 by the correspondence image of image enhancement.
Step S2.2:Deep learning training is carried out to the label image using this network, obtains corresponding to the label
Original image and after image enhancement operation picture characteristic information.
Step S2.3:The characteristic information is trained using stochastic gradient descent method, obtains corresponding to monitoring figure
It seem the no deep learning network model by enhancing operation.
Specifically, the deep learning network model in the embodiment of the present invention is using based on orientation consistency convolutional Neural
Network, concrete structure is referring to Fig. 3, including convolutional layer, normalization layer, active coating, pond layer, full articulamentum.Wherein convolutional layer packet
Include general convolutional layer, restrictive convolutional layer and orientation consistency convolutional layer, restrictive convolutional layer structure can join figure four in detail, in layer
Each position weights pass through the following formula hard constraints:
W (0.0)=- 1,
∑L, m ≠ 0W (l, m)=1,
By the constraint to convolution filter different location weights, picture material can be reduced to greatest extent to result
It influences, optimizes network structure, improve accuracy rate.Orientation consistency convolutional layer structure joins Fig. 5 in detail, it is contemplated that deep learning was trained
Ignore the content information of image in journey to greatest extent, the weights of each position of convolution filter meet central symmetry and axial symmetry,
It is so more effective that extract information needed, over-fitting is utmostly avoided, reduces parameter amount.Active coating selects the activation with parameter
Layer, the active coating with parameter is the enhanced edition of conventional activation layer, can effectively improve the accuracy rate of method.Pond layer includes maximum pond
Change layer and average pond layer, its use in a network is selected according to experimental result.Convolutional Neural net based on orientation consistency
Network includes six groupings, and defines the Classification Loss that loss function is overall training sample.
After the advance trained convolutional neural networks based on orientation consistency are obtained, testing image is put into network
Be detected, obtain testing image whether the probability Jing Guo image enhancement operation, finally judge it for original image or process
The image of image enhancement operation.
The embodiment of the present invention additionally provides a kind of image detecting system based on orientation consistency convolutional neural networks, described
System is used for the method for performing the above-mentioned image enhancement detection based on orientation consistency convolutional neural networks, referring to Fig. 6, the system
System includes:
Acquisition module:For testing image to be cut out to the image for 256*256 particular sizes;
Computing module, for using the advance trained convolutional neural networks model based on orientation consistency, calculating to be treated
Altimetric image is the probability of original image and is probability after image enhancement operation;
Judge templet, for comparing probability, it is original image or by image enhancement operation to judge the testing image
Image afterwards.
Specifically, an acquiring unit is included in above-mentioned acquisition module:
Acquiring unit:It inputs testing image to carry out, is cut out by center, obtain the image of 256*256 fixed sizes.
After testing image is cut out, need to calculate probability whether image passes through image enhancement operation, therefore, this
What inventive embodiments provided further includes computing module based on orientation consistency convolutional neural networks device, which includes:
Network establishes unit and computing unit.Wherein:
Network establishes unit:Based on the supervised learning method of label label, the original image in image data base is marked
For label 0, by the correspondence picture of image enhancement labeled as label 1, a large amount of pairs of training images are obtained, using based on direction
The convolutional neural networks mould of consistency carries out deep learning to the label image, obtains corresponding original image and by image
The feature of image, is trained the characteristic information using stochastic gradient descent method after enhancing operation, obtains corresponding to institute
State the deep learning network model of detection target;
Computing unit:For utilizing trained orientation consistency convolutional neural networks model, it is former to obtain testing image
The probability of beginning image and be probability after image enhancement operation.
After probability whether testing image is obtained by image enhancement operation, need to treat mapping by the judgement of probability size
Seem it is no by image enhancement operation, it is therefore, provided in an embodiment of the present invention to be based on orientation consistency convolutional neural networks device
Judgment module is further included, which includes a judging unit:
Judging unit:For comparing probability, it is original image or by image enhancement operation to judge the testing image
Image afterwards.
Image enhancement detection device provided in an embodiment of the present invention based on orientation consistency convolutional neural networks, it is and existing
Compared using the image enhancement detection method of deep learning, select a width testing image first, it is big to be cut into 256*256
It is small, then the image of particular size is input to trained in advance based on orientation consistency convolutional neural networks model, calculating
Testing image passes through image enhancement operation and the probability without image enhancement operation.Increase by comparing testing image by image
Strong operation and the probability without image enhancement operation, it is final to judge that testing image whether by image enhancement operation, uses
Orientation consistency convolutional neural networks structure, effectively reduces training parameter and data volume, avoids over-fitting, relatively significantly improves
Detection accuracy.
Obviously, the above embodiment of the present invention be only to clearly illustrate example of the present invention, and not be pair
The restriction of embodiments of the present invention.For those of ordinary skill in the art, may be used also on the basis of the above description
To make other variations or changes in different ways.There is no necessity and possibility to exhaust all the enbodiments.It is all this
All any modification, equivalent and improvement made within the spirit and principle of invention etc., should be included in the claims in the present invention
Protection domain within.
Claims (8)
1. the image enhancement detection method based on orientation consistency convolutional neural networks, which is characterized in that include the following steps:
S1:A width testing image is selected, and is cut to fixed dimension, selection center is cut and cuts;
S2:Image after cutting is input to trained based on orientation consistency convolutional neural networks model in advance, calculating is treated
Altimetric image passes through image enhancement operation and the probability without image enhancement operation;
S3:Compare testing image by image enhancement operation and the probability size without image enhancement operation, it is final to judge to treat
Whether altimetric image passes through image enhancement operation.
2. the image enhancement detection method according to claim 1 based on orientation consistency convolutional neural networks, feature
It is, fixed dimension described in step S1 is 256*256.
3. the image enhancement detection method according to claim 1 based on orientation consistency convolutional neural networks, feature
It is, it is trained based on orientation consistency convolutional neural networks model in advance that step S2 further includes acquisition, includes the following steps:
S2.1:Based on the supervised learning method of label label, by the original image in image data base labeled as label 0, pass through
The correspondence image of image enhancement is labeled as 1;
S2.2:Deep learning training is carried out to the label image, obtains corresponding to the label original image and by image
The characteristic information of picture after enhancing operation;
S2.3:The characteristic information is trained using stochastic gradient descent method, obtains whether passing through corresponding to monitoring image
Cross the deep learning network model of enhancing operation.
4. the image enhancement detection method according to claim 3 based on orientation consistency convolutional neural networks, feature
It is, the deep learning network model is 5 layers of structure convolutional neural networks, including convolutional layer, normalization layer, active coating, pond
Change layer and full articulamentum;
Wherein convolutional layer includes general convolutional layer, restrictive convolutional layer and orientation consistency convolutional layer, in restrictive convolutional layer
Each position weights pass through the following formula hard constraints:
W (0.0)=- 1, ∑L, m ≠ 0W (l, m)=1;
Active coating is the active coating with parameter;
Pond layer includes maximum pond layer and average pond layer.
5. the image enhancement detection method according to claim 4 based on orientation consistency convolutional neural networks, peculiar sign
It is, the convolutional neural networks further include loss function, and wherein loss function is the Classification Loss of overall training sample.
6. according to claim 1-5 any one of them systems, which is characterized in that including acquisition module, computing module and judgement
Module;
Wherein acquisition module is used to testing image being cut into the image of particular size;
Computing module is used to, using the advance trained convolutional neural networks model based on orientation consistency, calculate testing image
It is the probability of original image and the probability after image enhancement operation;
For judgment module for comparing probability, it is original image or the figure after image enhancement operation to judge the testing image
Picture.
7. the image enhancement detecting system according to claim 6 based on orientation consistency convolutional neural networks, feature
It is, the acquisition module includes acquiring unit;Wherein acquiring unit is cut by center, obtained for inputting testing image
The image of fixed size.
8. the image enhancement detecting system according to claim 6 based on orientation consistency convolutional neural networks, feature
It is, the computing module network establishes unit and computing unit;
Wherein network establishes unit for establishing deep learning network model;Based on the supervised learning method of label label, will scheme
As the original image in database is labeled as label 0, by the correspondence picture of image enhancement labeled as label 1, and using being based on
The convolutional neural networks of orientation consistency carry out deep learning to the label image, obtain corresponding original image and by figure
The feature of image, is trained the characteristic information using stochastic gradient descent method, is corresponded to after image intensifying operation
The deep learning network model of the detection target;
Computing unit is used to utilize trained orientation consistency convolutional neural networks model, and it is original image to obtain testing image
Probability and be probability after image enhancement operation.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711484389.1A CN108257115A (en) | 2018-04-13 | 2018-04-13 | Image enhancement detection method and system based on orientation consistency convolutional neural networks |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711484389.1A CN108257115A (en) | 2018-04-13 | 2018-04-13 | Image enhancement detection method and system based on orientation consistency convolutional neural networks |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108257115A true CN108257115A (en) | 2018-07-06 |
Family
ID=62725437
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711484389.1A Pending CN108257115A (en) | 2018-04-13 | 2018-04-13 | Image enhancement detection method and system based on orientation consistency convolutional neural networks |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108257115A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109146058A (en) * | 2018-07-27 | 2019-01-04 | 中国科学技术大学 | With the constant ability of transformation and the consistent convolutional neural networks of expression |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104021567A (en) * | 2014-06-26 | 2014-09-03 | 福州大学 | Gaussian blur falsification detection method of image based on initial digital law |
CN104424641A (en) * | 2013-09-07 | 2015-03-18 | 无锡华御信息技术有限公司 | Detection method for image fuzzy tampering |
CN107527337A (en) * | 2017-08-07 | 2017-12-29 | 杭州电子科技大学 | A kind of object video based on deep learning removes altering detecting method |
CN107679572A (en) * | 2017-09-29 | 2018-02-09 | 深圳大学 | A kind of image discriminating method, storage device and mobile terminal |
-
2018
- 2018-04-13 CN CN201711484389.1A patent/CN108257115A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104424641A (en) * | 2013-09-07 | 2015-03-18 | 无锡华御信息技术有限公司 | Detection method for image fuzzy tampering |
CN104021567A (en) * | 2014-06-26 | 2014-09-03 | 福州大学 | Gaussian blur falsification detection method of image based on initial digital law |
CN107527337A (en) * | 2017-08-07 | 2017-12-29 | 杭州电子科技大学 | A kind of object video based on deep learning removes altering detecting method |
CN107679572A (en) * | 2017-09-29 | 2018-02-09 | 深圳大学 | A kind of image discriminating method, storage device and mobile terminal |
Non-Patent Citations (1)
Title |
---|
QIANGCHANG WANG ET AL.: "Multiscale Rotation-Invariant Convolutional Neural Networks for Lung Texture Classification", 《IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109146058A (en) * | 2018-07-27 | 2019-01-04 | 中国科学技术大学 | With the constant ability of transformation and the consistent convolutional neural networks of expression |
CN109146058B (en) * | 2018-07-27 | 2022-03-01 | 中国科学技术大学 | Convolutional neural network with transform invariant capability and consistent expression |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US20210319561A1 (en) | Image segmentation method and system for pavement disease based on deep learning | |
CN108876780B (en) | Bridge crack image crack detection method under complex background | |
CN110222701B (en) | Automatic bridge disease identification method | |
CN105095856B (en) | Face identification method is blocked based on mask | |
CN109615116A (en) | A kind of telecommunication fraud event detecting method and detection system | |
CN110992349A (en) | Underground pipeline abnormity automatic positioning and identification method based on deep learning | |
CN107742099A (en) | A kind of crowd density estimation based on full convolutional network, the method for demographics | |
CN107491752A (en) | Ship board character recognition method, device in a kind of natural scene based on deep learning | |
CN107610113A (en) | The detection method and device of Small object based on deep learning in a kind of image | |
CN107833213A (en) | A kind of Weakly supervised object detecting method based on pseudo- true value adaptive method | |
CN105426870A (en) | Face key point positioning method and device | |
CN108647574A (en) | Floating material image detection model generating method, recognition methods and equipment | |
CN110490842A (en) | A kind of steel strip surface defect detection method based on deep learning | |
CN109242047A (en) | Bank card number detection and recognition methods based on K-means++ cluster and residual error network class | |
CN109146873A (en) | A kind of display screen defect intelligent detecting method and device based on study | |
CN111814873A (en) | Method for distinguishing drainage pipeline defect types and automatically identifying defect grades | |
CN116028499A (en) | Detection information generation method, electronic device, and computer-readable medium | |
CN103693532A (en) | Method of detecting violence in elevator car | |
CN108573238A (en) | A kind of vehicle checking method based on dual network structure | |
CN110889338A (en) | Unsupervised railway track bed foreign matter detection and sample construction method and unsupervised railway track bed foreign matter detection and sample construction device | |
CN108257115A (en) | Image enhancement detection method and system based on orientation consistency convolutional neural networks | |
CN110415236A (en) | A kind of method for detecting abnormality of the complicated underground piping based on double-current neural network | |
CN108280388A (en) | The method and apparatus and type of face detection method and device of training face detection model | |
CN110197483A (en) | Deep basal pit crack detection method based on vision signal | |
KR102311558B1 (en) | System and method for detecting structure damage using artificial intelligence, and a recording medium having computer readable program for executing the method |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180706 |
|
RJ01 | Rejection of invention patent application after publication |