CN108333183A - A kind of color based on DCGAN and DCNN knits shirt cut-parts defect inspection method - Google Patents
A kind of color based on DCGAN and DCNN knits shirt cut-parts defect inspection method Download PDFInfo
- Publication number
- CN108333183A CN108333183A CN201810095070.8A CN201810095070A CN108333183A CN 108333183 A CN108333183 A CN 108333183A CN 201810095070 A CN201810095070 A CN 201810095070A CN 108333183 A CN108333183 A CN 108333183A
- Authority
- CN
- China
- Prior art keywords
- defect
- parts
- dcgan
- color
- network model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
- G01N2021/8854—Grading and classifying of flaws
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
- G01N2021/8854—Grading and classifying of flaws
- G01N2021/8861—Determining coordinates of flaws
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
- G01N2021/8883—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges involving the calculation of gauges, generating models
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Treatment Of Fiber Materials (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
Abstract
The invention discloses a kind of colors based on DCGAN and DCNN to knit shirt cut-parts defect inspection method, the original sample library provided according to enterprise, progress DCGAN confrontation generations obtain new color and knit defect image, have expanded defect sample, can effectively avoid the overfitting problem generated in training process；By adjusting hyper parameter, network structure etc., a variety of models are had trained, are suitable for the network model that color knits shirt cut-parts defects detection by comparing selection；Final optimization network model is embedded in TETSON TX2, the detecting system of complete set is formed, shirt case of surface defects can be knitted to color and is positioned and is classified, the demand of industrial production detection is met.
Description
Technical field
The invention belongs to yarn dyed fabric cut-parts defect detection technical fields, and in particular to a kind of color based on DCGAN and DCNN
Knit shirt cut-parts defect inspection method.
Background technology
China is the world's largest fabric clothing producting and exporting country, and derivative textile and garment enterprise is that a labour is close
Collection degree is higher and the larger industry of external dependence degree.As the development of the degree of automation and people are various to yarn dyed fabric flower pattern
Change desired increase so that novel spinning equipment, fabric manufacture equipment continue to bring out.But fabric manufacture equipment failure and spinner behaviour
Make qualification difference, fabric surface can there is the faults of various different shapes, the presence of fault can cause the beauty of fabric
And a series of problems, such as quality, therefore, defect detection is a highly important process in weaving process.
It is one of strongest dress-goods of China's export ability to earn foreign exchange that color, which knits shirt in recent years, is pasted by the cut-parts of different parts
It serves as a contrast, sew, and fault existing for general cut-parts is manually browsed detection by the skilled worker Jing Guo special training,
This traditional detection method based on artificial vision, the influence of the factors such as examined personnel's physiology, psychology, that there are efficiency is low,
The shortcomings of subjectivity is strong.And it is existing it has been proposed that come some solve yarn dyed fabric defect detections method it is less or mainly in
Lily fabric, reason are that yarn dyed fabric flower pattern is complicated, and the background of yarn dyed fabric flower pattern is equally ever-changing, is also deposited between fault
Similar in class class, different situation in class can not meet industrial real-time and accuracy requirement simultaneously.
Invention content
The purpose of the present invention is to provide a kind of colors based on DCGAN and DCNN to knit shirt cut-parts defect inspection method, energy
It is enough shirt case of surface defects is knitted to color to be positioned and classified, improve experiment accuracy rate.
The technical scheme is that a kind of color based on DCGAN and DCNN knits shirt cut-parts defect inspection method, specifically
Implement according to the following steps：
Step 1, structure knit the sample database of shirt cut-parts defect image sample about color, and are located in advance to the sample database
Reason；
Step 2, design color knit the network model of shirt cut-parts defects detection；
Step 3 trains the network model designed in step 2 using pretreated sample database in step 1, and then to color
The network model for knitting shirt cut-parts defects detection optimizes, and obtains optimization network model；
Step 4 builds detecting system hardware platform using the optimization network model in step 3, by the detecting system hardware
Platform connects display device, and defect sample image to be detected is inputted into display device, is examined by detecting system hardware platform
Survey defect classification and defective locations.
The features of the present invention also characterized in that：
Step 1 is specially：
345 step 1.1, establishment colors knit shirt cut-parts defect image sample database；
Step 1.2 carries out the defects of sample database image the expansion of the sample based on DCGAN, obtains expansion defect image；
Step 1.3, using expansion the defects of defect image and original sample library picture construction new samples library, and to new samples
The defect area in library carries out artificial mark defect classification and defect coordinate；
Step 1.4, by the defect image random division after mark be 72% training set, 10% test set and 18%
Verification collection.
In step 1.2 to the defects of sample database image carry out the sample based on DCGAN expansion the specific steps are：First,
Generation model in DCGAN is learnt to knit shirt cut-parts defect characteristic to color using deconvolution, utilizes the defect characteristic learnt
New defect picture is generated, by comparing difference of the formula differentiation to image in new defect picture and original sample library, and by sentencing
Other result obtains expansion defect image.
Comparing formula is specially：
P in formula (1)data(x) it is that authentic specimen is distributed, PG(x) it is false sample distribution.
When differentiating that result D* (x) value is 0.5, then new defect picture is expansion defect image.
Step 1.3 carries out artificial mark detailed process to the defect area in new samples library：Work is marked using Labellmg
Have to color knit the defects of shirt cut-parts with rectangle frame carry out coordinate information, defect key point and defect classification mark.
The input defect image for the network model that step 2 design color knits shirt cut-parts defects detection is dimensioned to 416*
416, convolutional layer has 22 layers, and maximum pond layer is of five storeys, and using LeakyRelu as activation primitive, network model iterations are
2W times, learning rate is set as 0.01, and it is 0.001 and 0.0001 to change learning rate respectively when iterations are 2000,7500.
Step 4 optimization network model builds detecting system hardware platform and is specially：Optimization network model is embedded in JETSON
TX2, JETSON TX2 are a stylobates in the AI single module embedded platforms of NVIDA Pascal frameworks.
The invention has the advantages that
(1) a kind of color based on DCGAN and DCNN of the present invention knits shirt cut-parts defect inspection method, is provided according to enterprise
Original sample library, progress DCGAN confrontation generations obtain new color and knit defect image, expanded defect sample, can effectively avoid
The overfitting problem generated in training process；
(2) a kind of color based on DCGAN and DCNN of the present invention knit shirt cut-parts defect inspection method by adjusting hyper parameter,
Network structure etc. has trained a variety of models, is suitable for the network model that color knits shirt cut-parts defects detection by comparing selection；
(3) a kind of color based on DCGAN and DCNN of the present invention knits shirt cut-parts defect inspection method by final optimization net
Network model insertion forms the detecting system of complete set in TETSON TX2, color can be knitted shirt case of surface defects into
Row positioning and classification, meet the demand of industrial production detection.
Description of the drawings
Fig. 1 is that a kind of color based on DCGAN and DCNN of the present invention knits shirt cut-parts defect inspection method component flow chart；
Fig. 2 (a), Fig. 2 (b), Fig. 2 (c), Fig. 2 (d) are that a kind of color based on DCGAN and DCNN knits the inspection of shirt cut-parts defect
The real-time testing result figure of survey method；
Fig. 3 is that a kind of color based on DCGAN and DCNN knits training pattern generation figure in shirt cut-parts defect inspection method.
Specific implementation mode
The following describes the present invention in detail with reference to the accompanying drawings and specific embodiments.
A kind of color based on DCGAN and DCNN knits shirt cut-parts defect inspection method, as shown in Figure 1, specifically according to following
Step is implemented：
Step 1, structure knit the sample database of shirt cut-parts defect image sample about color, and are located in advance to the sample database
Reason；
Specially：
345 step 1.1, establishment colors knit shirt cut-parts defect image sample database；
Step 1.2 carries out the defects of sample database image the expansion of the sample based on DCGAN, obtains expansion defect image；
To the defects of sample database image carry out the sample based on DCGAN expansion the specific steps are：First, in DCGAN
Generation model, learn to color to knit shirt cut-parts defect characteristic using deconvolution, be generated newly using defect characteristic learn
Defect picture by comparing difference of the formula differentiation to image in new defect picture and original sample library, and is obtained by differentiation result
To expansion defect image.
Comparing formula is specially：
P in formula (1)data(x) it is that authentic specimen is distributed, PG(x) it is false sample distribution.
When differentiating that result D* (x) value is 0.5, then new defect picture is expansion defect image.
Confrontation is carried out using G generators and D arbiters and generates mutually balance so that PdataWith PGThe overall situation is reached when equal
Optimal, arbiter D differentiates that the output phase is same to authentic specimen and false sample at this time.Confrontation generates the picture that network generates
It is different from artwork, but specific features are close, are essentially different with the method for general extension picture number.
Step 1.3 merges expansion defect image with the defects of original sample library image, creates new samples library, and
Artificial mark defect classification and defect coordinate are carried out to the defect area in new samples library；
Carrying out artificial mark detailed process to the defect area in new samples library is：Color is knitted using Labellmg annotation tools
The defect of shirt cut-parts with rectangle frame carry out coordinate information, defect key point and defect classification mark.Wherein there are 7 for defect
Class, including Belt yarn, Knot tying, Hole, BrokenEnd, NettingMultiple, ThickBar, ThinBar.
Step 1.4, by the defect image random division after mark be 72% training set, 10% test set and 18%
Verification collection.
Step 2, design color knit the network model of shirt cut-parts defects detection, the input defect image size of the network model
It is set as 416*416, convolutional layer there are 22 layers, and maximum pond layer is of five storeys, and the first layer of network model is convolutional layer, and the second layer is
Maximum pond layer, third layer are convolutional layer, and the 4th layer is maximum pond layer, and layer 5 to layer 7 is convolutional layer, and the 8th layer is
Maximum pond layer, the 9th layer to eleventh floor is convolutional layer, and Floor 12 is maximum pond layer, and the 13rd layer to 17 layers are convolution
Layer, 18 layers are maximum pond layer, and 19 layers to 25 layers are convolutional layer.Using Leaky Relu as activation primitive, net
Network model iterations are 2W times, and learning rate is set as 0.01, and changes learning rate respectively when iterations are 2000,7500
For 0.001 and 0.0001；
The training pattern of generation is detected by following methods, detection the specific steps are：Piece image is divided into
S*S grid is fallen if the center of some target in this grid, then this grid is just responsible for predicting this target, each
B prediction block of grid forecasting, each prediction block will also be attached to other than returning the position of itself and predict a confidence level
Value, this confidence value represent the whether accurately double of the confidence level containing target and prediction block prediction in predicted frame
Information, intermediate value are that calculation is as follows：
If there is target is fallen in next grid, first item takes 1, and it is the frame and actual conditions of prediction otherwise to take 0. Section 2
Between IOU values, in addition to this also need to prediction classification information and coordinate.The loss function of whole network is as follows：
Shown in testing result such as Fig. 2 (a), Fig. 2 (b), Fig. 2 (c), Fig. 2 (d).
Step 3 trains the network model designed in step 2 using pretreated sample database in step 1, and then to color
The network model for knitting shirt cut-parts defects detection optimizes, and obtains optimization network model, as shown in Figure 3；
Step 4 builds detecting system hardware platform using the optimization network model in step 3, specially：Network will be optimized
Model insertion JETSON TX2, JETSON TX2 are a stylobates in the AI single module embedded platforms of NVIDA Pascal frameworks,
Complete detecting system hardware platform；The detecting system hardware platform is connected into display device, is inputted into display device to be checked
The defect sample image of survey detects defect classification and defective locations by detecting system hardware platform；It can be by its high property
Can, the characteristic of low energy consumption, handle complex data on onboard terminal device, meet industrial low cost, high real-time is wanted
It asks.
By the above-mentioned means, a kind of color based on DCGAN and DCNN of the present invention knits shirt cut-parts defect inspection method, according to
The original sample library that enterprise provides carries out DCGAN confrontation and generates obtaining new color and knitting defect image, expanded defect sample, can be with
Effectively avoid the overfitting problem generated in training process；By adjusting hyper parameter, network structure etc., a variety of moulds are had trained
Type is suitable for the network model that color knits shirt cut-parts defects detection by comparing selection；Final optimization network model is embedded in
In TETSON TX2, the detecting system of complete set is formed, shirt case of surface defects can be knitted to color and is positioned and is divided
Class meets the demand of industrial production detection.
Claims (8)
1. a kind of color based on DCGAN and DCNN knits shirt cut-parts defect inspection method, which is characterized in that specifically according to following step
It is rapid to implement：
Step 1, structure knit the sample database of shirt cut-parts defect image sample about color, and are pre-processed to the sample database；
Step 2, design color knit the network model of shirt cut-parts defects detection；
Step 3 trains the network model designed in step 2 using pretreated sample database in step 1, and then knits lining to color
The network model of shirt cut-parts defects detection optimizes, and obtains optimization network model；
Step 4 builds detecting system hardware platform using the optimization network model in step 3, by the detecting system hardware platform
Display device is connected, defect sample image to be detected is inputted into display device, detects and lacks by detecting system hardware platform
Fall into classification and defective locations.
2. a kind of color based on DCGAN and DCNN knits shirt cut-parts defect inspection method as described in claim 1, feature exists
In the step 1 is specially：
345 step 1.1, establishment colors knit shirt cut-parts defect image sample database；
Step 1.2 carries out the defects of sample database image the expansion of the sample based on DCGAN, obtains expansion defect image；
Step 1.3, using expansion the defects of defect image and original sample library picture construction new samples library, and to new samples library
Defect area carries out artificial mark defect classification and defect coordinate；
Defect image random division after mark is 72% training set, 10% test set and 18% tests by step 1.4
Card collection.
3. a kind of color based on DCGAN and DCNN knits shirt cut-parts defect inspection method as claimed in claim 2, feature exists
In, described in step 1.2 to the defects of sample database image carry out the sample based on DCGAN expansion the specific steps are：First, exist
Generation model in DCGAN is learnt to knit shirt cut-parts defect characteristic to color using deconvolution, be given birth to using the defect characteristic learnt
The defect picture of Cheng Xin, by comparing difference of the formula differentiation to image in new defect picture and original sample library, and by differentiating
As a result expansion defect image is obtained.
4. a kind of color based on DCGAN and DCNN knits shirt cut-parts defect inspection method as claimed in claim 3, feature exists
In the comparison formula is specially：
P in formula (1)data(x) it is that authentic specimen is distributed, PG(x) it is false sample distribution.
5. a kind of color based on DCGAN and DCNN knits shirt cut-parts defect inspection method as claimed in claim 4, feature exists
In when differentiation result D* (x) value is 0.5, then new defect picture is expansion defect image.
6. a kind of color based on DCGAN and DCNN knits shirt cut-parts defect inspection method as claimed in claim 2, feature exists
In carrying out artificial mark detailed process to the defect area in new samples library described in step 1.3 is：Using Labellmg annotation tools
To color knit the defects of shirt cut-parts with rectangle frame carry out coordinate information, defect key point and defect classification mark.
7. a kind of color knits shirt cut-parts defect inspection method as described in claim 1, which is characterized in that design color described in step 2
The input defect image for knitting the network model of shirt cut-parts defects detection is dimensioned to 416*416, and convolutional layer has 22 layers, maximum
Pond layer is of five storeys, and using Leaky Relu as activation primitive, network model iterations are 2W times, and learning rate is set as
0.01, and it is 0.001 and 0.0001 to change learning rate respectively when iterations are 2000,7500.
8. a kind of color based on DCGAN and DCNN knits shirt cut-parts defect inspection method as described in claim 1, feature exists
In optimization network model described in step 4 builds detecting system hardware platform and is specially：Optimization network model is embedded in JETSON
TX2, the JETSON TX2 are a stylobates in the AI single module embedded platforms of NVIDA Pascal frameworks.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810095070.8A CN108333183B (en) | 2018-01-31 | 2018-01-31 | Yarn-dyed shirt cut piece defect detection method based on DCGAN and DCNN |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810095070.8A CN108333183B (en) | 2018-01-31 | 2018-01-31 | Yarn-dyed shirt cut piece defect detection method based on DCGAN and DCNN |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108333183A true CN108333183A (en) | 2018-07-27 |
| CN108333183B CN108333183B (en) | 2021-03-16 |
Family
ID=62926924
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810095070.8A Active CN108333183B (en) | 2018-01-31 | 2018-01-31 | Yarn-dyed shirt cut piece defect detection method based on DCGAN and DCNN |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108333183B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109598287A (en) * | 2018-10-30 | 2019-04-09 | 中国科学院自动化研究所 | The apparent flaws detection method that confrontation network sample generates is generated based on depth convolution |
| CN111402197A (en) * | 2020-02-09 | 2020-07-10 | 西安工程大学 | Detection method for yarn-dyed fabric cut piece defect area |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104850858A (en) * | 2015-05-15 | 2015-08-19 | 华中科技大学 | Injection-molded product defect detection and recognition method |
| JP6150248B2 (en) * | 2013-01-30 | 2017-06-21 | 公立大学法人 富山県立大学 | Fabric defect inspection method and apparatus |
| CN107169956A (en) * | 2017-04-28 | 2017-09-15 | 西安工程大学 | Yarn dyed fabric defect detection method based on convolutional neural networks |
| CN107180392A (en) * | 2017-05-18 | 2017-09-19 | 北京科技大学 | A kind of electric power enterprise tariff recovery digital simulation method |
| CN107220649A (en) * | 2017-05-27 | 2017-09-29 | 江苏理工学院 | A kind of plain color cloth defects detection and sorting technique |
| CN107316295A (en) * | 2017-07-02 | 2017-11-03 | 苏州大学 | A kind of fabric defects detection method based on deep neural network |
| CN107392255A (en) * | 2017-07-31 | 2017-11-24 | 深圳先进技术研究院 | Generation method, device, computing device and the storage medium of minority class picture sample |
| CN107423701A (en) * | 2017-07-17 | 2017-12-01 | 北京智慧眼科技股份有限公司 | The non-supervisory feature learning method and device of face based on production confrontation network |
| CN107481231A (en) * | 2017-08-17 | 2017-12-15 | 广东工业大学 | A kind of handware defect classifying identification method based on depth convolutional neural networks |
| CN107563385A (en) * | 2017-09-02 | 2018-01-09 | 西安电子科技大学 | License plate character recognition method based on depth convolution production confrontation network |
| CN107590518A (en) * | 2017-08-14 | 2018-01-16 | 华南理工大学 | A kind of confrontation network training method of multiple features study |
-
2018
- 2018-01-31 CN CN201810095070.8A patent/CN108333183B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP6150248B2 (en) * | 2013-01-30 | 2017-06-21 | 公立大学法人 富山県立大学 | Fabric defect inspection method and apparatus |
| CN104850858A (en) * | 2015-05-15 | 2015-08-19 | 华中科技大学 | Injection-molded product defect detection and recognition method |
| CN107169956A (en) * | 2017-04-28 | 2017-09-15 | 西安工程大学 | Yarn dyed fabric defect detection method based on convolutional neural networks |
| CN107180392A (en) * | 2017-05-18 | 2017-09-19 | 北京科技大学 | A kind of electric power enterprise tariff recovery digital simulation method |
| CN107220649A (en) * | 2017-05-27 | 2017-09-29 | 江苏理工学院 | A kind of plain color cloth defects detection and sorting technique |
| CN107316295A (en) * | 2017-07-02 | 2017-11-03 | 苏州大学 | A kind of fabric defects detection method based on deep neural network |
| CN107423701A (en) * | 2017-07-17 | 2017-12-01 | 北京智慧眼科技股份有限公司 | The non-supervisory feature learning method and device of face based on production confrontation network |
| CN107392255A (en) * | 2017-07-31 | 2017-11-24 | 深圳先进技术研究院 | Generation method, device, computing device and the storage medium of minority class picture sample |
| CN107590518A (en) * | 2017-08-14 | 2018-01-16 | 华南理工大学 | A kind of confrontation network training method of multiple features study |
| CN107481231A (en) * | 2017-08-17 | 2017-12-15 | 广东工业大学 | A kind of handware defect classifying identification method based on depth convolutional neural networks |
| CN107563385A (en) * | 2017-09-02 | 2018-01-09 | 西安电子科技大学 | License plate character recognition method based on depth convolution production confrontation network |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109598287A (en) * | 2018-10-30 | 2019-04-09 | 中国科学院自动化研究所 | The apparent flaws detection method that confrontation network sample generates is generated based on depth convolution |
| CN109598287B (en) * | 2018-10-30 | 2021-06-08 | 中国科学院自动化研究所 | Appearance flaw detection method for resisting network sample generation based on deep convolution generation |
| CN111402197A (en) * | 2020-02-09 | 2020-07-10 | 西安工程大学 | Detection method for yarn-dyed fabric cut piece defect area |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108333183B (en) | 2021-03-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105654121B (en) | A kind of complicated jacquard fabric defect inspection method based on deep learning | |
| CN107392896A (en) | A kind of Wood Defects Testing method and system based on deep learning | |
| CN108520114A (en) | A kind of textile cloth defect detection model and its training method and application | |
| CN111402197B (en) | Detection method for colored fabric cut-parts defect area | |
| CN106804042A (en) | The clustering method in weak covering problem region and Bus stop planning method | |
| CN110349146A (en) | The building method of fabric defect identifying system based on lightweight convolutional neural networks | |
| CN103542937B (en) | Digital printing color difference inspecting method | |
| CN109859207A (en) | A kind of defect inspection method of high density flexible substrate | |
| CN110533086A (en) | The semi-automatic mask method of image data | |
| CN104318482A (en) | Comprehensive assessment system and method of smart distribution network | |
| CN108333183A (en) | A kind of color based on DCGAN and DCNN knits shirt cut-parts defect inspection method | |
| CN109358306A (en) | One kind being based on the intelligent electric energy meter health degree trend forecasting method of GM (1,1) | |
| CN107291830A (en) | A kind of creation method of equipment knowledge base | |
| CN110490858A (en) | A kind of fabric defect Pixel-level classification method based on deep learning | |
| CN109409425A (en) | A kind of fault type recognition method based on neighbour's constituent analysis | |
| CN107633309B (en) | It is a kind of complexity former maintenance policy determine method and system | |
| CN110473806A (en) | The deep learning identification of photovoltaic cell sorting and control method and device | |
| CN102411667A (en) | Method for predicating overdying formula of colored fabric | |
| CN109472790A (en) | A kind of machine components defect inspection method and system | |
| CN106021724A (en) | Energy efficiency evaluation method of machine tool product manufacturing system based on AHM and entropy method | |
| CN104679945B (en) | System comprehensive estimation method based on colored Petri network | |
| CN109918444A (en) | Training/verifying/management method/system, medium and equipment of model result | |
| CN108572639A (en) | A kind of dynamic process monitoring method rejected based on principal component autocorrelation | |
| CN106960188A (en) | Weather image sorting technique and device | |
| He et al. | Research on Fabric defect detection based on deep fusion DenseNet-SSD network |
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 |