CN108460761A - Method and apparatus for generating information - Google Patents
Method and apparatus for generating information Download PDFInfo
- Publication number
- CN108460761A CN108460761A CN201810201427.6A CN201810201427A CN108460761A CN 108460761 A CN108460761 A CN 108460761A CN 201810201427 A CN201810201427 A CN 201810201427A CN 108460761 A CN108460761 A CN 108460761A
- Authority
- CN
- China
- Prior art keywords
- image
- product
- pending
- sample
- subgraph
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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 embodiment of the present application discloses the method and apparatus for generating information.One specific implementation mode of this method includes：Acquisition includes the pending image of target product image, wherein target product image includes product defects image；Cutting is carried out to pending image, obtains at least two pending subgraphs, wherein pending subgraph includes target product subgraph；At least two pending subgraphs are inputted to convolutional neural networks trained in advance respectively, obtain the description information of the product defects image in the target product subgraph included by pending subgraph, wherein, convolutional neural networks are used to characterize the correspondence of image and the description information of the product defects image in the product image included by image.This embodiment improves the accuracys that information generates.
Description
Technical field
The invention relates to field of computer technology, the method and apparatus for more particularly, to generating information.
Background technology
During production, the defect of product can inevitably be caused (such as to split due to various physics or human factor
Line, cut etc.), generated defect has great influence to product quality and Product Safety etc..Currently, the relevant technologies people
Member's generally use mode for artificially observing determines the position of the product and existing defect of existing defects in the product.
Invention content
The embodiment of the present application proposes the method and apparatus for generating information.
In a first aspect, the embodiment of the present application provides a kind of method for generating information, this method includes：Acquisition includes
The pending image of target product image, wherein target product image includes product defects image；Pending image is cut
Point, obtain at least two pending subgraphs, wherein pending subgraph includes target product subgraph；At least two are waited for
Processing subgraph inputs convolutional neural networks trained in advance respectively, obtains the target product subgraph included by pending subgraph
The description information of product defects image as in, wherein convolutional neural networks are used to characterize image and the product included by image
The correspondence of the description information of product defects image in image.
In some embodiments, this method further includes：Pending image is inputted into convolutional neural networks, obtains pending figure
As the description information of the product defects image in included target product image.
In some embodiments, training obtains convolutional neural networks as follows：It includes sample product to obtain multiple
The pending image of sample of image, wherein sample product image includes sample product defect image；It obtains having demarcated, multiple
The sample product defect image in the sample product image included by the pending image of each sample in the pending image of sample
Description information；Using machine learning method, using the pending image of each sample in the pending image of multiple samples as defeated
Enter, it will be in the sample product image included by the pending image of each sample in pending images of sample demarcate, multiple
Sample product defect image description information as output, training obtain convolutional neural networks.
In some embodiments, multiple pending images of sample including sample product image are obtained, including：It obtains multiple
Sample product defect image；For each sample product defect image in multiple sample product defect images, which is produced
Product defect image is merged with the pending image of sample to be fused of sample product image preset including to be fused, is generated
The pending image of sample including sample product image, wherein sample product image to be fused does not include product defects image.
In some embodiments, multiple sample product defect images are obtained, including：Obtain multiple initial data；It will be multiple
Initial data inputs defect trained in advance and generates model respectively, obtains multiple product defects images, wherein defect generates model
Correspondence for characterize data and product defects image；The multiple product defects images obtained are determined as multiple samples
Product defects image.
In some embodiments, training obtains defect generation model as follows：Obtain training sample, wherein instruction
It includes multiple sample initial data and multiple product defects images to be entered to practice sample；Extract the production confrontation net pre-established
Network, wherein it includes generating network and differentiating network that production, which fights network, generates network for utilizing inputted data to generate
Image differentiates network for determining whether inputted image makes a living into the image that network is exported；Using machine learning method,
Using multiple sample initial data as the input for generating network, by the image for generating network output and multiple product defects to be entered
Generation network after training is determined as the input for differentiating network to generating network and differentiating that network is trained by image
Defect generates model.
In some embodiments, multiple sample product defect images are obtained, including：Obtain multiple original images；Multiple
Product defects image is generated on each original image in original image so that the product defects image generated, which meets, presets item
Part；The multiple product defects images generated are determined as multiple sample product defect images.
Second aspect, the embodiment of the present application provide a kind of device for generating information, which includes：It obtains single
Member is configured to the pending image that acquisition includes target product image, wherein target product image includes product defects figure
Picture；Cutting unit is configured to carry out cutting to pending image, obtains at least two pending subgraphs, wherein pending
Subgraph includes target product subgraph；First generation unit is configured to respectively input at least two pending subgraphs
Trained convolutional neural networks in advance, obtain the product defects image in the target product subgraph included by pending subgraph
Description information, wherein convolutional neural networks are used to characterize the product defects figure in the product image included by image and image
The correspondence of the description information of picture.
In some embodiments, which further includes：Second generation unit is configured to pending image inputting convolution
Neural network obtains the description information of the product defects image in the target product image included by pending image.
In some embodiments, training obtains convolutional neural networks as follows：It includes sample product to obtain multiple
The pending image of sample of image, wherein sample product image includes sample product defect image；It obtains having demarcated, multiple
The sample product defect image in the sample product image included by the pending image of each sample in the pending image of sample
Description information；Using machine learning method, using the pending image of each sample in the pending image of multiple samples as defeated
Enter, it will be in the sample product image included by the pending image of each sample in pending images of sample demarcate, multiple
Sample product defect image description information as output, training obtain convolutional neural networks.
The third aspect, the embodiment of the present application provide a kind of server, including：One or more processors；Storage device,
For storing one or more programs, when one or more programs are executed by one or more processors so that one or more
The method that processor realizes any embodiment in the above-mentioned method for generating information.
Fourth aspect, the embodiment of the present application provide a kind of computer-readable medium, are stored thereon with computer program, should
The method that any embodiment in the above-mentioned method for generating information is realized when program is executed by processor.
Method and apparatus provided by the embodiments of the present application for generating information include target product image by obtaining
Pending image, wherein target product image includes product defects image；Then cutting is carried out to pending image, obtained extremely
Few two pending subgraphs, wherein pending subgraph includes target product subgraph；Then by least two pending sons
Image inputs convolutional neural networks trained in advance respectively, obtains in the target product subgraph included by pending subgraph
The description information of product defects image, to which local message is significantly more to be waited for by the progress cutting of pending image, obtaining
Subgraph is handled, and by pending subgraph input convolutional neural networks trained in advance, obtains more structurally sound, product and is wrapped
The description information of the product defects included improves the accuracy of information generation.
Description of the drawings
By reading a detailed description of non-restrictive embodiments in the light of the attached drawings below, the application's is other
Feature, objects and advantages will become more apparent upon：
Fig. 1 is that this application can be applied to exemplary system architecture figures therein；
Fig. 2 is the flow chart according to one embodiment of the method for generating information of the application；
Fig. 3 is the schematic diagram of the flow that pending subgraph is obtained to the progress cutting of pending image of the application；
Fig. 4 is the schematic diagram according to an application scenarios of the method for generating information of the application；
Fig. 5 is the flow chart according to another embodiment of the method for generating information of the application；
Fig. 6 is the structural schematic diagram according to one embodiment of the device for generating information of the application；
Fig. 7 is adapted for the structural schematic diagram of the computer system of the server for realizing the embodiment of the present application.
Specific implementation mode
The application is described in further detail with reference to the accompanying drawings and examples.It is understood that this place is retouched
The specific embodiment stated is used only for explaining related invention, rather than the restriction to the invention.It also should be noted that in order to
Convenient for description, is illustrated only in attached drawing and invent relevant part with related.
It should be noted that in the absence of conflict, the features in the embodiments and the embodiments of the present application can phase
Mutually combination.The application is described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.
Fig. 1 shows the implementation of the method for generating information or the device for generating information that can apply the application
The exemplary system architecture 100 of example.
As shown in Figure 1, system architecture 100 may include terminal device 101,102,103, network 104 and server 105.
Network 104 between terminal device 101,102,103 and server 105 provide communication link medium.Network 104 can be with
Including various connection types, such as wired, wireless communication link or fiber optic cables etc..
User can be interacted by network 104 with server 105 with using terminal equipment 101,102,103, to receive or send out
Send message etc..Camera and various telecommunication customer end applications, such as image can be installed on terminal device 101,102,103
Handling implement, searching class application, instant messaging tools, mailbox client, U.S. figure software etc..
Terminal device 101,102,103 can be hardware, can also be software.When terminal device 101,102,103 is hard
Can be the various executive agents with display screen, including but not limited to smart mobile phone, tablet computer, e-book reading when part
(Moving Picture Experts Group Audio Layer III, dynamic image expert compress mark for device, MP3 player
Quasi- audio level 3), MP4 (Moving Picture Experts Group Audio Layer IV, dynamic image expert compression
Standard audio level 4) player, pocket computer on knee and desktop computer etc..When terminal device 101,102,103 is
When software, it may be mounted in above-mentioned cited executive agent.Its may be implemented into multiple softwares or software module (such as with
To provide Distributed Services), single software or software module can also be implemented as.It is not specifically limited herein.
Server 105 can be to provide the server of various services, such as to transmitted by terminal device 101,102,103
The netscape messaging server Netscape that pending image is handled.Netscape messaging server Netscape can count the pending image etc. received
According to carrying out the processing such as analyzing, and handling result (such as information for characterizing product defects image) is fed back into terminal device.
It should be noted that server can be hardware, can also be software.When server is hardware, may be implemented
At the distributed apparatus cluster that multiple equipment forms, individual equipment can also be implemented as.When server is software, can distinguish
It is implemented as multiple softwares or software module (such as providing Distributed Services), single software or software mould can also be implemented as
Block.It is not specifically limited herein.
It should be understood that the number of the terminal device, network and server in Fig. 1 is only schematical.According to realization need
It wants, can have any number of terminal device, network and server.
With continued reference to Fig. 2, the flow of one embodiment of the method for generating information according to the application is shown
200.The method for being used to generate information, includes the following steps：
Step 201, acquisition includes the pending image of target product image.
In the present embodiment, the method for generating information runs executive agent (such as service shown in FIG. 1 thereon
Device) the pending image including target product image can be obtained, wherein and target product image includes product defects image.
Here, target product image is the image of target product.Target product can be the production of the product defects to be determined included by it
Product.Illustratively, target product can be steel billet product, glassware, ceramic etc..Product defects are included by product
Fail meet consumption or using product prerequisite safety requirements defect (such as crackle, cut etc.).Specifically, above-mentioned
Executive agent, which can obtain, is pre-stored within local including target product image pending image；Alternatively, above-mentioned execution master
Body can obtain terminal (such as terminal device shown in FIG. 1 101,102,103) transmission including target product image waits locating
Manage image.Herein, terminal can first shoot target product, and acquisition includes the pending figure of target product image
Picture；Further, the pending image obtained can be sent to above-mentioned executive agent by terminal.
Step 202, cutting is carried out to pending image, obtains at least two pending subgraphs.
In the present embodiment, based on the pending image obtained in step 201, above-mentioned executive agent (such as it is shown in FIG. 1
Server) cutting can be carried out to pending image, obtain at least two pending subgraphs, wherein pending subgraph can
To include target product subgraph.Target product subgraph is the topography of target product image.It should be noted that target
Product subgraph may include product defects image.
In the present embodiment, pending image can be rectangle, square, circle etc..It is above-mentioned that pending image is carried out
Cutting obtains at least two pending subgraphs and is specifically as follows with various slit modes to the progress cutting of pending image, obtains
Obtain at least two pending subgraphs, wherein pending subgraph may include target product subgraph.Illustratively, it waits locating
Reason image is rectangle, can carry out cutting to pending image on the basis of the midpoint on each side of pending image, obtain four
The identical pending subgraph of shape size；Alternatively, pending image is rectangle, it can be with the four of each side of pending image
/ mono- position, the position of half and 3/4ths position on the basis of to pending image carry out cutting, obtain
The identical pending subgraph of 16 shape sizes.It should be noted that the slit mode of pending image can be by technology
Personnel predefine and are stored in above-mentioned executive agent.
Illustratively, as shown in figure 3, carrying out cutting to pending image for the application obtains the one of pending subgraph
The schematic diagram of a flow.In figure, 301 be pending image, and 3011 be target product (such as steel billet product) image, and 30111 are
Product defects (such as crackle) image, 3012,3013,3014, the 3015 four pending subgraphs obtained for cutting.It can see
Go out, herein, the slit mode of pending image 301 is on the basis of the midpoint on 301 each side of pending image to pending
Image 301 carries out cutting.
Optionally, above-mentioned executive agent can carry out pending image the cutting for having overlapping, obtain at least two and wait locating
Manage subgraph.Illustratively, pending image is rectangle, and above-mentioned executive agent can first determine included by pending image
, true edge for determining cutting benchmark, and then on the basis of 3/5ths of true edge position to pending image into
Row cutting obtains two images, and the larger image of area in obtained two images is determined as pending subgraph；So
Afterwards, for above-mentioned pending image, above-mentioned executive agent can be further using 2/5ths position of above-mentioned true edge as base
Standard carries out cutting to pending image, obtains two images, and the larger image of area in obtained two images is determined
For pending subgraph.The cutting that above-mentioned executive agent carries out having overlapping to pending image as a result, obtains with overlay region
Two pending subgraphs in domain.
Step 203, at least two pending subgraphs are inputted to convolutional neural networks trained in advance respectively, obtains and waits locating
Manage the description information of the product defects image in the target product subgraph included by subgraph.
In the present embodiment, the pending subgraph obtained based on step 202, above-mentioned executive agent can will at least two
A pending subgraph inputs convolutional neural networks trained in advance respectively, obtains the target product included by pending subgraph
The description information of product defects image in subgraph.Wherein, in the target product subgraph included by pending subgraph
The description information of product defects image can include but is not limited at least one of following：Word, number, symbol.Specifically, waiting locating
The description information of the product defects image in target product subgraph included by reason subgraph can be for characterizing target production
Product subgraph whether include product defects image information (such as " o "；" x ", wherein symbol " o ", which can be used for characterizing, includes,
Symbol " x " can be used for characterizing), or for for characterizing position of the product defects image in target product subgraph
Information (such as the coordinate of the pixel included by product defects image in target product subgraph), or for for characterizing
The information (such as " crackle " of the classification of product defects indicated by product defects image；" cut ").It should be noted that at this
In, the pending subgraph for inputting above-mentioned convolutional neural networks is at least two, correspondingly, the pending subgraph institute obtained
Including target product subgraph in product defects image description information be at least two.
In the present embodiment, convolutional neural networks can be used for characterizing image and the production in the product image included by image
The correspondence of the description information of product defect image.Above-mentioned convolutional neural networks may include convolutional layer, extracted region network and
Grader, wherein convolutional layer can be used for extracting characteristics of image, and extracted region network can be used for being extracted according to convolutional layer
Characteristics of image generates those suspected defects region, and the characteristics of image that grader can be used for being generated according to convolutional layer is to extracted region net
Each those suspected defects region that network is generated carries out Accurate Analysis, and generates the product defects in the product image included by image
The description information of image.It should be noted that above-mentioned executive agent can profit in various manners (such as Training, without prison
Supervise and instruct practice etc. modes) training obtain above-mentioned convolutional neural networks.
In some optional realization methods of the present embodiment, above-mentioned convolutional neural networks can train as follows
It obtains：
First, multiple pending images of sample including sample product image are obtained, wherein sample product image can wrap
Include sample product defect image.
Then, the sample included by the pending image of each sample in the pending image of sample demarcated, multiple is obtained
The description information (such as location information, classification information etc.) of sample product defect image in this product image.
Finally, using machine learning method, using the pending image of each sample in the pending image of multiple samples as
Input, by the sample product image included by the pending image of each sample in pending images of sample demarcated, multiple
In sample product defect image description information as output, training obtain convolutional neural networks.
In some optional realization methods of the present embodiment, it includes sample product that can obtain as follows multiple
The pending image of sample of image：
It is possible, firstly, to obtain multiple sample product defect images.
Then, it for each sample product defect image in acquired multiple sample product defect images, can incite somebody to action
The pending image of sample to be fused of the sample product defect image and sample product image preset including to be fused is melted
It closes, generation includes the pending image of sample of sample product image, wherein sample product image to be fused does not include product defects
Image.It should be noted that image co-registration is the known technology for being widely used at present and studying, details are not described herein.
In some optional realization methods of the present embodiment, multiple sample product defects can be obtained as follows
Image：
Step 2031, multiple initial data are obtained.
Wherein, multiple initial data can be the predetermined data for being used for input model and generating image, such as with
In feature vector, the random noise of characteristics of image etc. of characterization product defects image.
Step 2032, multiple initial data are inputted into defect trained in advance respectively and generates model, obtained multiple products and lack
Fall into image.
Wherein, defect generates the correspondence that model can be used for characterize data and product defects image.Specifically, conduct
Example, defect are generated model and can be pre-established based on the statistics to a large amount of data and product defects image for technical staff
, be stored with multiple data with the mapping table of the correspondence of product defects image；Can also be to advance with engineering
Learning method is trained rear institute based on training sample to the model (such as convolutional neural networks) for generating product defects image
Obtained model.
In some optional realization methods of the present embodiment, drawbacks described above generates model and can train as follows
It obtains：
First, training sample is obtained.Wherein, training sample may include multiple sample initial data and multiple productions to be entered
Product defect image.Sample initial data can be predetermined for training defect to generate the data of model.Product to be entered
Defect image can be the image being pre-stored in above-mentioned executive agent, or be terminal (such as terminal device shown in FIG. 1
101,102, the 103) image sent.
Then, the production confrontation network (Generative Adversarial Nets, GAN) pre-established is extracted.Example
Such as, above-mentioned production confrontation network can be that depth convolution generates confrontation network (Deep Convolutional Generative
Adversarial Network, DCGAN).
Wherein, production confrontation network may include generating network and differentiating that network, above-mentioned generation network can be used for profit
Image is generated with the data inputted, above-mentioned differentiation network is determined for whether inputted image is above-mentioned generation network
The image exported.It should be noted that above-mentioned generation network can be the various graders for generating image；Above-mentioned differentiation
Network can be convolutional neural networks (such as various convolutional neural networks structures comprising full articulamentum, wherein full articulamentum can
To realize classification feature).In addition, above-mentioned differentiation network can also be other model structures for realizing classification feature, such as
Support vector machines (Support Vector Machine, SVM).It should be noted that the image that above-mentioned generation network is exported
It can be expressed with the matrix of RGB triple channels.Herein, network is differentiated if it is determined that the image of input is that above-mentioned generation network institute is defeated
The image (carrying out self-generating data) gone out, then can export 1；If it is determined that the image of input is not the figure that above-mentioned generation network is exported
As (coming from truthful data, i.e., above-mentioned product defects image to be entered), then 0 can be exported.It should be noted that differentiating network
It can be redefined for exporting other numerical value, word or symbol, be not limited to above-mentioned 1 and 0.
Finally, network will be generated using multiple sample initial data as the input for generating network using machine learning method
The image of output and multiple product defects images to be entered are as the input for differentiating network, to generating network and differentiating that network carries out
Generation network after training is determined as defect and generates model by training.
Specifically, the ginseng for generating network and differentiating any network (can be described as first network) in network can be fixed first
Number, optimizes the network (can be described as the second network) of unlocked parameter；The parameter for fixing the second network again, to first network
It is improved.Above-mentioned iteration is constantly carried out, so that the image of the indistinguishable input of differentiation network whether is generated network and is generated,
Until final convergence.At this point, the image and product defects image to be entered that above-mentioned generation network is generated are close, above-mentioned differentiation net
Network can not accurately distinguish truthful data and generate data (i.e. accuracy rate is 50%), and then can determine generation network at this time
Model is generated for defect.
Step 2033, the multiple product defects images obtained are determined as multiple sample product defect images.
In some optional realization methods of the present embodiment, multiple sample product defects can be obtained as follows
Image：
It is possible, firstly, to obtain multiple original images, wherein original image can be for generating sample product defect image
Arbitrary image.
It is then possible to generate product defects image on each original image in above-mentioned multiple original images so that institute
The product defects image of generation meets preset condition.
Wherein, the use that preset condition can be pre-established for technical staff based on the statistics to large-tonnage product defect image
In the condition for generating product defects image.As an example, preset condition can include but is not limited to it is at least one of following：Product lacks
There are pixel differences for pixel included by sunken image；There are random fluctuations for the color value of pixel included by product defects image；Production
Continuous pixels included by product defect image are uninterrupted；It deposits position of the pixel in original image included by product defects image
In random fluctuation.
Finally, the multiple product defects images generated can be determined as multiple sample product defect images.
It is a signal according to the application scenarios of the method for generating information of the present embodiment with continued reference to Fig. 4, Fig. 4
Figure.In the application scenarios of Fig. 4, what server 401 can obtain the transmission of terminal device 402 first includes target product image
The pending image 403 of (steel billet product image), wherein target product image (steel billet product image) may include product defects
Image (crack image)；Then server 401 can carry out cutting to pending image 403, obtain two pending subgraphs
4031 and 4032, wherein pending subgraph 4031,4032 may include target product subgraph；Last server 401 can be with
Pending subgraph 4031,4032 is inputted to convolutional neural networks trained in advance respectively, is obtained included by pending subgraph
Target product subgraph in product defects image description information 4041,4042, wherein convolutional neural networks can be used for
Characterize the correspondence of image and the description information of the product defects image in the product image included by image.
The method that above-described embodiment of the application provides by obtain include target product image pending image,
In, target product image includes product defects image；Then cutting is carried out to pending image, obtains at least two pending sons
Image, wherein pending subgraph includes target product subgraph；Then at least two pending subgraphs are inputted in advance respectively
First trained convolutional neural networks, obtain the product defects image in the target product subgraph included by pending subgraph
Description information, to by carrying out cutting to pending image, obtain the significantly more pending subgraph of local message, and will
Pending subgraph input convolutional neural networks trained in advance, obtain product defects more structurally sound, included by product
Description information improves the accuracy of information generation.
With further reference to Fig. 5, it illustrates the flows 500 of another embodiment of the method for generating information.The use
In the flow 500 for the method for generating information, include the following steps：
Step 501, acquisition includes the pending image of target product image.
In the present embodiment, the method for generating information runs executive agent (such as service shown in FIG. 1 thereon
Device) the pending image including target product image can be obtained, wherein and target product image includes product defects image.
Here, target product image is the image of target product.Target product can be the production of the product defects to be determined included by it
Product.Illustratively, target product can be steel billet product, glassware, ceramic etc..Product defects are included by product
Fail meet consumption or using product prerequisite safety requirements defect (such as crackle, cut etc.).Specifically, above-mentioned
Executive agent, which can obtain, is pre-stored within local including target product image pending image；Alternatively, above-mentioned execution master
Body can obtain terminal (such as terminal device shown in FIG. 1 101,102,103) transmission including target product image waits locating
Manage image.Herein, terminal can first shoot target product, and acquisition includes the pending figure of target product image
Picture；Further, the pending image obtained can be sent to above-mentioned executive agent by terminal.
Step 502, cutting is carried out to pending image, obtains at least two pending subgraphs.
In the present embodiment, based on the pending image obtained in step 501, above-mentioned executive agent (such as it is shown in FIG. 1
Server) cutting can be carried out to pending image, obtain at least two pending subgraphs, wherein pending subgraph can
To include target product subgraph.It should be noted that target product subgraph may include product defects image.
In the present embodiment, pending image can be rectangle, square, circle etc..It is above-mentioned that pending image is carried out
Cutting obtains at least two pending subgraphs and is specifically as follows with various slit modes to the progress cutting of pending image, obtains
Obtain at least two pending subgraphs.
Step 503, at least two pending subgraphs are inputted to convolutional neural networks trained in advance respectively, obtains and waits locating
Manage the description information of the product defects image in the target product subgraph included by subgraph.
In the present embodiment, the pending subgraph obtained based on step 502, above-mentioned executive agent can will at least two
A pending subgraph inputs convolutional neural networks trained in advance respectively, obtains the target product included by pending subgraph
The description information of product defects image in subgraph.Wherein, in the target product subgraph included by pending subgraph
The description information of product defects image can include but is not limited at least one of following：Word, number, symbol.Specifically, waiting locating
The description information of the product defects image in target product subgraph included by reason subgraph can be for characterizing target production
Product subgraph whether include product defects image information (such as " o "；" x ", wherein symbol " o ", which can be used for characterizing, includes,
Symbol " x " can be used for characterizing), or for for characterizing position of the product defects image in target product subgraph
Information (such as the coordinate of the pixel included by product defects image in target product subgraph), or for for characterizing
The information (such as " crackle " of the classification of product defects indicated by product defects image；" cut ").It should be noted that at this
In, the pending subgraph for inputting above-mentioned convolutional neural networks is at least two, correspondingly, the pending subgraph institute obtained
Including target product subgraph in product defects image description information be at least two.
In the present embodiment, convolutional neural networks can be used for characterizing image and the production in the product image included by image
The correspondence of the description information of product defect image.Above-mentioned convolutional neural networks may include convolutional layer, extracted region network and
Grader, wherein convolutional layer can be used for extracting characteristics of image, and extracted region network can be used for being extracted according to convolutional layer
Characteristics of image generates those suspected defects region, and the characteristics of image that grader can be used for being generated according to convolutional layer is to extracted region net
Each those suspected defects region that network is generated carries out Accurate Analysis, and generate, the product in the product image included by image lacks
Fall into the description information of image.It should be noted that above-mentioned executive agent can profit (such as Training, nothing in various manners
The modes such as supervised training) training obtain above-mentioned convolutional neural networks.
Step 504, pending image is inputted into convolutional neural networks, obtains the target product figure included by pending image
The description information of product defects image as in.
In the present embodiment, the executive agent that the method for generating information is run thereon can be defeated by pending image
Enter above-mentioned convolutional neural networks, obtains the description letter of the product defects image in the target product image included by pending image
Breath.
Herein, the description information of the product defects image in the target product image included by pending image can wrap
It includes but is not limited at least one of following：Word, number, symbol.Specifically, in target product image included by pending image
The description information of product defects image can be whether to include the information of product defects image for characterizing target product image
(such as " o "；" x ", wherein symbol " o ", which can be used for characterizing, includes, and symbol " x ", which can be used for characterizing, does not include), or to use
Information (such as the pixel included by product defects image in position of the characterization product defects image in target product image
Coordinate in target product image), or be the letter of the classification for characterizing the product defects indicated by product defects image
Breath (such as " crackle "；" cut ").
From figure 5 it can be seen that compared with the corresponding embodiments of Fig. 2, the method for generating information in the present embodiment
Flow 500 highlight by pending image input convolutional neural networks the step of.The scheme of the present embodiment description can be with as a result,
The related data for being more used for analysis product defect image is introduced, to realize more fully information generation.
With further reference to Fig. 6, as the realization to method shown in above-mentioned each figure, this application provides one kind for generating letter
One embodiment of the device of breath, the device embodiment is corresponding with embodiment of the method shown in Fig. 2, which can specifically answer
For in various executive agents.
As shown in fig. 6, the device 600 for generating information of the present embodiment includes：Acquiring unit 601, cutting unit 602
With the first generation unit 603.Wherein, acquiring unit 601 is configured to the pending image that acquisition includes target product image,
In, target product image includes product defects image；Cutting unit 602 is configured to carry out cutting to pending image, obtains
At least two pending subgraphs, wherein pending subgraph includes target product subgraph；First generation unit 603 configures
For at least two pending subgraphs to be inputted to convolutional neural networks trained in advance respectively, obtains pending subgraph and wrapped
The description information of product defects image in the target product subgraph included, wherein convolutional neural networks for characterize image with
The correspondence of the description information of product defects image in product image included by image.
In the present embodiment, acquiring unit 601 can obtain the pending image including target product image, wherein mesh
It includes product defects image to mark product image.Herein, target product image is the image of target product.Target product can be
The product of product defects to be determined included by it.Product defects are to fail to meet consumption or using product institute included by product
The defect (such as crackle, cut etc.) of prerequisite safety requirements.It is prestored specifically, acquiring unit 601 can obtain
In the local pending image for including target product image；Alternatively, acquiring unit 601 can obtain terminal (such as Fig. 1 institutes
The terminal device 101 that shows, 102,103) the pending image for including target product image that sends.Herein, terminal can be first
First target product is shot, acquisition includes the pending image of target product image；Further, terminal can will be obtained
The pending image obtained is sent to acquiring unit 601.
In the present embodiment, the pending image obtained based on acquiring unit 601, cutting unit 602 can be to pending
Image carries out cutting, obtains at least two pending subgraphs, wherein pending subgraph may include target product subgraph
Picture.Target product subgraph can be the topography of target product image.It should be noted that target product subgraph can be with
Including product defects image.
In the present embodiment, pending image can be rectangle, square, circle etc..It is above-mentioned that pending image is carried out
Cutting obtains at least two pending subgraphs and is specifically as follows with various slit modes to the progress cutting of pending image, obtains
Obtain at least two pending subgraphs.
In the present embodiment, the pending subgraph obtained based on cutting unit 602, the first generation unit 603 can be with
At least two pending subgraphs are inputted to convolutional neural networks trained in advance respectively, are obtained included by pending subgraph
The description information of product defects image in target product subgraph.Wherein, target product included by pending subgraph
The description information of product defects image in image can include but is not limited at least one of following：Word, number, symbol.Tool
Body, the description information of the product defects image in target product subgraph included by pending subgraph can be for table
Whether sign target product subgraph includes the information of product defects image, or to be produced in target for characterizing product defects image
The information of position in product subgraph, or the letter for the classification for characterizing the product defects indicated by product defects image
Breath.It should be noted that herein, the pending subgraph for inputting above-mentioned convolutional neural networks is at least two, correspondingly, institute
The description information of product defects image in the target product subgraph included by pending subgraph obtained is at least two.
In the present embodiment, convolutional neural networks can be used for characterizing image and the production in the product image included by image
The correspondence of the description information of product defect image.Above-mentioned convolutional neural networks may include convolutional layer, extracted region network and
Grader, wherein convolutional layer can be used for extracting characteristics of image, and extracted region network can be used for being extracted according to convolutional layer
Characteristics of image generates those suspected defects region, and the characteristics of image that grader can be used for being generated according to convolutional layer is to extracted region net
Each those suspected defects region that network is generated carries out Accurate Analysis, and generates the product defects in the product image included by image
The description information of image.
In some optional realization methods of the present embodiment, the device 600 for generating information can also include：Second
Generation unit (not shown) is configured to pending image inputting convolutional neural networks, obtains pending image and wrapped
The description information of product defects image in the target product image included.
In some optional realization methods of the present embodiment, convolutional neural networks can be trained as follows
It arrives：Obtain multiple pending images of sample including sample product image, wherein sample product image includes sample product defect
Image；Obtain the sample product figure included by the pending image of each sample in the pending image of sample demarcated, multiple
The description information of sample product defect image as in；It, will be every in the pending image of multiple samples using machine learning method
A pending image of sample is as input, by the pending image of each sample in pending images of sample demarcated, multiple
The description information of sample product defect image in included sample product image obtains convolutional Neural net as output, training
Network.
In some optional realization methods of the present embodiment, it is pending to obtain multiple samples including sample product image
Image, including：Obtain multiple sample product defect images；Each sample product in multiple sample product defect images is lacked
Image is fallen into, the sample to be fused of the sample product defect image and sample product image preset including to be fused is pending
Image is merged, and generation includes the pending image of sample of sample product image, wherein sample product image to be fused does not wrap
Include product defects image.
In some optional realization methods of the present embodiment, multiple sample product defect images are obtained, including：It obtains more
A initial data；Multiple initial data are inputted into defect trained in advance respectively and generate model, obtain multiple product defects images,
Wherein, defect generates the correspondence that model can be used for characterize data and product defects image；The multiple products that will be obtained
Defect image is determined as multiple sample product defect images.
In some optional realization methods of the present embodiment, defect generates model can be trained as follows
It arrives：Obtain training sample, wherein training sample includes multiple sample initial data and multiple product defects images to be entered；It carries
The production pre-established is taken to fight network, wherein production confrontation network may include generating network and differentiating network, generate
Network can be used for utilizing inputted data to generate image, differentiate that network is determined for whether inputted image makes a living
The image exported at network；It will be given birth to using multiple sample initial data as the input for generating network using machine learning method
At the image and multiple product defects images to be entered of network output as the input for differentiating network, to generating network and differentiating net
Network is trained, and the generation network after training, which is determined as defect, generates model.
In some optional realization methods of the present embodiment, multiple sample product defect images are obtained, including：It obtains more
A original image；Product defects image is generated on each original image in multiple original images so that the product generated
Defect image meets preset condition；The multiple product defects images generated are determined as multiple sample product defect images.
The device 600 that above-described embodiment of the application provides includes target product image by the acquisition of acquiring unit 601
Pending image, wherein target product image includes product defects image；Then cutting unit 602 carries out pending image
Cutting obtains at least two pending subgraphs, wherein pending subgraph includes target product subgraph；Then the first life
At least two pending subgraphs are inputted to convolutional neural networks trained in advance respectively at unit 603, obtain pending subgraph
As the description information of the product defects image in included target product subgraph, to by being cut to pending image
Point, obtain the significantly more pending subgraph of local message, and the convolutional Neural that the input of pending subgraph is trained in advance
Network obtains information more structurally sound, for characterizing the product defects included by product, improves the accurate of information generation
Property.
As shown in fig. 7, computer system 700 includes central processing unit (CPU) 701, it can be read-only according to being stored in
Program in memory (ROM) 702 or be loaded into the program in random access storage device (RAM) 703 from storage section 708 and
Execute various actions appropriate and processing.In RAM 703, also it is stored with system 700 and operates required various programs and data.
CPU 701, ROM 702 and RAM 703 are connected with each other by bus 704.Input/output (I/O) interface 705 is also connected to always
Line 704.
It is connected to I/O interfaces 705 with lower component：Importation 706 including keyboard, mouse etc.；It is penetrated including such as cathode
The output par, c 707 of spool (CRT), liquid crystal display (LCD) etc. and loud speaker etc.；Storage section 708 including hard disk etc.；
And the communications portion 709 of the network interface card including LAN card, modem etc..Communications portion 709 via such as because
The network of spy's net executes communication process.Driver 710 is also according to needing to be connected to I/O interfaces 705.Detachable media 711, such as
Disk, CD, magneto-optic disk, semiconductor memory etc. are mounted on driver 710, as needed in order to be read from thereon
Computer program be mounted into storage section 708 as needed.
Particularly, in accordance with an embodiment of the present disclosure, it may be implemented as computer above with reference to the process of flow chart description
Software program.For example, embodiment of the disclosure includes a kind of computer program product comprising be carried on computer-readable medium
On computer program, which includes the program code for method shown in execution flow chart.In such reality
It applies in example, which can be downloaded and installed by communications portion 709 from network, and/or from detachable media
711 are mounted.When the computer program is executed by central processing unit (CPU) 701, limited in execution the present processes
Above-mentioned function.It should be noted that computer-readable medium described herein can be computer-readable signal media or
Computer readable storage medium either the two arbitrarily combines.Computer readable storage medium for example can be --- but
Be not limited to --- electricity, magnetic, optical, electromagnetic, infrared ray or semiconductor system, device or device, or arbitrary above combination.
The more specific example of computer readable storage medium can include but is not limited to：Electrical connection with one or more conducting wires,
Portable computer diskette, hard disk, random access storage device (RAM), read-only memory (ROM), erasable type may be programmed read-only deposit
Reservoir (EPROM or flash memory), optical fiber, portable compact disc read-only memory (CD-ROM), light storage device, magnetic memory
Part or above-mentioned any appropriate combination.In this application, computer readable storage medium can any be included or store
The tangible medium of program, the program can be commanded the either device use or in connection of execution system, device.And
In the application, computer-readable signal media may include the data letter propagated in a base band or as a carrier wave part
Number, wherein carrying computer-readable program code.Diversified forms may be used in the data-signal of this propagation, including but not
It is limited to electromagnetic signal, optical signal or above-mentioned any appropriate combination.Computer-readable signal media can also be computer
Any computer-readable medium other than readable storage medium storing program for executing, the computer-readable medium can send, propagate or transmit use
In by instruction execution system, device either device use or program in connection.Include on computer-readable medium
Program code can transmit with any suitable medium, including but not limited to：Wirelessly, electric wire, optical cable, RF etc., Huo Zheshang
Any appropriate combination stated.
Flow chart in attached drawing and block diagram, it is illustrated that according to the system of the various embodiments of the application, method and computer journey
The architecture, function and operation in the cards of sequence product.In this regard, each box in flowchart or block diagram can generation
A part for a part for one module, program segment, or code of table, the module, program segment, or code includes one or more uses
The executable instruction of the logic function as defined in realization.It should also be noted that in some implementations as replacements, being marked in box
The function of note can also occur in a different order than that indicated in the drawings.For example, two boxes succeedingly indicated are actually
It can be basically executed in parallel, they can also be executed in the opposite order sometimes, this is depended on the functions involved.Also it to note
Meaning, the combination of each box in block diagram and or flow chart and the box in block diagram and or flow chart can be with holding
The dedicated hardware based system of functions or operations as defined in row is realized, or can use specialized hardware and computer instruction
Combination realize.
Being described in unit involved in the embodiment of the present application can be realized by way of software, can also be by hard
The mode of part is realized.Described unit can also be arranged in the processor, for example, can be described as：A kind of processor packet
Include acquiring unit, cutting unit and the first generation unit.Wherein, the title of these units is not constituted to this under certain conditions
The restriction of unit itself, for example, acquiring unit is also described as " obtaining the unit of pending image ".
As on the other hand, present invention also provides a kind of computer-readable medium, which can be
Included in device described in above-described embodiment；Can also be individualism, and without be incorporated the device in.Above-mentioned calculating
Machine readable medium carries one or more program, when said one or multiple programs are executed by the device so that should
Device：Acquisition includes the pending image of target product image, wherein target product image includes product defects image；It treats
It handles image and carries out cutting, obtain at least two pending subgraphs, wherein pending subgraph includes target product subgraph
Picture；At least two pending subgraphs are inputted to convolutional neural networks trained in advance respectively, are obtained for characterizing pending son
The information of the product defects image in target product subgraph included by image, wherein convolutional neural networks are used for phenogram
As the correspondence with the information for characterizing the product defects image in the product image included by image.
Above description is only the preferred embodiment of the application and the explanation to institute's application technology principle.People in the art
Member should be appreciated that invention scope involved in the application, however it is not limited to technology made of the specific combination of above-mentioned technical characteristic
Scheme, while should also cover in the case where not departing from foregoing invention design, it is carried out by above-mentioned technical characteristic or its equivalent feature
Other technical solutions of arbitrary combination and formation.Such as features described above has similar work(with (but not limited to) disclosed herein
Can technical characteristic replaced mutually and the technical solution that is formed.
Claims (12)
1. a kind of method for generating information, including：
Acquisition includes the pending image of target product image, wherein the target product image includes product defects image；
Cutting is carried out to the pending image, obtains at least two pending subgraphs, wherein pending subgraph includes mesh
Mark product subgraph；
Described at least two pending subgraphs are inputted to convolutional neural networks trained in advance respectively, obtain pending subgraph
The description information of product defects image in included target product subgraph, wherein the convolutional neural networks are used for table
Levy the correspondence of image and the description information of the product defects image in the product image included by image.
2. according to the method described in claim 1, wherein, the method further includes：
The pending image is inputted into the convolutional neural networks, obtains the target product figure included by the pending image
The description information of product defects image as in.
3. method according to claim 1 or 2, wherein training obtains the convolutional neural networks as follows：
Obtain multiple pending images of sample including sample product image, wherein sample product image includes that sample product lacks
Fall into image；
Obtain the sample product included by the pending image of each sample in the pending image of sample demarcated, the multiple
The description information of sample product defect image in image；
Using machine learning method, using the pending image of each sample in the pending image of the multiple sample as input,
It will be in the sample product image included by the pending image of each sample in the pending image of sample demarcate, the multiple
Sample product defect image description information as output, training obtain convolutional neural networks.
4. described to obtain multiple pending figures of sample including sample product image according to the method described in claim 3, wherein
Picture, including：
Obtain multiple sample product defect images；
For each sample product defect image in the multiple sample product defect image, by the sample product defect image
It is merged with the pending image of sample to be fused of sample product image preset including to be fused, generation includes sample production
The pending image of sample of product image, wherein sample product image to be fused does not include product defects image.
It is described to obtain multiple sample product defect images 5. according to the method described in claim 4, wherein, including：
Obtain multiple initial data；
The multiple initial data is inputted into defect trained in advance respectively and generates model, obtains multiple product defects images,
In, the defect generates the correspondence that model is used for characterize data and product defects image；
The multiple product defects images obtained are determined as multiple sample product defect images.
6. according to the method described in claim 5, wherein, the defect generates model, and training obtains as follows：
Obtain training sample, wherein the training sample includes multiple sample initial data and multiple product defects figures to be entered
Picture；
Extract the production confrontation network pre-established, wherein the production confrontation network is including generating network and differentiating net
Network, the network that generates is used to utilize inputted data to generate image, described to differentiate network for determining inputted image
Whether by the image for generating network and exporting；
Using machine learning method, using the multiple sample initial data as the input for generating network, by the generation
The image and the multiple product defects image to be entered of network output are as the input for differentiating network, to the generation net
Network and the differentiation network are trained, and the generation network after training, which is determined as defect, generates model.
7. to go the method described in 4 according to right, wherein the multiple sample product defect images of acquisition, including：
Obtain multiple original images；
Product defects image is generated on each original image in the multiple original image so that the product defects generated
Image meets preset condition；
The multiple product defects images generated are determined as multiple sample product defect images.
8. a kind of device for generating information, including：
Acquiring unit is configured to the pending image that acquisition includes target product image, wherein the target product image packet
Include product defects image；
Cutting unit is configured to carry out cutting to the pending image, obtains at least two pending subgraphs, wherein
Pending subgraph includes target product subgraph；
First generation unit is configured to inputting described at least two pending subgraphs into convolutional Neural trained in advance respectively
Network obtains the description information of the product defects image in the target product subgraph included by pending subgraph, wherein institute
Convolutional neural networks are stated for characterizing image and the description information of the product defects image in the product image included by image
Correspondence.
9. device according to claim 8, wherein described device further includes：
Second generation unit is configured to the pending image inputting the convolutional neural networks, obtain described pending
The description information of the product defects image in target product image included by image.
10. device according to claim 8 or claim 9, wherein training obtains the convolutional neural networks as follows：
Obtain multiple pending images of sample including sample product image, wherein sample product image includes that sample product lacks
Fall into image；
Obtain the sample product included by the pending image of each sample in the pending image of sample demarcated, the multiple
The description information of sample product defect image in image；
Using machine learning method, using the pending image of each sample in the pending image of the multiple sample as input,
It will be in the sample product image included by the pending image of each sample in the pending image of sample demarcate, the multiple
Sample product defect image description information as output, training obtain convolutional neural networks.
11. a kind of server, including：
One or more processors；
Storage device, for storing one or more programs,
When one or more of programs are executed by one or more of processors so that one or more of processors are real
The now method as described in any in claim 1-7.
12. a kind of computer-readable medium, is stored thereon with computer program, wherein the program is realized when being executed by processor
Method as described in any in claim 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810201427.6A CN108460761A (en) | 2018-03-12 | 2018-03-12 | Method and apparatus for generating information |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810201427.6A CN108460761A (en) | 2018-03-12 | 2018-03-12 | Method and apparatus for generating information |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108460761A true CN108460761A (en) | 2018-08-28 |
Family
ID=63217142
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810201427.6A Pending CN108460761A (en) | 2018-03-12 | 2018-03-12 | Method and apparatus for generating information |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108460761A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109410253A (en) * | 2018-11-06 | 2019-03-01 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating information |
| CN109614979A (en) * | 2018-10-11 | 2019-04-12 | 北京大学 | A kind of data augmentation method and image classification method based on selection with generation |
| CN109949286A (en) * | 2019-03-12 | 2019-06-28 | 北京百度网讯科技有限公司 | Method and apparatus for output information |
| CN110675359A (en) * | 2019-06-29 | 2020-01-10 | 创新奇智(南京)科技有限公司 | Defect sample generation method and system for steel coil surface and electronic equipment |
| CN111652297A (en) * | 2020-05-25 | 2020-09-11 | 哈尔滨市科佳通用机电股份有限公司 | Fault picture generation method for image detection model training |
| CN111768381A (en) * | 2020-06-29 | 2020-10-13 | 北京百度网讯科技有限公司 | Part defect detection method and device and electronic equipment |
| CN111784567A (en) * | 2020-07-03 | 2020-10-16 | 北京字节跳动网络技术有限公司 | Method, apparatus, electronic device, and computer-readable medium for converting an image |
| CN112488984A (en) * | 2019-09-11 | 2021-03-12 | 中信戴卡股份有限公司 | Method and device for acquiring defect picture generation network and defect picture generation method |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105118044A (en) * | 2015-06-16 | 2015-12-02 | 华南理工大学 | Method for automatically detecting defects of wheel-shaped cast product |
| CN105335689A (en) * | 2014-08-06 | 2016-02-17 | 阿里巴巴集团控股有限公司 | Character recognition method and apparatus |
| CN105891215A (en) * | 2016-03-31 | 2016-08-24 | 浙江工业大学 | Welding visual detection method and device based on convolutional neural network |
| CN106709486A (en) * | 2016-11-11 | 2017-05-24 | 南京理工大学 | Automatic license plate identification method based on deep convolutional neural network |
| CN106803071A (en) * | 2016-12-29 | 2017-06-06 | 浙江大华技术股份有限公司 | Object detecting method and device in a kind of image |
| CN106886795A (en) * | 2017-02-17 | 2017-06-23 | 北京维弦科技有限责任公司 | Object identification method based on the obvious object in image |
| CN107358262A (en) * | 2017-07-13 | 2017-11-17 | 京东方科技集团股份有限公司 | The sorting technique and sorter of a kind of high-definition picture |
| CN107358636A (en) * | 2017-06-16 | 2017-11-17 | 华南理工大学 | A kind of rarefaction defect image generating method based on textures synthesis |
| 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 |
| 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 |
| US20180012359A1 (en) * | 2016-07-06 | 2018-01-11 | Marinko Venci Sarunic | Systems and Methods for Automated Image Classification and Segmentation |
-
2018
- 2018-03-12 CN CN201810201427.6A patent/CN108460761A/en active Pending
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105335689A (en) * | 2014-08-06 | 2016-02-17 | 阿里巴巴集团控股有限公司 | Character recognition method and apparatus |
| CN105118044A (en) * | 2015-06-16 | 2015-12-02 | 华南理工大学 | Method for automatically detecting defects of wheel-shaped cast product |
| CN105891215A (en) * | 2016-03-31 | 2016-08-24 | 浙江工业大学 | Welding visual detection method and device based on convolutional neural network |
| US20180012359A1 (en) * | 2016-07-06 | 2018-01-11 | Marinko Venci Sarunic | Systems and Methods for Automated Image Classification and Segmentation |
| CN106709486A (en) * | 2016-11-11 | 2017-05-24 | 南京理工大学 | Automatic license plate identification method based on deep convolutional neural network |
| CN106803071A (en) * | 2016-12-29 | 2017-06-06 | 浙江大华技术股份有限公司 | Object detecting method and device in a kind of image |
| CN106886795A (en) * | 2017-02-17 | 2017-06-23 | 北京维弦科技有限责任公司 | Object identification method based on the obvious object in image |
| CN107358636A (en) * | 2017-06-16 | 2017-11-17 | 华南理工大学 | A kind of rarefaction defect image generating method based on textures synthesis |
| CN107358262A (en) * | 2017-07-13 | 2017-11-17 | 京东方科技集团股份有限公司 | The sorting technique and sorter of a kind of high-definition picture |
| 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 |
| 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 |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109614979A (en) * | 2018-10-11 | 2019-04-12 | 北京大学 | A kind of data augmentation method and image classification method based on selection with generation |
| CN109614979B (en) * | 2018-10-11 | 2023-05-02 | 北京大学 | Data augmentation method and image classification method based on selection and generation |
| CN109410253A (en) * | 2018-11-06 | 2019-03-01 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating information |
| CN109949286A (en) * | 2019-03-12 | 2019-06-28 | 北京百度网讯科技有限公司 | Method and apparatus for output information |
| US11341626B2 (en) | 2019-03-12 | 2022-05-24 | Beijing Baidu Netcom Science And Technology Co., Ltd. | Method and apparatus for outputting information |
| CN110675359A (en) * | 2019-06-29 | 2020-01-10 | 创新奇智(南京)科技有限公司 | Defect sample generation method and system for steel coil surface and electronic equipment |
| CN112488984A (en) * | 2019-09-11 | 2021-03-12 | 中信戴卡股份有限公司 | Method and device for acquiring defect picture generation network and defect picture generation method |
| CN111652297A (en) * | 2020-05-25 | 2020-09-11 | 哈尔滨市科佳通用机电股份有限公司 | Fault picture generation method for image detection model training |
| CN111652297B (en) * | 2020-05-25 | 2021-05-25 | 哈尔滨市科佳通用机电股份有限公司 | Fault picture generation method for image detection model training |
| CN111768381A (en) * | 2020-06-29 | 2020-10-13 | 北京百度网讯科技有限公司 | Part defect detection method and device and electronic equipment |
| CN111784567A (en) * | 2020-07-03 | 2020-10-16 | 北京字节跳动网络技术有限公司 | Method, apparatus, electronic device, and computer-readable medium for converting an image |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108460761A (en) | Method and apparatus for generating information | |
| CN108776808A (en) | A kind of method and apparatus for detecting ladle corrosion defect | |
| CN108520324A (en) | Method and apparatus for generating information | |
| CN108510472B (en) | Method and apparatus for handling image | |
| CN109582982A (en) | Method and apparatus for translated speech | |
| CN108229485A (en) | For testing the method and apparatus of user interface | |
| CN108764487A (en) | For generating the method and apparatus of model, the method and apparatus of information for identification | |
| CN109446990A (en) | Method and apparatus for generating information | |
| CN108038052A (en) | Automatic test management method, device, terminal device and storage medium | |
| CN108986169A (en) | Method and apparatus for handling image | |
| CN109410253B (en) | For generating method, apparatus, electronic equipment and the computer-readable medium of information | |
| CN108256591A (en) | For the method and apparatus of output information | |
| CN109308469A (en) | Method and apparatus for generating information | |
| CN109389640A (en) | Image processing method and device | |
| CN108197652A (en) | For generating the method and apparatus of information | |
| CN109214501A (en) | The method and apparatus of information for identification | |
| CN108509892A (en) | Method and apparatus for generating near-infrared image | |
| CN109344752A (en) | Method and apparatus for handling mouth image | |
| CN109634833A (en) | A kind of Software Defects Predict Methods and device | |
| CN108629611A (en) | Method and apparatus for distributing resource | |
| CN108389172A (en) | Method and apparatus for generating information | |
| CN108509921A (en) | Method and apparatus for generating information | |
| CN108491823A (en) | Method and apparatus for generating eye recognition model | |
| CN108491825A (en) | information generating method and device | |
| CN108427941A (en) | Method, method for detecting human face and device for generating Face datection model |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180828 |