CN108846419A - Single page high load image-recognizing method, device, computer equipment and storage medium - Google Patents
Single page high load image-recognizing method, device, computer equipment and storage medium Download PDFInfo
- Publication number
- CN108846419A CN108846419A CN201810513879.8A CN201810513879A CN108846419A CN 108846419 A CN108846419 A CN 108846419A CN 201810513879 A CN201810513879 A CN 201810513879A CN 108846419 A CN108846419 A CN 108846419A
- Authority
- CN
- China
- Prior art keywords
- convolutional neural
- neural networks
- single page
- local
- picture
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Abstract
The embodiment of the present application discloses a kind of single page high load image-recognizing method, device, computer equipment and storage medium.This method includes：Obtain the multiple convolutional neural networks models for being used to carry out image recognition stored in the local single page application of user terminal；It obtains picture to be identified, and using the picture to be identified as the input of multiple convolutional neural networks models in the local single page application, obtains and each convolutional neural networks model result to be processed correspondingly；If the local picture recognition result is to identify successfully, the local corresponding keyword of picture recognition result is shown；If local picture recognition result is recognition failures, picture to be identified is uploaded to background server, receives the backstage recognition result fed back by background server.This method is realized and is identified by convolutional neural networks model to picture in the local single page application of user terminal, is quickly obtained recognition result, is identified it is not necessary that picture is directly uploaded to background server, improve recognition efficiency.
Description
Technical field
This application involves image identification technical field more particularly to a kind of single page high load image-recognizing methods, device, meter
Calculate machine equipment and storage medium.
Background technique
Currently, the image recognition application of mainstream, figure can just be carried out by requiring to upload to picture into enterprise's background server
As identification, such as Baidu's picture searching.It cannot be uploaded in client application neural network recognization picture, if network speed is poor
Picture speed is slower, will influence the speed of image recognition result feedback.
Summary of the invention
This application provides a kind of single page high load image-recognizing method, device, computer equipment and storage mediums, it is intended to
The image recognition application for solving mainstream in the prior art, figure can just be carried out by requiring to upload to picture into enterprise's background server
The problem of as identification, if network speed is poor, uploading pictures speed is slower, will influence the speed of image recognition result feedback.
In a first aspect, this application provides a kind of single page high load image-recognizing methods comprising：
Obtain the multiple convolutional neural networks moulds for being used to carry out image recognition stored in the local single page application of user terminal
Type；
Picture to be identified is obtained, and using the picture to be identified as multiple convolutional Neural nets in the local single page application
The input of network model obtains and each convolutional neural networks model result to be processed correspondingly；
If the local picture recognition result is to identify successfully, by the local corresponding key of picture recognition result
Word is shown；
If local picture recognition result is recognition failures, picture to be identified is uploaded to background server, is received by rear
The backstage recognition result of platform server feedback.
Second aspect, this application provides a kind of single page high load pattern recognition devices comprising：
Model acquiring unit, it is multiple for carrying out image recognition for being stored in obtaining the local single page application of user terminal
Convolutional neural networks model；
Local picture recognition unit, for obtaining picture to be identified, and using the picture to be identified as described local single
The input of multiple convolutional neural networks models, obtains to be processed correspondingly with each convolutional neural networks model in page application
As a result；
Display unit, if being to identify successfully for the local picture recognition result, by the local picture recognition
As a result corresponding keyword is shown；
Backstage recognition unit, if being recognition failures for local picture recognition result, after picture to be identified is uploaded to
Platform server receives the backstage recognition result fed back by background server.
The third aspect, the application provide a kind of computer equipment again, including memory, processor and are stored in described deposit
On reservoir and the computer program that can run on the processor, the processor realize this when executing the computer program
The described in any item single page high load image-recognizing methods provided are provided.
Fourth aspect, present invention also provides a kind of storage mediums, wherein the storage medium is stored with computer program,
The computer program includes program instruction, and described program instruction makes the processor execute the application when being executed by a processor
The described in any item single page high load image-recognizing methods provided.
The application provides a kind of single page high load image-recognizing method, device, computer equipment and storage medium.This method
It is multiple for carrying out the convolutional neural networks model of image recognition by being stored in obtaining the local single page application of user terminal；It obtains
Take picture to be identified, and using the picture to be identified as in the local single page application multiple convolutional neural networks models it is defeated
Enter, obtains and each convolutional neural networks model result to be processed correspondingly；If the local picture recognition result is
It identifies successfully, the local corresponding keyword of picture recognition result is shown；If local picture recognition result is
Picture to be identified is uploaded to background server, receives the backstage recognition result fed back by background server by recognition failures.The party
Method is realized and is identified by convolutional neural networks model to picture in the local single page application of user terminal, and knowledge is quickly obtained
Not as a result, identifying it is not necessary that picture is directly uploaded to background server, recognition efficiency is improved.
Detailed description of the invention
Technical solution in ord to more clearly illustrate embodiments of the present application, below will be to needed in embodiment description
Attached drawing is briefly described, it should be apparent that, the accompanying drawings in the following description is some embodiments of the present application, general for this field
For logical technical staff, without creative efforts, it is also possible to obtain other drawings based on these drawings.
Fig. 1 is a kind of schematic flow diagram of single page high load image-recognizing method provided by the embodiments of the present application；
Fig. 2 is a kind of another schematic flow diagram of single page high load image-recognizing method provided by the embodiments of the present application；
Fig. 3 is a kind of sub-process schematic diagram of single page high load image-recognizing method provided by the embodiments of the present application；
Fig. 4 is a kind of another sub-process schematic diagram of single page high load image-recognizing method provided by the embodiments of the present application；
Fig. 5 is a kind of another sub-process schematic diagram of single page high load image-recognizing method provided by the embodiments of the present application；
Fig. 6 is a kind of schematic block diagram of single page high load pattern recognition device provided by the embodiments of the present application；
Fig. 7 is a kind of another schematic block diagram of single page high load pattern recognition device provided by the embodiments of the present application；
Fig. 8 is a kind of subelement schematic block diagram of single page high load pattern recognition device provided by the embodiments of the present application；
Fig. 9 is a kind of schematic frame of another subelement of single page high load pattern recognition device provided by the embodiments of the present application
Figure；
Figure 10 is that a kind of another subelement of single page high load pattern recognition device provided by the embodiments of the present application is schematic
Block diagram；
Figure 11 is a kind of schematic block diagram of computer equipment provided by the embodiments of the present application.
Specific embodiment
Below in conjunction with the attached drawing in the embodiment of the present application, technical solutions in the embodiments of the present application carries out clear, complete
Site preparation description, it is clear that described embodiment is some embodiments of the present application, instead of all the embodiments.Based on this Shen
Please in embodiment, every other implementation obtained by those of ordinary skill in the art without making creative efforts
Example, shall fall in the protection scope of this application.
It should be appreciated that ought use in this specification and in the appended claims, term " includes " and "comprising" instruction
Described feature, entirety, step, operation, the presence of element and/or component, but one or more of the other feature, whole is not precluded
Body, step, operation, the presence or addition of element, component and/or its set.
It is also understood that mesh of the term used in this present specification merely for the sake of description specific embodiment
And be not intended to limit the application.As present specification and it is used in the attached claims, unless on
Other situations are hereafter clearly indicated, otherwise " one " of singular, "one" and "the" are intended to include plural form.
It will be further appreciated that the term "and/or" used in present specification and the appended claims is
Refer to any combination and all possible combinations of one or more of associated item listed, and including these combinations.
Referring to Fig. 1, Fig. 1 is a kind of exemplary flow of single page high load image-recognizing method provided by the embodiments of the present application
Figure.This method is applied in the terminals such as desktop computer, laptop computer, tablet computer.As shown in Figure 1, the method comprising the steps of
S101~S104.
S101, obtain stored in the local single page application of user terminal it is multiple for carrying out the convolutional Neural nets of image recognition
Network model.
In the present embodiment, the convolutional neural networks model stored in local single page application is the peace in downloading single page application
When dress packet, the convolutional Neural model that can be used for image recognition carried in installation kit, which, which is stored in, passes through
In the application of local single page constructed by AngularJS (being the preceding end instruction based on javascript).User terminal is mounted with the peace
After dress packet, image recognition can be carried out by the convolutional neural networks model stored in the application of local single page.
Wherein, AngularJS is only needed HTML (i.e. hypertext markup language), CSS (i.e. cascading style sheets) and
JavaScript (being a kind of literal translation formula scripting language) can create single page application in client, and target is so that exploitation and survey
Examination is easier, the performance of enhancing MVC Web application.
So-called single page application refers to integrating multiple functions or even whole system on one page with regard to only one page
Face, all business functions are all its submodules, are articulated in main interface by specific mode.User terminal in the application
It after opening is clicked in local single page application, is similar to and enters a single machine Webpage, have on the single machine Webpage simple
Operation interface such as adds the operating space of picture to be identified, and the viewing area of output recognition result.
The convolutional neural networks model stored in local single page application is to be led on enterprise's background server by that will preselect
The picture of the image data set (such as described image data set) entered inputs convolutional neural networks, and is trained and obtains.Its
In, it is the current maximum data of image recognition in the world that ImageNet, which is a computer vision system identification project name,
Library, it is the computer scientist of U.S. Stamford, simulates the image recognition database that the identifying system of the mankind is established.
ImageNet data set (namely described image data set) has more than 1,400 ten thousand width pictures, covers a classification more than 20,000；Wherein have more than
Million picture has the mark of specific classification mark and objects in images position, and the size of the ImageNet data set is about
1TB。
In one embodiment, as shown in Fig. 2, further including before the step S101：
S1001, the installation kit that the local single page including convolutional neural networks model is applied is obtained；
S1002, the installation kit of the local single page application is decompressed, obtains including multiple convolutional neural networks models
Local single page application.
In the present embodiment, the installation kit of local single page application (is stored with the convolution for image recognition in the installation kit
Neural network model) in included convolutional neural networks model, be that a large amount of image knowledge is carried out on enterprise's background server
Not Xun Lian and obtain.Input training convolutional neural networks are used as by described image data set i.e. on enterprise's background server,
Obtained convolutional neural networks model includes multiple kinds of class models, such as has the convolutional neural networks of the identification vehicles
Model, the convolutional neural networks model etc. of insighted object of not moving, after the completion of the convolutional neural networks model training of above-mentioned multiple types,
The installation kit (such as apk installation kit or iOs installation kit) of single page application can be introduced directly into.When user has downloaded the installation of single page application
Bao Hou, decompression installs the installation kit of single page application on local terminal, and can be obtained one, there is the single page of identification image to answer
With being identified it is not necessary that picture to be identified to be uploaded on enterprise's background server.
In one embodiment, as shown in figure 3, the step S1001 includes：
S10011, the training data training using the image data set imported in advance as the convolutional neural networks model
The convolutional neural networks model, the convolutional neural networks model after being trained；
S10012, the convolutional neural networks model after the training is packed into the installation kit of local single page application.
In the present embodiment, above-mentioned training process both may be selected to carry out at enterprise's background server end, can also be in user terminal
It carries out.Such as when user terminal is trained, the type (such as vehicles, animal, plant etc.) of images to be recognized can choose,
Directly according to the type of selected images to be recognized, the corresponding Image that obtains is in data set, and the conduct in the local of user terminal
The training data of convolutional neural networks obtains convolutional neural networks model in user terminal training, and by the convolution after the training
Neural network model is packed into the installation kit of local single page application, for this terminal itself use, or for other downloadings
Installation kit that the user terminal uploads (installation kit be include convolutional neural networks model the application of local single page installation kit).
Wherein, in order to improve the accuracy to picture recognition, the convolutional Neural net importeding into the installation kit of single page application
The type of network model is The more the better, but the control of the size for the installation kit for generally applying entire single page is in 200M or so
Meet basic picture recognition requirement, namely import the convolutional neural networks model of 10-20 kind (such as identification can be met and handed over
Logical tool, animal, the image of the classifications such as plant).
In one embodiment, as shown in figure 4, the step S10011 includes：
S10011a, the picture of described image data set is subjected to convolution according to filter, obtains low-level image feature；
S10011b, low-level image feature is sampled according to sampling window, obtains pond feature；
S10011c, low-level image feature and pond feature are attached by full articulamentum, obtain convolutional neural networks mould
Type.
In the present embodiment, convolutional neural networks are made of convolutional layer, pond layer, full articulamentum.Wherein convolutional layer and pond
Change layer cooperation, form multiple convolution groups, successively extract feature, completes classification eventually by several full articulamentums.
Convolutional network is inherently a kind of mapping for being input to output, it can learn largely to input between output
Mapping relations, without the accurate mathematic(al) representation between any output and input, as long as with known mode to volume
Product network is trained, and network just has the mapping ability between inputoutput pair.What convolutional network executed is to have tutor's training,
So its sample set be by shaped like：The vector of (input vector, ideal output vector) is to composition.All these vectors pair, all
The practical " RUN " result for the system that network will simulate should be derived from.They can be acquires from actual motion system
Come.Before starting training, all power should all be initialized with some different small random numbers." small random number " is used to
Guarantee that network will not enter saturation state because weight is excessive, so as to cause failure to train；" difference " is used to guarantee that network can be with
Normally learn.In fact, if network impotentia learns with identical several deinitialization weight matrixs.
For example, having the image (its original size is 20x20) of a 5x5, with the filter (convolution kernel) of a 3x3 to figure
As carrying out convolution, the low-level image feature of 3x3 has been obtained；This process can be understood as filtering using a filter (convolution kernel)
Each zonule of image, to obtain the characteristic value of these zonules.During hands-on, the value of convolution kernel is to learn
It is acquired during practising.In a particular application, often there are multiple convolution kernels, it is believed that each convolution kernel represents a kind of figure
As mode, if some image block and the value that this convolution nuclear convolution goes out are big, then it is assumed that this image block is in close proximity to this convolution kernel.
If devising 6 convolution kernels, it is possible to understand that：There are 6 kinds of base layer texture modes on this image, that is, with 6 kinds of basic schemas
A sub-picture can be depicted.
Since original image is 20x20, carry out down-sampling to it, sampling window 10x10, finally by its down-sampling at
For the characteristic pattern of a 2x2 size.Why carry out above-mentioned pond, be because even doing the convolution that is over, image still very it is big (because
It is smaller for convolution kernel), so in order to reduce data dimension, with regard to carrying out down-sampling.During pond, even if reducing perhaps
Most evidences, the statistical attribute of feature still are able to description image, and due to reducing data dimension, efficiently avoid intending
It closes.In practical applications, pond is divided under maximum value down-sampling (Max-Pooling) and average value according to the method for down-sampling
It samples (Mean-Pooling).
Full articulamentum is mainly fitted feature again, the loss of characteristic information is reduced, such as by low-level image feature and pond
Feature carries out after connecting entirely, obtains convolutional neural networks model.
S102, picture to be identified is obtained, and using the picture to be identified as multiple convolution in the local single page application
The input of neural network model obtains and each convolutional neural networks model result to be processed correspondingly.
In one embodiment, as shown in figure 5, the step S102 includes：
S1021, picture to be identified is inputted to each convolutional neural networks included in multiple convolutional neural networks models
Model；
S1022, calculate the real output value of each convolutional neural networks model using as with each convolutional neural networks mould
Type result to be processed correspondingly.
In the present embodiment, such as by the picture of a cat convolutional neural networks model is inputted, passes through convolutional neural networks mould
After type calculates, obtaining real output value is 1 (1 indicates the cat class in animal major class).Due to passing through a convolutional neural networks mould
There may be errors for the real output value that type obtains, subsequent to be averaged to each real output value, seek variance or ask
The operations such as standard deviation are to reduce the error of real output value.Such as count the result to be processed, obtain local figure
After piece recognition result, then according to picture recognition as a result, obtaining the picture recognition result with picture recognition result.For example, it is also possible to
Being arranged when picture recognition result is 2 indicates to identify successfully and indicate the dog class in animal major class, when setting real output value is greater than 2
Indicate recognition failures.
Due to being the direct convolutional neural networks model adjusted in local single page application, backstage is uploaded to without by picture
Server calls the model of background server to be identified, without considering that uploading pictures are slow to background server network speed and drop
Low recognition efficiency.
S103, the result to be processed is counted, obtains local picture recognition result.
In the present embodiment, one-to-one to be processed as a result, then may be used with each convolutional neural networks model when obtaining
Multiple results to be processed are counted and (such as carries out averaging operation or asks variance operation or seek standard difference operation), are obtained
One representative local picture recognition result.In such a way that this Multiple recognition seeks statistical value, can effectively it avoid
The error of single operation.
In one embodiment, the result to be processed is counted, obtains the specific reality of local picture recognition result
Existing mode is, to carrying out the result to be processed to average operation or asking variance operation or seek standard difference operation, obtains
Local picture recognition result.
If S104, the local picture recognition result are to identify successfully, and the local picture recognition result is corresponding
Keyword shown.
In the present embodiment, when local picture recognition result is to identify successfully, directly in local single page application
Region as the result is shown show the corresponding keyword of picture recognition result.Such as the picture of a cat is inputted into convolutional neural networks mould
Type, after being calculated by convolutional neural networks model, obtaining picture recognition result is the 1 (picture of the cat class in 1 expression animal major class
Recognition result), then according to picture recognition as a result, obtaining keyword (such as cat) corresponding with picture recognition result.By in local
Single page application in directly display recognition result, facilitate user to check.
In one embodiment, as shown in Figure 1, further including after step S103：
If S105, local picture recognition result are recognition failures, picture to be identified is uploaded to background server, is received
The backstage recognition result fed back by background server.
In the present embodiment, when convolutional neural networks model mounted in local single page application cannot accurately identify to
When identifying picture, then the picture to be identified can be uploaded to enterprise's background server, in the convolutional Neural with richer type
It is identified on enterprise's background server of network model, and feeds back correct recognition result after the completion of identification, it is this to have both
Local identification and backstage are known otherwise, it is ensured that the efficiency and accuracy rate of picture recognition.
Picture is carried out by convolutional neural networks model in the local single page application of user terminal as it can be seen that this method is realized
Identification, is quickly obtained recognition result, identifies it is not necessary that picture is directly uploaded to background server, improves identification effect
Rate.
The embodiment of the present application also provides a kind of single page high load pattern recognition device, the single page high load pattern recognition device
For executing any embodiment of aforementioned single page high load image-recognizing method.Specifically, referring to Fig. 6, Fig. 6 is the application reality
A kind of schematic block diagram of single page high load pattern recognition device of example offer is provided.Single page high load pattern recognition device 100 can
Be configured at desktop computer, tablet computer, laptop computer, etc. in terminals.
As shown in fig. 6, single page high load pattern recognition device 100 includes model acquiring unit 101, local picture recognition list
Member 102, statistic unit 103 and display unit 104.
Model acquiring unit 101, it is multiple for carrying out image for being stored in obtaining the local single page application of user terminal
The convolutional neural networks model of identification.
In the present embodiment, the convolutional neural networks model stored in local single page application is the peace in downloading single page application
When dress packet, the convolutional Neural model that can be used for image recognition carried in installation kit, which, which is stored in, passes through
In the application of local single page constructed by AngularJS (being the preceding end instruction based on javascript).User terminal is mounted with the peace
After dress packet, image recognition can be carried out by the convolutional neural networks model stored in the application of local single page.
Wherein, AngularJS is only needed HTML (i.e. hypertext markup language), CSS (i.e. cascading style sheets) and
JavaScript (being a kind of literal translation formula scripting language) can create single page application in client, and target is so that exploitation and survey
Examination is easier, the performance of enhancing MVC Web application.
So-called single page application refers to integrating multiple functions or even whole system on one page with regard to only one page
Face, all business functions are all its submodules, are articulated in main interface by specific mode.User terminal in the application
It after opening is clicked in local single page application, is similar to and enters a single machine Webpage, have on the single machine Webpage simple
Operation interface such as adds the operating space of picture to be identified, and the viewing area of output recognition result.
The convolutional neural networks model stored in local single page application is to be led on enterprise's background server by that will preselect
The picture of the image data set (such as described image data set) entered inputs convolutional neural networks, and is trained and obtains.Its
In, it is the current maximum data of image recognition in the world that ImageNet, which is a computer vision system identification project name,
Library, it is the computer scientist of U.S. Stamford, simulates the image recognition database that the identifying system of the mankind is established.
ImageNet data set (namely described image data set) has more than 1,400 ten thousand width pictures, covers a classification more than 20,000；Wherein have more than
Million picture has the mark of specific classification mark and objects in images position, and the size of the ImageNet data set is about
1TB。
In one embodiment, as shown in fig. 7, the single page high load pattern recognition device 100 further includes：
Installation kit acquiring unit 1001, the installation applied for obtaining the local single page including convolutional neural networks model
Packet；
Installation unit 1002 is decompressed, for decompressing the installation kit of the local single page application, obtains including multiple
The local single page application of convolutional neural networks model.
In the present embodiment, the installation kit of local single page application (is stored with the convolution for image recognition in the installation kit
Neural network model) in included convolutional neural networks model, be that a large amount of image knowledge is carried out on enterprise's background server
Not Xun Lian and obtain.Input training convolutional neural networks are used as by described image data set i.e. on enterprise's background server,
Obtained convolutional neural networks model includes multiple kinds of class models, such as has the convolutional neural networks of the identification vehicles
Model, the convolutional neural networks model etc. of insighted object of not moving, after the completion of the convolutional neural networks model training of above-mentioned multiple types,
The installation kit (such as apk installation kit or iOs installation kit) of single page application can be introduced directly into.When user has downloaded the installation of single page application
Bao Hou, decompression installs the installation kit of single page application on local terminal, and can be obtained one, there is the single page of identification image to answer
With being identified it is not necessary that picture to be identified to be uploaded on enterprise's background server.
In one embodiment, as shown in figure 8, the installation kit acquiring unit 1001 includes：
Model training unit 10011, for using the image data set imported in advance as the convolutional neural networks mould
The training data training convolutional neural networks model of type, the convolutional neural networks model after being trained；
Model loading unit 10012 is answered for the convolutional neural networks model after the training to be packed into local single page
In installation kit.
In the present embodiment, above-mentioned training process both may be selected to carry out at enterprise's background server end, can also be in user terminal
It carries out.Such as when user terminal is trained, the type (such as vehicles, animal, plant etc.) of images to be recognized can choose,
Directly according to the type of selected images to be recognized, the corresponding Image that obtains is in data set, and the conduct in the local of user terminal
The training data of convolutional neural networks obtains convolutional neural networks model in user terminal training, and by the convolution after the training
Neural network model is packed into the installation kit of local single page application, for this terminal itself use, or for other downloadings
Installation kit that the user terminal uploads (installation kit be include convolutional neural networks model the application of local single page installation kit).
Wherein, in order to improve the accuracy to picture recognition, the convolutional Neural net importeding into the installation kit of single page application
The type of network model is The more the better, but the control of the size for the installation kit for generally applying entire single page is in 200M or so
Meet basic picture recognition requirement, namely import the convolutional neural networks model of 10-20 kind (such as identification can be met and handed over
Logical tool, animal, the image of the classifications such as plant).
In one embodiment, as shown in figure 9, the model training unit 10011 includes：
Convolution layer unit 10011a is obtained on earth for the picture of described image data set to be carried out convolution according to filter
Layer feature；
Pond unit 10011b obtains pond feature for sampling according to sampling window to low-level image feature；
Full connection unit 10011c is rolled up for low-level image feature to be attached with pond feature by full articulamentum
Product neural network model.
In the present embodiment, convolutional neural networks are made of convolutional layer, pond layer, full articulamentum.Wherein convolutional layer and pond
Change layer cooperation, form multiple convolution groups, successively extract feature, completes classification eventually by several full articulamentums.
Convolutional network is inherently a kind of mapping for being input to output, it can learn largely to input between output
Mapping relations, without the accurate mathematic(al) representation between any output and input, as long as with known mode to volume
Product network is trained, and network just has the mapping ability between inputoutput pair.What convolutional network executed is to have tutor's training,
So its sample set be by shaped like：The vector of (input vector, ideal output vector) is to composition.All these vectors pair, all
The practical " RUN " result for the system that network will simulate should be derived from.They can be acquires from actual motion system
Come.Before starting training, all power should all be initialized with some different small random numbers." small random number " is used to
Guarantee that network will not enter saturation state because weight is excessive, so as to cause failure to train；" difference " is used to guarantee that network can be with
Normally learn.In fact, if network impotentia learns with identical several deinitialization weight matrixs.
For example, having the image (its original size is 20x20) of a 5x5, with the filter (convolution kernel) of a 3x3 to figure
As carrying out convolution, the low-level image feature of 3x3 has been obtained；This process can be understood as filtering using a filter (convolution kernel)
Each zonule of image, to obtain the characteristic value of these zonules.During hands-on, the value of convolution kernel is to learn
It is acquired during practising.In a particular application, often there are multiple convolution kernels, it is believed that each convolution kernel represents a kind of figure
As mode, if some image block and the value that this convolution nuclear convolution goes out are big, then it is assumed that this image block is in close proximity to this convolution kernel.
If devising 6 convolution kernels, it is possible to understand that：There are 6 kinds of base layer texture modes on this image, that is, with 6 kinds of basic schemas
A sub-picture can be depicted.
Since original image is 20x20, carry out down-sampling to it, sampling window 10x10, finally by its down-sampling at
For the characteristic pattern of a 2x2 size.Why carry out above-mentioned pond, be because even doing the convolution that is over, image still very it is big (because
It is smaller for convolution kernel), so in order to reduce data dimension, with regard to carrying out down-sampling.During pond, even if reducing perhaps
Most evidences, the statistical attribute of feature still are able to description image, and due to reducing data dimension, efficiently avoid intending
It closes.In practical applications, pond is divided under maximum value down-sampling (Max-Pooling) and average value according to the method for down-sampling
It samples (Mean-Pooling).
Full articulamentum is mainly fitted feature again, the loss of characteristic information is reduced, such as by low-level image feature and pond
Feature carries out after connecting entirely, obtains convolutional neural networks model.
Local picture recognition unit 102, for obtaining picture to be identified, and using the picture to be identified as the local
The input of multiple convolutional neural networks models, obtains with each convolutional neural networks model correspondingly wait locate in single page application
Manage result.
In one embodiment, as shown in Figure 10, the local picture recognition unit 102 includes：
Picture input unit 1021, it is included every for inputting picture to be identified in multiple convolutional neural networks models
One convolution neural network model；
Computing unit 1022, for calculate the real output value of each convolutional neural networks model using as with each convolution
Neural network model result to be processed correspondingly.
In the present embodiment, such as by the picture of a cat convolutional neural networks model is inputted, passes through convolutional neural networks mould
After type calculates, obtaining real output value is 1 (1 indicates the cat class in animal major class).Due to passing through a convolutional neural networks mould
There may be errors for the real output value that type obtains, subsequent to be averaged to each real output value, seek variance or ask
The operations such as standard deviation are to reduce the error of real output value.Such as count the result to be processed, obtain local figure
After piece recognition result, then according to picture recognition as a result, obtaining the picture recognition result with picture recognition result.For example, it is also possible to
Being arranged when picture recognition result is 2 indicates to identify successfully and indicate the dog class in animal major class, when setting real output value is greater than 2
Indicate recognition failures.
Due to being the direct convolutional neural networks model adjusted in local single page application, backstage is uploaded to without by picture
Server calls the model of background server to be identified, without considering that uploading pictures are slow to background server network speed and drop
Low recognition efficiency.
Statistic unit 103 obtains local picture recognition result for counting the result to be processed.
In the present embodiment, one-to-one to be processed as a result, then may be used with each convolutional neural networks model when obtaining
Multiple results to be processed are counted and (such as carries out averaging operation or asks variance operation or seek standard difference operation), are obtained
One representative local picture recognition result.In such a way that this Multiple recognition seeks statistical value, can effectively it avoid
The error of single operation.
In one embodiment, the result to be processed is counted, obtains the specific reality of local picture recognition result
Existing mode is, to carrying out the result to be processed to average operation or asking variance operation or seek standard difference operation, obtains
Local picture recognition result.
Display unit 104 knows the local picture if being to identify successfully for the local picture recognition result
The corresponding keyword of other result is shown.
In the present embodiment, when local picture recognition result is to identify successfully, directly in local single page application
Region as the result is shown show the corresponding keyword of picture recognition result.Such as the picture of a cat is inputted into convolutional neural networks mould
Type, after being calculated by convolutional neural networks model, obtaining picture recognition result is the 1 (picture of the cat class in 1 expression animal major class
Recognition result), then according to picture recognition as a result, obtaining keyword (such as cat) corresponding with picture recognition result.By in local
Single page application in directly display recognition result, facilitate user to check.
In one embodiment, as shown in fig. 6, the single page high load pattern recognition device 100 further includes：
Picture to be identified is uploaded to by backstage recognition unit 105 if being recognition failures for local picture recognition result
Background server receives the backstage recognition result fed back by background server.
In the present embodiment, when convolutional neural networks model mounted in local single page application cannot accurately identify to
When identifying picture, then the picture to be identified can be uploaded to enterprise's background server, in the convolutional Neural with richer type
It is identified on enterprise's background server of network model, and feeds back correct recognition result after the completion of identification, it is this to have both
Local identification and backstage are known otherwise, it is ensured that the efficiency and accuracy rate of picture recognition.
Picture is carried out by convolutional neural networks model in the local single page application of user terminal as it can be seen that the device is realized
Identification, is quickly obtained recognition result, identifies it is not necessary that picture is directly uploaded to background server, improves identification effect
Rate.
Above-mentioned single page high load pattern recognition device can be implemented as a kind of form of computer program, the computer program
It can be run in computer equipment as shown in figure 11.
Figure 11 is please referred to, Figure 11 is a kind of schematic block diagram of computer equipment provided by the embodiments of the present application.The calculating
500 equipment of machine equipment can be terminal.The terminal can be tablet computer, laptop, desktop computer, personal digital assistant
Equal electronic equipments.
Refering to fig. 11, which includes processor 502, memory and the net connected by system bus 501
Network interface 505, wherein memory may include non-volatile memory medium 503 and built-in storage 504.
The non-volatile memory medium 503 can storage program area 5031 and computer program 5032.The computer program
5032 include program instruction, which is performed, and processor 502 may make to execute a kind of single page high load image recognition
Method.
The processor 502 supports the operation of entire computer equipment 500 for providing calculating and control ability.
The built-in storage 504 provides environment for the operation of the computer program 5032 in non-volatile memory medium 503, should
When computer program 5032 is executed by processor 502, processor 502 may make to execute a kind of single page high load image recognition side
Method.
The network interface 505 such as sends the task dispatching of distribution for carrying out network communication.Those skilled in the art can manage
It solves, structure shown in Figure 11, only the block diagram of part-structure relevant to application scheme, is not constituted to the application side
The restriction for the computer equipment 500 that case is applied thereon, specific computer equipment 500 may include more than as shown in the figure
Or less component, perhaps combine certain components or with different component layouts.
Wherein, the processor 502 is for running computer program 5032 stored in memory, to realize following function
Energy：Obtain the multiple convolutional neural networks models for being used to carry out image recognition stored in the local single page application of user terminal；It obtains
Take picture to be identified, and using the picture to be identified as in the local single page application multiple convolutional neural networks models it is defeated
Enter, obtains and each convolutional neural networks model result to be processed correspondingly；The result to be processed is counted, is obtained
To local picture recognition result；If the local picture recognition result is to identify successfully, by the local picture recognition
As a result corresponding keyword is shown.
In one embodiment, processor 502 also performs the following operations：Obtaining includes the local single of convolutional neural networks model
The installation kit of page application；The installation kit of the local single page application is decompressed, obtains including multiple convolutional neural networks moulds
The local single page application of type.
In one embodiment, processor 502 also performs the following operations：Using described in the image data set conduct imported in advance
The training data training convolutional neural networks model of convolutional neural networks model, the convolutional neural networks mould after being trained
Type；Convolutional neural networks model after the training is packed into the installation kit of local single page application.
In one embodiment, processor 502 also performs the following operations：By the picture of described image data set according to filter
Convolution is carried out, low-level image feature is obtained；Low-level image feature is sampled according to sampling window, obtains pond feature；By low-level image feature
It is attached with pond feature by full articulamentum, obtains convolutional neural networks model.
In one embodiment, processor 502 also performs the following operations：Picture to be identified is inputted into multiple convolutional neural networks
Included each convolutional neural networks model in model；Calculate the real output value of each convolutional neural networks model using as
With each convolutional neural networks model result to be processed correspondingly.
In one embodiment, processor 502 also performs the following operations：The result to be processed is subjected to fortune of averaging
It calculates or asks variance operation or seek standard difference operation, obtain local picture recognition result.
In one embodiment, processor 502 also performs the following operations：If the local picture recognition result is that identification is lost
It loses, picture to be identified is uploaded to background server, receives the backstage recognition result fed back by background server.
It will be understood by those skilled in the art that the embodiment of computer equipment shown in Figure 11 is not constituted to computer
The restriction of equipment specific composition, in other embodiments, computer equipment may include components more more or fewer than diagram, or
Person combines certain components or different component layouts.For example, in some embodiments, computer equipment can only include depositing
Reservoir and processor, in such embodiments, the structure and function of memory and processor are consistent with embodiment illustrated in fig. 11,
Details are not described herein.
It should be appreciated that in the embodiment of the present application, processor 502 can be central processing unit (Central
Processing Unit, CPU), which can also be other general processors, digital signal processor (Digital
Signal Processor, DSP), specific integrated circuit (Application Specific Integrated Circuit,
ASIC), ready-made programmable gate array (Field-Programmable GateArray, FPGA) or other programmable logic devices
Part, discrete gate or transistor logic, discrete hardware components etc..Wherein, general processor can be microprocessor or
The processor is also possible to any conventional processor etc..
A kind of storage medium is provided in another embodiment of the application.The storage medium can be computer-readable storage
Medium.The storage medium is stored with computer program, and wherein computer program includes program instruction.The program instruction is by processor
It is realized when execution：Obtain the multiple convolutional neural networks for being used to carry out image recognition stored in the local single page application of user terminal
Model；Picture to be identified is obtained, and using the picture to be identified as multiple convolutional neural networks in the local single page application
The input of model obtains and each convolutional neural networks model result to be processed correspondingly；By the result to be processed into
Row statistics, obtains local picture recognition result；If the local picture recognition result is to identify successfully, will be described local
The corresponding keyword of picture recognition result is shown.
In one embodiment, realization when which is executed by processor：Obtaining includes convolutional neural networks model
The installation kit of local single page application；The installation kit of the local single page application is decompressed, obtains including multiple convolutional Neurals
The local single page application of network model.
In one embodiment, realization when which is executed by processor：Made using the image data set imported in advance
The convolutional neural networks model is trained for the training data of the convolutional neural networks model, the convolutional Neural after being trained
Network model；Convolutional neural networks model after the training is packed into the installation kit of local single page application.
In one embodiment, realization when which is executed by processor：By the picture of described image data set according to
Filter carries out convolution, obtains low-level image feature；Low-level image feature is sampled according to sampling window, obtains pond feature；The bottom of by
Layer feature is attached with pond feature by full articulamentum, and convolutional neural networks model is obtained.
In one embodiment, realization when which is executed by processor：Picture to be identified is inputted into multiple convolution minds
Through each convolutional neural networks model included in network model；Calculate the real output value of each convolutional neural networks model
Using as with each convolutional neural networks model result to be processed correspondingly.
In one embodiment, realization when which is executed by processor：The result to be processed is averaging
Value operation asks variance operation or seeks standard difference operation, obtains local picture recognition result.
In one embodiment, realization when which is executed by processor：If the local picture recognition result is
Picture to be identified is uploaded to background server, receives the backstage recognition result fed back by background server by recognition failures.
The storage medium can be the internal storage unit of aforementioned device, such as the hard disk or memory of equipment.It is described to deposit
Storage media is also possible to the plug-in type hard disk being equipped on the External memory equipment of the equipment, such as the equipment, intelligent storage
Block (Smart Media Card, SMC), secure digital (Secure Digital, SD) card, flash card (Flash Card) etc..
Further, the storage medium can also both including the equipment internal storage unit and also including External memory equipment.
It is apparent to those skilled in the art that for convenience of description and succinctly, foregoing description is set
The specific work process of standby, device and unit, can refer to corresponding processes in the foregoing method embodiment, and details are not described herein.
Those of ordinary skill in the art may be aware that unit described in conjunction with the examples disclosed in the embodiments of the present disclosure and algorithm
Step can be realized with electronic hardware, computer software, or a combination of the two, in order to clearly demonstrate hardware and software
Interchangeability generally describes each exemplary composition and step according to function in the above description.These functions are studied carefully
Unexpectedly the specific application and design constraint depending on technical solution are implemented in hardware or software.Professional technician
Each specific application can be used different methods to achieve the described function, but this realization is it is not considered that exceed
The scope of the present invention.
In several embodiments provided herein, it should be understood that disclosed unit and method, it can be with
It realizes by another way.For example, the apparatus embodiments described above are merely exemplary, for example, the unit
It divides, only a kind of logical function partition, there may be another division manner in actual implementation, can also will have identical function
The unit set of energy can be combined or can be integrated into another system at a unit, such as multiple units or components, or
Some features can be ignored or not executed.In addition, shown or discussed mutual coupling or direct-coupling or communication link
Connect can be through some interfaces, the indirect coupling or communication connection of device or unit, be also possible to electricity, it is mechanical or other
Form connection.
The unit as illustrated by the separation member may or may not be physically separated, aobvious as unit
The component shown may or may not be physical unit, it can and it is in one place, or may be distributed over multiple
In network unit.Some or all of unit therein can be selected to realize the embodiment of the present invention according to the actual needs
Purpose.
It, can also be in addition, the functional units in various embodiments of the present invention may be integrated into one processing unit
It is that each unit physically exists alone, is also possible to two or more units and is integrated in one unit.It is above-mentioned integrated
Unit both can take the form of hardware realization, can also realize in the form of software functional units.
If the integrated unit is realized in the form of SFU software functional unit and sells or use as independent product
When, it can store in one storage medium.Based on this understanding, technical solution of the present invention is substantially in other words to existing
The all or part of part or the technical solution that technology contributes can be embodied in the form of software products, should
Computer software product is stored in a storage medium, including some instructions are used so that a computer equipment (can be
Personal computer, server or network equipment etc.) execute all or part of step of each embodiment the method for the present invention
Suddenly.And storage medium above-mentioned includes：USB flash disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), magnetic disk or
The various media that can store program code such as person's CD.
The above description is merely a specific embodiment, but scope of protection of the present invention is not limited thereto, any
Those familiar with the art in the technical scope disclosed by the present invention, can readily occur in various equivalent modifications or replace
It changes, these modifications or substitutions should be covered by the protection scope of the present invention.Therefore, protection scope of the present invention should be with right
It is required that protection scope subject to.
Claims (10)
1. a kind of single page high load image-recognizing method, which is characterized in that including：
Obtain the multiple convolutional neural networks models for being used to carry out image recognition stored in the local single page application of user terminal；
Picture to be identified is obtained, and using the picture to be identified as multiple convolutional neural networks moulds in the local single page application
The input of type obtains and each convolutional neural networks model result to be processed correspondingly；
The result to be processed is counted, local picture recognition result is obtained；
If the local picture recognition result be identify successfully, by the local corresponding keyword of picture recognition result into
Row display.
2. single page high load image-recognizing method according to claim 1, which is characterized in that the sheet for obtaining user terminal
Before the multiple convolutional neural networks models for carrying out image recognition stored in the application of ground single page, including：
Obtain the installation kit that the local single page including convolutional neural networks model is applied；
The installation kit of the local single page application is decompressed, obtain include multiple convolutional neural networks models local single page
Using.
3. single page high load image-recognizing method according to claim 2, which is characterized in that described obtain includes convolution mind
The installation kit of local single page application through network model, including：
The training data training convolution mind using the image data set imported in advance as the convolutional neural networks model
Convolutional neural networks model through network model, after being trained；
Convolutional neural networks model after the training is packed into the installation kit of local single page application.
4. single page high load image-recognizing method according to claim 3, which is characterized in that the use imported in advance
The training data training convolutional neural networks model of the image data set as the convolutional neural networks model, is trained
Convolutional neural networks model afterwards, including：
The picture of described image data set is subjected to convolution according to filter, obtains low-level image feature；
Low-level image feature is sampled according to sampling window, obtains pond feature；
Low-level image feature is attached with pond feature by full articulamentum, convolutional neural networks model is obtained.
5. single page high load image-recognizing method according to claim 1, which is characterized in that described by the figure to be identified
Input of the piece as multiple convolutional neural networks models in the local single page application, obtains and each convolutional neural networks model
It is one-to-one to be processed as a result, including：
Picture to be identified is inputted to each convolutional neural networks model included in multiple convolutional neural networks models；
The real output value of each convolutional neural networks model is calculated to correspond as with each convolutional neural networks model
Result to be processed.
6. single page high load image-recognizing method according to claim 5, which is characterized in that described by the knot to be processed
Fruit is counted, and obtains local picture recognition as a result, including：
It carries out the result to be processed to average operation or asks variance operation or seek standard difference operation, obtain local figure
Piece recognition result.
According to real output value, picture recognition result corresponding with real output value is obtained.
7. single page high load image-recognizing method according to claim 1, which is characterized in that described by the figure to be identified
Piece further includes after obtaining local picture recognition result as the input of convolutional neural networks model in local single page application：
If the local picture recognition result is recognition failures, picture to be identified is uploaded to background server, is received by rear
The backstage recognition result of platform server feedback.
8. a kind of single page high load pattern recognition device, which is characterized in that including：
Model acquiring unit, it is multiple for carrying out the volumes of image recognition for being stored in obtaining the local single page application of user terminal
Product neural network model；
Local picture recognition unit is answered for obtaining picture to be identified, and using the picture to be identified as the local single page
The input of multiple convolutional neural networks models in obtains and each convolutional neural networks model knot to be processed correspondingly
Fruit；
Statistic unit obtains local picture recognition result for counting the result to be processed；
Display unit, if being to identify successfully for the local picture recognition result, by the local picture recognition result
Corresponding keyword is shown.
9. a kind of computer equipment, including memory, processor and it is stored on the memory and can be on the processor
The computer program of operation, which is characterized in that the processor is realized when executing the computer program as in claim 1-7
Described in any item single page high load image-recognizing methods.
10. a kind of storage medium, which is characterized in that the storage medium is stored with computer program, the computer program packet
Program instruction is included, described program instruction executes the processor such as any one of claim 1-7 institute
The single page high load image-recognizing method stated.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810513879.8A CN108846419A (en) | 2018-05-25 | 2018-05-25 | Single page high load image-recognizing method, device, computer equipment and storage medium |
| PCT/CN2018/102471 WO2019223154A1 (en) | 2018-05-25 | 2018-08-27 | Single-page high-load image recognition method, device, computer apparatus, and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810513879.8A CN108846419A (en) | 2018-05-25 | 2018-05-25 | Single page high load image-recognizing method, device, computer equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108846419A true CN108846419A (en) | 2018-11-20 |
Family
ID=64213520
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810513879.8A Pending CN108846419A (en) | 2018-05-25 | 2018-05-25 | Single page high load image-recognizing method, device, computer equipment and storage medium |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN108846419A (en) |
| WO (1) | WO2019223154A1 (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111767902A (en) * | 2020-07-31 | 2020-10-13 | 上海中通吉网络技术有限公司 | Method, device and equipment for identifying dangerous goods of security check machine and storage medium |
| CN111917837A (en) * | 2020-07-13 | 2020-11-10 | 西安即刻易用网络科技有限公司 | Web micro application program publishing system and implementation method thereof |
| CN111913757A (en) * | 2020-07-13 | 2020-11-10 | 西安即刻易用网络科技有限公司 | Single-page application program loading method and system |
| CN112241648A (en) * | 2019-07-16 | 2021-01-19 | 杭州海康威视数字技术股份有限公司 | Image processing system and image device |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110969143A (en) * | 2019-12-19 | 2020-04-07 | 深圳壹账通智能科技有限公司 | Evidence obtaining method and system based on image recognition, computer equipment and storage medium |
| CN111652262A (en) * | 2020-03-19 | 2020-09-11 | 深圳市彬讯科技有限公司 | Image object recognition method and device, computer equipment and storage medium |
| CN111598226B (en) * | 2020-05-18 | 2023-07-25 | 山东浪潮科学研究院有限公司 | Convolutional network visualization method and device for image recognition |
| CN111680775B (en) * | 2020-06-09 | 2022-02-01 | 河北科技师范学院 | Image recognition module based on deep neural network |
| CN111797921A (en) * | 2020-07-02 | 2020-10-20 | 中国工商银行股份有限公司 | Image data comparison method and device |
| CN112330816B (en) * | 2020-10-19 | 2024-03-26 | 杭州易现先进科技有限公司 | AR identification processing method and device and electronic device |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105512676A (en) * | 2015-11-30 | 2016-04-20 | 华南理工大学 | Food recognition method at intelligent terminal |
| CN105825214A (en) * | 2016-03-14 | 2016-08-03 | 东南大学 | Tesseract engine based character recognition method and device |
| CN107491726A (en) * | 2017-07-04 | 2017-12-19 | 重庆邮电大学 | A kind of real-time expression recognition method based on multi-channel parallel convolutional neural networks |
| CN107657281A (en) * | 2017-09-28 | 2018-02-02 | 辽宁工程技术大学 | A kind of image-recognizing method based on improved convolutional neural networks |
| CN107704853A (en) * | 2017-11-24 | 2018-02-16 | 重庆邮电大学 | A kind of recognition methods of the traffic lights based on multi-categorizer |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR101729976B1 (en) * | 2011-06-09 | 2017-04-25 | 한국전자통신연구원 | Image recognition apparatus and method for recognizing image thereof |
| JP6706788B2 (en) * | 2015-03-06 | 2020-06-10 | パナソニックＩｐマネジメント株式会社 | Image recognition method, image recognition device and program |
| CN106293972B (en) * | 2016-08-12 | 2019-12-10 | 金蝶软件（中国）有限公司 | single-page application page communication method and device |
| CN106375378B (en) * | 2016-08-25 | 2020-08-28 | 北京每刻风物科技有限公司 | Application deployment method and system based on local area network client server structure |
-
2018
- 2018-05-25 CN CN201810513879.8A patent/CN108846419A/en active Pending
- 2018-08-27 WO PCT/CN2018/102471 patent/WO2019223154A1/en active Application Filing
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105512676A (en) * | 2015-11-30 | 2016-04-20 | 华南理工大学 | Food recognition method at intelligent terminal |
| CN105825214A (en) * | 2016-03-14 | 2016-08-03 | 东南大学 | Tesseract engine based character recognition method and device |
| CN107491726A (en) * | 2017-07-04 | 2017-12-19 | 重庆邮电大学 | A kind of real-time expression recognition method based on multi-channel parallel convolutional neural networks |
| CN107657281A (en) * | 2017-09-28 | 2018-02-02 | 辽宁工程技术大学 | A kind of image-recognizing method based on improved convolutional neural networks |
| CN107704853A (en) * | 2017-11-24 | 2018-02-16 | 重庆邮电大学 | A kind of recognition methods of the traffic lights based on multi-categorizer |
Non-Patent Citations (1)
| Title |
|---|
| 杨晓宇: "基于深度学习智能标注图片关键字系统的设计与实现", 《中国优秀硕士学位论文全文数据库》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112241648A (en) * | 2019-07-16 | 2021-01-19 | 杭州海康威视数字技术股份有限公司 | Image processing system and image device |
| CN111917837A (en) * | 2020-07-13 | 2020-11-10 | 西安即刻易用网络科技有限公司 | Web micro application program publishing system and implementation method thereof |
| CN111913757A (en) * | 2020-07-13 | 2020-11-10 | 西安即刻易用网络科技有限公司 | Single-page application program loading method and system |
| CN111767902A (en) * | 2020-07-31 | 2020-10-13 | 上海中通吉网络技术有限公司 | Method, device and equipment for identifying dangerous goods of security check machine and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2019223154A1 (en) | 2019-11-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108846419A (en) | Single page high load image-recognizing method, device, computer equipment and storage medium | |
| Westerlund | New simple tests for panel cointegration | |
| CN109165573B (en) | Method and device for extracting video feature vector | |
| WO2020253852A1 (en) | Image identification method and device, identification model training method and device, and storage medium | |
| CN107239318A (en) | Using conversion method, device and equipment | |
| CN106796716A (en) | Apparatus and method for providing super-resolution for low-resolution image | |
| CN110880036A (en) | Neural network compression method and device, computer equipment and storage medium | |
| CN107516090A (en) | Integrated face identification method and system | |
| CN109471795A (en) | It is grouped test method, device, computer equipment and storage medium | |
| CN109298840A (en) | Data integrating method, server and storage medium based on data lake | |
| CN110060750A (en) | Medical data method for pushing, system, computer equipment and readable storage medium storing program for executing | |
| CN110377218B (en) | Data processing method, data processing device, computer equipment and storage medium | |
| CN110321761A (en) | A kind of Activity recognition method, terminal device and computer readable storage medium | |
| CN110298007A (en) | User behavior statistical method, device, electronic equipment and computer readable storage medium | |
| CN109640068A (en) | Information forecasting method, device, equipment and the storage medium of video frame | |
| DE112020005789T5 (en) | HIERARCHICAL PARTITIONING OF OPERATORS | |
| CN111985197A (en) | Template generation method based on medical information | |
| CN112215293A (en) | Plant disease and insect pest identification method and device and computer equipment | |
| CN111126626A (en) | Training method, device, server, platform and storage medium | |
| CN111507407B (en) | Training method and device for image classification model | |
| DE102021107510A1 (en) | TRAINING OF A NEURAL NETWORK UNDER MEMORY RESTRICTION | |
| WO2023185541A1 (en) | Model training method and related device | |
| CN112801685A (en) | Information pushing method and device, computer equipment and storage medium | |
| CN111026945A (en) | Multi-platform crawler scheduling method and device and storage medium | |
| CN111859933A (en) | Training method, recognition method, device and equipment of Malay recognition 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: 20181120 |