CN110275820A - Page compatibility test method, system and equipment - Google Patents
Page compatibility test method, system and equipment Download PDFInfo
- Publication number
- CN110275820A CN110275820A CN201810215223.8A CN201810215223A CN110275820A CN 110275820 A CN110275820 A CN 110275820A CN 201810215223 A CN201810215223 A CN 201810215223A CN 110275820 A CN110275820 A CN 110275820A
- Authority
- CN
- China
- Prior art keywords
- screenshot
- measured
- sample
- page
- training
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000010998 test method Methods 0.000 title claims abstract description 25
- 238000012360 testing method Methods 0.000 claims abstract description 166
- 238000000034 method Methods 0.000 claims abstract description 96
- 238000012549 training Methods 0.000 claims description 219
- 238000000605 extraction Methods 0.000 claims description 50
- 230000008859 change Effects 0.000 claims description 38
- 230000008569 process Effects 0.000 claims description 33
- 238000013527 convolutional neural network Methods 0.000 claims description 20
- 238000012545 processing Methods 0.000 claims description 14
- 238000002372 labelling Methods 0.000 claims description 11
- 239000012141 concentrate Substances 0.000 claims description 8
- 238000011068 loading method Methods 0.000 claims description 7
- 230000004044 response Effects 0.000 claims description 6
- 239000000284 extract Substances 0.000 claims description 4
- 241000208340 Araliaceae Species 0.000 claims description 3
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 claims description 3
- 235000003140 Panax quinquefolius Nutrition 0.000 claims description 3
- 235000008434 ginseng Nutrition 0.000 claims description 3
- 238000005086 pumping Methods 0.000 claims 3
- 238000012423 maintenance Methods 0.000 abstract description 16
- 238000010586 diagram Methods 0.000 description 28
- 230000006870 function Effects 0.000 description 18
- 238000004422 calculation algorithm Methods 0.000 description 14
- 238000004590 computer program Methods 0.000 description 13
- 239000000047 product Substances 0.000 description 11
- 238000003860 storage Methods 0.000 description 9
- 230000002159 abnormal effect Effects 0.000 description 6
- 238000013473 artificial intelligence Methods 0.000 description 6
- 235000013399 edible fruits Nutrition 0.000 description 5
- 210000002569 neuron Anatomy 0.000 description 5
- 230000000644 propagated effect Effects 0.000 description 5
- 238000004891 communication Methods 0.000 description 4
- 238000013461 design Methods 0.000 description 4
- 230000005611 electricity Effects 0.000 description 4
- 230000005236 sound signal Effects 0.000 description 4
- 230000000007 visual effect Effects 0.000 description 4
- 210000004556 brain Anatomy 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000003068 static effect Effects 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- VYZAMTAEIAYCRO-UHFFFAOYSA-N Chromium Chemical compound [Cr] VYZAMTAEIAYCRO-UHFFFAOYSA-N 0.000 description 1
- 244000035744 Hura crepitans Species 0.000 description 1
- 238000001545 Page's trend test Methods 0.000 description 1
- BUGBHKTXTAQXES-UHFFFAOYSA-N Selenium Chemical compound [Se] BUGBHKTXTAQXES-UHFFFAOYSA-N 0.000 description 1
- 241000282485 Vulpes vulpes Species 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000000429 assembly Methods 0.000 description 1
- 230000000712 assembly Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 210000004027 cell Anatomy 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 230000008451 emotion Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 210000002364 input neuron Anatomy 0.000 description 1
- 238000003475 lamination Methods 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 210000004205 output neuron Anatomy 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 235000008954 quail grass Nutrition 0.000 description 1
- 229910052711 selenium Inorganic materials 0.000 description 1
- 239000011669 selenium Substances 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
- 230000011664 signaling Effects 0.000 description 1
- 239000013589 supplement Substances 0.000 description 1
- 238000003786 synthesis reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/36—Preventing errors by testing or debugging software
- G06F11/3668—Software testing
- G06F11/3672—Test management
- G06F11/3688—Test management for test execution, e.g. scheduling of test suites
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/958—Organisation or management of web site content, e.g. publishing, maintaining pages or automatic linking
-
- 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 provides a kind of page compatibility test method, system and equipment.Wherein, method includes: at least one screenshot to be measured for obtaining the page to be measured of display in a browser；The abstract characteristics of each screenshot to be measured are extracted from least one described screenshot to be measured respectively；According to the abstract characteristics of each screenshot to be measured, compatibility of the page to be measured in the browser is determined.Technical solution provided by the embodiments of the present application, by way of extracting the abstract characteristics of at least one screenshot to be measured of the page to be measured, to identify whether multiple screenshots of the page normally show in a browser, with the compatibility in the determination page to be measured also browser, realize the automatic test of page compatibility, and do not influenced by page adjustment, maintenance cost is low, and accuracy rate is high.
Description
Technical field
This application involves field of computer technology more particularly to a kind of page compatibility test methods, system and equipment.
Background technique
Increasing with the network user, the web browser type designed to meet the needs of different users is also got over
Carrying out more, common browsers includes IE (Internet Explorer), Firefox (red fox), Chrome (Google's browsing
Device), sogou browser, 360 browsers etc., the kernel of these browsers is different, thus the support for various webpages
Property is also not quite similar, and thus just brings page compatibility issue.
Page compatibility test, or browser compatibility is made to test, it is for the same front end page in different viewing
The whether consistent test of bandwagon effect in device.Since there are many quantity of browser, if carrying out the compatibility of each browser manually
Property test, i.e., manually check whether the page normal on each browser, then duplicate workload is very big, need to cover PC
Multiple operating systems and browser at end and mobile terminal.
Summary of the invention
In view of the above problems, the application is proposed to solve the above problems or at least be partially solved the page of the above problem
Face compatibility test method, system and equipment.
Then, in one embodiment of the invention, a kind of page compatibility test method is provided.This method, comprising:
Obtain at least one screenshot to be measured of the page to be measured of display in a browser；
The abstract characteristics of each screenshot to be measured are extracted from least one described screenshot to be measured respectively；
According to the abstract characteristics of each screenshot to be measured, compatibility of the page to be measured in the browser is determined.
In another embodiment of the application, a kind of page compatibility test method is provided.This method is suitable for service
End, comprising:
Receive at least one screenshot to be measured for the page to be measured being shown in a browser that client uploads；
The abstract characteristics of each screenshot to be measured are extracted from least one described screenshot to be measured respectively；
According to the abstract characteristics of each screenshot to be measured, compatibility of the page to be measured in the browser is determined；
Definitive result is fed back into the client.
In the another embodiment of the application, a kind of page compatibility test method is provided.This method is suitable for client
End, comprising:
The page to be measured is loaded in a browser；
Subregion carries out screenshot to the page to be measured, to obtain at least one screenshot to be measured；
At least one described screenshot to be measured is uploaded to server-side, to determine the page to be measured in institute by the server-side
State the compatibility in browser；
Wherein it is determined that the abstract characteristics of each screenshot to be measured are respectively from described according to the abstract characteristics for being each screenshot to be measured
It is extracted at least one screenshot to be measured.
In the another embodiment of the application, a kind of page compatibility test system is provided.The system includes:
Client, for loading the page to be measured in a browser；Subregion carries out screenshot to the page to be measured, with obtain to
A few screenshot to be measured；At least one described screenshot to be measured is uploaded to server-side；
Server-side, at least one to be measured section of the page to be measured being shown in a browser for receiving client upload
Figure；The abstract characteristics of each screenshot to be measured are extracted from least one described screenshot to be measured respectively；It will be according to each screenshot to be measured
Abstract characteristics determine compatibility of the page to be measured in the browser, and definitive result is fed back into the client
End.
In the another embodiment of the application, a kind of electronic equipment is provided.The electronic equipment include: first memory and
First processor, wherein
The first memory, for storing program；
The first processor is coupled with the first memory, for executing the institute stored in the first memory
Program is stated, to be used for:
Obtain at least one screenshot to be measured of the page to be measured of display in a browser；
The abstract characteristics of each screenshot to be measured are extracted from least one described screenshot to be measured respectively；
According to the abstract characteristics of each screenshot to be measured, compatibility of the page to be measured in the browser is determined.
In the another embodiment of the application, a kind of server device is provided.The server device includes: the second storage
Device and second processor, wherein
The second memory, for storing program；
The second processor is coupled with the second memory, for executing the institute stored in the second memory
Program is stated, to be used for:
Receive at least one screenshot to be measured for the page to be measured being shown in a browser that client uploads；
The abstract characteristics of each screenshot to be measured are extracted from least one described screenshot to be measured respectively；
According to the abstract characteristics of each screenshot to be measured, compatibility of the page to be measured in the browser is determined；
Definitive result is fed back into the client.
In the another embodiment of the application, a kind of client device is provided.The client device includes: third storage
Device and third processor, wherein
The third memory, for storing program；
The third processor is coupled with the third memory, for executing the institute stored in the third memory
Program is stated, to be used for:
The page to be measured is loaded in a browser；
Subregion carries out screenshot to the page to be measured, to obtain at least one screenshot to be measured；
At least one described screenshot to be measured is uploaded to server-side, to determine the page to be measured in institute by the server-side
State the compatibility in browser；
Wherein it is determined that the abstract characteristics of each screenshot to be measured are respectively from described according to the abstract characteristics for being each screenshot to be measured
It is extracted at least one screenshot to be measured.
Technical solution provided by the embodiments of the present application, the abstract spy of at least one screenshot to be measured by extracting the page to be measured
The mode of sign, to identify whether multiple screenshots of the page normally show in a browser, in a browser with the determination page to be measured
Whether have compatibility, realize the automatic test of page compatibility, and do not influenced by page adjustment, maintenance cost is low, quasi-
True rate is high.
Detailed description of the invention
The drawings described herein are used to provide a further understanding of the present application, constitutes part of this application, this Shen
Illustrative embodiments and their description please are not constituted an undue limitation on the present application for explaining the application.In the accompanying drawings:
Fig. 1 is the flow diagram for the page compatibility test method that one embodiment of the application provides；
Fig. 2 is that the application is declarative semantics label and the corresponding exemplary photo provided；
Image data is through convolutional neural networks knot in the page compatibility test method that Fig. 3 provides for one embodiment of the application
Each result schematic diagram that each layer of structure obtains after calculating；
Fig. 4 is the structural schematic diagram for the page compatibility test system that one embodiment of the application provides；
Fig. 5 is the flow diagram for the page compatibility test method that another embodiment of the application provides；
Fig. 6 is the flow diagram for the page compatibility test method that the another embodiment of the application provides；
Fig. 7 is the flow diagram for the page compatibility test method that the another embodiment of the application provides；
Fig. 8 is the structural schematic diagram for the page device for testing compatibility that one embodiment of the application provides；
Fig. 9 is the structural schematic diagram for the page device for testing compatibility that another embodiment of the application provides；
Figure 10 is the structural schematic diagram for the page device for testing compatibility that the another embodiment of the application provides；
Figure 11 is the structural schematic diagram for the electronic equipment that one embodiment of the application provides；
Figure 12 is the structural schematic diagram for the server device that one embodiment of the application provides；
Figure 13 is the structural schematic diagram for the client device that one embodiment of the application provides.
Specific embodiment
In order to make those skilled in the art more fully understand application scheme, below in conjunction in the embodiment of the present application
Attached drawing, the technical scheme in the embodiment of the application is clearly and completely described.
In some processes described in the description of the present application, claims and above-mentioned attached drawing, contain according to spy
Multiple operations that fixed sequence occurs, these operations can not be executed according to its sequence what appears in this article or be executed parallel.
Serial number of operation such as 101,102 etc. is only used for distinguishing each different operation, and it is suitable that serial number itself does not represent any execution
Sequence.In addition, these processes may include more or fewer operations, and these operations can be executed in order or be held parallel
Row.It should be noted that the description such as herein " first ", " second ", be for distinguishing different message, equipment, module etc.,
Sequencing is not represented, " first " and " second " is not also limited and is different type.
The automation tools or scheme of page compatibility test at present are broadly divided into two classes.
First kind scheme or tool need to be manually entered parameter and configuration, such as Browsershots, IE Tester,
Spoon Browser Sandbox etc..This scheme is typically based on the virtual machine for being mounted with a variety of browsers in advance,
Such as Windows virtual machine.Virtual machine provides some configurable software interfaces and inputs or configured in be tested for user
Hold, such as the format of page elements, the font of the page, the page presentation etc. under different resolution.Good content to be measured is configured, is used
After family provides the chained address of the page to be measured, under the test respective operations system (such as Windows) that virtual machine can automate
Installed multiple browsers compatibility.And for mobile phone operating system, such as IOS system or Android system, lead to
Often provided without suitable virtual machine.Pattern of the method that existing this crawl page elements of virtual machine are tested for the page
Problem is often invalid, it may appear that page elements are normal, but certain browser lower page pattern entanglements the problem of.
Second class scheme or tool are the comparisons based on image pixel, for example, Selenium.This tool usually requires people
Work writes procedure script to record the content of each test page, is visited again by playback the page when testing every time later
It asks, and carries out the comparison of Pixel-level with the page screenshot recorded to current page screenshot, realize the simultaneous of the page by this method
Capacitive test.This scheme page problematic for some patterns is effective, but is easy by running environment, such as net
The influence of network transmission, the machine of different resolution and page adjustment, not only stability is inadequate, but also also needs after page adjustment
A page verifying is first carried out manually, records script again again later.And exactly front end page usually changes relatively frequently, it is this
The method the cost of manual maintenance for recording script is also higher.
Part or all of present in existing measuring technology in order to overcome the problems, such as, the embodiment of the present application provides a kind of applicability
Extensively, the cost of manual maintenance low page compatibility test method, system and equipment.Technical solution provided by the embodiments of the present application
Main thought is: visual page-images being expressed as to the abstract characteristics of high-rise rank, are determined based on the abstract characteristics of the page
The compatibility of the page.It not will receive the influence of page adjustment using technical solution provided by the embodiments of the present application, therefore just keep away yet
The problem of existing second class scheme or tool need to will rewrite script after the adjustment page every time is exempted from, so as to greatly subtract
Few maintenance cost, improves tested performance；On the other hand it can also largely solve existing first kind scheme or tool exists
Style sheet entanglement the problem of being unable to test out.
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.Obviously, described embodiments are only a part of embodiments of the present application, instead of all the embodiments.It is based on
Embodiment in the application, those skilled in the art's every other implementation obtained without making creative work
Example, shall fall in the protection scope of this application.
Fig. 1 shows the flow diagram of the page compatibility test method of one embodiment of the application offer.The present embodiment
The executing subject of the technical solution of offer can be client or server-side.Wherein, client can be integrated in terminal
One hardware with embedded program is also possible to install an application software in the terminal, can also be and be embedded in end
The tool software etc. in operating system is held, the embodiment of the present application is not especially limited this.The terminal can be mobile phone, plate electricity
Brain, PDA (PersonalDigital Assistant, personal digital assistant), POS (Point of Sales, point-of-sale terminal),
Any terminal device such as vehicle-mounted computer.Server-side can be General Server, cloud, virtual server etc., the embodiment of the present application
This is not especially limited.As shown in Figure 1, which comprises
101, at least one screenshot to be measured of the page to be measured of display in a browser is obtained.
102, the abstract characteristics of each screenshot to be measured are extracted from least one described screenshot to be measured respectively.
103, according to the abstract characteristics of each screenshot to be measured, compatibility of the page to be measured in the browser is determined
Property.
In above-mentioned 101, as soon as the full page is shown in screen, screenshot one to be measured；And for the longer page, i.e. user needs
It to slide (such as mobile phone) or mouse rollovers (such as PC machine) manually to check the page for not showing content, need subregional to the page
Progress screenshot, the screenshot to be measured of this kind of page just not only one.What needs to be explained here is that at least one in the embodiment of the present application
Screenshot to be measured need to cover the entire area of the page to be measured.A part of region content for such as missing the page to be measured, because there are do not test
Screenshot to be measured so that compatible definitive result inaccuracy.In a kind of example of specific implementation, it can be mounted with different viewing
Dispose automatic screenshot tool in the equipment (such as client or server-side) of device, with call the interface of browser in equipment access to
It surveys the page and traverses all areas of the page to be measured, the page to be measured is intercepted with subregion and obtains at least one screenshot to be measured.
People is by understanding picture material in the meanings such as object, event described in image and the emotion of expression.And
These meanings can simply be interpreted as the higher level of abstraction feature of image.This higher level of abstraction is characterized in can not be directly from image
Visual signature (such as color, unity and coherence in writing, shape feature) obtain.Bottom visual signature only describes the letter of image in a certain respect
Breath, ability to express are limited.It follows that the key for accurately identifying image is to extract the abstract spy with compared with high rule complexity
Sign.Objects in images essence, constant information can be described with the feature compared with strongly expressed power.Skill provided by the embodiments of the present application
Art scheme can be used extracts abstract characteristics based on artificial intelligence learning algorithm from image.Human perception system is a kind of specific
Hierarchical structure, the mankind are transmitting layer by layer, a constantly abstract process to the processing of visual information, pass through the table layer by layer to data
It reaches, the essential information being hidden in inside data can be better described.Therefore, the artificial intelligence learning algorithm of human brain is simulated, it can be from
It is extracted in image more comprehensively, the stronger rear lane feature of ability to express.
Following method extraction abstract characteristics can be used in each screenshot to be measured in step 102 in the embodiment of the present application as a result,.
Below by taking a screenshot (the i.e. first screenshot to be measured) to be measured in screenshot how to be measured as an example, abstract characteristics extraction process is said
It is bright.The abstract characteristics of the described first screenshot to be measured are extracted from the first screenshot to be measured, comprising:
1021, the described first screenshot to be measured is pre-processed to obtain image data.
1022, using described image data as the input for extracting model, execute the extraction model obtain described first to
Survey the abstract characteristics of screenshot.
Wherein, the first screenshot to be measured is any one screenshot to be measured at least one screenshot to be measured.The extraction model
It is to be constructed based on artificial intelligence learning algorithm.Specifically, in a kind of achievable scheme, artificial intelligence learning algorithm
Convolutional neural networks algorithm can be selected.Convolutional neural networks, Convolutional Neural Network (CNN): being a kind of
Feedforward neural network, by one or more convolutional layers, pond layer (pooling layer) and full articulamentum (corresponding classical mind
Through network) composition, while also including shared weight (shared weights) and shared biasing (shared bias).This knot
Structure enables convolutional neural networks to utilize the two-dimensional structure of input data.Compared with other deep learning structures, it artificial
Neuron can respond the surrounding cells in a part of coverage area, can provide more preferably result in image procossing.
In above-mentioned 1021, to the first screenshot to be measured carry out pretreated purpose be screenshot to be measured is converted to extract model can
With the format of processing.For example, carrying out pretreatment to the described first screenshot to be measured in above-mentioned 1021 includes: by described first to be measured section
Figure is adjusted to the triple channel rgb image data being sized.For example, the first screenshot to be measured is adjusted to the 3 of 224 × 224 sizes
Channel rgb image data.
Artificial intelligence learning algorithm can just obtain workable extraction model in above-mentioned 1022 after need to training.Training process is just
It is the parameter (such as shared weight and shared biasing) in adjustment algorithm, so that it is more accurate to extract model output.Wherein, algorithm training
Process will be described in detail in subsequent content.
The value for extracting that the abstract characteristics that model obtains are usually a numeralization is executed in above-mentioned 1022.For the ease of compatibility
Property determination, specific implementation the abstract characteristics of numeralization value can be converted to semantic label, semantic label can also simply understand
At semantic feature.Picture as shown in Figure 2, and corresponding semantic label includes: " aircraft ", " meadow ", " sky " etc..Semanteme mark
There are one-to-one relationships with the value after its numeralization for label.And abstract characteristics are the values of a numeralization, are thus abstracted
Corresponding relationship between feature and semantic label.
Therefore, it above-mentioned 103 can be respectively corresponded according to each screenshot to be measured of the corresponding relationship of abstract characteristics and semantic label acquisition
Semantic label, then compatibility of the page to be measured in the browser is determined based on the corresponding semantic label of each screenshot to be measured
Property.Specifically, step 103 can be used the following two kinds mode and realize in above-described embodiment.
Mode one:
1031, according to the corresponding relationship of abstract characteristics and semantic label, the abstract characteristics point of each screenshot to be measured are obtained
Not corresponding semantic label.
1032, it by determining whether the corresponding semantic label of each screenshot to be measured belongs to exception class label, determines
Compatibility of the page to be measured in the browser.
Specifically, step 1032 includes: to have a screenshot to be measured in the corresponding semantic label of each screenshot to be measured
Corresponding semantic label belongs to exception class label, it is determined that the page to be measured does not have compatibility in the browser.I.e. only
As soon as corresponding semantic label of screenshot to be measured for the page to be measured is wanted to belong to exception class label, illustrate that the page to be measured is shown
It is abnormal, do not have compatibility.Only when the corresponding semantic label of all screenshots to be measured is not admitted to exception class label, just determine
The page to be measured has compatibility in the browser.
Wherein, an exception class list of labels can be preset when specific implementation, have collected a variety of exception class in the list
Label.It is described to determine by way of whether inquiring the corresponding semantic label of screenshot to be measured in the exception class list of labels
Whether the corresponding semantic label of the screenshot to be measured belongs to exception class label.In the specific implementation, know which screenshot to be measured is corresponding
Semantic label belongs to exception class label, can know which block region display of the page to be measured is abnormal；And it can be according to semantic label institute
The abnormal type of category, determines which kind of the display abnormal problem of this block of the page to be measured belongs to；These information can help page
Face designer improves the page.
Mode two,
1031 ', the corresponding standard semantic tally set of the page to be measured is obtained.
1032 ', by comparing the semantic feature and the standard semantic feature set of each screenshot to be measured, determine it is described to
Survey whether the page normally shows in the browser.
Substantially, after the design is completed, corresponding semantic label determines each page.The page is completed in design
Afterwards, it is at least one semantic label of the page configuration that staff is manual, to obtain standard semantic tally set.Therefore, exist
The semantic label and standard semantic tally set of all screenshots to be measured of the comparison page to be measured, test judgement can be passed through when specific implementation
Whether the page to be measured normally shows in the browser, to obtain the compatibility of the page to be measured.Specifically, if by comparing institute
There is one of following situation in the semantic feature and standard semantic feature set for stating each screenshot to be measured, then determine the page to be measured
It is shown in the browser abnormal.
The semantic feature for having a screenshot to be measured in situation 1, at least one screenshot to be measured is not standard semantic feature set
Subset.
For example, the standard semantic feature set of some page includes { A, B, C, D, E, F }.There are two the page to be measured includes
Screenshot to be measured.Wherein, the semantic feature of the first screenshot to be measured is { C, D }；The semantic feature of second screenshot to be measured is { A, B, E, G }
Or { A, B, E, F, G }；The semantic feature of first screenshot to be measured is the subset of standard semantic feature set, and the second screenshot to be measured
Semantic feature is not the subset of standard semantic feature set.
Situation 2, the union of the semantic feature of each screenshot to be measured are the subsets of standard semantic feature set.
For example, the standard semantic feature set of some page includes { A, B, C, D, E, F }.There are two the page to be measured includes
Screenshot to be measured.Wherein, the semantic feature of the first screenshot to be measured is { C, D }；The semantic feature of second screenshot to be measured is { A, B, E }.
Although the first screenshot to be measured and the second screenshot to be measured are the subsets of standard semantic feature set, the semanteme of the first screenshot to be measured is special
Seek peace the second screenshot to be measured semantic feature union be standard semantic feature set subset, and concentrate lack semantic label " F ",
Instruction page lacks the corresponding picture material of the semantic label.
Technical solution provided by the embodiments of the present application, the abstract spy of at least one screenshot to be measured by extracting the page to be measured
The mode of sign, to identify whether multiple screenshots of the page normally show in a browser, in a browser with the determination page to be measured
Whether there is compatibility, realize the automatic test of page compatibility, and do not influenced by page adjustment, maintenance cost is low, quasi-
True rate is high.
Further, the convolutional neural networks knot similar to AlexNet can specifically be selected by model being extracted in above-described embodiment
Structure.This similar to AlexNet convolutional neural networks structure include 5 layers of convolutional layer, 3 layers of pond layer, 3 layers of full articulamentum because
Pond layer usually calculates in corresponding convolutional layer, does not calculate individually, so totally 8 layers of convolutional neural networks of the invention, 5 layers of volume
Lamination, 3 layers of full articulamentum, as shown in Figure 3.
Step 1022 i.e. in above-described embodiment, using described image data as the input for extracting model, execute the extraction
Model obtains the corresponding output of the described first screenshot to be measured as a result, can specifically realize using following steps:
S1, by described image data and the first convolution nuclear convolution, obtain the first convolution results.
Specifically, shown in Figure 2, image data is with 64 11 × 11 × 3 first convolution nuclear convolution output sizes
55 × 55 × 64 the first convolution results C1.
S2, first convolution results are carried out to obtain the second convolution knot again with the second convolution nuclear convolution after pondization operates
Fruit.
Wherein, pondization operation is to carry out down-sampled operation to the characteristic pattern of input, is sharply become to product in the first convolution results
The region of change is equalized.With continued reference to shown in Fig. 3, the first convolution results C1 after pondization operation and 192 5 × 5 ×
64 the second convolution nuclear convolution obtains the second convolution results of output C2 that size is 27 × 27 × 192.
S3, second convolution results are carried out to obtain third convolution knot again with third convolution nuclear convolution after pondization operates
Fruit.
With continued reference to the second convolution results C2 and 384 3 × 3 × 192 third convolution shown in Fig. 3, after pondization operation
Nuclear convolution obtains the output third convolution results C3 that size is 13 × 13 × 384.
S4, the third convolution results and Volume Four product nuclear convolution, obtain Volume Four product result.
With continued reference to shown in Fig. 3, Volume Four product nuclear convolution of the third convolution results C3 directly with 256 3 × 3 × 384 is obtained
The 4th layer of convolution results C4 for being 13 × 13 × 256 to size.
S5, the Volume Four is accumulated into result and the 5th convolution nuclear convolution, obtains the 5th convolution results.
With continued reference to shown in Fig. 3, the 4th layer of convolution results C4 and 256 3 × 3 × 256 the 5th convolution nuclear convolutions are obtained
The layer 5 convolution results C5 that size is 13 × 13 × 256.
S6, to the 5th convolution results carry out pondization operation, and to pondization operate after the 5th convolution results into
Row attended operation complete at least once, obtains the 6th full connection result.
With continued reference to the C5 ' that shown in Fig. 3, the output size after the operation of layer 5 convolution results C5 pondization is 6 × 6 × 256；
C5 ' obtains the first full articulamentum that size is 1 × 1 × 4096 (i.e. 4096) and exports FC6 using the first full articulamentum；
FC6 obtains the second full articulamentum that size is 4096 and exports FC7 using the second full articulamentum；
FC obtains the full articulamentum output FC7 ' of third as the 6th full connection result using the full articulamentum of third.
S7, classify to the 6th full connection result, obtain the abstract characteristics of the described first screenshot to be measured.
With continued reference to shown in Fig. 3, the 6th full connection result FC7 ' obtains the 8th layer entirely by a softmax classification layer
Articulamentum exports FC8, size 2, that is, the output result.
Image data passes through convolutional layer every time, i.e., with convolution nuclear convolution, in obtained output (C1-C5 as shown in Figure 3)
The calculation formula that each neuron is connect with the neuron of respective numbers in input image data may be expressed as:
Wherein,Indicate j-th of neuron in l layers of output characteristic pattern,Indicate all and neuronIt is connected
Input picture in neuron,Indicate withCorresponding weight,Indicate withCorresponding biasing, ReLU are indicated
Rectified Linear Unit activation primitive.Output neuron of the image data after full articulamentum and input neuron
Relationship it is similar with (1), unlike the full articulamentum of the last layer activation primitive be softmax.
What needs to be explained here is that: the embodiment of the present application is not construed as limiting the specific implementation of convolutional neural networks structure, on
State a kind of specific implementation for the only convolutional neural networks structure listed.Substantially, input picture size, the number of plies, each layer of function
Can, each layer parameter etc. can be all changed according to actual needs.
By content above it is found that the embodiment of the present application uses mould to improve the accuracy of High level feature extraction
The artificial intelligence learning algorithm of anthropomorphic brain, such as convolutional neural networks algorithm.This kind of learning algorithm need to be trained and test process,
Network output could be made more accurate.By taking convolutional neural networks algorithm as an example, which need to first pass through propagated forward will training sample
This is transmitted layer by layer, obtains the corresponding predicted value of training sample；Then backpropagation (Backpropagation is recycled
Algorithm, BP) error between predicted value and true value is back to every layer by algorithm layer by layer, finally calculate error cost function
To the local derviation of each layer parameter, the weight using stochastic gradient descent method to every layer is adjusted update, and each layer is each by modification
From weight so that network output it is more accurate.
It is obtained it follows that extracting model in technical solution provided by the embodiments of the present application and under type such as can be used:
104, training sample set and test sample collection are obtained.
105, training pattern is trained using the training sample set, to have been instructed model.
106, the model of having instructed is tested using the test sample collection, to obtain described having instructed the accurate of model
Rate.
107, using accuracy rate meet the requirements described in instructed model as the extraction model.
In above-mentioned 104, training sample set and test sample collection can be obtained by way of the intercepted samples page.For example,
Browser automation tools are disposed, on the standardized test machine for be mounted with different browsers to call each on test machine clear
The interface of device look at accesses each sample page and traverses page elements, while obtaining the subregion screenshot of each sample page.Often
All screenshots meeting batch of a browser is uploaded in database, and the information such as the quantity of programming count image and format.System
(such as server-side or client) can support the label for marking image online, facilitate subsequent training.Because being needed in subsequent training process
Comparison result using the corresponding label of image and training pattern output result carrys out the parameter in adjusting training model, so that training
Model output afterwards is more acurrate.When it is implemented, system can be that image carries out the mark of label or system mentions for user automatically
For a label configuration interface, user can be each image labeling label by label configuration interface；Etc. the embodiment of the present application pair
This is not especially limited.
A part of screenshot can be used as training sample in all screenshots of database purchase, and another part can be used as test specimens
This.For example, system is automatically separated out 10% screenshot as test sample, remaining 90% screenshot is as training sample, thus
Establish training sample set and test sample collection.
Wherein, the format of screenshot can be png or jpg format etc., and the embodiment of the present application is not especially limited this.
It follows that this step 104 may include following steps:
1041, multiple screenshot samples are obtained.
When it is implemented, sample page can be loaded in multiple and different browsers；The corresponding interface of each browser is called to visit
Ask the sample page shown in each browser, with subregion intercept the sample page shown in each browser obtain it is described
Multiple screenshot samples.
1042, semantic label is marked for each screenshot sample.
Wherein, label for labelling can be manually completed by user, can also system be automatically performed.For example, this step 1042 can be specific
Include:
In response to user for the label for labelling operation of the first screenshot sample triggering, it is associated with the label for labelling operation
At least one semantic label and the first screenshot sample being directed toward；Or
Using the first screenshot sample as input, executes image multi-tag marking model and obtain the first screenshot sample
At least one corresponding semantic label.
Wherein, image multi-tag marking model can be found in related content in the prior art, and details are not described herein again.
1043, part screenshot sample is isolated from the multiple screenshot sample as training sample, part screenshot sample
As test sample.
For example, random selects 10% as test sample from multiple screenshot samples, remaining 90% as training sample
This.
1044, the semantic label based on the training sample and the training sample, establishes the training sample set.
1045, the semantic label based on the test sample and the test sample, establishes the test sample collection.
Further, in order to improve the generalization ability of model, the i.e. accuracy rate of raising model in practical applications.Training sample
This concentration may also include following training sample, that is, above-mentioned steps 104 may also include that
1046, Image Adjusting is carried out to the screenshot sample as training sample.
1047, the training sample set is added to using the screenshot sample after Image Adjusting as newly-increased training sample.
Wherein, above-mentioned 1046 progress Image Adjusting includes:
Change the size of screenshot sample；And/or
Change the shape of screenshot sample；And/or
Intercept 85%~90% region in screenshot sample；And/or
Change the direction of screenshot sample；And/or
Change the color of screenshot sample.
Wherein, change the color of screenshot sample, comprising: change the brightness of screenshot sample；And/or change the full of screenshot sample
And degree；And/or change the contrast of screenshot sample.
What needs to be explained here is that: it can be in such a way that the pixel value of image be reduced 10% at random, to change screenshot sample
This size；By way of stretching or compressing the shape of figure, the shape of screenshot sample is adjusted；By being cut from image at random
The mode for taking the region of 85%~90% (such as 87.5%) carries out twice-screenshot to screenshot sample, to distinguish and adjust image slices
Element or compression；Direction changing screenshot sample by screenshot sample or so or by way of inverting upside down.
After training sample set is ready to, the size of image may be unmatched with training pattern at this time, therefore instruct
Before white silk, the training sample concentrated to training sample is needed to pre-process.Then, then using the data obtained after pretreatment as
The input of training pattern completes training and obtains extracting model needed for the embodiment of the present application.I.e. above-mentioned 105 step can be wrapped specifically
Include following steps:
1051, the first training sample concentrated to the training sample is pre-processed to obtain first sample data.
For example, training sample to be adjusted to the triple channel rgb image data being sized.Wherein, it is sized and training
Model is related.The planned network structure of training pattern is different, corresponding to be sized namely different.Than as shown in Figure 3 above
The convolutional neural networks model of structure, is sized and can be 224 × 224.Correspondingly, preprocessing process can specifically: will instruct
Practice the triple channel rgb image data that sample is adjusted to 224 × 224.
Need exist for supplement: in order to improve training speed when training pattern, using multi-thread concurrent processing mode.
For example, opening image using 100 thread concurrent processing 2100,2s is only needed through testing.In addition, in order to accelerate to train, specific
When implementation, also multiple images can be synthesized one " batch " (batch), and batch of data is converted into training pattern can quickly to handle
Format, training process can be accelerated 1 times or more in this way, and this processing method tests the speed for the pre- of trained model
Degree is also improved.
1052, using the first sample data as the input of training pattern, the training pattern is executed to obtain first
As a result.
1053, right according to the difference of the numeralization value of the corresponding semantic label of the training sample and first result
Parameter in the training pattern optimizes processing.
1054, the second training sample concentrated to the training sample pre-processes, and by pretreated described the
Input of two training samples as the training pattern executes the training pattern to obtain second as a result, until training sample
The numeralization value of the corresponding semantic label of the training sample of concentration meets preset with the difference for executing the result that training pattern obtains
Condition.
Wherein, when training sample executes the abstract characteristics semanteme corresponding with training sample that training pattern obtains as input
The difference of the numeralization value of label meets prerequisite, and the training pattern obtained at this time i.e. training is completed, and obtains described having instructed mould
Type.I.e. model training process can simply understand are as follows: training pattern starts iteration optimization after designing, according to the knot of each iteration
Fruit goes the parameters such as weight and the biasing of Optimized model, and the process of calculating is divided into propagated forward (i.e. above-mentioned steps 1052) and reversed biography
Broadcast (i.e. above-mentioned steps 1053 and 1054) two processes.Propagated forward is that training sample is inputted training pattern, uses parameter current
Calculate reality output；The process of backpropagation is the difference first calculated between reality output and ideal output, further according to difference
By output layer to input layer Reverse optimization network parameter, such iteration, until meeting the condition stopped, after meeting stop condition,
Training terminates to get to having instructed model.Wherein, ideal output is the value pre-set, i.e., each trained sample as input
This corresponding semantic label.Each semantic label has the value of a corresponding numeralization, defeated by comparing training sample
The value for entering the numeralization of result and semantic label that training pattern obtains, can be obtained difference.Next test sample is carried out
The test for extracting progress accuracy rate in model of training completion is input to after similarly pre-processing, this process only includes preceding to biography
It broadcasts, i.e., input picture is successively propagated in a network, obtains reality output as a result, comparing reality output and ideal output, synthesis again
All test samples obtain the accuracy rate of model.Same ideal output is the corresponding label of test sample；It is practical by comparison
It is whether accurate that extraction model output can be obtained in the value for exporting the numeralization of label corresponding with test sample.
In i.e. above-mentioned 106, can specifically it realize with the following method:
1061, for each test sample for respectively concentrating the test sample as input, execution is described to have instructed model to obtain
To the corresponding implementing result of each test sample.
1062, the number for the semantic label that the test sample concentrates the corresponding implementing result of test sample corresponding is calculated
The consistent quantity of the value of value accounts for the ratio of test sample total quantity, obtains the accuracy rate.
For example, it is 100 that the test sample, which concentrates the quantity of test sample, wherein the corresponding implementing result of test sample
The consistent quantity of the value of the numeralization of corresponding semantic label is 98；Then the accuracy rate for having instructed model is 98%.
System can be trained in training pattern multiple has instructed model.Here it is carried out using test sample collection to model has been instructed
The purpose of test can exactly select the satisfactory model of having instructed of accuracy rate as extraction model from multiple models.It is specific real
It is highest as extraction model to may be selected accuracy rate, or selects model of the accuracy rate greater than 97% as extraction model by Shi Shi.If
There are two or multiple accuracys rate for having instructed model meet the requirements, then can therefrom select one to instruct model as the extraction at random
Model.
It obtains extracting the extraction that the extraction model can be used to carry out the screenshot of the page to be measured abstract characteristics after model,
Process is similar with the process of test sample: opening the page to be measured with various browsers on a testing machine, tool is automatically to page to be measured
Face carries out screenshot, obtains screenshot to be measured.First screenshot to be measured is saved into database, is conveniently later used to the essence for extracting model
It adjusts, the abstract characteristics that model obtains each screenshot to be measured of the page to be measured are extracted in screenshot to be measured input after pretreatment later, right
It extracts and also only includes propagated forward for model.
In order to illustrate the technical effect of technical solution provided by the present application, inventor acquires the page-images of several applications
Carry out verification experimental verification.Total amount of images about 2100 in experiment, wherein the image 1400 normally shown is opened, abnormal show
Image 700 is opened.The image of the normal image of 140 displays and 70 display exceptions has therefrom been randomly selected as test sample
Collection, accounts for about the 10% of entire sample set, in addition 90% is used as training sample set.The Detection accuracy of experimental result, system is greater than
97%.
It should be noted that the executing subject of each step of above-described embodiment institute providing method may each be same equipment,
Alternatively, this method is also by distinct device as executing subject.For example, the executing subject of step 501 to step 503 can be equipment
A；For another example, step 501 and 502 executing subject can be equipment A, the executing subject of step 503 can be equipment B；Etc..
The technical solution that the application implements to provide does not need to provide the virtual machine dedicated for test, as long as fit
Hold multiple operating systems of (such as end PC and mobile terminal) or the page auto-browsing/traversal and screenshot function of client.Client
The screenshot for the page to be measured being truncated to is uploaded to server-side, server-side based on extract the technology of semantic feature to the page to be measured into
Row compatibility test, then test result is fed back into client.One embodiment of the application as shown in Figure 4 mentions as a result,
The structural schematic diagram of the page compatibility test system of confession.As shown in figure 4, provided in this embodiment, the system comprises clients
End and server-side.Wherein,
Client 201, for loading the page to be measured in a browser；Subregion carries out screenshot to the page to be measured, with
To at least one screenshot to be measured；At least one described screenshot to be measured is uploaded to server-side 202；
Server-side 202, for receive client 201 upload the page to be measured being shown in a browser at least one
Screenshot to be measured；The abstract characteristics of each screenshot to be measured are extracted from least one described screenshot to be measured respectively；According to described each to be measured
The abstract characteristics of screenshot determine compatibility of the page to be measured in the browser, and definitive result is fed back to the visitor
Family end 201.
Technical solution provided by the embodiments of the present application, the abstract spy of at least one screenshot to be measured by extracting the page to be measured
The mode of sign, to identify whether multiple screenshots of the page normally show in a browser, in a browser with the determination page to be measured
Whether have compatibility, realize the automatic test of page compatibility, and do not influenced by page adjustment, maintenance cost is low, quasi-
True rate is high.
Each component units in the page compatibility test system provided by the embodiments of the present application, such as client, server-side
Specific workflow and between Signalling exchange will be further described in following embodiment.
Fig. 5 outputs the flow diagram of the page compatibility test method of one embodiment of the application offer.The present embodiment
The method of offer is suitable for server-side.Wherein, the server-side can be General Server, cloud, virtual server etc.,
The embodiment of the present application is not especially limited this.As shown in figure 5, the page compatibility test method, comprising:
301, at least one screenshot to be measured for showing the page to be measured in a browser that client uploads is received.
302, the abstract characteristics of each screenshot to be measured are extracted from least one described screenshot to be measured respectively.
303, according to the abstract characteristics of each screenshot to be measured, compatibility of the page to be measured in the browser is determined
Property.
304, definitive result is fed back into the client.
In above-mentioned 301, client by calling itself screenshot function, to the page to be measured being shown in a browser into
Row screenshot；Then the screenshot to be measured being truncated to is uploaded to server-side, to be based on these screenshots to be measured to page to be measured by server-side
Compatibility of the face in the browser is determined.Using such framework, without providing the virtual machine dedicated for test, only
Cooperate each operating system of different clients, can easily carry out page compatibility test.
Above-mentioned 302 and 303 specific implementation can be found in the related content in above-mentioned embodiment illustrated in fig. 1, no longer superfluous herein
It states.
Technical solution provided by the embodiments of the present application, the abstract spy of at least one screenshot to be measured by extracting the page to be measured
The mode of sign, to identify whether multiple screenshots of the page normally show in a browser, in a browser with the determination page to be measured
Whether have compatibility, realize the automatic test of page compatibility, and do not influenced by page adjustment, maintenance cost is low, quasi-
True rate is high.
Further, page compatibility test method provided by the embodiments of the present application may also include the steps of:
304, training sample set and test sample collection are obtained.
305, training pattern is trained using the training sample set, to have been instructed model.
306, the model of having instructed is tested using the test sample collection, to obtain described having instructed the accurate of model
Rate.
307, using accuracy rate meet the requirements described in instructed model as the extraction model.
Above-mentioned 305~306 can related content in design parameter embodiment shown in FIG. 1, details are not described herein again.
In above-mentioned 304, the acquisition modes of training sample set and test sample collection can specifically:
3041, multiple screenshot samples that client uploads are received.
When it is implemented, calling the screenshot function of itself with right after client can be loaded with sample page in a browser
Sample page carries out screenshot；Then server-side is uploaded to using the image being truncated to as screenshot sample.
3042, semantic label is marked for each screenshot sample.
Wherein, the mark of semantic label can be user and be marked manually by client, and be uploaded to server-side；Or language
The mark of adopted label is realized automatically by server-side.I.e. this step 3042 may particularly include:
After receiving at least one semantic label that the user that the client uploads specifies for the first screenshot sample,
It is associated at least one described semantic label and the first screenshot sample；Or
Using the first screenshot sample as input, executes image multi-tag marking model and obtain the first screenshot sample
At least one corresponding semantic label.
3043, part screenshot sample is isolated from the multiple screenshot sample as training sample, part screenshot sample
As test sample.
3045, the semantic label based on the training sample and the training sample, establishes the training sample set.
3046, the semantic label based on the test sample and the test sample, establishes the test sample collection.
With above-mentioned embodiment illustrated in fig. 1, the training sample set in the present embodiment, which can also be added with, carries out figure to training sample
The newly-increased training sample being used as screenshot sample adjusted.I.e. above-mentioned 304 may also include that
3047, Image Adjusting is carried out to the screenshot sample as training sample.
Specifically, Image Adjusting includes: the size for changing screenshot sample；And/or change the shape of screenshot sample；And/or
Intercept 85%~90% region in screenshot sample；And/or change the direction of screenshot sample；And/or change the color of screenshot sample
It is color.
Wherein, change the color of screenshot sample, comprising: change the brightness of screenshot sample；And/or change the full of screenshot sample
And degree；And/or change the contrast of screenshot sample.
More detailed content in relation to Image Adjusting, reference can be made to above-mentioned embodiment shown in FIG. 1, details are not described herein again.
3048, the training sample set is added to using the screenshot sample after Image Adjusting as newly-increased training sample.
Fig. 6 shows the flow diagram for the page compatibility test method that the another embodiment of the application provides.This implementation
The method that example provides is suitable for client, which can be integrated in one in terminal with embedded program
Hardware is also possible to install an application software in the terminal, can also be that the tool being embedded in terminal operating system is soft
Part etc., the embodiment of the present invention is not construed as limiting this.The terminal can be include mobile phone, tablet computer, PDA (Personal
Digital Assistant, personal digital assistant), POS (Point of Sales, point-of-sale terminal), vehicle-mounted computer etc. it is any eventually
End equipment.Specifically, as shown in fig. 6, the method provided in this embodiment, comprising:
401, the page to be measured is loaded in a browser.
402, subregion carries out screenshot to the page to be measured, to obtain at least one screenshot to be measured.
403, at least one described screenshot to be measured is uploaded to server-side, to determine the page to be measured by the server-side
Compatibility in the browser.
Wherein it is determined that the abstract characteristics of each screenshot to be measured are respectively from described according to the abstract characteristics for being each screenshot to be measured
It is extracted at least one screenshot to be measured.
In above-mentioned 401, the page to be measured, which can be user and input after network address or clickthrough to load in a browser, to be browsed
In device.
In above-mentioned 402, shot operation executes after can be the screenshot instruction for receiving server-side transmission, is also possible to
It is loaded in browser and executes after the page to be measured automatically, the embodiment of the present application is not especially limited this.
Server-side determines the compatibility of the page to be measured in a browser based on the abstract characteristics of each screenshot to be measured in above-mentioned 403
Property and each screenshot to be measured abstract characteristics how to extract can be found in the various embodiments described above in related content, herein no longer
It repeats.
Technical solution provided by the embodiments of the present application, the abstract spy of at least one screenshot to be measured by extracting the page to be measured
The mode of sign, to identify whether multiple screenshots of the page normally show in a browser, in a browser with the determination page to be measured
Whether have compatibility, realize the automatic test of page compatibility, and do not influenced by page adjustment, maintenance cost is low, quasi-
True rate is high.
Further, client can show the server-side feedback after the definitive result for receiving server-side feedback
Definitive result；And/or the semantic definitive result for exporting the server-side feedback；And/or definitive result is that the screenshot to be measured exists
When not having compatibility in the browser, outputting alarm prompt.
Further, client can also be provided for server-side and be extracted for training in addition to uploading the page to be measured to server-side
The training sample and test sample of model.The method i.e. provided by the embodiments of the present application may also include that
404, sample page is loaded in multiple and different browsers.
405, it calls the corresponding interface of each browser to access the sample page shown in each browser, is intercepted with subregion
The sample page shown in each browser obtains the multiple screenshot sample.
406, the multiple screenshot sample is uploaded to server-side, to use multiple screenshot samples to instruction by the server-side
White silk model, which is trained and tests, to be obtained extracting model.
Wherein, the abstract characteristics of each screenshot to be measured are to execute the extraction mould using each screenshot to be measured as ginseng is entered
What type obtained.
Wherein, the training and test process of extracting model can be found in the corresponding contents in the various embodiments described above, herein no longer
It repeats.
As shown in the above, training pattern also needs the mark that semantic label is carried out to screenshot sample in training.Wherein,
The mark of semantic label can be marked by user by client manually, then by client upload service end；It can also be by server-side
Automatic marking.If the mark of semantic label marks realization using user manually, technical solution provided in this embodiment can also be wrapped
Include following steps:
It is that each screenshot sample marks by user in response to user for the label for labelling operation of each screenshot sample triggering
Semantic label be uploaded to the server-side, with by the server-side be based on each screenshot sample and the corresponding semanteme of each screenshot sample
Label establishes training sample set and test sample collection, and is utilized respectively the training sample set and the test sample collection to described
Training pattern, which is trained and tests, obtains the extraction model.
Technical solution provided by the embodiments of the present application can simply understand are as follows: be collected under various browsers by automation tools
Page-images and carry out statistics, be saved in database；The reasonable feature for extracting model structure study page-images of design,
And save the highest model of accuracy rate；Training will be saved and the extraction model tested is deployed in system, for new how clear
The page-images of device of looking at are tested.As shown in fig. 7,
501, client uploads multiple screenshot samples to server-side and user is the semantic label that each screenshot sample is specified.
502, server-side is based on multiple screenshot samples and establishes training sample set and test sample collection.
503, server-side is trained training pattern using the training sample that training sample is concentrated, and obtains multiple having instructed mould
Type.
504, the test sample that server-side is concentrated using test sample is tested to having instructed model respectively, and from it is multiple
Select the highest model of test accuracy rate as extraction model in instruction model.
505, client uploads at least one screenshot to be measured of the page to be measured to server-side.
506, server-side pre-processes each screenshot to be measured, and using pretreated each screenshot to be measured as extraction mould
The input of type executes the extraction model and obtains the abstract characteristics of each screenshot to be measured.
507, server-side obtains the abstract characteristics point of each screenshot to be measured according to the corresponding relationship of abstract characteristics and semantic label
Not corresponding semantic label.
508, server-side is determined by determining whether the corresponding semantic label of each screenshot to be measured belongs to exception class label
The compatibility of the page to be measured in a browser.
509, server-side is to client feedback definitive result.
510, client exports definitive result.
Above-mentioned 501~510 specific implementation can be found in the corresponding contents in the various embodiments described above, herein no longer specific limit
It is fixed.
Firstly, not needing to provide the virtual machine dedicated for test using technical solution provided by the embodiments of the present application, only
Cooperate multiple operating systems at the end PC and mobile terminal or the page auto-browsing/traversal and screenshot function of client, so that it may
It is convenient to carry out page compatibility test.On the other hand, this programme does not need manual compiling procedure script, and one is applied
For the page, once model training is completed, only need to complete to test using the automatic screenshot of program and detection every time.
Since convolutional neural networks are the extractions to image high-level characteristic, rather than to image pixel scale comparison, so even if
The page has frequent change, as long as global feature style is constant, then also not needing again screenshot mark or training pattern.Another side
Face, this programme not will receive environment influence, even if the pixel of the page to be measured is varied, page-size to be measured is varied, by
In the extraction that convolutional neural networks are to image high-level characteristic, rather than to image pixel scale comparison, this programme for
Page test to be measured is unaffected.Even if re -training model, the cost of this programme is also again well below other schemes
The cost for recording script need to only input new image pattern, so that it may small at 3 because model training systems have been prepared for
When within training generate and new model and dispose, and trained process does not need manual intervention, and model does not need manually yet
Maintenance, cost of labor are far below the recording of script.In addition, this programme identification single image time about 0.2s, comparison manually into
Row page compatibility test single image is in 2s or more, if it is considered that the switching time of image and the page, there are about 2 for recognition efficiency
The promotion of the order of magnitude, and by increasing training data, the accuracy rate of model can be improved to 99% or more.
Fig. 8 shows the structural schematic diagram of the page device for testing compatibility of one embodiment of the application offer.Such as Fig. 8 institute
Show, described device provided in this embodiment includes:
First obtains module 601, for obtaining at least one screenshot to be measured of the page to be measured of display in a browser；
First extraction module 602, for extracting the abstract of each screenshot to be measured from least one described screenshot to be measured respectively
Feature；
First determining module 603 determines the page to be measured in institute for the abstract characteristics according to each screenshot to be measured
State the compatibility in browser.
Further, first extraction module 602 is also used to: being pre-processed the first screenshot to be measured to obtain image
Data；Using described image data as the input for extracting model, executes the extraction model and obtain the described first screenshot to be measured
Abstract characteristics.Wherein, the first screenshot to be measured is any one screenshot to be measured at least one described screenshot to be measured.
Further, first extraction module 602 is also used to: the described first screenshot to be measured is adjusted to be sized
Triple channel rgb image data.
Wherein, convolutional neural networks model can be selected in first extraction module 602.
Further, first extraction module 602 is also used to:
By described image data and the first convolution nuclear convolution, the first convolution results are obtained；
First convolution results are carried out to obtain the second convolution results again with the second convolution nuclear convolution after pondization operates；
Second convolution results are carried out to obtain third convolution results again with third convolution nuclear convolution after pondization operates；
The third convolution results and Volume Four product nuclear convolution, obtain Volume Four product result；
By Volume Four product result and the 5th convolution nuclear convolution, the 5th convolution results are obtained；
To the 5th convolution results carry out pondization operation, and to pondization operate after the 5th convolution results carry out to
Few primary full attended operation, obtains the 6th full connection result；
Classify to the 6th full connection result, obtains the abstract characteristics of the described first screenshot to be measured.
Further, first determining module 603 is also used to: according to the corresponding relationship of abstract characteristics and semantic label,
Obtain the corresponding semantic label of abstract characteristics of each screenshot to be measured；By determining that each screenshot to be measured respectively corresponds
Semantic label whether belong to exception class label, determine compatibility of the page to be measured in the browser.
Further, first determining module 603 is also used to: in the corresponding semantic label of each screenshot to be measured
There is the corresponding semantic label of a screenshot to be measured to belong to exception class label, it is determined that the page to be measured in the browser not
Tool compatibility.
Further, page device for testing compatibility provided in this embodiment further include:
Second obtains module, for obtaining training sample set and test sample collection；
Training module, for being trained using the training sample set to training pattern, to have been instructed model；
Test module, for being tested using the test sample collection the model of having instructed, to obtain described instructed
The accuracy rate of model；
Choose module, for using accuracy rate meet the requirements described in instructed model as the extraction model.
Further, the second acquisition module is also used to: obtaining multiple screenshot samples；It is marked for each screenshot sample semantic
Label；A part of screenshot sample is isolated from the multiple screenshot sample as training sample, another part screenshot sample is made
For test sample；Semantic label based on the training sample and the training sample, establishes the training sample set；Based on institute
The semantic label for stating test sample and the test sample establishes the test sample collection.
Further, the second acquisition module is also used to: loading sample page in multiple and different browsers；It calls each
The corresponding interface of browser accesses the sample page shown in each browser, and the institute shown in each browser is intercepted with subregion
It states sample page and obtains the multiple screenshot sample.
Further, the second acquisition module is also used to: in response to user for the first screenshot sample triggering
Label for labelling operation is associated at least one semantic label and the first screenshot sample that the label for labelling operation is directed toward；Or
For person using the first screenshot sample as input, it is corresponding that execution image multi-tag marking model obtains the first screenshot sample
At least one semantic label.
Further, the second acquisition module is also used to: carrying out Image Adjusting to the screenshot sample as training sample；
The training sample set is added to using the screenshot sample after Image Adjusting as newly-increased training sample.
Further, the second acquisition module is also used to: changing the size of screenshot sample；And/or change screenshot sample
Shape；85%~90% region in interception screenshot sample and/or；And/or change the direction of screenshot sample；And/or change
The color of screenshot sample.
Further, the second acquisition module is also used to: changing the brightness of screenshot sample；And/or change screenshot sample
Saturation degree；And/or change the contrast of screenshot sample.
Further, the training module is also used to: the first training sample concentrated to the training sample is located in advance
Reason, obtains first sample data；Using the first sample data as the input of training pattern, the training pattern is executed to obtain
To the first result；According to the difference of the numeralization value of the corresponding semantic label of the training sample and first result, to institute
It states the parameter in training pattern and optimizes processing；The second training sample concentrated to the training sample pre-processes, and
Using pretreated second training sample as the input of the training pattern, the training pattern is executed to obtain second
As a result, until the knot that the numeralization value and execution training pattern of the corresponding semantic label of training sample that training sample is concentrated obtain
The difference of fruit meets prerequisite and has instructed model to get to described.
Further, the test module is also used to: each test sample for respectively concentrating the test sample is as defeated
Enter, execution is described to have instructed model to obtain the corresponding implementing result of each test sample；It calculates the test sample and concentrates test specimens
The consistent quantity of value of the numeralization of the corresponding semantic label of this corresponding implementing result accounts for the ratio of test sample total quantity
Example, obtains the accuracy rate.
What needs to be explained here is that: page device for testing compatibility provided by the above embodiment can be realized shown in above-mentioned Fig. 1
It is real that the principle of technical solution described in embodiment of the method, above-mentioned each module or unit specific implementation can be found in above-mentioned correlation method
The corresponding contents in example are applied, details are not described herein again.
Technical solution provided by the embodiments of the present application, the abstract spy of at least one screenshot to be measured by extracting the page to be measured
The mode of sign, to identify whether multiple screenshots of the page normally show in a browser, in a browser with the determination page to be measured
Whether have compatibility, realize the automatic test of page compatibility, and do not influenced by page adjustment, maintenance cost is low, quasi-
True rate is high.
Fig. 9 shows the structural schematic diagram of the page device for testing compatibility of one embodiment of the application offer.Such as Fig. 9 institute
Show, described device provided in this embodiment includes:
Receiving module 701, at least one of the display page to be measured in a browser for receiving client upload wait for
Survey screenshot；
Second extraction module 702, for extracting the abstract of each screenshot to be measured from least one described screenshot to be measured respectively
Feature；
Second determining module 703 determines the page to be measured in institute for the abstract characteristics according to each screenshot to be measured
State the compatibility in browser；
Definitive result is fed back to the client by feedback module 704.
Further, second extraction module 702 is also used to pre-process to obtain image the first screenshot to be measured
Data；Using described image data as the input for extracting model, executes the extraction model and obtain the described first screenshot to be measured
Abstract characteristics.Wherein, the described first screenshot to be measured is any one screenshot to be measured at least one described screenshot to be measured.
Further, second extraction module 702 is also used to be adjusted to the described first screenshot to be measured to be sized
Triple channel rgb image data.
Further, described second model is extracted as convolutional neural networks model.
Further, second extraction module 702 is also used to obtain described image data and the first convolution nuclear convolution
First convolution results；First convolution results are carried out to obtain the second convolution again with the second convolution nuclear convolution after pondization operates
As a result；Second convolution results are carried out to obtain third convolution results again with third convolution nuclear convolution after pondization operates；It is described
Third convolution results and Volume Four product nuclear convolution, obtain Volume Four product result；By Volume Four product result and the 5th convolution kernel
Convolution obtains the 5th convolution results；Pondization operation is carried out to the 5th convolution results, and to the described 5th after pondization operation
Convolution results carry out full attended operation at least once, obtain the 6th full connection result；The 6th full connection result is divided
Class obtains the abstract characteristics of the described first screenshot to be measured.
Further, second determining module 703 is also used to the corresponding relationship according to abstract characteristics and semantic label,
Obtain the corresponding semantic label of abstract characteristics of each screenshot to be measured；By determining the corresponding language of each screenshot to be measured
Whether adopted label belongs to exception class label, determines compatibility of the page to be measured in the browser.
Further, second determining module 703 is also used in the corresponding semantic label of each screenshot to be measured
There is the corresponding semantic label of a screenshot to be measured to belong to exception class label, it is determined that the page to be measured in the browser not
Tool compatibility.
Further, the page device for testing compatibility further include:
Module is obtained, for obtaining training sample set and test sample collection；
Training module, for being trained using the training sample set to training pattern, to have been instructed model；
Test module, for being tested using the test sample collection the model of having instructed, to obtain described instructed
The accuracy rate of model；
Selecting module, for using accuracy rate meet the requirements described in instructed model as the extraction model.
Further, the multiple screenshot samples for obtaining module and being also used to receive client upload；For each screenshot sample
Mark semantic label；A part of screenshot sample is isolated from the multiple screenshot sample as training sample, another part is cut
Pattern this as test sample；Semantic label based on the training sample and the training sample, establishes the training sample
Collection；Semantic label based on the test sample and the test sample, establishes the test sample collection.
Further, the module that obtains is also used to: receiving the user that the client uploads is first screenshot
After at least one specified semantic label of sample, it is associated at least one described semantic label and the first screenshot sample；Or
Using the first screenshot sample as input, executes image multi-tag marking model and obtain the first screenshot sample
At least one corresponding semantic label.
Further, the module that obtains is also used to carry out Image Adjusting to the screenshot sample as training sample；It will figure
As screenshot sample adjusted is added to the training sample set as newly-increased training sample.
Further, the module that obtains is also used to: changing the size of screenshot sample；And/or change the shape of screenshot sample
Shape；85%~90% region in interception screenshot sample and/or；And/or change the direction of screenshot sample；And/or change screenshot
The color of sample.
Further, the training module is also used to pre-process the first training sample that the training sample is concentrated
To obtain first sample data；Using the first sample data as the input of training pattern, the training pattern is executed to obtain
To the first result；According to the difference of the numeralization value of the corresponding semantic label of the training sample and first result, to institute
It states the parameter in training pattern and optimizes processing；The second training sample concentrated to the training sample pre-processes, and
Using pretreated second training sample as the input of the training pattern, the training pattern is executed to obtain second
As a result, until the knot that the numeralization value and execution training pattern of the corresponding semantic label of training sample that training sample is concentrated obtain
The difference of fruit meets prerequisite and has instructed model to get to described.
Further, each test sample that the test module is also used to respectively concentrate the test sample is as defeated
Enter, execution is described to have instructed model to obtain the corresponding implementing result of each test sample；It calculates the test sample and concentrates test specimens
The consistent quantity of value of the numeralization of the corresponding semantic label of this corresponding implementing result accounts for the ratio of test sample total quantity
Example, obtains the accuracy rate.
What needs to be explained here is that: page device for testing compatibility provided by the above embodiment can be realized shown in above-mentioned Fig. 5
It is real that the principle of technical solution described in embodiment of the method, above-mentioned each module or unit specific implementation can be found in above-mentioned correlation method
The corresponding contents in example are applied, details are not described herein again.
Technical solution provided by the embodiments of the present application, the abstract spy of at least one screenshot to be measured by extracting the page to be measured
The mode of sign, to identify whether multiple screenshots of the page normally show in a browser, in a browser with the determination page to be measured
Whether have compatibility, realize the automatic test of page compatibility, and do not influenced by page adjustment, maintenance cost is low, quasi-
True rate is high.
Figure 10 shows the structural schematic diagram of the page device for testing compatibility of one embodiment of the application offer.Such as Figure 10 institute
Show, described device includes:
Loading module 801, for loading the page to be measured in a browser；
Screen capture module 802 carries out screenshot to the page to be measured for subregion, to obtain at least one screenshot to be measured；
Uploading module 803, at least one described screenshot to be measured to be uploaded to server-side, to be determined by the server-side
Compatibility of the page to be measured in the browser；
Wherein it is determined that the abstract characteristics of each screenshot to be measured are respectively from described according to the abstract characteristics for being each screenshot to be measured
It is extracted at least one screenshot to be measured.
Further, described device further include:
Output module, for showing the definitive result of the server-side feedback；And/or the semantic output server-side feedback
Definitive result；And/or definitive result is the screenshot to be measured when not having compatibility in the browser, outputting alarm mentions
Show.
Further, described device further include:
The loading module 801 is also used to load sample page in multiple and different browsers；
The screen capture module 802 is also used to that the corresponding interface of each browser is called to access the sample shown in each browser
This page intercepts the sample page shown in each browser with subregion and obtains the multiple screenshot sample；
The uploading module 803 is also used to the multiple screenshot sample being uploaded to server-side, to be made by the server-side
Training pattern is trained and is tested with multiple screenshot samples and obtains extracting model；
Wherein, the abstract characteristics of each screenshot to be measured are to execute the extraction mould using each screenshot to be measured as ginseng is entered
What type obtained.
Further, the uploading module 803 is also used to be directed to the label for labelling of each screenshot sample triggering in response to user
The semantic label that user is each screenshot sample mark is uploaded to the server-side, to be based on by the server-side by operation
Each screenshot sample and the corresponding semantic label of each screenshot sample establish training sample set and test sample collection, and are utilized respectively described
Training sample set and the test sample collection, which are trained and test to the training pattern, obtains the extraction model.
What needs to be explained here is that: page device for testing compatibility provided by the above embodiment can be realized shown in above-mentioned Fig. 6
It is real that the principle of technical solution described in embodiment of the method, above-mentioned each module or unit specific implementation can be found in above-mentioned correlation method
The corresponding contents in example are applied, details are not described herein again.
Technical solution provided by the embodiments of the present application, the abstract spy of at least one screenshot to be measured by extracting the page to be measured
The mode of sign, to identify whether multiple screenshots of the page normally show in a browser, in a browser with the determination page to be measured
Whether have compatibility, realize the automatic test of page compatibility, and do not influenced by page adjustment, maintenance cost is low, quasi-
True rate is high.
Figure 11 shows the structural schematic diagram of the electronic equipment of one embodiment of the application offer.As shown in figure 11, the electricity
Sub- equipment includes: first memory 901 and first processor 902, wherein
The first memory 901, for storing program；
The first processor 902 is coupled with the first memory 901, for executing in the first memory 901
The described program of storage, to be used for:
Obtain at least one screenshot to be measured of the page to be measured of display in a browser；
The abstract characteristics of each screenshot to be measured are extracted from least one described screenshot to be measured respectively；
According to the abstract characteristics of each screenshot to be measured, compatibility of the page to be measured in the browser is determined.
Technical solution provided by the embodiments of the present application, the abstract spy of at least one screenshot to be measured by extracting the page to be measured
The mode of sign, to identify whether multiple screenshots of the page normally show in a browser, in a browser with the determination page to be measured
Whether have compatibility, realize the automatic test of page compatibility, and do not influenced by page adjustment, maintenance cost is low, quasi-
True rate is high.
Above-mentioned first memory 901 can be configured to store various other data to support the operation in equipment beyond the clouds.
The example of these data includes the instruction of any application or method operated in equipment beyond the clouds.First memory
901 can be realized by any kind of volatibility or non-volatile memory device or their combination, such as static random-access
Memory (SRAM), electrically erasable programmable read-only memory (EEPROM), Erasable Programmable Read Only Memory EPROM (EPROM),
Programmable read only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, disk or CD.
Above-mentioned first processor 902 is when executing the program in first memory 901, other than function above, also
Other functions can be achieved, for details, reference can be made to the descriptions of previous embodiments.
Further, as shown in figure 11, electronic equipment further include: the first communication component 903, first the 904, first electricity of display
Other components such as source component 905, the first audio component 906.Members are only schematically provided in Figure 11, are not meant to electronics
Equipment only includes component shown in Figure 11.
Correspondingly, the embodiment of the present application also provides a kind of computer readable storage medium for being stored with computer program, institute
State can be realized when computer program is computer-executed the various embodiments described above offer page compatibility test method step or
Function.
Figure 12 shows the structural schematic diagram of the server device of one embodiment of the application offer.As shown in figure 12, described
Server device includes: second memory 1001 and second processor 1002, wherein
The second memory 1001, for storing program；
The second processor 1002 is coupled with the second memory 1001, for executing the second memory
The described program stored in 1001, to be used for:
Receive at least one screenshot to be measured for the page to be measured being shown in a browser that client uploads；
The abstract characteristics of each screenshot to be measured are extracted from least one described screenshot to be measured respectively；
According to the abstract characteristics of each screenshot to be measured, compatibility of the page to be measured in the browser is determined；
Definitive result is fed back into the client.
Technical solution provided by the embodiments of the present application, the abstract spy of at least one screenshot to be measured by extracting the page to be measured
The mode of sign, to identify whether multiple screenshots of the page normally show in a browser, in a browser with the determination page to be measured
Whether have compatibility, realize the automatic test of page compatibility, and do not influenced by page adjustment, maintenance cost is low, quasi-
True rate is high.
Above-mentioned second memory 1001 can be configured to store various other data to support the operation in equipment beyond the clouds.
The example of these data includes the instruction of any application or method operated in equipment beyond the clouds.Second memory
1001 can be realized by any kind of volatibility or non-volatile memory device or their combination, such as static random-access
Memory (SRAM), electrically erasable programmable read-only memory (EEPROM), Erasable Programmable Read Only Memory EPROM (EPROM),
Programmable read only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, disk or CD.
Above-mentioned second processor 1002 is when executing the program in second memory 1001, other than function above,
It can also be achieved other functions, for details, reference can be made to the descriptions of previous embodiments.
Further, as shown in figure 12, server device further include: the second communication component 1003, second display 1004,
Other components such as two power supply modules 1005, the second audio component 1006.Members are only schematically provided in Figure 12, are not intended to
Server device only include component shown in Figure 12.
Correspondingly, the embodiment of the present application also provides a kind of computer readable storage medium for being stored with computer program, institute
State can be realized when computer program is computer-executed the various embodiments described above offer page compatibility test method step or
Function.
Figure 13 shows the structural schematic diagram for the client device that the another embodiment of the application provides.As shown in figure 13, institute
Stating client device includes: third memory 1101 and third processor 1102, wherein
The third memory 1101, for storing program；
The third processor 1102 is coupled with the third memory 1101, for executing the third memory
The described program stored in 1101, to be used for:
The page to be measured is loaded in a browser；
Subregion carries out screenshot to the page to be measured, to obtain at least one screenshot to be measured；
At least one described screenshot to be measured is uploaded to server-side, to determine the page to be measured in institute by the server-side
State the compatibility in browser；
Wherein it is determined that the abstract characteristics of each screenshot to be measured are respectively from described according to the abstract characteristics for being each screenshot to be measured
It is extracted at least one screenshot to be measured.
Technical solution provided by the embodiments of the present application, the abstract spy of at least one screenshot to be measured by extracting the page to be measured
The mode of sign, to identify whether multiple screenshots of the page normally show in a browser, in a browser with the determination page to be measured
Whether have compatibility, realize the automatic test of page compatibility, and do not influenced by page adjustment, maintenance cost is low, quasi-
True rate is high.
Above-mentioned third memory 1101 can be configured to store various other data to support the operation in equipment beyond the clouds.
The example of these data includes the instruction of any application or method operated in equipment beyond the clouds.Third memory
1101 can be realized by any kind of volatibility or non-volatile memory device or their combination, such as static random-access
Memory (SRAM), electrically erasable programmable read-only memory (EEPROM), Erasable Programmable Read Only Memory EPROM (EPROM),
Programmable read only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, disk or CD.
Above-mentioned third processor 1102 is when executing the program in third memory 1101, other than function above,
It can also be achieved other functions, for details, reference can be made to the descriptions of previous embodiments.
Further, as shown in figure 13, client device further include: third communication component 1103, third display 1104,
Other components such as three power supply modules 1105, third audio component 1106.Members are only schematically provided in Figure 13, are not intended to
Client device only include component shown in Figure 13.
Correspondingly, the embodiment of the present application also provides a kind of computer readable storage medium for being stored with computer program, institute
State can be realized when computer program is computer-executed the various embodiments described above offer page compatibility test method step or
Function.
Display in Figure 11, Figure 12 and Figure 13, may include screen, and screen may include liquid crystal display
(LCD) and touch panel (TP).If screen includes touch panel, screen may be implemented as touch screen, be used by oneself with receiving
The input signal at family.Touch panel includes one or more touch sensors to sense the hand on touch, slide, and touch panel
Gesture.The touch sensor can not only sense the boundary of a touch or slide action, but also detect and the touch or sliding
Operate relevant duration and pressure.
Power supply module in Figure 11, Figure 12 and Figure 13, the various assemblies for power supply module corresponding device provide electric power.Electricity
Source component may include power-supply management system, one or more power supplys and other with for power supply module corresponding device generate, management
With the distribution associated component of electric power.
Audio component in Figure 11, Figure 12 and Figure 13, is configured as output and/or input audio signal.For example, audio
Component includes a microphone (MIC), when audio component corresponding device is in operation mode, as call model, logging mode and
When speech recognition mode, microphone is configured as receiving external audio signal.The received audio signal can be deposited further
Storage is sent in memory or via communication component.In some embodiments, audio component further includes a loudspeaker, for exporting
Audio signal.
It should be understood by those skilled in the art that, the embodiment of the present invention can provide as method, system or computer program
Product.Therefore, complete hardware embodiment, complete software embodiment or reality combining software and hardware aspects can be used in the present invention
Apply the form of example.Moreover, it wherein includes the computer of computer usable program code that the present invention, which can be used in one or more,
The computer program implemented in usable storage medium (including but not limited to magnetic disk storage, CD-ROM, optical memory etc.) produces
The form of product.
The present invention be referring to according to the method for the embodiment of the present invention, the process of equipment (system) and computer program product
Figure and/or block diagram describe.It should be understood that every one stream in flowchart and/or the block diagram can be realized by computer program instructions
The combination of process and/or box in journey and/or box and flowchart and/or the block diagram.It can provide these computer programs
Instruct the processor of general purpose computer, special purpose computer, Embedded Processor or other programmable data processing devices to produce
A raw machine, so that being generated by the instruction that computer or the processor of other programmable data processing devices execute for real
The device for the function of being specified in present one or more flows of the flowchart and/or one or more blocks of the block diagram.
These computer program instructions, which may also be stored in, is able to guide computer or other programmable data processing devices with spy
Determine in the computer-readable memory that mode works, so that it includes referring to that instruction stored in the computer readable memory, which generates,
Enable the manufacture of device, the command device realize in one box of one or more flows of the flowchart and/or block diagram or
The function of being specified in multiple boxes.
These computer program instructions also can be loaded onto a computer or other programmable data processing device, so that counting
Series of operation steps are executed on calculation machine or other programmable devices to generate computer implemented processing, thus in computer or
The instruction executed on other programmable devices is provided for realizing in one or more flows of the flowchart and/or block diagram one
The step of function of being specified in a box or multiple boxes.
It should also be noted that, the terms "include", "comprise" or its any other variant are intended to nonexcludability
It include so that the process, method, commodity or the equipment that include a series of elements not only include those elements, but also to wrap
Include other elements that are not explicitly listed, or further include for this process, method, commodity or equipment intrinsic want
Element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that including described want
There is also other identical elements in the process, method of element, commodity or equipment.
The above description is only an example of the present application, is not intended to limit this application.For those skilled in the art
For, various changes and changes are possible in this application.All any modifications made within the spirit and principles of the present application are equal
Replacement, improvement etc., should be included within the scope of the claims of this application.
Claims (38)
1. a kind of page compatibility test method characterized by comprising
Obtain at least one screenshot to be measured of the page to be measured of display in a browser；
The abstract characteristics of each screenshot to be measured are extracted from least one described screenshot to be measured respectively；
According to the abstract characteristics of each screenshot to be measured, compatibility of the page to be measured in the browser is determined.
2. the method according to claim 1, wherein extracting the described first screenshot to be measured from the first screenshot to be measured
Abstract characteristics, comprising:
Described first screenshot to be measured is pre-processed to obtain image data；
Using described image data as the input for extracting model, executes the extraction model and obtain the pumping of the described first screenshot to be measured
As feature；
Wherein, the described first screenshot to be measured is any one screenshot to be measured at least one described screenshot to be measured.
3. according to the method described in claim 2, it is characterized in that, being pre-processed to the described first screenshot to be measured, comprising:
Described first screenshot to be measured is adjusted to the triple channel rgb image data being sized.
4. according to the method described in claim 2, it is characterized in that, the extraction model is convolutional neural networks model.
5. according to the method described in claim 4, it is characterized in that, being held using described image data as the input for extracting model
The row extraction model obtains the abstract characteristics of the described first screenshot to be measured, comprising:
By described image data and the first convolution nuclear convolution, the first convolution results are obtained；
First convolution results are carried out to obtain the second convolution results again with the second convolution nuclear convolution after pondization operates；
Second convolution results are carried out to obtain third convolution results again with third convolution nuclear convolution after pondization operates；
The third convolution results and Volume Four product nuclear convolution, obtain Volume Four product result；
By Volume Four product result and the 5th convolution nuclear convolution, the 5th convolution results are obtained；
Pondization operation is carried out to the 5th convolution results, and at least one is carried out to the 5th convolution results after pondization operation
Secondary full attended operation, obtains the 6th full connection result；
Classify to the 6th full connection result, obtains the abstract characteristics of the described first screenshot to be measured.
6. the method according to any one of claims 1 to 5, which is characterized in that according to the abstract of each screenshot to be measured
Feature determines compatibility of the page to be measured in the browser, comprising:
According to the corresponding relationship of abstract characteristics and semantic label, the corresponding language of abstract characteristics of each screenshot to be measured is obtained
Adopted label；
By determining whether the corresponding semantic label of each screenshot to be measured belongs to exception class label, the page to be measured is determined
Compatibility of the face in the browser.
7. according to the method described in claim 6, it is characterized in that, described by determining that each screenshot to be measured is corresponding
Whether semantic label belongs to exception class label, determines compatibility of the page to be measured in the browser, comprising:
There is the corresponding semantic label of a screenshot to be measured to belong to exception class in the corresponding semantic label of each screenshot to be measured
Label, it is determined that the page to be measured does not have compatibility in the browser.
8. the method according to any one of claim 2 to 5, which is characterized in that further include:
Obtain training sample set and test sample collection；
Training pattern is trained using the training sample set, to have been instructed model；
The model of having instructed is tested using the test sample collection, to obtain the accuracy rate for having instructed model；
Using accuracy rate meet the requirements described in instructed model as the extraction model.
9. according to the method described in claim 8, it is characterized in that, the acquisition training sample set and test sample collection, comprising:
Obtain multiple screenshot samples；
Semantic label is marked for each screenshot sample；
A part of screenshot sample is isolated from the multiple screenshot sample as training sample, another part screenshot sample conduct
Test sample；
Semantic label based on the training sample and the training sample, establishes the training sample set；
Semantic label based on the test sample and the test sample, establishes the test sample collection.
10. according to the method described in claim 9, it is characterized in that, described obtain multiple screenshot samples, comprising:
Sample page is loaded in multiple and different browsers；
It calls the corresponding interface of each browser to access the sample page shown in each browser, each browser is intercepted with subregion
The sample page of middle display obtains the multiple screenshot sample.
11. according to the method described in claim 9, it is characterized in that, for the first screenshot sample in the multiple screenshot sample
Mark semantic label, comprising:
In response to user for the label for labelling operation of the first screenshot sample triggering, it is associated with the label for labelling operation and is directed toward
At least one semantic label and the first screenshot sample；Or
Using the first screenshot sample as input, it is corresponding that execution image multi-tag marking model obtains the first screenshot sample
At least one semantic label.
12. the method according to any one of claim 9 to 11, which is characterized in that the acquisition training sample set also wraps
It includes:
Image Adjusting is carried out to the screenshot sample as training sample；
The training sample set is added to using the screenshot sample after Image Adjusting as newly-increased training sample.
13. according to the method for claim 12, which is characterized in that carry out image tune to the screenshot sample as training sample
It is whole, comprising:
Change the size of screenshot sample；And/or
Change the shape of screenshot sample；And/or
Intercept 85%~90% region in screenshot sample；And/or
Change the direction of screenshot sample；And/or
Change the color of screenshot sample.
14. according to the method for claim 13, which is characterized in that change the color of screenshot sample, comprising:
Change the brightness of screenshot sample；And/or
Change the saturation degree of screenshot sample；And/or
Change the contrast of screenshot sample.
15. according to the method described in claim 8, it is characterized in that, being instructed using the training sample set to training pattern
Practice, to have been instructed model, comprising:
The first training sample concentrated to the training sample pre-processes, and obtains first sample data；
Using the first sample data as the input of training pattern, the training pattern is executed to obtain the first result；
According to the difference of the numeralization value of the corresponding semantic label of the training sample and first result, to the trained mould
Parameter in type optimizes processing；
The second training sample concentrated to the training sample pre-processes, and by pretreated second training sample
As the input of the training pattern, the training pattern is executed to obtain second as a result, until the training that training sample is concentrated
The numeralization value of the corresponding semantic label of sample meets prerequisite with the difference for executing the result that training pattern obtains to get arriving
It is described to have instructed model.
16. according to the method described in claim 8, it is characterized in that, using the test sample collection to it is described instructed model into
Row test, to obtain the accuracy rate for having instructed model, comprising:
For each test sample that the test sample is concentrated respectively as input, execution is described to have instructed model to obtain each test specimens
This corresponding implementing result；
Calculate the value of the numeralization for the semantic label that the test sample concentrates the corresponding implementing result of test sample corresponding
Consistent quantity accounts for the ratio of test sample total quantity, obtains the accuracy rate.
17. a kind of page compatibility test method characterized by comprising
Receive at least one screenshot to be measured for showing the page to be measured in a browser that client uploads；
The abstract characteristics of each screenshot to be measured are extracted from least one described screenshot to be measured respectively；
According to the abstract characteristics of each screenshot to be measured, compatibility of the page to be measured in the browser is determined；
Definitive result is fed back into the client.
18. according to the method for claim 17, which is characterized in that extract described first to be measured section from the first screenshot to be measured
The abstract characteristics of figure, comprising:
Described first screenshot to be measured is pre-processed, image data is obtained；
Using described image data as the input for extracting model, executes the extraction model and obtain the pumping of the described first screenshot to be measured
As feature；
Wherein, the described first screenshot to be measured is any one screenshot to be measured at least one described screenshot to be measured.
19. according to the method for claim 18, which is characterized in that pre-processed to the described first screenshot to be measured, comprising:
Described first screenshot to be measured is adjusted to the triple channel rgb image data being sized.
20. according to the method for claim 18, which is characterized in that the extraction model is convolutional neural networks model.
21. according to the method for claim 20, which is characterized in that using described image data as extraction model input,
It executes the extraction model and obtains the abstract characteristics of the described first screenshot to be measured, comprising:
By described image data and the first convolution nuclear convolution, the first convolution results are obtained；
First convolution results are carried out to obtain the second convolution results again with the second convolution nuclear convolution after pondization operates；
Second convolution results are carried out to obtain third convolution results again with third convolution nuclear convolution after pondization operates；
The third convolution results and Volume Four product nuclear convolution, obtain Volume Four product result；
By Volume Four product result and the 5th convolution nuclear convolution, the 5th convolution results are obtained；
Pondization operation is carried out to the 5th convolution results, and at least one is carried out to the 5th convolution results after pondization operation
Secondary full attended operation, obtains the 6th full connection result；
Classify to the 6th full connection result, obtains the abstract characteristics of the described first screenshot to be measured.
22. method described in any one of 7 to 21 according to claim 1, which is characterized in that according to the pumping of each screenshot to be measured
As feature, compatibility of the page to be measured in the browser is determined, comprising:
According to the corresponding relationship of abstract characteristics and semantic label, the corresponding semantic mark of abstract characteristics of each screenshot to be measured is obtained
Label；
By determining whether the corresponding semantic label of each screenshot to be measured belongs to exception class label, the page to be measured is determined
Compatibility of the face in the browser.
23. according to the method for claim 22, which is characterized in that described by determining that each screenshot to be measured respectively corresponds
Semantic label whether belong to exception class label, determine compatibility of the page to be measured in the browser, comprising:
There is the corresponding semantic label of a screenshot to be measured to belong to exception class in the corresponding semantic label of each screenshot to be measured
Label, it is determined that the page to be measured does not have compatibility in the browser.
24. method described in any one of 8 to 21 according to claim 1, which is characterized in that further include:
Obtain training sample set and test sample collection；
Training pattern is trained using the training sample set, to have been instructed model；
The model of having instructed is tested using the test sample collection, to obtain the accuracy rate for having instructed model；
Using accuracy rate meet the requirements described in instructed model as the extraction model.
25. according to the method for claim 24, which is characterized in that the acquisition training sample set and test sample collection, packet
It includes:
Receive multiple screenshot samples that client uploads；
Semantic label is marked for each screenshot sample；
A part of screenshot sample is isolated from the multiple screenshot sample as training sample, another part screenshot sample conduct
Test sample；
Semantic label based on the training sample and the training sample, establishes the training sample set；
Semantic label based on the test sample and the test sample, establishes the test sample collection.
26. according to the method for claim 25, which is characterized in that for the first screenshot sample in the multiple screenshot sample
Mark semantic label, comprising:
Receiving the user that the client uploads is association after at least one semantic label that the first screenshot sample is specified
At least one described semantic label and the first screenshot sample；Or
Using the first screenshot sample as input, it is corresponding that execution image multi-tag marking model obtains the first screenshot sample
At least one semantic label.
27. according to the method for claim 25, which is characterized in that the acquisition training sample set, further includes:
Image Adjusting is carried out to the screenshot sample as training sample；
The training sample set is added to using the screenshot sample after Image Adjusting as newly-increased training sample.
28. according to the method for claim 27, which is characterized in that carry out image tune to the screenshot sample as training sample
It is whole, comprising:
Change the size of screenshot sample；And/or
Change the shape of screenshot sample；And/or
Intercept 85%~90% region in screenshot sample；And/or
Change the direction of screenshot sample；And/or
Change the color of screenshot sample.
29. according to the method for claim 24, which is characterized in that instructed using the training sample set to training pattern
Practice, to have been instructed model, comprising:
The first training sample concentrated to the training sample is pre-processed to obtain first sample data；
Using the first sample data as the input of training pattern, the training pattern is executed to obtain the first result；
According to the difference of the numeralization value of the corresponding semantic label of the training sample and first result, to the trained mould
Parameter in type optimizes processing；
The second training sample concentrated to the training sample pre-processes, and by pretreated second training sample
As the input of the training pattern, the training pattern is executed to obtain second as a result, until the training that training sample is concentrated
The numeralization value of the corresponding semantic label of sample meets prerequisite with the difference for executing the result that training pattern obtains to get arriving
It is described to have instructed model.
30. according to the method for claim 24, which is characterized in that using the test sample collection to it is described instructed model into
Row test, to obtain the accuracy rate for having instructed model, comprising:
For each test sample that the test sample is concentrated respectively as input, execution is described to have instructed model to obtain each test specimens
This corresponding implementing result；
Calculate the value of the numeralization for the semantic label that the test sample concentrates the corresponding implementing result of test sample corresponding
Consistent quantity accounts for the ratio of test sample total quantity, obtains the accuracy rate.
31. a kind of page compatibility test method characterized by comprising
The page to be measured is loaded in a browser；
Subregion carries out screenshot to the page to be measured, to obtain at least one screenshot to be measured；
At least one described screenshot to be measured is uploaded to server-side, to determine the page to be measured described clear by the server-side
The compatibility look in device；
Wherein it is determined that according to be each screenshot to be measured abstract characteristics, the abstract characteristics of each screenshot to be measured be respectively from it is described at least
It is extracted in one screenshot to be measured.
32. according to the method for claim 31, which is characterized in that further include:
Show the definitive result of the server-side feedback；And/or
Semanteme exports the definitive result of the server-side feedback；And/or
When definitive result is that the screenshot to be measured does not have compatibility in the browser, outputting alarm prompt.
33. the method according to claim 31 or 32, which is characterized in that further include:
Sample page is loaded in multiple and different browsers；
It calls the corresponding interface of each browser to access the sample page shown in each browser, each browser is intercepted with subregion
The sample page of middle display obtains the multiple screenshot sample；
The multiple screenshot sample is uploaded to server-side, with by the server-side using multiple screenshot samples to training pattern into
Row training and test obtain extracting model；
Wherein, the abstract characteristics of each screenshot to be measured are obtained using each screenshot to be measured as the ginseng execution extraction model is entered
It arrives.
34. according to the method for claim 33, which is characterized in that further include:
It is the language of each screenshot sample mark by user in response to user for the label for labelling operation of each screenshot sample triggering
Adopted label is uploaded to the server-side, to be based on each screenshot sample and the corresponding semantic label of each screenshot sample by the server-side
Training sample set and test sample collection are established, and is utilized respectively the training sample set and the test sample collection to the training
Model, which is trained and tests, obtains the extraction model.
35. a kind of page compatibility test system characterized by comprising
Client, for loading the page to be measured in a browser；Subregion carries out screenshot to the page to be measured, to obtain at least one
A screenshot to be measured；At least one described screenshot to be measured is uploaded to server-side；
Server-side, at least one screenshot to be measured of the page to be measured being shown in a browser for receiving client upload；
The abstract characteristics of each screenshot to be measured are extracted from least one described screenshot to be measured respectively；According to the abstract of each screenshot to be measured
Feature determines compatibility of the page to be measured in the browser, and definitive result is fed back to the client.
36. a kind of electronic equipment characterized by comprising first memory and first processor, wherein
The first memory, for storing program；
The first processor is coupled with the first memory, for executing the journey stored in the first memory
Sequence, to be used for:
Obtain at least one screenshot to be measured of the page to be measured of display in a browser；
The abstract characteristics of each screenshot to be measured are extracted from least one described screenshot to be measured respectively；
According to the abstract characteristics of each screenshot to be measured, compatibility of the page to be measured in the browser is determined.
37. a kind of server device characterized by comprising second memory and second processor, wherein
The second memory, for storing program；
The second processor is coupled with the second memory, for executing the journey stored in the second memory
Sequence, to be used for:
Receive at least one screenshot to be measured for the page to be measured being shown in a browser that client uploads；
The abstract characteristics of each screenshot to be measured are extracted from least one described screenshot to be measured respectively；
According to the abstract characteristics of each screenshot to be measured, compatibility of the page to be measured in the browser is determined；
Definitive result is fed back into the client.
38. a kind of client device characterized by comprising third memory and third processor, wherein
The third memory, for storing program；
The third processor is coupled with the third memory, for executing the journey stored in the third memory
Sequence, to be used for:
The page to be measured is loaded in a browser；
Subregion carries out screenshot to the page to be measured, to obtain at least one screenshot to be measured；
At least one described screenshot to be measured is uploaded to server-side, to determine the page to be measured described clear by the server-side
The compatibility look in device；
Wherein it is determined that according to be each screenshot to be measured abstract characteristics, the abstract characteristics of each screenshot to be measured be respectively from it is described at least
It is extracted in one screenshot to be measured.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810215223.8A CN110275820B (en) | 2018-03-15 | 2018-03-15 | Page compatibility testing method, system and equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810215223.8A CN110275820B (en) | 2018-03-15 | 2018-03-15 | Page compatibility testing method, system and equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110275820A true CN110275820A (en) | 2019-09-24 |
| CN110275820B CN110275820B (en) | 2023-11-21 |
Family
ID=67958108
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810215223.8A Active CN110275820B (en) | 2018-03-15 | 2018-03-15 | Page compatibility testing method, system and equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110275820B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110807007A (en) * | 2019-09-30 | 2020-02-18 | 支付宝(杭州)信息技术有限公司 | Target detection model training method, device and system and storage medium |
| CN110851349A (en) * | 2019-10-10 | 2020-02-28 | 重庆金融资产交易所有限责任公司 | Page abnormal display detection method, terminal equipment and storage medium |
| CN113506291A (en) * | 2021-07-29 | 2021-10-15 | 上海幻电信息科技有限公司 | Compatibility testing method and device |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110093773A1 (en) * | 2009-10-19 | 2011-04-21 | Browsera LLC | Automated application compatibility testing |
| CN106326091A (en) * | 2015-06-24 | 2017-01-11 | 深圳市腾讯计算机系统有限公司 | Browser webpage compatibility detection method and system |
| CN106407119A (en) * | 2016-09-28 | 2017-02-15 | 浪潮软件集团有限公司 | Browser compatibility testing method based on automatic testing |
| CN106681917A (en) * | 2016-12-21 | 2017-05-17 | 南京大学 | Method for automatically evaluating front ends on basis of neural networks |
| CN106886491A (en) * | 2017-01-17 | 2017-06-23 | 博彦科技股份有限公司 | Webpage compatibility test method and device |
| CN107729249A (en) * | 2017-10-18 | 2018-02-23 | 北京奇虎科技有限公司 | Browser compatibility method of testing, device, system and electronic equipment |
-
2018
- 2018-03-15 CN CN201810215223.8A patent/CN110275820B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110093773A1 (en) * | 2009-10-19 | 2011-04-21 | Browsera LLC | Automated application compatibility testing |
| CN106326091A (en) * | 2015-06-24 | 2017-01-11 | 深圳市腾讯计算机系统有限公司 | Browser webpage compatibility detection method and system |
| CN106407119A (en) * | 2016-09-28 | 2017-02-15 | 浪潮软件集团有限公司 | Browser compatibility testing method based on automatic testing |
| CN106681917A (en) * | 2016-12-21 | 2017-05-17 | 南京大学 | Method for automatically evaluating front ends on basis of neural networks |
| CN106886491A (en) * | 2017-01-17 | 2017-06-23 | 博彦科技股份有限公司 | Webpage compatibility test method and device |
| CN107729249A (en) * | 2017-10-18 | 2018-02-23 | 北京奇虎科技有限公司 | Browser compatibility method of testing, device, system and electronic equipment |
Non-Patent Citations (6)
| Title |
|---|
| ALI MESBAH ET.AL: "Automated cross-browser compatibility testing", 《2011 33RD INTERNATIONAL CONFERENCE ON SOFTWARE ENGINEERING (ICSE)》 * |
| ALI MESBAH ET.AL: "Automated cross-browser compatibility testing", 《2011 33RD INTERNATIONAL CONFERENCE ON SOFTWARE ENGINEERING (ICSE)》, 10 October 2011 (2011-10-10), pages 561 - 570 * |
| 吴国全 等: "基于捕获/重放的Web应用跨浏览器兼容性检测", 《计算机研究与发展》 * |
| 吴国全 等: "基于捕获/重放的Web应用跨浏览器兼容性检测", 《计算机研究与发展》, 15 March 2017 (2017-03-15), pages 623 - 632 * |
| 王欢欢等: "Web应用界面跨浏览器兼容性的自动检测方法", 《计算机科学》 * |
| 王欢欢等: "Web应用界面跨浏览器兼容性的自动检测方法", 《计算机科学》, 15 November 2015 (2015-11-15), pages 453 - 458 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110807007A (en) * | 2019-09-30 | 2020-02-18 | 支付宝(杭州)信息技术有限公司 | Target detection model training method, device and system and storage medium |
| CN110851349A (en) * | 2019-10-10 | 2020-02-28 | 重庆金融资产交易所有限责任公司 | Page abnormal display detection method, terminal equipment and storage medium |
| CN110851349B (en) * | 2019-10-10 | 2023-12-26 | 岳阳礼一科技股份有限公司 | Page abnormity display detection method, terminal equipment and storage medium |
| CN113506291A (en) * | 2021-07-29 | 2021-10-15 | 上海幻电信息科技有限公司 | Compatibility testing method and device |
| CN113506291B (en) * | 2021-07-29 | 2024-03-26 | 上海幻电信息科技有限公司 | Compatibility testing method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110275820B (en) | 2023-11-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Sun et al. | MEAN-SSD: A novel real-time detector for apple leaf diseases using improved light-weight convolutional neural networks | |
| CN110378420A (en) | A kind of image detecting method, device and computer readable storage medium | |
| CN110321477A (en) | Information recommendation method, device, terminal and storage medium | |
| CN110348441A (en) | VAT invoice recognition methods, device, computer equipment and storage medium | |
| CN109002766A (en) | A kind of expression recognition method and device | |
| US10169853B2 (en) | Score weights for user interface (UI) elements | |
| CN110009614A (en) | Method and apparatus for output information | |
| CN110378206B (en) | Intelligent image examination system and method | |
| CN110516671A (en) | Training method, image detecting method and the device of neural network model | |
| CN109815955A (en) | Topic householder method and system | |
| CN109345553A (en) | A kind of palm and its critical point detection method, apparatus and terminal device | |
| CN109978870A (en) | Method and apparatus for output information | |
| CN110275820A (en) | Page compatibility test method, system and equipment | |
| WO2020228283A1 (en) | Feature extraction method and apparatus, and computer readable storage medium | |
| CN108229418A (en) | Human body critical point detection method and apparatus, electronic equipment, storage medium and program | |
| CN109241030A (en) | Robot manipulating task data analytics server and robot manipulating task data analysing method | |
| CN113537496A (en) | Deep learning model visual construction system and application and design method thereof | |
| CN109033472A (en) | Picture retrieval method and device, computer equipment and computer-readable medium | |
| CN110580489A (en) | Data object classification system, method and equipment | |
| CN111368768A (en) | Human body key point-based employee gesture guidance detection method | |
| CN117132763A (en) | Power image anomaly detection method, device, computer equipment and storage medium | |
| US20210195095A1 (en) | Systems and methods for guiding image sensor angle settings in different environments | |
| CN106033431A (en) | A method and a server for product information processing | |
| CN110363248A (en) | The computer identification device and method of mobile crowdsourcing test report based on image | |
| Mashuri et al. | Smart Victims Detection in Natural Disaster using Deep Learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |