CN107992894A - Image-recognizing method, device and computer-readable recording medium - Google Patents
Image-recognizing method, device and computer-readable recording medium Download PDFInfo
- Publication number
- CN107992894A CN107992894A CN201711318139.0A CN201711318139A CN107992894A CN 107992894 A CN107992894 A CN 107992894A CN 201711318139 A CN201711318139 A CN 201711318139A CN 107992894 A CN107992894 A CN 107992894A
- Authority
- CN
- China
- Prior art keywords
- image
- resolution ratio
- default
- default resolution
- layer
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/29—Graphical models, e.g. Bayesian networks
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
The disclosure is directed to a kind of image-recognizing method, device and computer-readable recording medium, the described method includes：Obtain the resolution ratio of the first image to be identified；When the resolution ratio of the first image is more than the first default resolution ratio or presets resolution ratio less than second, according to the first image, by specifying the 3rd image of the convolutional layer included in multiple dimensioned layer or the generation of warp lamination with the 3rd default resolution ratio, first default resolution ratio is N times of the 3rd default resolution ratio, 3rd default resolution ratio is N times of the second default resolution ratio, and N is more than 1；By specifying grader that the 3rd image is identified.The feature of the pixel of the 3rd image is more accurate in the disclosure, it that is to say, 3rd image is smaller compared to the distortion factor of the first image, and the 3rd the resolution ratio of image be to specify grader accurately to carry out the resolution ratio of image recognition, thus by specifying grader that accurately the 3rd image can be identified, the first image is accurately identified so as to also be achieved that.
Description
Technical field
This disclosure relates to technical field of image processing, more particularly to a kind of image-recognizing method, device and computer-readable
Storage medium.
Background technology
With the development of image processing techniques, the disaggregated model such as depth convolutional neural networks model has become image
Indispensable instrument during identification.Before image is identified by disaggregated model, first the disaggregated model can be carried out
Training, the disaggregated model that training is completed are properly termed as grader, image can be then identified using the grader.
In correlation technique, when image being identified using grader, the image directly can be inputted into the grader, it
Afterwards, the convolutional layer that is included in the grader, active coating, pond layer, full articulamentum etc. can be handled the image, and this point
The class probability layer included in class device can export final image recognition result.
The content of the invention
To overcome problem present in correlation technique, the disclosure provides a kind of image-recognizing method, device and computer can
Read storage medium.
According to the first aspect of the embodiment of the present disclosure, there is provided a kind of image-recognizing method, the described method includes：
Obtain the resolution ratio of the first image to be identified；
When the resolution ratio of described first image is more than the first default resolution ratio or presets resolution ratio less than second, according to institute
The first image is stated, by specifying the convolutional layer that includes in multiple dimensioned layer or the generation of warp lamination the with the 3rd default resolution ratio
Three images, the described first default resolution ratio are N times of the described 3rd default resolution ratio, and the described 3rd to preset resolution ratio be described the
N times of two default resolution ratio, the N is more than 1；
By specifying grader that the 3rd image is identified.
Alternatively, it is described according to described first image, by specifying the convolutional layer or warp lamination that are included in multiple dimensioned layer
Threeth image of the generation with the 3rd default resolution ratio, including：
According to the resolution ratio of the first image, described first image is scaled with the described first default resolution ratio or described
Second image of the second default resolution ratio；
The convolutional layer included in multiple dimensioned layer or warp lamination is specified to generate the corresponding tool of second image by described
There is the 3rd image of the described 3rd default resolution ratio.
Alternatively, the resolution ratio according to the first image, described first image is scaled and is preset with described first
Second image of resolution ratio or the second default resolution ratio, including：
When the resolution ratio of described first image is more than the described first default resolution ratio, described first image is scaled tool
There is the second image of the described first default resolution ratio；
When the resolution ratio of described first image is less than the described second default resolution ratio, described first image is scaled tool
There is the second image of the described second default resolution ratio.
Alternatively, it is described to specify the convolutional layer included in multiple dimensioned layer or warp lamination to generate second figure by described
As corresponding 3rd image with the described 3rd default resolution ratio, including：
When the resolution ratio of second image presets resolution ratio for described first, wrapped by described specify in multiple dimensioned layer
The convolutional layer contained generates corresponding 3rd image with the described 3rd default resolution ratio of second image；
When the resolution ratio of second image presets resolution ratio for described second, wrapped by described specify in multiple dimensioned layer
The warp lamination contained generates corresponding 3rd image with the described 3rd default resolution ratio of second image.
Alternatively, after the resolution ratio for obtaining the first image to be identified, further include：
When the resolution ratio of described first image is less than or equal to the described first default resolution ratio and more than or equal to described the
During two default resolution ratio, described first image is scaled the 3rd image with the described 3rd default resolution ratio.
Alternatively, it is described according to described first image, by specifying the convolutional layer or warp lamination that are included in multiple dimensioned layer
Before the 3rd image of the generation with the 3rd default resolution ratio, further include：
Obtain multiple pre-set image collection, it is all default to concentrate that each pre-set image collection includes for the multiple pre-set image
Image belongs to same category；
Trained multiple dimensioned layer and disaggregated model are treated using the multiple pre-set image collection to be trained, and obtain the finger
Fixed multiple dimensioned layer and the specified grader.
According to the second aspect of the embodiment of the present disclosure, there is provided a kind of pattern recognition device, described device include：
First acquisition module, for obtaining the resolution ratio of the first image to be identified；
Generation module, for being more than the first default resolution ratio or less than second default point when the resolution ratio of described first image
During resolution, according to described first image, by specifying the convolutional layer included in multiple dimensioned layer or the generation of warp lamination to have the 3rd
3rd image of default resolution ratio, the described first default resolution ratio are N times of the described 3rd default resolution ratio, and the described 3rd is default
Resolution ratio is N times of the described second default resolution ratio, and the N is more than 1；
Identification module, for by specifying grader that the 3rd image is identified.
Alternatively, the generation module includes：
Submodule is scaled, for the resolution ratio according to the first image, described first image is scaled with described first
Second image of default resolution ratio or the second default resolution ratio；
Submodule is generated, for specifying in multiple dimensioned layer the convolutional layer that includes or warp lamination to generate described the by described
Corresponding 3rd image with the described 3rd default resolution ratio of two images.
Alternatively, the scaling submodule is additionally operable to：
When the resolution ratio of described first image is more than the described first default resolution ratio, described first image is scaled tool
There is the second image of the described first default resolution ratio；
When the resolution ratio of described first image is less than the described second default resolution ratio, described first image is scaled tool
There is the second image of the described second default resolution ratio.
Alternatively, the generation submodule is additionally operable to：
When the resolution ratio of second image presets resolution ratio for described first, wrapped by described specify in multiple dimensioned layer
The convolutional layer contained generates corresponding 3rd image with the described 3rd default resolution ratio of second image；
When the resolution ratio of second image presets resolution ratio for described second, wrapped by described specify in multiple dimensioned layer
The warp lamination contained generates corresponding 3rd image with the described 3rd default resolution ratio of second image.
Alternatively, described device further includes：
Zoom module, for when the resolution ratio of described first image is less than or equal to the described first default resolution ratio and is more than
Or during equal to the described second default resolution ratio, described first image is scaled the 3rd figure with the described 3rd default resolution ratio
Picture.
Alternatively, described device further includes：
Second acquisition module, for obtaining multiple pre-set image collection, the multiple pre-set image concentrates each pre-set image
All pre-set images that collecting includes belong to same category；
Training module, is instructed for treating trained multiple dimensioned layer and disaggregated model using the multiple pre-set image collection
Practice, obtain the specified multiple dimensioned layer and the specified grader.
According to the third aspect of the embodiment of the present disclosure, there is provided a kind of pattern recognition device, described device include：
Processor；
For storing the memory of processor-executable instruction；
Wherein, the step of processor is configured as performing above-mentioned first aspect the method.
According to the fourth aspect of the embodiment of the present disclosure, there is provided a kind of computer-readable recording medium, it is described computer-readable
The step of instruction is stored with storage medium, above-mentioned first aspect the method is realized when described instruction is executed by processor.
The technical scheme provided by this disclosed embodiment can include the following benefits：
In the embodiments of the present disclosure, the resolution ratio of the first image to be identified can be obtained, then when point of the first image
When resolution is more than the first default resolution ratio or resolution ratio is preset less than second, i.e., when the resolution ratio of the first image and the 3rd default point
, can be according to the first image, by specifying the convolutional layer included in multiple dimensioned layer or warp lamination to generate when resolution difference is larger
The 3rd image with the 3rd default resolution ratio, is identified the 3rd image finally by specified grader.Due to the 3rd figure
Seem to generate to obtain by convolution mode or deconvolution mode, so the feature of the pixel in the 3rd image is more accurate,
It is that the 3rd image is smaller compared to the distortion factor of the first image, and since the resolution ratio of the 3rd image is to specify grader can
Accurately to carry out the resolution ratio of image recognition, therefore by specifying grader that accurately the 3rd image can be identified, so that
Also the identification to the first image is just accurately realized, the accuracy of image recognition is higher.
It should be appreciated that the general description and following detailed description of the above are only exemplary and explanatory, not
The disclosure can be limited.
Brief description of the drawings
Attached drawing herein is merged in specification and forms the part of this specification, shows the implementation for meeting the present invention
Example, and for explaining the principle of the present invention together with specification.
Fig. 1 is a kind of flow chart of image-recognizing method according to an exemplary embodiment.
Fig. 2A is the flow chart of another image-recognizing method according to an exemplary embodiment.
Fig. 2 B are a kind of schematic diagrames of the network structure of image recognition according to an exemplary embodiment.
Fig. 3 A are the block diagrams of the first pattern recognition device according to an exemplary embodiment.
Fig. 3 B are a kind of block diagrams of generation module according to an exemplary embodiment.
Fig. 3 C are the block diagrams of second of pattern recognition device according to an exemplary embodiment.
Fig. 3 D are the block diagrams of the third pattern recognition device according to an exemplary embodiment.
Fig. 4 is the block diagram of the 4th kind of pattern recognition device according to an exemplary embodiment.
Embodiment
Here exemplary embodiment will be illustrated in detail, its example is illustrated in the accompanying drawings.Following description is related to
During attached drawing, unless otherwise indicated, the same numbers in different attached drawings represent the same or similar key element.Following exemplary embodiment
Described in embodiment do not represent and the consistent all embodiments of the present invention.On the contrary, they be only with it is such as appended
The example of the consistent apparatus and method of some aspects being described in detail in claims, of the invention.
In order to make it easy to understand, before detailed explanation is carried out to the embodiment of the present disclosure, first to the embodiment of the present disclosure
The application scenarios being related to are introduced.
With the development of image processing techniques, the disaggregated model such as depth convolutional neural networks model has become image
Indispensable instrument during identification.Before image is identified by disaggregated model, first the disaggregated model can be carried out
Training, the disaggregated model that training is completed are properly termed as grader, image can be then identified using the grader.It is related
In technology, when image being identified using grader, the image directly can be inputted into the grader, then the grader can
To export final image recognition result.It is that first extraction should from the pixel of the image since grader is when identifying image
The feature of image, the image is identified further according to the feature of the image, so grader is typically only capable to recognition resolution certain
Resolving range in image, thus if the image directly is inputted the grader, when the resolution ratio of the image with should
When the resolving range difference that grader can identify is larger, image recognition result inaccuracy may result in.For this reason, the disclosure carries
For a kind of image-recognizing method, by the adjustment of the resolution ratio to image, to improve the accuracy of image recognition.
The image-recognizing method that the disclosure provides can be applied to image recognition scene.For example, may in daily life
There are many different classes of trademark images, in order to check situation of these trademark images with the presence or absence of infringement, it will usually pass through
These trademark images are identified in grader, and the resolution ratio of different trademark image may differ bigger, at this time can be with
The image-recognizing method provided using the disclosure, to ensure to accurately identify trademark image.
Next the image-recognizing method that the embodiment of the present disclosure provides will be described in detail with reference to attached drawing.
Fig. 1 is a kind of flow chart of image-recognizing method according to an exemplary embodiment, as shown in Figure 1, the party
Method comprises the following steps.
In a step 101, the resolution ratio of the first image to be identified is obtained.
In a step 102, when the resolution ratio of the first image is more than the first default resolution ratio or less than the second default resolution ratio
When, according to the first image, by specifying the convolutional layer included in multiple dimensioned layer or the generation of warp lamination that there is the 3rd default resolution
3rd image of rate, the first default resolution ratio are N times of the 3rd default resolution ratio, and the 3rd default resolution ratio is the second default resolution
N times of rate, N are more than 1.
In step 103, by specifying grader that the 3rd image is identified.
In the embodiments of the present disclosure, the resolution ratio of the first image to be identified can be obtained, then when point of the first image
When resolution is more than the first default resolution ratio or resolution ratio is preset less than second, i.e., when the resolution ratio of the first image and the 3rd default point
, can be according to the first image, by specifying the convolutional layer included in multiple dimensioned layer or warp lamination to generate when resolution difference is larger
The 3rd image with the 3rd default resolution ratio, is identified the 3rd image finally by specified grader.Due to the 3rd figure
Seem to generate to obtain by convolution mode or deconvolution mode, so the feature of the pixel in the 3rd image is more accurate,
It is that the 3rd image is smaller compared to the distortion factor of the first image, and since the resolution ratio of the 3rd image is to specify grader can
Accurately to carry out the resolution ratio of image recognition, therefore by specifying grader that accurately the 3rd image can be identified, so that
Also the identification to the first image is just accurately realized, the accuracy of image recognition is higher.
Alternatively, according to the first image, by specifying the convolutional layer included in multiple dimensioned layer or the generation of warp lamination to have
3rd image of the 3rd default resolution ratio, including：
It is with the first default resolution ratio or the second default resolution by the first image scaling according to the resolution ratio of the first image
Second image of rate；
By specify in multiple dimensioned layer the convolutional layer that includes or warp lamination generate the second image it is corresponding have it is the 3rd pre-
If the 3rd image of resolution ratio.
Alternatively, it is with the first default resolution ratio or second by the first image scaling according to the resolution ratio of the first image
Second image of default resolution ratio, including：
It is with first default point by the first image scaling when the resolution ratio of the first image is more than the first default resolution ratio
Second image of resolution；
It is with second default point by the first image scaling when the resolution ratio of the first image is less than the second default resolution ratio
Second image of resolution.
Alternatively, the second image is corresponding by specifying in multiple dimensioned layer the convolutional layer that includes or the generation of warp lamination has
3rd image of the 3rd default resolution ratio, including：
When the resolution ratio of the second image presets resolution ratio for first, pass through and specify the convolutional layer included in multiple dimensioned layer to give birth to
Into corresponding 3rd image with the 3rd default resolution ratio of the second image；
When the resolution ratio of the second image presets resolution ratio for second, pass through the warp lamination specified and included in multiple dimensioned layer
Generate corresponding 3rd image with the 3rd default resolution ratio of the second image.
Alternatively, after the resolution ratio for obtaining the first image to be identified, further include：
When the resolution ratio of the first image is less than or equal to the first default resolution ratio and more than or equal to the second default resolution ratio
When, it is the 3rd image with the 3rd default resolution ratio by the first image scaling.
Alternatively, according to the first image, by specifying the convolutional layer included in multiple dimensioned layer or the generation of warp lamination to have
Before 3rd image of the 3rd default resolution ratio, further include：
Multiple pre-set image collection are obtained, the plurality of pre-set image concentrates all default figures that each pre-set image collection includes
As belonging to same category；
Trained multiple dimensioned layer and disaggregated model are treated using the plurality of pre-set image collection to be trained, and obtain specifying more rulers
Spend layer and specified grader.
Above-mentioned all optional technical solutions, can form the alternative embodiment of the disclosure according to any combination, and the disclosure is real
Example is applied no longer to repeat this one by one.
Fig. 2A is a kind of flow chart of image-recognizing method according to an exemplary embodiment, below in conjunction with Fig. 2A
The image-recognizing method provided Fig. 1 embodiments carries out expansion explanation.As shown in Figure 2 A, this method comprises the following steps.
In step 201, the resolution ratio of the first image to be identified is obtained.
It should be noted that the resolution ratio of image refers to how many pixel in image per inch, the resolution ratio of image
It can determine the measure of precision of image detail with the size of representative image, size etc., the resolution ratio of image, that is to say, image
Resolution ratio is higher, and the pixel that image is included is more, and image is more clear, and the resolution ratio of image is lower, and image is included
Pixel it is fewer, image is fuzzyyer.
In step 202, when the resolution ratio of the first image is more than the first default resolution ratio or less than the second default resolution ratio
When, according to the first image, by specifying the convolutional layer included in multiple dimensioned layer or the generation of warp lamination that there is the 3rd default resolution
3rd image of rate.
It should be noted that the first default resolution ratio, the second default resolution ratio and the 3rd default resolution ratio can bases
Different demands are configured in advance, and the first default resolution ratio is N times of the 3rd default resolution ratio, and the 3rd to preset resolution ratio be the
N times of two default resolution ratio, N is more than 1.For example, N is 4, it is assumed that the first default resolution ratio is 448*448, then the 3rd presets at this time
Resolution ratio is just 224*224, and the second default resolution ratio is just 112*112.
In addition, specify multiple dimensioned layer to be configured in advance, and it is to be used to contract to image progress for N times to specify multiple dimensioned layer
The layer put, that is, specify multiple dimensioned layer can by it includes convolutional layer by N times of image down, or by it includes warp
Lamination is by N times of image magnification.
Furthermore convolutional layer is opposite to the processing procedure of image with warp lamination, and convolutional layer can divide the image into more
A region, is then a feature by the feature extraction of multiple pixels in each region, that is to say, can be with by convolutional layer
The number for the pixel that image includes is reduced, the resolution ratio of image is reduced, realizes the diminution to image；Warp lamination can incite somebody to action
The feature of each pixel in image is reduced to multiple features, that is to say, can increase what image included by convolutional layer
The number of pixel, increases the resolution ratio of image, realizes the amplification to image.
Wherein, the process of realizing of step 202 can be：It is tool by the first image scaling according to the resolution ratio of the first image
There is the second image of the first default resolution ratio or the second default resolution ratio, then by specifying the convolutional layer included in multiple dimensioned layer
Or warp lamination generates corresponding 3rd image with the 3rd default resolution ratio of the second image.
It should be noted that after the second image is inputted specified multiple dimensioned layer, the image for specifying multiple dimensioned layer output is
The corresponding image of second image, and due to specifying the convolutional layer included in multiple dimensioned layer or warp lamination can be by image scaling N
Times, thus the image for specifying multiple dimensioned layer to export at this time is the 3rd image with the 3rd default resolution ratio.
Wherein, it is pre- with the first default resolution ratio or second by the first image scaling according to the resolution ratio of the first image
If the process of realizing of the second image of resolution ratio can be：, will when the resolution ratio of the first image is more than the first default resolution ratio
First image scaling is the second image with the first default resolution ratio；When the resolution ratio of the first image is less than the second default resolution
It is the second image with the second default resolution ratio by the first image scaling during rate.
For example, the resolution ratio of the first image is 552*552, it is assumed that the first default resolution ratio is 448*448, due to 552*
552 are more than 448*448, therefore can be the second image that resolution ratio is 448*448 by the first image scaling.
In another example the resolution ratio of the first image is 56*56, it is assumed that the second default resolution ratio is 112*112, due to 56*56
Less than 112*112, therefore can be the second image that resolution ratio is 112*112 by the first image scaling.
Wherein, by specifying in multiple dimensioned layer the convolutional layer that includes or warp lamination to generate, the second image is corresponding to have the
The processes of realizing of 3rd image of three default resolution ratio can be：When the resolution ratio of the second image presets resolution ratio for first,
The convolutional layer included in multiple dimensioned layer is specified to generate corresponding 3rd figure with the 3rd default resolution ratio of the second image by this
Picture；When the resolution ratio of the second image presets resolution ratio for second, the warp lamination included in multiple dimensioned layer is specified to give birth to by this
Into corresponding 3rd image with the 3rd default resolution ratio of the second image.
It should be noted that since the 3rd default resolution ratio is less than the first default resolution ratio, when point of the second image
Resolution for the first default resolution ratio when, can by specifying the convolutional layer included in multiple dimensioned layer by N times of the second image down,
To generate corresponding 3rd image with the 3rd default resolution ratio of the second image.And since the second default resolution ratio is less than the 3rd
Default resolution ratio, therefore when the resolution ratio of the second image presets resolution ratio for second, can be wrapped by specifying in multiple dimensioned layer
The warp lamination contained is by N times of the second image magnification, and to generate, the second image is corresponding to have the 3rd of the 3rd default resolution ratio
Image.
It should be noted that convolutional layer and warp lamination are usually provided with convolution kernel and step when realizing the processing to image
Long, wherein when step-length is equal with the length of side of convolution kernel, the area of convolution kernel is exactly the scaling multiple to image.Therefore, into one
Step ground, by specifying the 3rd figure of the convolutional layer included in multiple dimensioned layer or the generation of warp lamination with the 3rd default resolution ratio
Before picture, the convolution kernel in the convolutional layer and warp lamination that are included in multiple dimensioned layer can also will be specified to be disposed as N*N, step-length
It is disposed as N.
For example, N is 4, then the convolution kernel for specifying the convolutional layer included in multiple dimensioned layer and warp lamination can be respectively provided with
For 2*2, step-length is disposed as 2, and the convolutional layer can be 4 times by image down at this time, which can be by image magnification 4
Times.
What deserves to be explained is according to the first image in the embodiment of the present disclosure, by specifying the convolution included in multiple dimensioned layer
The 3rd image of layer or the generation of warp lamination with the 3rd default resolution ratio, is in order to subsequently through the realization pair of specified grader
Accurately identifying for 3rd image, that is to say, the 3rd default resolution ratio specifies grader to be accurately identified to image
Resolution ratio.
In practical application, if differing larger between the resolution ratio of images to be recognized and the 3rd default resolution ratio, directly
Images to be recognized is zoomed into the image with the 3rd default resolution ratio, it is more likely that because of excessively scaling image can be caused to lose
Very, accuracy during follow-up image recognition is then influenced.Thus in the embodiments of the present disclosure, can be first by the first image scaling
The second image being closer to the resolution ratio with the first image, avoids image fault, and then by specifying in multiple dimensioned layer
Comprising convolutional layer or warp lamination generate corresponding 3rd image of the second image, due to being to use convolution or deconvolution at this time
Second image scaling is the 3rd image by mode, and the feature of the pixel in the 3rd image because obtained from is more accurate, so that
It is possible to prevente effectively from image fault, and then the first image can be accurately identified subsequently being realized according to the 3rd image.
In step 203, when the resolution ratio of the first image is less than or equal to the first default resolution ratio and more than or equal to the
It is the 3rd image with the 3rd default resolution ratio by the first image scaling during two default resolution ratio.
For example, the resolution ratio of the first image is 336*336, it is assumed that the first default resolution ratio be 448*448, and second default divides
Resolution is 112*112, and the 3rd default resolution ratio is 224*224, since 336*336 is less than 448*448 and is more than 112*112, because
First image scaling can be the 3rd image that resolution ratio is 224*224 by this.
What deserves to be explained is due to the 3rd default resolution bits in the first default resolution ratio and the second default resolution ratio it
Between, therefore when the resolution bits of the first image are between the first default resolution ratio and the second default resolution ratio, the first image
Differ smaller between resolution ratio and the 3rd default resolution ratio, thus can be at this time pre- with the 3rd directly by the first image scaling
If the 3rd image of resolution ratio, so as to not only avoid image fault, and resource can be saved to avoid unnecessary operation.
In step 204, by specifying grader that the 3rd image is identified.
It should be noted that specifying grader to be configured in advance, and grader is specified to be used to know image
Not, image recognition result is obtained.And due in the embodiment of the present disclosure the 3rd image contract indeed through to the first image
Put to obtain, thus the image recognition result obtained after the 3rd image is identified by specified grader is the first image
Image recognition result.
In addition, in the embodiment of the present disclosure, since the 3rd image is smaller compared to the distortion factor of the first image, and the 3rd is default
Resolution ratio is to specify grader accurately to carry out the resolution ratio of image recognition, thus by specify grader to the 3rd image into
Row identification, you can accurately realize the identification to the first image, the accuracy of image recognition is higher.
It is illustrated with reference to Fig. 2 B image-recognizing methods provided the embodiment of the present disclosure.
Referring to Fig. 2 B, it is assumed that the first default resolution ratio is 448*448, and the second default resolution ratio is 112*112, and the 3rd is default
Resolution ratio is 224*224.When the resolution ratio of the first image is more than 448*448, by the first image scaling into resolution ratio be 448*
448 the second image, then by specifying the 3rd figure that the convolutional layer included in multiple dimensioned layer generation resolution ratio is 224*224
Picture；When the resolution ratio of the first image is greater than or equal to 112*112 and is less than or equal to 448*448, the first image is directly contracted
Put into the 3rd image that resolution ratio is 224*224；When the resolution ratio of the first image is less than 112*112, by the first image scaling
Into the second image that resolution ratio is 112*112, the warp lamination then included by specifying in multiple dimensioned layer generates resolution ratio and is
The 3rd image of 224*224.Afterwards, the 3rd image is inputted and specifies grader, by specifying grader to export the 3rd image
Image recognition result, the image recognition result of the 3rd image is the image recognition result of the first image.
Further, according to the first image, generated by the convolutional layer or warp lamination that are included in specified multiple dimensioned layer
Before the 3rd image with the 3rd default resolution ratio, specified multiple dimensioned layer and specified grader can also be generated, wherein, generation
When specifying multiple dimensioned layer and specified grader, multiple pre-set image collection can be obtained, instruction is treated using the plurality of pre-set image collection
Experienced multiple dimensioned layer and disaggregated model is trained, and obtains specifying multiple dimensioned layer and specified grader.
It should be noted that the plurality of pre-set image collection can be configured in advance, and the plurality of pre-set image is concentrated often
All pre-set images that a pre-set image collection includes belong to same category, that is to say, the plurality of pre-set image collection includes pre-
If image is the image with classification logotype, and all pre-set images that each pre-set image collection includes have identical class
Do not identify.
In addition, in order to enable specified multiple dimensioned layer and specified grader that training obtains possess more preferable robustness, this is more
A pre-set image, which is concentrated, to be included all pre-set images and may each be to have passed through in advance after cutting, upset etc. operate to obtain, this public affairs
Embodiment is opened to be not construed as limiting this.
Wherein, trained multiple dimensioned layer and disaggregated model are treated using the plurality of pre-set image collection to be trained, is referred to
Determining the process of realizing of multiple dimensioned layer and specified grader can be：Selected in the multiple images included from the plurality of pre-set image collection
Go out a pre-set image, following processing is performed to the pre-set image selected, until having handled the plurality of pre-set image collection includes
Each pre-set image untill：A pre-set image, and the figure that the multiple dimensioned layer is exported are inputted to multiple dimensioned layer to be trained
As being input in disaggregated model to be trained, by the disaggregated model output image classification, then by presetting loss function meter
Calculate the penalty values of the pre-set image, and each layer that the disaggregated model includes is returned into the penalty values anti-pass and this is multiple dimensioned
Layer, and the penalty values are substituted into the local derviation letter of the parameters in each layer and the multiple dimensioned layer that the disaggregated model includes
Number, you can determine the specific error amount of the local derviation value of parameters, i.e. parameters, the local derviation value based on parameters is to each
A parameter is updated, and completes the once adjustment to parameters.In this way, constantly inputting pre-set image, repeat the above,
The parameters of the multiple dimensioned layer and the disaggregated model will constantly be learnt, and can be achieved after repeatedly renewal will be each
Parameter adjustment is target component, so as to complete to train, obtains this and specifies multiple dimensioned layer and specified grader.
It should be noted that the default loss function can be configured in advance according to different demands, the embodiment of the present disclosure
This is not construed as limiting.
In addition, in the embodiment of the present disclosure, multiple dimensioned layer and disaggregated model to be trained are to be carried out at the same time training, to be referred to
Fixed multiple dimensioned layer and specified grader, when can so cause the image for specifying grader identification to specify multiple dimensioned layer output
Accuracy higher.
Certainly, in practical applications, using the plurality of pre-set image collection treat trained multiple dimensioned layer and disaggregated model into
Row training, when obtaining specifying multiple dimensioned layer and specified grader, in addition to above-mentioned training method, it is also possible to which there are other training sides
Formula, for example, end2end (end-to-end) training method, from lower rising unsupervised learning method, top-down supervised learning method
Deng the disclosure is without limitation.
In the embodiments of the present disclosure, the resolution ratio of the first image to be identified can be obtained, then when point of the first image
When resolution is more than the first default resolution ratio or resolution ratio is preset less than second, i.e., when the resolution ratio of the first image and the 3rd default point
, can be according to the first image, by specifying the convolutional layer included in multiple dimensioned layer or warp lamination to generate when resolution difference is larger
The 3rd image with the 3rd default resolution ratio, is identified the 3rd image finally by specified grader.Due to the 3rd figure
Seem to generate to obtain by convolution mode or deconvolution mode, so the feature of the pixel in the 3rd image is more accurate,
It is that the 3rd image is smaller compared to the distortion factor of the first image, and since the resolution ratio of the 3rd image is to specify grader can
Accurately to carry out the resolution ratio of image recognition, therefore by specifying grader that accurately the 3rd image can be identified, so that
Also the identification to the first image is just accurately realized, the accuracy of image recognition is higher.
Fig. 3 A are a kind of block diagrams of pattern recognition device according to an exemplary embodiment.Referring to Fig. 3 A, the device
Including the first acquisition module 301, generation module 302 and identification module 303.
First acquisition module 301, for obtaining the resolution ratio of the first image to be identified.
Generation module 302, for being more than the first default resolution ratio or less than second default point when the resolution ratio of the first image
During resolution, according to the first image, by specifying the convolutional layer included in multiple dimensioned layer or the generation of warp lamination to have the 3rd to preset
3rd image of resolution ratio, the first default resolution ratio are N times of the 3rd default resolution ratio, and the 3rd default resolution ratio is second default
N times of resolution ratio, N are more than 1.
Identification module 303, for by specifying grader that the 3rd image is identified.
Alternatively, include referring to Fig. 3 B, generation module 302：
Submodule 3021 is scaled, is to be preset with first by the first image scaling for the resolution ratio according to the first image
Second image of resolution ratio or the second default resolution ratio.
Submodule 3022 is generated, for by specifying the convolutional layer included in multiple dimensioned layer or warp lamination to generate the second figure
As corresponding 3rd image with the 3rd default resolution ratio.
Alternatively, which is additionally operable to：
It is with first default point by the first image scaling when the resolution ratio of the first image is more than the first default resolution ratio
Second image of resolution；
It is with second default point by the first image scaling when the resolution ratio of the first image is less than the second default resolution ratio
Second image of resolution.
Alternatively, which is additionally operable to：
When the resolution ratio of the second image presets resolution ratio for first, pass through and specify the convolutional layer included in multiple dimensioned layer to give birth to
Into corresponding 3rd image with the 3rd default resolution ratio of the second image；
When the resolution ratio of the second image presets resolution ratio for second, pass through the warp lamination specified and included in multiple dimensioned layer
Generate corresponding 3rd image with the 3rd default resolution ratio of the second image.
Alternatively, further included referring to Fig. 3 C, the device：
Zoom module 304, the resolution ratio for working as the first image are less than or equal to the first default resolution ratio and are more than or wait
It is the 3rd image with the 3rd default resolution ratio by the first image scaling when the second default resolution ratio.
Alternatively, further included referring to Fig. 3 D, the device：
Second acquisition module 305, for obtaining multiple pre-set image collection, the plurality of pre-set image concentrates each pre-set image
All pre-set images that collecting includes belong to same category.
Training module 306, carries out for treating trained multiple dimensioned layer and disaggregated model using the plurality of pre-set image collection
Training, obtains specifying multiple dimensioned layer and specified grader.
In the embodiments of the present disclosure, the resolution ratio of the first image to be identified can be obtained, then when point of the first image
When resolution is more than the first default resolution ratio or resolution ratio is preset less than second, i.e., when the resolution ratio of the first image and the 3rd default point
, can be according to the first image, by specifying the convolutional layer included in multiple dimensioned layer or warp lamination to generate when resolution difference is larger
The 3rd image with the 3rd default resolution ratio, is identified the 3rd image finally by specified grader.Due to the 3rd figure
Seem to generate to obtain by convolution mode or deconvolution mode, so the feature of the pixel in the 3rd image is more accurate,
It is that the 3rd image is smaller compared to the distortion factor of the first image, and since the resolution ratio of the 3rd image is to specify grader can
Accurately to carry out the resolution ratio of image recognition, therefore by specifying grader that accurately the 3rd image can be identified, so that
Also the identification to the first image is just accurately realized, the accuracy of image recognition is higher.
On the device in above-described embodiment, wherein modules perform the concrete mode of operation in related this method
Embodiment in be described in detail, explanation will be not set forth in detail herein.
Fig. 4 is a kind of block diagram of pattern recognition device 400 according to an exemplary embodiment.For example, device 400 can
To be mobile phone, computer, digital broadcast terminal, messaging devices, game console, tablet device, Medical Devices, are good for
Body equipment, personal digital assistant etc..
With reference to Fig. 4, device 400 can include following one or more assemblies：Processing component 402, memory 404, power supply
Component 406, multimedia component 408, audio component 410, the interface 412 of input/output (I/O), sensor component 414, and
Communication component 416.
The integrated operation of the usual control device 400 of processing component 402, such as with display, call, data communication, phase
The operation that machine operates and record operation is associated.Processing component 402 can refer to including one or more processors 420 to perform
Order, to complete all or part of step of above-mentioned method.In addition, processing component 402 can include one or more modules, just
Interaction between processing component 402 and other assemblies.For example, processing component 402 can include multi-media module, it is more to facilitate
Interaction between media component 408 and processing component 402.
Memory 404 is configured as storing various types of data to support the operation in device 400.These data are shown
Example includes the instruction of any application program or method for being operated on device 400, and contact data, telephone book data, disappears
Breath, picture, video etc..Memory 404 can be by any kind of volatibility or non-volatile memory device or their group
Close and realize, as static RAM (SRAM), electrically erasable programmable read-only memory (EEPROM) are erasable to compile
Journey read-only storage (EPROM), programmable read only memory (PROM), read-only storage (ROM), magnetic memory, flash
Device, disk or CD.
Power supply module 406 provides power supply for the various assemblies of device 400.Power supply module 406 can include power management system
System, one or more power supplys, and other components associated with generating, managing and distributing power supply for device 400.
Multimedia component 408 is included in the screen of one output interface of offer between described device 400 and user.One
In a little embodiments, screen can include liquid crystal display (LCD) and touch panel (TP).If screen includes touch panel, screen
Curtain may be implemented as touch-screen, to receive input signal from the user.Touch panel includes one or more touch sensings
Device is to sense the gesture on touch, slip and touch panel.The touch sensor can not only sense touch or sliding action
Border, but also detect and the duration and pressure associated with the touch or slide operation.In certain embodiments, more matchmakers
Body component 408 includes a front camera and/or rear camera.When device 400 is in operator scheme, such as screening-mode or
During video mode, front camera and/or rear camera can receive exterior multi-medium data.Each front camera and
Rear camera can be a fixed optical lens system or have focusing and optical zoom capabilities.
Audio component 410 is configured as output and/or input audio signal.For example, audio component 410 includes a Mike
Wind (MIC), when device 400 is in operator scheme, during such as call model, logging mode and speech recognition mode, microphone by with
It is set to reception external audio signal.The received audio signal can be further stored in memory 404 or via communication set
Part 416 is sent.In certain embodiments, audio component 410 further includes a loudspeaker, for exports audio signal.
I/O interfaces 412 provide interface between processing component 402 and peripheral interface module, and above-mentioned peripheral interface module can
To be keyboard, click wheel, button etc..These buttons may include but be not limited to：Home button, volume button, start button and lock
Determine button.
Sensor component 414 includes one or more sensors, and the state for providing various aspects for device 400 is commented
Estimate.For example, sensor component 414 can detect opening/closed mode of device 400, and the relative positioning of component, for example, it is described
Component is the display and keypad of device 400, and sensor component 414 can be with 400 1 components of detection device 400 or device
Position change, the existence or non-existence that user contacts with device 400,400 orientation of device or acceleration/deceleration and device 400
Temperature change.Sensor component 414 can include proximity sensor, be configured to detect without any physical contact
Presence of nearby objects.Sensor component 414 can also include optical sensor, such as CMOS or ccd image sensor, for into
As being used in application.In certain embodiments, which can also include acceleration transducer, gyro sensors
Device, Magnetic Sensor, pressure sensor or temperature sensor.
Communication component 416 is configured to facilitate the communication of wired or wireless way between device 400 and other equipment.Device
400 can access the wireless network based on communication standard, such as WiFi, 2G or 3G, or combinations thereof.In an exemplary implementation
In example, communication component 416 receives broadcast singal or broadcast related information from external broadcasting management system via broadcast channel.
In one exemplary embodiment, the communication component 416 further includes near-field communication (NFC) module, to promote junction service.Example
Such as, in NFC module radio frequency identification (RFID) technology can be based on, Infrared Data Association (IrDA) technology, ultra wide band (UWB) technology,
Bluetooth (BT) technology and other technologies are realized.
In the exemplary embodiment, device 400 can be believed by one or more application application-specific integrated circuit (ASIC), numeral
Number processor (DSP), digital signal processing appts (DSPD), programmable logic device (PLD), field programmable gate array
(FPGA), controller, microcontroller, microprocessor or other electronic components are realized, real shown in above-mentioned Fig. 1 or Fig. 2A for performing
The method that example offer is provided.
In the exemplary embodiment, a kind of non-transitorycomputer readable storage medium including instructing, example are additionally provided
Such as include the memory 404 of instruction, above-metioned instruction can be performed to complete the above method by the processor 420 of device 400.For example,
The non-transitorycomputer readable storage medium can be ROM, random access memory (RAM), CD-ROM, tape, floppy disk
With optical data storage devices etc..
A kind of non-transitorycomputer readable storage medium, when the instruction in the storage medium is by the processing of mobile terminal
When device performs so that mobile terminal is able to carry out a kind of image-recognizing method, the described method includes：
Obtain the resolution ratio of the first image to be identified；
When the resolution ratio of the first image is more than the first default resolution ratio or presets resolution ratio less than second, according to the first figure
Picture, by specifying the 3rd image of the convolutional layer included in multiple dimensioned layer or the generation of warp lamination with the 3rd default resolution ratio,
First default resolution ratio is N times of the 3rd default resolution ratio, and the 3rd default resolution ratio is N times of the second default resolution ratio, and N is more than
1；
By specifying grader that the 3rd image is identified.
Alternatively, according to the first image, by specifying the convolutional layer included in multiple dimensioned layer or the generation of warp lamination to have
3rd image of the 3rd default resolution ratio, including：
It is with the first default resolution ratio or the second default resolution by the first image scaling according to the resolution ratio of the first image
Second image of rate；
By specify in multiple dimensioned layer the convolutional layer that includes or warp lamination generate the second image it is corresponding have it is the 3rd pre-
If the 3rd image of resolution ratio.
Alternatively, it is with the first default resolution ratio or second by the first image scaling according to the resolution ratio of the first image
Second image of default resolution ratio, including：
It is with first default point by the first image scaling when the resolution ratio of the first image is more than the first default resolution ratio
Second image of resolution；
It is with second default point by the first image scaling when the resolution ratio of the first image is less than the second default resolution ratio
Second image of resolution.
Alternatively, the second image is corresponding by specifying in multiple dimensioned layer the convolutional layer that includes or the generation of warp lamination has
3rd image of the 3rd default resolution ratio, including：
When the resolution ratio of the second image presets resolution ratio for first, pass through and specify the convolutional layer included in multiple dimensioned layer to give birth to
Into corresponding 3rd image with the 3rd default resolution ratio of the second image；
When the resolution ratio of the second image presets resolution ratio for second, pass through the warp lamination specified and included in multiple dimensioned layer
Generate corresponding 3rd image with the 3rd default resolution ratio of the second image.
Alternatively, after the resolution ratio for obtaining the first image to be identified, further include：
When the resolution ratio of the first image is less than or equal to the first default resolution ratio and more than or equal to the second default resolution ratio
When, it is the 3rd image with the 3rd default resolution ratio by the first image scaling.
Alternatively, according to the first image, by specifying the convolutional layer included in multiple dimensioned layer or the generation of warp lamination to have
Before 3rd image of the 3rd default resolution ratio, further include：
Multiple pre-set image collection are obtained, the plurality of pre-set image concentrates all default figures that each pre-set image collection includes
As belonging to same category；
Trained multiple dimensioned layer and disaggregated model are treated using the plurality of pre-set image collection to be trained, and obtain specifying more rulers
Spend layer and specified grader.
In the embodiments of the present disclosure, the resolution ratio of the first image to be identified can be obtained, then when point of the first image
When resolution is more than the first default resolution ratio or resolution ratio is preset less than second, i.e., when the resolution ratio of the first image and the 3rd default point
, can be according to the first image, by specifying the convolutional layer included in multiple dimensioned layer or warp lamination to generate when resolution difference is larger
The 3rd image with the 3rd default resolution ratio, is identified the 3rd image finally by specified grader.Due to the 3rd figure
Seem to generate to obtain by convolution mode or deconvolution mode, so the feature of the pixel in the 3rd image is more accurate,
It is that the 3rd image is smaller compared to the distortion factor of the first image, and since the resolution ratio of the 3rd image is to specify grader can
Accurately to carry out the resolution ratio of image recognition, therefore by specifying grader that accurately the 3rd image can be identified, so that
Also the identification to the first image is just accurately realized, the accuracy of image recognition is higher.
Those skilled in the art will readily occur to the present invention its after considering specification and putting into practice invention disclosed herein
Its embodiment.This application is intended to cover the present invention any variations, uses, or adaptations, these modifications, purposes or
Person's adaptive change follows the general principle of the present invention and including the undocumented common knowledge in the art of the disclosure
Or conventional techniques.Description and embodiments are considered only as exemplary, and true scope and spirit of the invention are by following
Claim is pointed out.
It should be appreciated that the invention is not limited in the precision architecture for being described above and being shown in the drawings, and
And various modifications and changes may be made without departing from the scope thereof.The scope of the present invention is only limited by appended claim.
Claims (14)
- A kind of 1. image-recognizing method, it is characterised in that the described method includes：Obtain the resolution ratio of the first image to be identified；When the resolution ratio of described first image is more than the first default resolution ratio or during less than the second default resolution ratio, according to described the One image, by specifying the 3rd figure of the convolutional layer included in multiple dimensioned layer or the generation of warp lamination with the 3rd default resolution ratio Picture, the described first default resolution ratio are N times of the described 3rd default resolution ratio, and the described 3rd default resolution ratio is described second pre- If N times of resolution ratio, the N is more than 1；By specifying grader that the 3rd image is identified.
- It is 2. multiple dimensioned by specifying according to the method described in claim 1, it is characterized in that, described according to described first image Threeth image of convolutional layer or warp the lamination generation included in layer with the 3rd default resolution ratio, including：According to the resolution ratio of the first image, described first image is scaled with the described first default resolution ratio or described second Second image of default resolution ratio；By it is described specify in multiple dimensioned layer the convolutional layer that includes or warp lamination to generate second image is corresponding to have institute State the 3rd image of the 3rd default resolution ratio.
- 3. according to the method described in claim 2, it is characterized in that, the resolution ratio according to the first image, by described first Image scaling is the second image with the described first default resolution ratio or the second default resolution ratio, including：When the resolution ratio of described first image is more than the described first default resolution ratio, described first image is scaled with institute State the second image of the first default resolution ratio；When the resolution ratio of described first image is less than the described second default resolution ratio, described first image is scaled with institute State the second image of the second default resolution ratio.
- 4. according to the method described in claim 2, it is characterized in that, described specify the convolution included in multiple dimensioned layer by described Layer or warp lamination generate corresponding 3rd image with the described 3rd default resolution ratio of second image, including：When the resolution ratio of second image presets resolution ratio for described first, pass through what is included in the specified multiple dimensioned layer Convolutional layer generates corresponding 3rd image with the described 3rd default resolution ratio of second image；When the resolution ratio of second image presets resolution ratio for described second, pass through what is included in the specified multiple dimensioned layer Warp lamination generates corresponding 3rd image with the described 3rd default resolution ratio of second image.
- 5. according to the method described in claim 1, it is characterized in that, the resolution ratio for obtaining the first image to be identified it Afterwards, further include：When the resolution ratio of described first image is less than or equal to the described first default resolution ratio and pre- more than or equal to described second If during resolution ratio, described first image is scaled the 3rd image with the described 3rd default resolution ratio.
- 6. according to any methods of claim 1-5, it is characterised in that it is described according to described first image, by specifying Before the 3rd image of convolutional layer or warp the lamination generation included in multiple dimensioned layer with the 3rd default resolution ratio, further include：Multiple pre-set image collection are obtained, the multiple pre-set image concentrates all pre-set images that each pre-set image collection includes Belong to same category；Trained multiple dimensioned layer and disaggregated model are treated using the multiple pre-set image collection to be trained, and are obtained described specified more Scale layer and the specified grader.
- 7. a kind of pattern recognition device, it is characterised in that described device includes：First acquisition module, for obtaining the resolution ratio of the first image to be identified；Generation module, for being more than the first default resolution ratio or less than the second default resolution ratio when the resolution ratio of described first image When, according to described first image, by specifying the convolutional layer included in multiple dimensioned layer or the generation of warp lamination to have the 3rd to preset 3rd image of resolution ratio, the described first default resolution ratio are N times of the described 3rd default resolution ratio, and the described 3rd presets resolution Rate is N times of the described second default resolution ratio, and the N is more than 1；Identification module, for by specifying grader that the 3rd image is identified.
- 8. device according to claim 7, it is characterised in that the generation module includes：Submodule is scaled, for the resolution ratio according to the first image, described first image is scaled and is preset with described first Second image of resolution ratio or the second default resolution ratio；Submodule is generated, for specifying the convolutional layer included in multiple dimensioned layer or warp lamination to generate second figure by described As corresponding 3rd image with the described 3rd default resolution ratio.
- 9. device according to claim 8, it is characterised in that the scaling submodule is additionally operable to：When the resolution ratio of described first image is more than the described first default resolution ratio, described first image is scaled with institute State the second image of the first default resolution ratio；When the resolution ratio of described first image is less than the described second default resolution ratio, described first image is scaled with institute State the second image of the second default resolution ratio.
- 10. device according to claim 8, it is characterised in that the generation submodule is additionally operable to：When the resolution ratio of second image presets resolution ratio for described first, pass through what is included in the specified multiple dimensioned layer Convolutional layer generates corresponding 3rd image with the described 3rd default resolution ratio of second image；When the resolution ratio of second image presets resolution ratio for described second, pass through what is included in the specified multiple dimensioned layer Warp lamination generates corresponding 3rd image with the described 3rd default resolution ratio of second image.
- 11. device according to claim 7, it is characterised in that described device further includes：Zoom module, for when the resolution ratio of described first image is less than or equal to the described first default resolution ratio and is more than or waits When the described second default resolution ratio, described first image is scaled the 3rd image with the described 3rd default resolution ratio.
- 12. according to any devices of claim 7-11, it is characterised in that described device further includes：Second acquisition module, for obtaining multiple pre-set image collection, the multiple pre-set image concentrates each pre-set image Ji Bao All pre-set images included belong to same category；Training module, is trained for treating trained multiple dimensioned layer and disaggregated model using the multiple pre-set image collection, Obtain the specified multiple dimensioned layer and the specified grader.
- 13. a kind of pattern recognition device, it is characterised in that described device includes：Processor；For storing the memory of processor-executable instruction；Wherein, the processor is configured as the step of perform claim requires any one method described in 1-6.
- 14. a kind of computer-readable recording medium, instruction is stored with the computer-readable recording medium, it is characterised in that The step of any one method described in claim 1-6 is realized when described instruction is executed by processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711318139.0A CN107992894B (en) | 2017-12-12 | 2017-12-12 | Image recognition method, image recognition device and computer-readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711318139.0A CN107992894B (en) | 2017-12-12 | 2017-12-12 | Image recognition method, image recognition device and computer-readable storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107992894A true CN107992894A (en) | 2018-05-04 |
| CN107992894B CN107992894B (en) | 2022-02-08 |
Family
ID=62037195
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711318139.0A Active CN107992894B (en) | 2017-12-12 | 2017-12-12 | Image recognition method, image recognition device and computer-readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107992894B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111429458A (en) * | 2020-03-20 | 2020-07-17 | 北京创世云科技有限公司 | Image restoration method and device and electronic equipment |
| CN112149756A (en) * | 2020-10-14 | 2020-12-29 | 深圳前海微众银行股份有限公司 | Model training method, image recognition method, device, equipment and storage medium |
| CN112965604A (en) * | 2021-03-29 | 2021-06-15 | 深圳市优必选科技股份有限公司 | Gesture recognition method and device, terminal equipment and computer readable storage medium |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1852388A (en) * | 2006-05-12 | 2006-10-25 | 北京中星微电子有限公司 | Image minifying-amplifying device |
| CN101325040A (en) * | 2008-07-16 | 2008-12-17 | 宇龙计算机通信科技(深圳)有限公司 | Mobile terminal capable of adjusting resolution and method for adjusting resolution of the mobile terminal |
| CN103745234A (en) * | 2014-01-23 | 2014-04-23 | 东北大学 | Band steel surface defect feature extraction and classification method |
| CN104361559A (en) * | 2014-11-12 | 2015-02-18 | 惠州Tcl移动通信有限公司 | Image magnifying method and system based on small image memory |
| CN105825243A (en) * | 2015-01-07 | 2016-08-03 | 阿里巴巴集团控股有限公司 | Method and device for certificate image detection |
| CN106250838A (en) * | 2016-07-27 | 2016-12-21 | 乐视控股（北京）有限公司 | vehicle identification method and system |
| CN106407903A (en) * | 2016-08-31 | 2017-02-15 | 四川瞳知科技有限公司 | Multiple dimensioned convolution neural network-based real time human body abnormal behavior identification method |
| CN106530227A (en) * | 2016-10-27 | 2017-03-22 | 北京小米移动软件有限公司 | Image restoration method and device |
| CN106778773A (en) * | 2016-11-23 | 2017-05-31 | 北京小米移动软件有限公司 | The localization method and device of object in picture |
| CN106919920A (en) * | 2017-03-06 | 2017-07-04 | 重庆邮电大学 | Scene recognition method based on convolution feature and spatial vision bag of words |
| CN107341517A (en) * | 2017-07-07 | 2017-11-10 | 哈尔滨工业大学 | The multiple dimensioned wisp detection method of Fusion Features between a kind of level based on deep learning |
-
2017
- 2017-12-12 CN CN201711318139.0A patent/CN107992894B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1852388A (en) * | 2006-05-12 | 2006-10-25 | 北京中星微电子有限公司 | Image minifying-amplifying device |
| CN101325040A (en) * | 2008-07-16 | 2008-12-17 | 宇龙计算机通信科技(深圳)有限公司 | Mobile terminal capable of adjusting resolution and method for adjusting resolution of the mobile terminal |
| CN103745234A (en) * | 2014-01-23 | 2014-04-23 | 东北大学 | Band steel surface defect feature extraction and classification method |
| CN104361559A (en) * | 2014-11-12 | 2015-02-18 | 惠州Tcl移动通信有限公司 | Image magnifying method and system based on small image memory |
| CN105825243A (en) * | 2015-01-07 | 2016-08-03 | 阿里巴巴集团控股有限公司 | Method and device for certificate image detection |
| CN106250838A (en) * | 2016-07-27 | 2016-12-21 | 乐视控股（北京）有限公司 | vehicle identification method and system |
| CN106407903A (en) * | 2016-08-31 | 2017-02-15 | 四川瞳知科技有限公司 | Multiple dimensioned convolution neural network-based real time human body abnormal behavior identification method |
| CN106530227A (en) * | 2016-10-27 | 2017-03-22 | 北京小米移动软件有限公司 | Image restoration method and device |
| CN106778773A (en) * | 2016-11-23 | 2017-05-31 | 北京小米移动软件有限公司 | The localization method and device of object in picture |
| CN106919920A (en) * | 2017-03-06 | 2017-07-04 | 重庆邮电大学 | Scene recognition method based on convolution feature and spatial vision bag of words |
| CN107341517A (en) * | 2017-07-07 | 2017-11-10 | 哈尔滨工业大学 | The multiple dimensioned wisp detection method of Fusion Features between a kind of level based on deep learning |
Non-Patent Citations (1)
| Title |
|---|
| 魏传义: "文本图像版面分析关键技术研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111429458A (en) * | 2020-03-20 | 2020-07-17 | 北京创世云科技有限公司 | Image restoration method and device and electronic equipment |
| CN112149756A (en) * | 2020-10-14 | 2020-12-29 | 深圳前海微众银行股份有限公司 | Model training method, image recognition method, device, equipment and storage medium |
| CN112965604A (en) * | 2021-03-29 | 2021-06-15 | 深圳市优必选科技股份有限公司 | Gesture recognition method and device, terminal equipment and computer readable storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107992894B (en) | 2022-02-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108629354B (en) | Target detection method and device | |
| CN107527059A (en) | Character recognition method, device and terminal | |
| CN107944447B (en) | Image classification method and device | |
| CN106202330A (en) | The determination methods of junk information and device | |
| CN106778773A (en) | The localization method and device of object in picture | |
| CN105528606A (en) | Region identification method and device | |
| CN106778531A (en) | Face detection method and device | |
| CN108010060A (en) | Object detection method and device | |
| CN108062547A (en) | Character detecting method and device | |
| CN107563994A (en) | The conspicuousness detection method and device of image | |
| CN106228556A (en) | Image quality analysis method and device | |
| CN107992848A (en) | Obtain the method, apparatus and computer-readable recording medium of depth image | |
| CN105354560A (en) | Fingerprint identification method and device | |
| CN105301183B (en) | Detection of Air Quality method and apparatus | |
| CN107967459A (en) | convolution processing method, device and storage medium | |
| CN107729880A (en) | Method for detecting human face and device | |
| CN107463903A (en) | Face key independent positioning method and device | |
| CN104050645A (en) | Image processing method and device | |
| CN108717542A (en) | Identify the method, apparatus and computer readable storage medium of character area | |
| CN105487773A (en) | Screen capturing method and device | |
| CN106446946A (en) | Image recognition method and device | |
| CN105095366A (en) | Method and device for processing character messages | |
| CN107220614A (en) | Image-recognizing method, device and computer-readable recording medium | |
| CN107992894A (en) | Image-recognizing method, device and computer-readable recording medium | |
| CN104077563A (en) | Human face recognition method and device |
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 |