CN111260623B - Picture evaluation method, device, equipment and storage medium - Google Patents
Picture evaluation method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN111260623B CN111260623B CN202010038652.XA CN202010038652A CN111260623B CN 111260623 B CN111260623 B CN 111260623B CN 202010038652 A CN202010038652 A CN 202010038652A CN 111260623 B CN111260623 B CN 111260623B
- Authority
- CN
- China
- Prior art keywords
- picture
- evaluated
- similarity
- pixel
- standard
- 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.)
- Active
Links
- 238000011156 evaluation Methods 0.000 title claims abstract description 106
- 238000012937 correction Methods 0.000 claims abstract description 94
- 239000011159 matrix material Substances 0.000 claims description 42
- 238000004364 calculation method Methods 0.000 claims description 18
- 238000013528 artificial neural network Methods 0.000 claims description 16
- 238000012015 optical character recognition Methods 0.000 claims description 8
- 238000013473 artificial intelligence Methods 0.000 abstract description 3
- 238000000034 method Methods 0.000 description 29
- 238000012545 processing Methods 0.000 description 11
- 238000012549 training Methods 0.000 description 9
- 238000012795 verification Methods 0.000 description 9
- 238000010586 diagram Methods 0.000 description 8
- 230000006870 function Effects 0.000 description 8
- 238000004891 communication Methods 0.000 description 6
- 238000002788 crimping Methods 0.000 description 5
- 238000012550 audit Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000003811 curling process Effects 0.000 description 2
- 238000003702 image correction Methods 0.000 description 2
- 238000000926 separation method Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 238000010276 construction Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 230000008707 rearrangement Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 239000011800 void material Substances 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/80—Geometric correction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/22—Image preprocessing by selection of a specific region containing or referencing a pattern; Locating or processing of specific regions to guide the detection or recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30168—Image quality inspection
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
The embodiment of the application discloses a picture evaluation method, a device, equipment and a storage medium, which relate to the technical field of artificial intelligence and comprise the following steps: acquiring a picture to be evaluated; performing deformation correction on the picture to be evaluated through a picture correction model to obtain a corrected picture; and calculating the similarity between the corrected picture and each standard picture in the standard picture library, and obtaining the evaluation score of the picture to be evaluated according to the similarity. By adopting the scheme, the technical problem that the auditing result of the picture cannot be obtained rapidly and efficiently in the prior art can be solved.
Description
Technical Field
The embodiment of the application relates to the technical field of artificial intelligence, in particular to a picture evaluation method, a device, equipment and a storage medium.
Background
Nowadays, intelligent technology is widely applied to various fields of people's life. For example, in an indoor shopping scene, a user can determine the position of the intelligent device through application software with a positioning function. For example, the smart watch determines its own position by the positioning function and displays the position, and at the same time, sends the position to the associated smart phone, so that the user of the smart phone confirms the position of the smart watch. In order to enable a user to accurately and comprehensively determine position information, store pictures of shops at the position can be synchronously displayed in a smart watch or a smart phone as display information. In the prior art, a picture serving as presentation information is usually required to be manually photographed and checked, so that a picture with high quality is displayed to a user. However, in order to ensure the success rate of the positioning result, a large number of pictures are stored in a background server of the application software with the positioning function, so that a large amount of human resources are required to be input, and the auditing period is long. Meanwhile, subjectivity exists in manual auditing, and accuracy of an image auditing result can be affected.
In summary, how to obtain the auditing result of the picture rapidly and efficiently becomes a problem to be solved urgently.
Disclosure of Invention
The application provides a picture evaluation method, device, equipment and storage medium, which are used for solving the technical problem that an audit result of a picture cannot be obtained rapidly and efficiently in the prior art.
In a first aspect, an embodiment of the present application provides a method for evaluating a picture, including:
acquiring a picture to be evaluated;
performing deformation correction on the picture to be evaluated through a picture correction model to obtain a corrected picture;
and calculating the similarity between the corrected picture and each standard picture in the standard picture library, and obtaining the evaluation score of the picture to be evaluated according to the similarity.
Further, the performing deformation correction on the image to be evaluated through the image correction model to obtain a corrected image includes:
inputting the picture to be evaluated into a picture correction model;
instructing the picture correction model to fold and/or curl each pixel point in the picture to be evaluated so as to obtain at least one distorted picture;
the picture correction model is instructed to determine a pixel matrix of a distorted picture according to the picture to be evaluated and the distorted picture, wherein the pixel matrix comprises gray values of all pixel points in the distorted picture, moving directions and moving distances of all pixel points on an x-axis, and moving directions and moving distances of all pixel points on a y-axis;
And indicating the picture correction model to input the distorted picture containing the pixel matrix into a neural network containing cavity convolution so as to obtain a corrected picture.
Further, after the indicating the picture correction model determines the pixel matrix of the distorted picture according to the picture to be evaluated and the distorted picture, the method further includes:
indicating the picture correction model to determine blank pixel points in the distorted picture based on the pixel matrix;
and indicating the picture correction model to conduct neighbor interpolation based on the neighbor pixel points of the blank pixel points so as to obtain gray values of the blank pixel points, and updating the pixel matrix.
Further, the calculating the similarity between the corrected picture and each standard picture in the standard picture library, and obtaining the evaluation score of the picture to be evaluated according to the similarity includes:
acquiring a pixel area containing text information in the corrected picture, wherein the text information is obtained by utilizing optical character recognition;
calculating the similarity between the pixel region and each standard picture in the standard picture library;
and selecting the maximum similarity from the calculated similarities, and taking the maximum similarity as the evaluation score of the picture to be evaluated.
Further, the calculating the similarity between the pixel region and each standard picture in the standard picture library includes:
obtaining local similarity between the pixel region and a standard picture in a standard picture library by using a structural similarity algorithm;
and calculating the average similarity of the local similarity by combining the resolution dimension of the picture to be evaluated to obtain the similarity between the pixel region and the corresponding standard picture.
Further, the calculating the average similarity of the local similarity by combining the resolution dimension of the picture to be evaluated includes:
confirming a resolution dimension of a sub-pixel region corresponding to the local similarity in the picture to be evaluated;
searching for the weight corresponding to the resolution dimension;
and calculating average similarity according to each local similarity and the corresponding weight.
Further, after calculating the similarity between the corrected picture and each standard picture in the standard picture library and obtaining the evaluation score of the picture to be evaluated according to the similarity, the method further includes:
if the evaluation score is greater than or equal to a score threshold, confirming the picture to be evaluated as a valid picture;
Uploading the effective picture and text information contained in the effective picture to a database, so that a user can check the effective picture and the text information through the database.
In a second aspect, an embodiment of the present application further provides a device for evaluating a picture, including:
the image acquisition module is used for acquiring an image to be evaluated;
the picture correction module is used for carrying out deformation correction on the picture to be evaluated through a picture correction model so as to obtain a corrected picture;
the score calculation module is used for calculating the similarity between the corrected picture and each standard picture in the standard picture library and obtaining the evaluation score of the picture to be evaluated according to the similarity.
In a third aspect, an embodiment of the present application further provides a picture evaluation apparatus, including:
one or more processors;
a memory for storing one or more programs;
the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the picture evaluation method as described in the first aspect.
In a fourth aspect, embodiments of the present application also provide a storage medium containing computer-executable instructions, which when executed by a computer processor, are for performing the picture evaluation method as described in the first aspect.
According to the picture evaluation method, device, equipment and storage medium, the picture to be evaluated is obtained, the picture to be evaluated is subjected to deformation correction by using the picture correction model to obtain the corrected picture, and then the similarity of the corrected picture and each standard picture in the standard picture library is calculated, so that the technical means of obtaining the evaluation score of the picture to be evaluated can be achieved, the technical problem that the auditing result of the picture cannot be obtained quickly and efficiently in the prior art can be solved, the picture is automatically evaluated by using the picture correction and the picture similarity calculation, a large amount of manpower resources can be saved, the auditing period is shortened, the subjectivity of the manually-audited picture can be prevented, and the quality of picture auditing is guaranteed.
Drawings
Fig. 1 is a flowchart of a picture evaluation method according to an embodiment of the present application;
fig. 2 is a flowchart of a picture evaluation method according to an embodiment of the present application;
FIG. 3 is a schematic illustration of the folding and/or crimping process provided in accordance with one embodiment of the present application;
fig. 4 is a schematic structural diagram of a picture evaluation device according to an embodiment of the present application;
fig. 5 is a schematic structural diagram of a picture evaluation apparatus according to an embodiment of the present application.
Detailed Description
The present application is described in further detail below with reference to the drawings and examples. It is to be understood that the specific embodiments described herein are for purposes of illustration and not limitation. It should be further noted that, for convenience of description, only some, but not all of the structures related to the present application are shown in the drawings.
The picture evaluation method provided in the embodiment may be performed by a picture evaluation apparatus, which may be implemented by means of software and/or hardware and integrated in a picture evaluation device. The picture evaluation device can be intelligent devices with data processing and analyzing capabilities, such as a tablet computer and a desktop computer.
Fig. 1 is a flowchart of a picture evaluation method according to an embodiment of the present application. Referring to fig. 1, the picture evaluation method specifically includes:
and 110, acquiring a picture to be evaluated.
The picture to be evaluated is a picture which is manually shot and needs to be evaluated by picture evaluation equipment, and the content of the picture to be evaluated can be set according to actual setting. For example, in a shop search scene, the picture to be evaluated may be a store picture; under a scenic spot searching scene, the picture to be evaluated can be a scenic spot picture; under the tour showing scene, the pictures to be evaluated can be scenic spot pictures, site pictures and the like. In the embodiment, taking a picture to be evaluated as a store picture as an example, at this time, the picture to be evaluated displays the name and/or logo of the store. In the embodiment, description will be made taking, as an example, a name and logo of a shop displayed in a picture to be evaluated.
Further, the embodiment of the obtaining manner of the picture to be evaluated is not limited. For example, the picture evaluation apparatus is provided with an image pickup device (such as a camera or the like), and a picture to be evaluated is acquired by the image pickup device. For another example, the image evaluation device may perform data communication with other devices (such as a mobile phone, a camera, a tablet computer, etc.) configured with the image capturing apparatus, so as to obtain the image to be evaluated. As another example, the other device configured with the image capturing apparatus uploads the picture to be evaluated to a specific storage location (for example Yun Pan) of the internet, and the picture evaluation device may acquire the picture to be evaluated through the specific storage location of the internet.
And 120, performing deformation correction on the picture to be evaluated through a picture correction model to obtain a corrected picture.
When a user shoots a picture to be evaluated, the shot picture is deformed to a certain extent compared with the actual content due to the limitation of shooting angle and shooting level. For example, the shop name of the shop is usually located above the shop entrance, and when the user takes the shop picture, the image capturing device needs to be tilted to include the complete shop name, and at this time, the picture to be evaluated taken by the image capturing device will show the effect of looking up to take, that is, compared with the actual shop, the effect that the shop image in the picture to be evaluated will be wider at the top and narrower at the bottom. Therefore, in order to facilitate effective evaluation of the image to be evaluated, the deformation in the image to be evaluated needs to be corrected first, in the embodiment, the correction process is referred to as deformation correction, and the result obtained after the deformation correction is referred to as a corrected image. The corrected picture is equal to the picture to be evaluated in size.
Specifically, the deformation correction can be realized by a picture correction model. The specific embodiment of the construction mode of the picture correction model is not limited. For example, the distortion transformation is used as one of deformation correction, and can be implemented by at least one operation of folding, curling, moving pixels, and the like, at this time, calculation rules corresponding to each operation can be set in the picture correction model, such as a folding formula, a curling formula, and a moving formula of the pixels are set in the picture correction model, and formula parameters can be modified in combination with the picture to be evaluated, so that the deformation correction is implemented by the formulas. For another example, a neural network is used to train a large number of known pictures to be evaluated and corrected pictures in an artificial intelligence manner to obtain a picture correction model. For another example, a picture correction model is obtained by combining a calculation rule and a neural network.
Optionally, in the deformation correction process, blank pixels may appear in the corrected picture, and in this case, in order to ensure the quality of the corrected picture, a neighbor interpolation method may be used to assign pixel values to the blank pixels.
And 130, calculating the similarity between the corrected picture and each standard picture in the standard picture library, and obtaining the evaluation score of the picture to be evaluated according to the similarity.
In one embodiment, a large number of standard pictures are stored in a standard picture library. And the names and/or the logo of the stores are displayed in each standard picture, and the names and/or the logo in the standard picture are regular patterns. It will be appreciated that if the names and/or corresponding logo of multiple stores are the same, a standard picture may be shared. Furthermore, the collection range of the standard pictures in the standard picture library can be set according to actual conditions. For example, in the province, the names and/or the corresponding logo of each store in the province are collected, and standard pictures of each store are obtained and form a standard picture library. For another example, the names and/or the corresponding logo of each store in the whole country are collected by taking the country as a unit, so that standard pictures of each store are obtained and a standard picture library is formed. It should be noted that the embodiment of the standard image capturing manner is not limited. In the embodiment, description is made by taking a standard picture including a name and logo as an example.
Specifically, the higher the similarity between the name and logo of the shop in the corrected picture and a certain standard picture, the clearer and more accurate the name and logo of the shop in the corrected picture are, and correspondingly, the better the quality of the picture to be evaluated is and the higher the evaluation score is. Otherwise, the lower the similarity between the corrected picture and the standard picture is, the more inaccurate the names and logo of shops in the corrected picture are, and correspondingly, the lower the quality of the picture to be evaluated is and the lower the evaluation score is. Therefore, in the embodiment, the evaluation score of the picture to be evaluated is obtained in such a manner that the similarity between the corrected picture and each standard picture in the standard picture library is calculated.
Exemplary embodiments of the mode of calculating the similarity between the corrected picture and the standard picture are not limited. For example, the similarity between the corrected picture and the standard picture is calculated using the structural similarity (structural similarity index, SSIM). As another example, the similarity between the modified picture and the standard picture is calculated based on cosine similarity. Further, each standard picture has a similarity calculated with the corrected picture, and the higher the similarity is, the more similar the corrected picture is to the corresponding standard picture, and the more accurate the corrected picture is, so that the maximum similarity is selected as the evaluation score of the picture to be evaluated in all the calculated similarities.
Optionally, after the evaluation score of the picture to be evaluated is obtained, the score to be evaluated may be compared with a preset score threshold. The score threshold may be set according to the actual situation. When the evaluation score is greater than or equal to the score threshold, the quality of the picture to be evaluated is good, and the picture to be evaluated can be considered to pass the audit and can be used later. When the evaluation score is smaller than the score threshold, the quality of the score to be evaluated is poor, and the picture to be evaluated can be considered as failed in verification, and at the moment, the picture to be evaluated cannot be used later.
According to the technical means, the picture to be evaluated is obtained, the picture to be evaluated is subjected to deformation correction by using the picture correction model to obtain the corrected picture, and then the similarity between the corrected picture and each standard picture in the standard picture library is calculated, so that the evaluation score of the picture to be evaluated is obtained, the problem that the auditing result of the picture cannot be obtained rapidly and efficiently in the prior art can be solved, the picture is automatically evaluated by using picture correction and picture similarity calculation, a large amount of manpower resources can be saved, the auditing period is shortened, subjectivity of manual auditing of the picture can be prevented, and the quality of picture auditing is ensured.
Fig. 2 is a flowchart of a picture evaluation method according to an embodiment of the present application. The picture evaluation method provided in this embodiment is embodied on the basis of the above embodiment. In this embodiment, description will be made taking an example in which a picture to be evaluated is used for presentation to a user. Specifically, referring to fig. 2, the method for evaluating a picture provided in this embodiment includes:
step 201, obtaining a picture to be evaluated.
Step 202, inputting the picture to be evaluated into a picture correction model.
Specifically, the picture to be evaluated is used as the input of a picture correction model. It can be understood that the picture to be evaluated input into the picture correction model is not subjected to other processing and is an uncompressed undistorted image.
And 203, instructing the picture correction model to perform folding and/or curling processing on each pixel point in the picture to be evaluated so as to obtain at least one distorted picture.
In an embodiment, the distortion is divided into folding and curling, that is, the distortion picture may refer to a picture obtained by folding and/or curling a picture to be evaluated, and the number of the distortion pictures is at least one, and may be set according to actual situations.
Specifically, when the picture correction model performs folding processing on a picture to be evaluated, a folding formula may be used:wherein w is 1 To the extent of folding alpha 1 Is the super parameter in folding process, d 1 Is a reconciliation parameter in the folding process. d, d 1 Can be set according to actual conditions, and can refer to historical experience data. Alternatively, by adjusting alpha 1 Can obtain different w 1 Further obtaining the distorted pictures with different folding degrees. When the picture to be evaluated is curled, a curling formula can be utilized:wherein w is 2 To the extent of curling alpha 2 Is the super parameter in the curling treatment, d 2 Is a reconciliation parameter at the time of the crimping process. d, d 2 Can be set according to actual conditions, and can refer to historical experience data. Alternatively, by adjusting alpha 2 Can obtain different w 2 Further, distorted pictures with different curling degrees are obtained. In practical applications, the twisted picture can be obtained by adopting folding and/or curling treatment in combination with practical situations. It will be appreciated that when both folding and crimping processes are employed, alpha 1 And alpha 2 May be set to the same parameters or to different parameters.
Optionally, when obtaining the distorted picture, firstly converting the picture to be evaluated into a P-V thermodynamic diagram, and then passing through alpha 1 And/or alpha 2 The corresponding w is obtained by utilizing a folding formula and a curling formula 1 And/or w 2 . Thereafter, based on w 1 And/or w 2 And (3) twisting the P-V thermodynamic diagram, and mapping gray values on all pixel points in the picture to be evaluated to corresponding pixel points in the twisted P-V thermodynamic diagram to obtain a corresponding twisted picture. It will be appreciated that at this point the warped picture is a grey scale image.
Step 204, instructing the picture correction model to determine a pixel matrix of the distorted picture according to the picture to be evaluated and the distorted picture, where the pixel matrix includes a gray value of each pixel point in the distorted picture, a moving direction and a moving distance of each pixel point on an x-axis, and a moving direction and a moving distance of each pixel point on a y-axis.
In an embodiment, labels of all pixel points in the distorted picture are embodied through a pixel matrix. The pixel matrix can be obtained by comparing the picture to be evaluated with the distorted picture, specifically, the picture correction model finds out the pixel points corresponding to the pixel points in the distorted picture in the picture to be evaluated, and further obtains the pixel matrix. The pixel matrix is a three-dimensional matrix and is used for storing gray values of all pixel points in the distorted picture, and the moving direction and the moving distance of all pixel points on an x axis and the moving direction and the moving distance of all pixel points on a y axis. The moving direction and the moving distance refer to the direction and the pixel distance, respectively, in which the pixel point in the distorted picture is mapped to the picture to be evaluated and should be moved. At this time, the pixel matrix may include gray values, moving directions and moving distances of all the pixels in the warped picture, that is, each pixel in the warped picture may find a corresponding label in the pixel matrix.
Step 205, instruct the picture correction model to input the distorted picture including the pixel matrix into the neural network including the hole convolution, so as to obtain a corrected picture.
Specifically, the distorted picture and the pixel matrix are identified through the neural network, and then the corrected picture is output. At this time, the neural network can convert the pixel level classification problem into the pixel level regression problem by utilizing semantic learning, so as to correct the picture to be evaluated to obtain a corrected picture.
It will be appreciated that the specific structure of the neural network may be set according to the actual situation. In one embodiment, a neural network is employed that includes a hole convolution. Hole convolution (Dilated Convolution) is to inject holes into a standard convolution kernel to increase the receptive field (receptive field), and to reduce the distortion of the corrected picture to some extent. At this time, each pixel point can see the information of more surrounding pixels, so that a more accurate result is obtained. For example, when the hole convolution is 1, i.e., 1-delayed conv, the convolution layer operates as a normal convolution layer. When the hole convolution is 2, that is, when the size of the convolution kernel is 3×3, the receptive field of the convolution layer may reach 7×7, and at this time, the size of the convolution kernel and the receptive field have an exponential relationship. Therefore, the receptive field can be increased under the condition of not carrying out pooling loss information, the parameter number of the neural network is not increased, the occurrence of the overfitting phenomenon is reduced, and meanwhile, the training speed is increased. Alternatively, the neural network containing the hole convolution may be a multi-scale U-net network, and it may contain a parallel network and a serial network, where both the parallel network and the serial network contain the hole convolution.
It will be appreciated that the above-mentioned picture correction models may be obtained by training. I.e. a large amount of training data is pre-selected, wherein an original picture taken by a user and a corresponding modified picture are taken as a set of training data. Then, the picture correction model is trained based on a large amount of training data to adjust parameters (such as super parameters, void volume number, etc.) involved in the picture correction model. When the model training is stable, a certain amount of verification data can be collected again, wherein an original picture and a corresponding correction picture shot by a user serve as a group of verification data. The verification data and the training data are different, for example, a large amount of data is collected, 80% of which is the training data and 20% of which is the verification data. And then, verifying the picture correction model based on the verification data to determine the correction effect of the picture correction model. When the verification of the picture correction model is passed, the stable correction result of the picture correction model can be indicated to be applicable. In this embodiment, the verified picture correction model may be directly applied to obtain the corrected picture. It should be noted that, when the picture correction model is trained and verified, the processing procedure of the picture correction model is similar to the processing procedure when the picture correction model is applied in steps 203 to 205, except that parameters in the application procedure are fixed, and parameters in the training procedure need to be adjusted.
It can be understood that the picture correction model can be implemented in other manners, for example, by using deep lab-CRF as the back end of picture separation, so as to ensure the prediction accuracy of picture semantic separation.
And 206, acquiring a pixel area containing text information in the corrected picture, wherein the text information is obtained by utilizing optical character recognition.
Optical character recognition (Optical Character Recognition, OCR) refers to the process of determining the shape of a character in an image by detecting dark and light patterns and then translating the shape into computer text using a character recognition method. In an embodiment, the corrected picture is identified by OCR, so that text information contained in the corrected picture and a pixel area of the text information in the corrected picture can be determined. The text information may be information obtained by identifying a text included in the corrected picture, and the text information may include contents such as a text, a number, and a character. For example, the picture to be evaluated is a shop picture, and the text information may include the name and logo of the shop. The pixel area of the text information in the corrected picture refers to an area formed by pixel points for displaying the text information in the corrected picture. It is understood that a pixel region is a block region that is smaller than or equal to the modified picture size.
Step 207, calculating the similarity between the pixel region and each standard picture in the standard picture library.
Since the pixel region contains the name and logo of the store obtained by shooting and the standard picture contains the name and logo of the standard, the similarity between the corrected picture and the standard picture can be determined by calculating the similarity between the pixel region and the standard picture.
Specifically, a structural similarity algorithm may be used in calculating the similarity. At this time, the steps specifically include steps 2071 to 2072:
step 2071, obtaining the local similarity between the pixel region and the standard picture in the standard picture library by using a structural similarity algorithm.
Specifically, the SSIM obtains the similarity by comparing the brightness, contrast and structure of the corrected picture and the standard picture. In SSIM calculation, a window of size n×n is taken, where N may be set according to practical situations, and is usually a positive integer. And then, sliding the window on the pixel area and the standard picture respectively to traverse all the pixel area and the standard picture. The sliding amplitude of the window during sliding may be set according to practical situations, for example, the sliding amplitude is 1, that is, the window slides one pixel point at a time. Further, since the window needs to be slid for multiple times, all pixel areas and standard pictures can be traversed, after each time the window slides, the similarity between the sub-pixel areas of the window in the pixel areas and the areas of the window in the standard pictures needs to be calculated, and in the embodiment, the similarity is recorded as local similarity. That is, a new local similarity is obtained every time the window slides, in other words, a plurality of local similarities can be obtained between the pixel area and a standard picture. It should be noted that in a local similarity calculation process, the sub-pixel area of the window in the pixel area is the same as the pixel coordinate range of the area of the window in the standard picture. According to the mode, the local similarity between the pixel area and each standard picture can be calculated.
Step 2072, calculating the average similarity of the local similarities by combining the resolution dimensions of the pictures to be evaluated, so as to obtain the similarity between the pixel region and the corresponding standard picture.
For example, for each local similarity calculated by a standard picture and a pixel region, an average value of each local similarity is taken as a similarity between the standard picture and the pixel region, where the average value of each local similarity may be recorded as an MS-SSIM index. In practical application, the resolution of the pictures to be evaluated acquired by different image acquisition devices may be different, and the finally obtained evaluation score may be affected by the different resolution, so in order to ensure the accuracy of the evaluation score, in the embodiment, in calculating the similarity, the resolution dimension is added. I.e. the resolution of the current picture to be evaluated is taken into account when calculating the similarity. At this time, the step specifically includes steps 20721 to 20723:
step 20721, determining the resolution dimension in which the sub-pixel region corresponding to the local similarity falls in the image to be evaluated.
Specifically, different resolution ranges are divided in advance, and each resolution range corresponds to one resolution dimension. The embodiment of the division rule is not limited herein. When each local similarity is calculated, the window falls into a sub-pixel area in the pixel area, and in this step, the resolution corresponding to each sub-pixel area can be determined, and then the resolution range to which the resolution belongs can be determined, so that the resolution dimension corresponding to the resolution can be determined.
Step 20722, find the weight corresponding to the resolution dimension.
Specifically, weights corresponding to different resolution dimensions are preset, wherein the embodiment of the setting rule is not limited. The influence of different resolutions on the evaluation score can be balanced by the weights. In an embodiment, after determining a resolution dimension corresponding to each local similarity, obtaining a weight corresponding to each local similarity according to a corresponding relationship between the resolution dimension and the weight.
Step 20723, calculating average similarity according to each local similarity and the corresponding weight.
Specifically, when calculating the average similarity, the average similarity can be obtained by combining weights and using a weighted average mode, and the calculated average similarity is used as the similarity between the pixel region and the corresponding standard picture.
And step 208, selecting the maximum similarity from the calculated similarities, and taking the maximum similarity as the evaluation score of the picture to be evaluated.
Specifically, the higher the similarity is, the more similar the corrected picture is to the corresponding standard picture, the more accurate the text information in the corrected picture is, and the higher the quality of the corresponding picture to be evaluated is. Therefore, after calculating the similarity between the corrected picture and each standard picture in the embodiment, the maximum similarity is selected among all the similarities. And then taking the selected maximum similarity as the evaluation score of the picture to be evaluated.
It can be understood that in practical application, an evaluation score model may be constructed, and in this case, after the corrected picture is input into the evaluation score model, the evaluation score model may output the maximum similarity based on the above processing procedure.
Step 209, determining whether the evaluation score is greater than or equal to the score threshold, and if the evaluation score is greater than or equal to the score threshold, executing step 210. Otherwise, step 212 is performed.
Specifically, the score threshold may be set according to an actual situation, and used for dividing the quality of the picture to be evaluated. When the evaluation score is greater than or equal to the score threshold, it is indicated that the quality of the picture to be evaluated is relatively good, and step 210 is performed, and when the evaluation score is less than the score threshold, it is indicated that the quality of the picture to be evaluated is relatively poor, and step 212 is performed. For example, the score threshold is set to 0.5, if the evaluation score is greater than or equal to 0.5, step 210 is executed, otherwise, step 212 is executed.
And step 210, confirming the picture to be evaluated as a valid picture.
Specifically, when the evaluation score is greater than or equal to the score threshold, the picture to be evaluated is confirmed to be an effective picture, namely, the passing of the examination of the picture to be evaluated is confirmed, and the picture to be evaluated can be used later.
Step 211, uploading the effective picture and text information contained in the effective picture to a database, so that a user can check the effective picture and the text information through the database.
In one embodiment, the valid picture (i.e. the picture to be evaluated) is uploaded to a database, where the database may be a database configured by the picture evaluation device itself, or may be a database configured in other devices or cloud storage, and embodiments of the specific database form are not limited. Further, the user can view the data stored in the database by accessing the database.
Optionally, when uploading the effective picture to the database, text information (such as the name and/or logo of the store) in the effective picture is uploaded at the same time. It can be understood that other information can be synchronously uploaded in combination with the application scene of the database, so that a user can conveniently view text information and effective pictures. For example, assuming that the effective picture is a store picture, for a store display scene, the effective picture and text information may be stored in the database, and at this time, the user may view the corresponding effective picture by inputting the text information. For another example, for an indoor positioning scene, the database also stores specific geographic positions of stores, and at this time, the database can determine the store position nearest to the user by acquiring the position of the user, so as to feed back corresponding effective pictures and text information, thereby facilitating the user to view. It should be noted that, the user accessing the database means that the user performs data communication with the device configuring the database through the device used, thereby realizing access to the database. Because the effective pictures are all the pictures passing through the examination, the user only views the picture with small deformation.
And step 212, confirming the picture to be evaluated as an invalid picture.
Specifically, when the evaluation score is smaller than the score threshold, determining that the picture to be evaluated is an invalid picture, namely, a picture which is not passed by the audit. At the moment, the picture to be evaluated can be collected again and evaluated, and whether the picture to be evaluated passes the verification can be confirmed in a manual verification mode.
According to the method, the picture to be evaluated is obtained through the picture correction model, then the picture to be evaluated and the picture to be evaluated are compared to obtain the pixel matrix of the picture to be evaluated, namely, the label is added to the picture to be evaluated, then the picture to be evaluated and the pixel matrix are used as input of a neural network containing hole convolution to obtain the corrected picture, the local similarity between the pixel area of text information in the corrected picture and the standard picture is obtained through the structural similarity algorithm, the average similarity of the local similarity is calculated by combining the resolution dimension of the picture to be evaluated, the similarity between the pixel area and the standard picture is further obtained, after the similarity between all the standard pictures and the pixel area in the standard picture library is obtained, the maximum similarity is used as the evaluation score of the picture to be evaluated, and then if the evaluation score is larger than or equal to the score threshold, the picture to be evaluated is confirmed to be an effective picture, and the text information contained in the effective picture is uploaded to the database, so that a user can conveniently view the technical means of the effective icon and the text information, the problem that the picture cannot be obtained quickly and efficiently in the prior art is solved, the manual quality can be automatically verified, the quality can be prevented, and the subjective quality can be verified, and the quality of the picture can be verified is only is greatly shortened, and the quality is not be verified. Meanwhile, when the picture is corrected, a label is added to the distorted picture through the pixel matrix, so that the corrected picture can be obtained in a neural network mode, the mode has generalization, and the pixel level classification problem is converted into the pixel level regression problem. Meanwhile, the neural network comprises cavity convolution, so that the receptive field can be increased, and the accuracy of a processing result is ensured. In addition, when the similarity calculation is performed, the influence of different resolutions on the similarity is considered, so that the similarity can be ensured to be more accurate, namely the evaluation accuracy of the picture to be evaluated is improved. Furthermore, after the pictures are automatically evaluated, the effective pictures can be input into the database, so that the user can conveniently inquire, the database is not required to be manually input, and the human resource waste can be further reduced.
It will be appreciated that after the folding and/or crimping process, there may be some black dots or lines in the distorted picture. This is because there is a rounding operation when performing the folding and/or crimping process, which may leave two or two rows of pixels originally adjacent to each other by one row, thereby causing the pixels in the distorted picture to be empty, i.e., appear as black dots or black lines. For example, fig. 3 is a schematic diagram before and after the folding and/or curling process provided in one embodiment of the present application, referring to fig. 3, a region 21 is a pixel region in a picture to be evaluated, each square represents a pixel point, where a pixel value of a first column of pixels is a, a pixel value of a second column of pixels is B, and there is a rounding operation during folding and/or curling transformation, where the region 21 is changed into a region 22, as can be seen from fig. 3, the second column of pixels is blank pixels, and then a black line is displayed at a position of the column of pixels in the distorted picture. At this time, in order to prevent the influence of the black dot or the black line on the subsequent processing, in the embodiment, after setting the image correction model to perform the folding and/or curling process on each pixel point in the image to be evaluated, the method further includes:
And step 213, instructing the picture correction model to determine blank pixel points based on the pixel matrix.
Specifically, since the gray value of each pixel point in the distorted picture is recorded in the pixel matrix, at this time, a blank pixel point may be determined by looking at each pixel value in the pixel matrix. Since the blank pixel is displayed in black, that is, the gray value is 0, the pixel having the gray value of 0 can be determined as the blank pixel.
Step 214, instructing the picture correction model to perform neighbor interpolation based on the neighbor pixel points of the blank pixel points, so as to obtain gray values of the blank pixel points, and updating the pixel matrix.
Specifically, a neighbor interpolation method is adopted to obtain the gray value of the blank pixel point. The gray value of the blank pixel point is obtained by the gray value of the neighboring pixel point of the blank pixel point, wherein the gray value determining mode can be set according to the actual situation, for example, the gray value of a certain neighboring pixel point is selected as the gray value of the blank pixel point. For another example, the gray average value of the neighboring pixel points is calculated, and the gray value of the blank pixel point is obtained after rounding. Or the gray value of the blank pixel point is obtained by adopting other modes. Then, the gray value of each blank pixel point in the pixel matrix can be updated. And then the updated pixel matrix is input into the network model.
By means of the neighbor interpolation mode, blank pixel points in the distorted picture can be avoided, and further the accuracy of subsequent processing is guaranteed. It will be appreciated that, in addition to the neighbor interpolation, other ways of repairing blank pixels may be used, and the embodiment is not limited thereto.
Fig. 4 is a schematic structural diagram of a picture evaluation device according to an embodiment of the present application. Referring to fig. 4, the picture evaluation apparatus provided in this embodiment includes: a picture acquisition module 301, a picture correction module 302 and a score calculation module 303.
The image acquisition module 301 is configured to acquire an image to be evaluated; the picture correction module 302 is configured to perform deformation correction on the picture to be evaluated through a picture correction model, so as to obtain a corrected picture; and the score calculating module 303 is configured to calculate a similarity between the corrected picture and each standard picture in the standard picture library, and obtain an evaluation score of the picture to be evaluated according to the similarity.
According to the technical means, the picture to be evaluated is obtained, the picture to be evaluated is subjected to deformation correction by using the picture correction model to obtain the corrected picture, and then the similarity between the corrected picture and each standard picture in the standard picture library is calculated, so that the evaluation score of the picture to be evaluated is obtained, the technical problem that the auditing result of the picture cannot be obtained quickly and efficiently in the prior art can be solved, the picture is automatically evaluated by using picture correction and picture similarity calculation, a large amount of manpower resources can be saved, the auditing period is shortened, subjectivity of manual auditing pictures can be prevented, and the auditing quality of the picture is guaranteed.
On the basis of the above embodiment, the picture correction module 302 includes: the data input unit is used for inputting the picture to be evaluated into a picture correction model; the distortion picture acquisition unit is used for indicating the picture correction model to carry out folding and/or curling treatment on each pixel point in the picture to be evaluated so as to obtain at least one distortion picture; the matrix determining unit is used for indicating the picture correction model to determine a pixel matrix of the distorted picture according to the picture to be evaluated and the distorted picture, wherein the pixel matrix comprises gray values of all pixel points in the distorted picture, the moving direction and the moving distance of all pixel points on an x-axis and the moving direction and the moving distance of all pixel points on a y-axis; and the corrected picture acquisition unit is used for indicating the picture correction model to input the distorted picture containing the pixel matrix into the neural network containing the hole convolution so as to obtain a corrected picture.
On the basis of the above embodiment, the method further comprises: the blank pixel searching module is used for indicating the picture correction model to determine blank pixel points in the distorted picture based on the pixel matrix after determining the pixel matrix of the distorted picture according to the picture to be evaluated and the distorted picture; and the neighbor interpolation module is used for indicating the picture correction model to perform neighbor interpolation based on the neighbor pixel points of the blank pixel points so as to obtain gray values of the blank pixel points and update the pixel matrix.
On the basis of the above embodiment, the score calculation module 303 includes: the area acquisition unit is used for acquiring a pixel area containing text information in the corrected picture, wherein the text information is obtained by utilizing optical character recognition; the similarity calculation unit is used for calculating the similarity between the pixel area and each standard picture in the standard picture library; and the score selection unit is used for selecting the maximum similarity from the calculated similarities, and taking the maximum similarity as the evaluation score of the picture to be evaluated.
On the basis of the above embodiment, the similarity calculation unit includes: the local similarity calculation subunit is used for obtaining the local similarity between the pixel region and the standard picture in the standard picture library by utilizing a structural similarity algorithm; and the average similarity calculation subunit is used for calculating the average similarity of the local similarity by combining the resolution dimension of the picture to be evaluated so as to obtain the similarity between the pixel region and the corresponding standard picture.
On the basis of the above embodiment, the average similarity calculation subunit includes: the resolution dimension determining grandchild unit is used for determining the resolution dimension of the sub-pixel region corresponding to the local similarity in the picture to be evaluated; the weight searching grandson unit is used for searching the weight corresponding to the resolution dimension; and the computing grandson unit is used for computing average similarity according to each local similarity and the corresponding weight.
On the basis of the above embodiment, the method further comprises: the score comparison module is used for calculating the similarity between the corrected picture and each standard picture in the standard picture library, and determining the picture to be evaluated as an effective picture if the evaluation score is greater than or equal to a score threshold after obtaining the evaluation score of the picture to be evaluated according to the similarity; and the data uploading module is used for uploading the effective picture and the text information contained in the effective picture to a database so that a user can check the effective picture and the text information through the database.
The image evaluation device provided in this embodiment is included in the image evaluation apparatus, and may be used to execute the image evaluation method provided in any of the foregoing embodiments, and has corresponding functions and beneficial effects.
It should be noted that, in the above-mentioned embodiment of the image evaluation device, each unit and module included are only divided according to the functional logic, but not limited to the above-mentioned division, as long as the corresponding functions can be implemented; in addition, the specific names of the functional units are also only for distinguishing from each other, and are not used to limit the protection scope of the present application.
Fig. 5 is a schematic structural diagram of a picture evaluation apparatus according to an embodiment of the present application. Specifically, as shown in fig. 5, the picture evaluation apparatus includes a processor 40, a memory 41, an input device 42, an output device 43, and a communication device 44; the number of processors 40 in the picture evaluation device may be one or more, one processor 40 being taken as an example in fig. 5; the processor 40, the memory 41, the input means 42, the output means 43 and the communication means 44 in the picture evaluation device may be connected by a bus or other means, in fig. 5 by way of example.
The memory 41 is a computer-readable storage medium that can be used to store a software program, a computer-executable program, and modules such as program instructions/modules in the picture evaluation method in the embodiment of the present application (for example, the picture acquisition module 301, the picture correction module 302, and the score calculation module 303 in the picture evaluation apparatus). The processor 40 executes various functional applications of the picture evaluation apparatus and data processing by executing software programs, instructions, and modules stored in the memory 41, that is, implements the picture evaluation method provided by any of the above-described embodiments.
The memory 41 may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, at least one application program required for functions; the storage data area may store data created according to the use of the picture evaluation apparatus, and the like. In addition, memory 41 may include high-speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other non-volatile solid-state storage device. In some examples, memory 41 may further include memory remotely located relative to processor 40, which may be connected to the picture evaluation device via a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
The input means 42 may be used to receive input numeric or character information and to generate key signal inputs related to user settings and function control of the picture evaluation device, and may also be via an image capture device (e.g., a camera), an audio capture device (e.g., a microphone), etc. The output means 43 may comprise a display screen, a speaker or the like. The communication means 44 is for data communication with other devices.
The image evaluation device can be used for executing the image evaluation method provided by any embodiment, and has corresponding functions and beneficial effects.
The present embodiments also provide a storage medium containing computer executable instructions, which when executed by a computer processor, are for performing a picture evaluation method, the method comprising:
acquiring a picture to be evaluated;
performing deformation correction on the picture to be evaluated through a picture correction model to obtain a corrected picture;
and calculating the similarity between the corrected picture and each standard picture in the standard picture library, and obtaining the evaluation score of the picture to be evaluated according to the similarity.
Of course, the storage medium containing the computer executable instructions provided in the embodiments of the present application is not limited to the method operations described above, and may also perform the related operations in the picture evaluation method provided in any embodiment of the present application.
From the above description of embodiments, it will be clear to a person skilled in the art that the present application may be implemented by means of software and necessary general purpose hardware, but of course also by means of hardware, but in many cases the former is a preferred embodiment. Based on such understanding, the technical solution of the present application may be embodied essentially or in a part contributing to the prior art in the form of a software product, which may be stored in a computer readable storage medium, such as a floppy disk, a Read-Only Memory (ROM), a random access Memory (Random Access Memory, RAM), a FLASH Memory (FLASH), a hard disk, or an optical disk of a computer, where the instructions include a number of instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform the picture evaluation method described in the embodiments of the present application.
Note that the above is only a preferred embodiment of the present application and the technical principle applied. Those skilled in the art will appreciate that the present application is not limited to the particular embodiments described herein, but is capable of numerous obvious changes, rearrangements and substitutions as will now become apparent to those skilled in the art without departing from the scope of the present application. Therefore, while the present application has been described in connection with the above embodiments, the present application is not limited to the above embodiments, but may include many other equivalent embodiments without departing from the spirit of the present application, the scope of which is defined by the scope of the appended claims.
Claims (9)
1. A picture evaluation method, characterized by comprising:
acquiring a picture to be evaluated;
the picture to be evaluated is subjected to deformation correction through a picture correction model to obtain a corrected picture, wherein the picture correction model is instructed to carry out folding and/or curling treatment on each pixel point in the picture to be evaluated to obtain at least one distorted picture, the picture correction model is instructed to determine a pixel matrix of the distorted picture according to the picture to be evaluated and the distorted picture, the pixel matrix comprises gray values of each pixel point in the distorted picture, moving directions and moving distances of each pixel point on an x-axis and moving directions and moving distances of each pixel point on a y-axis, and the picture correction model is instructed to input the distorted picture containing the pixel matrix into a neural network containing cavity convolution to obtain the corrected picture;
And calculating the similarity between the corrected picture and each standard picture in the standard picture library, and obtaining the evaluation score of the picture to be evaluated according to the similarity.
2. The picture evaluation method according to claim 1, wherein after the indicating the picture correction model determines the pixel matrix of the warped picture from the picture to be evaluated and the warped picture, further comprising:
indicating the picture correction model to determine blank pixel points in the distorted picture based on the pixel matrix;
and indicating the picture correction model to conduct neighbor interpolation based on the neighbor pixel points of the blank pixel points so as to obtain gray values of the blank pixel points, and updating the pixel matrix.
3. The picture evaluation method according to claim 1, wherein the calculating the similarity between the corrected picture and each standard picture in the standard picture library, and obtaining the evaluation score of the picture to be evaluated according to the similarity, comprises:
acquiring a pixel area containing text information in the corrected picture, wherein the text information is obtained by utilizing optical character recognition;
calculating the similarity between the pixel region and each standard picture in the standard picture library;
And selecting the maximum similarity from the calculated similarities, and taking the maximum similarity as the evaluation score of the picture to be evaluated.
4. A picture evaluation method according to claim 3, wherein said calculating the similarity between the pixel region and each standard picture in the standard picture library comprises:
obtaining local similarity between the pixel region and a standard picture in a standard picture library by using a structural similarity algorithm;
and calculating the average similarity of the local similarity by combining the resolution dimension of the picture to be evaluated to obtain the similarity between the pixel region and the corresponding standard picture.
5. The picture evaluation method according to claim 4, wherein the calculating the average similarity of the local similarities in combination with the resolution dimension of the picture to be evaluated includes:
confirming a resolution dimension of a sub-pixel region corresponding to the local similarity in the picture to be evaluated;
searching for the weight corresponding to the resolution dimension;
and calculating average similarity according to each local similarity and the corresponding weight.
6. The picture evaluation method according to claim 1, wherein after calculating a similarity between the corrected picture and each standard picture in the standard picture library and obtaining the evaluation score of the picture to be evaluated according to the similarity, further comprising:
If the evaluation score is greater than or equal to a score threshold, confirming the picture to be evaluated as a valid picture;
uploading the effective picture and text information contained in the effective picture to a database, so that a user can check the effective picture and the text information through the database.
7. A picture evaluation device, comprising:
the image acquisition module is used for acquiring an image to be evaluated;
the picture correction module is used for carrying out deformation correction on the picture to be evaluated through a picture correction model to obtain a corrected picture, and comprises the steps of inputting the picture to be evaluated into the picture correction model, instructing the picture correction model to carry out folding and/or curling treatment on each pixel point in the picture to be evaluated to obtain at least one distorted picture, instructing the picture correction model to determine a pixel matrix of the distorted picture according to the picture to be evaluated and the distorted picture, wherein the pixel matrix comprises gray values of each pixel point in the distorted picture, moving directions and moving distances of each pixel point on an x-axis, and moving directions and moving distances of each pixel point on a y-axis, and instructing the picture correction model to input the distorted picture containing the pixel matrix into a neural network containing cavity convolution to obtain the corrected picture;
The score calculation module is used for calculating the similarity between the corrected picture and each standard picture in the standard picture library and obtaining the evaluation score of the picture to be evaluated according to the similarity.
8. A picture evaluation apparatus, characterized by comprising:
one or more processors;
a memory for storing one or more programs;
when the one or more programs are executed by the one or more processors, the one or more processors are caused to implement the picture evaluation method of any one of claims 1-6.
9. A storage medium containing computer executable instructions which, when executed by a computer processor, are for performing the picture evaluation method as claimed in any one of claims 1-6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010038652.XA CN111260623B (en) | 2020-01-14 | 2020-01-14 | Picture evaluation method, device, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010038652.XA CN111260623B (en) | 2020-01-14 | 2020-01-14 | Picture evaluation method, device, equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111260623A CN111260623A (en) | 2020-06-09 |
| CN111260623B true CN111260623B (en) | 2023-07-25 |
Family
ID=70943955
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010038652.XA Active CN111260623B (en) | 2020-01-14 | 2020-01-14 | Picture evaluation method, device, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111260623B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113075208A (en) * | 2021-03-24 | 2021-07-06 | 贵州省草业研究所 | Intelligent cattle and sheep fermented feed quality evaluation method and device based on picture collection |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109118470A (en) * | 2018-06-26 | 2019-01-01 | 腾讯科技（深圳）有限公司 | A kind of image quality evaluating method, device, terminal and server |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| FR2868168B1 (en) * | 2004-03-26 | 2006-09-15 | Cnes Epic | FINE MATCHING OF STEREOSCOPIC IMAGES AND DEDICATED INSTRUMENT WITH A LOW STEREOSCOPIC COEFFICIENT |
| US7848592B2 (en) * | 2006-07-31 | 2010-12-07 | Carestream Health, Inc. | Image fusion for radiation therapy |
| CN106940798A (en) * | 2017-03-08 | 2017-07-11 | 深圳市金立通信设备有限公司 | The modification method and terminal of a kind of Text region |
| CN109598299A (en) * | 2018-11-29 | 2019-04-09 | 微梦创科网络科技（中国）有限公司 | A kind of image similarity determines method, apparatus and electronic equipment |
| CN109858410A (en) * | 2019-01-18 | 2019-06-07 | 深圳壹账通智能科技有限公司 | Service evaluation method, apparatus, equipment and storage medium based on Expression analysis |
| CN109871841B (en) * | 2019-02-14 | 2023-04-21 | 腾讯科技（深圳）有限公司 | Image processing method, device, terminal and storage medium |
-
2020
- 2020-01-14 CN CN202010038652.XA patent/CN111260623B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109118470A (en) * | 2018-06-26 | 2019-01-01 | 腾讯科技（深圳）有限公司 | A kind of image quality evaluating method, device, terminal and server |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111260623A (en) | 2020-06-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11830230B2 (en) | Living body detection method based on facial recognition, and electronic device and storage medium | |
| EP3933693B1 (en) | Object recognition method and device | |
| CN108805047B (en) | Living body detection method and device, electronic equipment and computer readable medium | |
| CN111639744B (en) | Training method and device for student model and electronic equipment | |
| US20220261968A1 (en) | Image optimization method and apparatus, computer storage medium, and electronic device | |
| KR102629380B1 (en) | Method for Distinguishing a Real Three-Dimensional Object from a Two-Dimensional Spoof of the Real Object | |
| CN110009027B (en) | Image comparison method and device, storage medium and electronic device | |
| CN109993102B (en) | Similar face retrieval method, device and storage medium | |
| US10885660B2 (en) | Object detection method, device, system and storage medium | |
| CN109919971B (en) | Image processing method, image processing device, electronic equipment and computer readable storage medium | |
| CN115457531A (en) | Method and device for recognizing text | |
| US20220254134A1 (en) | Region recognition method, apparatus and device, and readable storage medium | |
| CN109446873A (en) | Hand-written script recognition methods, system and terminal device | |
| US20220237917A1 (en) | Video comparison method and apparatus, computer device, and storage medium | |
| CN111985465A (en) | Text recognition method, device, equipment and storage medium | |
| CN108846385A (en) | Based on convolution-deconvolution neural network image recognition, bearing calibration and device | |
| CN116977674A (en) | Image matching method, related device, storage medium and program product | |
| CN112950640A (en) | Video portrait segmentation method and device, electronic equipment and storage medium | |
| CN109564636B (en) | Training one neural network using another neural network | |
| CN111260623B (en) | Picture evaluation method, device, equipment and storage medium | |
| CN111401193A (en) | Method and device for obtaining expression recognition model and expression recognition method and device | |
| CN113822871A (en) | Target detection method and device based on dynamic detection head, storage medium and equipment | |
| KR102634186B1 (en) | Method for verifying the identity of a user by identifying an object by identifying an object in an image containing the user's biometric characteristics and isolating the part of the image containing the biometric characteristics from other parts of the image within an image that has a biometric characteristic of the user and separating a portion of the image comprising the biometric characteristic from other portions of the image) | |
| Chen et al. | Face super resolution based on parent patch prior for VLQ scenarios | |
| CN113128278B (en) | Image 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 |