Image analysis method and apparatus

Abstract

The one or more embodiments of the specification provide an image analysis method and device; the method can comprise: in response to an analysis instruction for a target image, determining a matching degree of the target image with a preset analysis standard in a preset dimension; outputting an image adjustment suggestion in a case where the matching degree is lower than a preset value, the image adjustment suggestion being used to indicate an adjustment processing in the preset dimension for the target image.

Description

TECHNICAL FIELD

[0001] One or more embodiments of the present specification relate to the technical field of computer technology, and particularly relate to an image analysis method and device. BACKGROUND

[0002] With the rapid popularization of cameras, video cameras and various electronic devices with shooting functions, people can obtain various images through electronic devices with shooting functions. People often analyze the quality of images according to personal aesthetic angles, aesthetic habits, etc. Such analysis is often subjective, and different people have large differences in the analysis results of the same image, so it is difficult to obtain a unified image analysis result. SUMMARY

[0003] Therefore, one or more embodiments of the present specification provide an image analysis method and device.

[0004] Specifically, one or more embodiments of the present specification provide technical solutions as follows:

[0005] According to a first aspect of one or more embodiments of the present specification, an image analysis method is provided, comprising:

[0006] In response to an analysis instruction for a target image, determining a matching degree of the target image with a preset analysis standard in a preset dimension;

[0007] Outputting an image adjustment suggestion in a case where the matching degree is lower than a preset value, the image adjustment suggestion being used to indicate an adjustment processing for the target image in the preset dimension.

[0008] According to a second aspect of one or more embodiments of the present specification, a training method of an image analysis model is provided, comprising:

[0009] Obtaining an original sample image and performing color adjustment on the original sample image to generate an expanded sample image;

[0010] Taking the original sample image and the expanded sample image as training samples to train an image analysis model, the image analysis model being used to analyze a matching degree of a target image with a preset analysis standard in a preset dimension.

[0011] According to a third aspect of one or more embodiments of the present specification, an image analysis result display page is provided, comprising:

[0012] A first display area is used to display a target image input by a user;

[0013] a second display area for displaying an analysis result of the target image, the analysis result comprising at least one of: a matching degree of the target image in a preset dimension, an image adjustment suggestion, an analysis score.

[0014] According to a fourth aspect of one or more embodiments of the present specification, an image analysis apparatus is provided, comprising:

[0015] a determination unit configured to determine a matching degree of a target image in a preset dimension with respect to a preset analysis standard in response to an analysis instruction of the target image.

[0016] an output unit configured to output an image adjustment suggestion for indicating an adjustment process of the target image in the preset dimension in a case where the matching degree is lower than a preset value.

[0017] According to a fifth aspect of one or more embodiments of the present specification, a training apparatus of an image analysis model is provided, comprising:

[0018] an expansion unit configured to obtain an original sample image and perform color adjustment on the original sample image to generate an expanded sample image;

[0019] a training unit configured to use the original sample image and the expanded sample image as training samples to train an image analysis model, the image analysis model being configured to analyze a matching degree of a target image in a preset dimension with respect to a preset analysis standard.

[0020] According to a sixth aspect of one or more embodiments of the present specification, an electronic device is provided. The electronic device comprises:

[0021] a processor;

[0022] a memory for storing processor-executable instructions;

[0023] wherein the processor implements the method of the first aspect or the second aspect by running the executable instructions.

[0024] According to a seventh aspect of one or more embodiments of the present specification, a computer-readable storage medium is provided, having stored thereon computer instructions which, when executed by a processor, implement the steps of the method of the first aspect or the second aspect. BRIEF DESCRIPTION OF DRAWINGS

[0025] Figure 1 is an architectural schematic diagram of an image analysis method provided by an exemplary embodiment of the present specification.

[0026] Figure 2 is a flowchart of an image analysis method provided by an exemplary embodiment of the present specification.

[0027] Figure 3 is a flowchart of an image analysis method provided by an example embodiment of the present specification.

[0028] Figure 4 is a schematic diagram of a sample image provided by an example embodiment of the present specification.

[0029] Figures 5-8 is a schematic diagram of an image analysis interface provided by an example embodiment of the present specification.

[0030] Figure 9 is a structural schematic diagram of an electronic device provided by an example embodiment of the present specification.

[0031] Figure 10 is a block diagram of an image analysis device provided by an example embodiment of the present specification.

[0032] Figure 11 is a block diagram of a training device of an image analysis model provided by an example embodiment of the present specification. DETAILED DESCRIPTION

[0033] The example embodiments will be described in detail herein with reference to the drawings. When the following description refers to arrangements in the drawings, identical numbers on different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following example embodiments do not represent all implementations consistent with one or more embodiments of the present specification. Instead, they are merely examples of apparatuses and methods consistent with some aspects of one or more embodiments of the present specification as detailed in the appended claims.

[0034] It should be noted that the steps of the methods in other embodiments are not necessarily performed in the order shown and described in the present specification. In some other embodiments, the steps included in the methods can be more or less than those described in the present specification. Furthermore, a single step described in the present specification can be broken down into multiple steps in other embodiments, and multiple steps described in the present specification can be combined into a single step in other embodiments.

[0035] Figure 1 is an architectural schematic diagram of an image analysis method shown in the present specification. As shown in Figure 1 , it can include a server 11, a network 12, and an electronic device 13.

[0036] The server 11 can be a physical server comprising a single host, or the server 11 can be a virtual server carried by a host cluster. In operation, the server 11 can be configured with an image analysis apparatus, which can be implemented in software and / or hardware, for determining a matching degree of a target image with respect to a preset analysis standard in a preset dimension, and outputting an image adjustment suggestion.

[0037] The electronic device 13 refers to a type of electronic device that can be used by a user. In practice, the user can also use other types of electronic devices such as a mobile phone, a tablet device, a notebook computer, a personal digital assistant (PDA), a wearable device (such as smart glasses, a smart watch, etc.), and the like, without limitation. In operation, the electronic device can display a target image and an analysis result corresponding to the target image.

[0038] The network 12 for interaction between the server 11 and the electronic device 13 can comprise various types of wired or wireless networks.

[0039] Figure 2 is a flowchart of an image analysis method shown in the present specification. As shown in Figure 2 , the method can be applied to a server (such as the server 11 shown in Figure 1 , etc.); the method can comprise the following steps:

[0040] Step 202: in response to an analysis instruction for a target image, determining a matching degree of the target image with respect to a preset analysis standard in a preset dimension.

[0041] In an embodiment, the server can determine the matching degree between the target image and the preset analysis standard in the preset dimension in response to the analysis instruction for the target image, wherein the server can obtain the target image in advance or the server can obtain the corresponding target image after receiving the analysis instruction. The preset maintenance can include an overall score, image aesthetics, subject prominence, layout balance, shooting effect, color harmony, detail condition, and image clarity, etc. The preset dimension can be set according to actual needs, without limitation in the present specification. The matching degree can be a corresponding text label, a corresponding score evaluation, a hierarchical evaluation, or a text label, without limitation in the present specification.

[0042] In an embodiment, the server can extract image features of the target image in each preset dimension, and then the server can input the extracted image features into a pre-trained image analysis model, and the server can obtain a matching degree of the target image in the preset dimension output by the image analysis model. Alternatively, the server can receive an analysis result provided by an external analysis object for the target image, and the analysis result can include a matching degree of the target image in the preset dimension, where the external analysis object can be another server or other electronic devices, etc. distinct from the server, and the image analysis model can be a convolutional neural network (CNN) model, a recurrent neural network (RNN) model, or a generative adversarial network (GAN) model, or other types of neural network models, which are not limited in the present specification.

[0043] In an embodiment, the pre-trained image analysis model can be obtained in the following manner: first, sample features of each sample image in a training sample set in a preset dimension can be extracted respectively, where the sample images have been labeled with actual matching degrees in the preset dimension with a preset analysis standard, and then the sample features can be input into the image analysis model, so that the predicted matching degrees of the sample images in the preset dimension with the preset analysis standard output by the image analysis model can be obtained. Then, the server can adjust the analysis parameters in the image analysis model according to the difference information between the actual matching degrees and the predicted matching degrees of the sample images, and obtain the trained image analysis model. Wherein, the difference information can be the difference between the actual matching degree and the predicted matching degree, the absolute value of the difference between the actual matching degree and the predicted matching degree, or other comparison results between the two, which are not limited in the present specification.

[0044] In an embodiment, the sample image in the training sample set can be labeled with a first actual matching degree in a first preset dimension, and the sample image can be labeled with a second actual matching degree in a second preset dimension associated with the first preset dimension, and the first actual matching degree matches the second actual matching degree. Wherein, the first preset dimension can be an overall score, and the second preset dimension can be a local score, and of course the first preset dimension and the second preset dimension can be set according to actual needs, which are not limited in the present specification.

[0045] In an embodiment, the sample image can include a raw image directly input by a user, and / or a retouched image, where the retouched image can be generated by color adjustment on the raw image. Wherein, the color adjustment on the raw image can include at least one of the following: color adjustment on a background part in the raw image, color adjustment on a subject part in the raw image, color adjustment on a preset part in the raw image, etc. Inputting both the raw image input by the user and the retouched raw image into the image analysis model can enable the image analysis model to be reinforced in the color dimension, which is conducive to improving the accuracy of the prediction matching degree of the image output by the image analysis model subsequently.

[0046] In step 204, an image adjustment suggestion is output in the case where the matching degree is lower than the preset value, and the image adjustment suggestion is used to indicate the adjustment processing on the target image in the preset dimension.

[0047] In an embodiment, in the case where the matching degree of the target image in the preset dimension with the preset analysis standard is lower than the preset value, the server can output an image adjustment suggestion, which can be used to indicate the adjustment processing on the target image in the preset dimension. The server can determine and output each adjustment suggestion corresponding to each matching degree of the target image according to the mapping relationship between the matching degree in the predefined preset dimension and the image adjustment suggestion.

[0048] In an embodiment, the server can also determine an analysis score calculation formula of the target image according to the weight value corresponding to each preset dimension. The server can input the matching degree of the target image in each preset dimension into the analysis score calculation formula, and the server can calculate the analysis score of the target image, and the server can also output and display the calculated analysis score.

[0049] As can be seen from the above technical solutions, in the present specification, the server can determine the matching degree of the target image in the preset dimension with the preset analysis standard in response to the analysis instruction for the target image, so as to determine the analysis result and the image adjustment suggestion of the corresponding target image according to the matching degree, wherein the matching degree of each image with the preset analysis standard is determined in the same preset dimension, which can ensure the objectivity of the matching degree of each image, is conducive to improving the objectivity and accuracy of the obtained image analysis result, and meanwhile, the user can improve the target image according to the image adjustment suggestion, which can improve the efficiency of improving the image and significantly improve the user experience.

[0050] For ease of understanding, the following takes the image analysis of a commodity main image as an example, and the technical solutions of the present specification are described in detail in combination with the following embodiments. Figure 3 The technical solutions of the present specification are described in detail. Figure 3is a flowchart of an image analysis method provided by an exemplary embodiment of the present specification, as shown in Figure 3 The method can include the following steps:

[0051] Step 302, processing the original image.

[0052] In this embodiment, the server can receive the original image input by the user, and the server can perform color expansion on the original image input by the user to obtain a color-adjusted image, wherein the original image and the color-adjusted image can be added to the training sample set. Wherein the color-adjusted image of the original image can also be referred to as an expanded sample image, and the image contained in the training sample set can be referred to as a training sample, and the present specification does not limit this. Of course, the server can also perform shape expansion and other forms of expansion on the original image input by the user to obtain a plurality of adjusted images, and the present specification does not limit this.

[0053] In this embodiment, the server can perform color expansion on the sample image A as shown in Figure 4 The server can perform overall expansion on the sample image A, for example, adjust the overall color of the sample image A to various colors with different RGB values such as red, yellow, blue, green, etc., to obtain a plurality of color-adjusted images. Alternatively, the server can expand the color of a part of the sample image A, for example, the server can expand the color of the background part of the sample image A, and adjust the color of the background part of the sample image A to various colors with different RGB values such as red, yellow, blue, green, etc., to obtain a plurality of color-adjusted images.

[0054] Of course, the server can adjust the overall color of the sample image to obtain a plurality of color-adjusted images and adjust the color of the background part of the sample image to obtain a plurality of color-adjusted images, and the server can also adjust the color of the pre-set part of the sample image to obtain a plurality of color-adjusted images, and the present specification does not limit this. The server can perform color expansion on part of the sample images in the training sample set, or the server can perform color adjustment on all sample images in the training sample set, or the server can not perform color adjustment on the sample images in the training sample set, that is, the training sample set can only contain the original image input by the user, and the present specification does not limit this.

[0055] In this embodiment, the sample images in the training sample set are all labeled with actual matching degrees with the preset analysis standard in the preset dimension. The server can perform color expansion on the sample images in the training sample set with actual matching degrees greater than a preset matching degree threshold, that is, the server performs color expansion on the sample images in the training sample set with higher actual matching degrees, so that a plurality of sample images after color adjustment with different actual matching degrees can be obtained, so that the distribution of the sample images corresponding to different actual matching degrees in the training sample set can be more uniform, and the accuracy of the image analysis model obtained subsequently according to the training sample set can be improved. Generally, the original sample images with higher actual matching degrees are expanded in color, and the actual matching degree of the sample images after color adjustment can be lower than that of the original sample images. Of course, the actual matching degree of the sample images after color adjustment can also be higher than that of the original sample images, which is not limited in this specification.

[0056] In this embodiment, the expansion of the original image input by the user can quickly increase the number of sample images in the training sample set, can avoid the influence of the user inputting a large number of original images on the reading rate of the server, and can improve the accuracy of the image analysis model trained subsequently according to the sample images.

[0057] Step 304, obtaining a training sample set.

[0058] In this embodiment, the training sample set contains a plurality of sample images, and the server can obtain the input training sample set. Each sample image is labeled by a labeler with an actual matching degree with a preset analysis standard in a preset dimension, which can be represented by a text label, a grade, or a score, etc. For example Figure 5As shown, the labeler can label the sample image A with a text label and a score indicating the actual matching degree of the sample image A in the preset dimensions. The preset dimensions include: picture type, subjective score, product appearance, shooting effect, image clarity, subject prominence, layout balance, and color harmony. The product appearance, shooting effect, image clarity, subject prominence, layout balance, and color harmony can be collectively referred to as the score in the objective dimension. Taking the image clarity as an example, the labeler labels according to the following preset analysis standard. If the image is clear and identifiable, the image clarity is determined to be "agree"; if the image is relatively clear and difficult to identify, the image clarity is determined to be "undecided"; if the image is not clear and difficult to identify, the image clarity is determined to be "disagree". In addition, the labeler can perform subjective scoring on the sample image A according to subjective feelings. The interval of the subjective score is 1-5 points. As shown, the subjective score of the sample image A is "2 points". Figure 5

[0059] In this embodiment, the labeler can directly select the "disagree", "undecided", and "agree" labels corresponding to the preset dimensions to label the actual matching degree of the sample image A in the preset dimensions. Of course, the labeler can also directly score or use other ways to label the sample image in the preset dimensions, which is not limited in this specification.

[0060] Step 306, screening the sample image in the training sample set.

[0061] In this embodiment, it is assumed that the "disagree" label is set to "-1 point", the "undecided" label is set to "0 point", and the "agree" label is set to "+1 point". It is assumed that the score of the sample image in each objective dimension is X points. The rule for converting the score to 0-5 points in the server is as follows: objective total score is 1 point: -6≤X≤-4; objective total score is 2 points: -3≤X≤-1; objective total score is 3 points: 0≤X≤3; objective total score is 4 points: 4≤X≤5; objective total score is 5 points: X=6. In addition, it is assumed that when the absolute value Y of the difference between the objective total score and the subjective score of the sample image is less than or equal to 2, it can be considered that the objective total score and the subjective score are matched, and the sample image can be retained in the training sample set; when the absolute value Y of the difference between the objective total score and the subjective score of the sample image is greater than 2, it can be considered that the objective total score and the subjective score are not matched, and the sample image can be deleted from the training sample set.

[0062] ​Then the server can calculate that the score sum value of the sample image A in the objective dimension is "0", and the objective total score corresponding to the sample image A is "3". The subjective score of the sample image A is 2, and the absolute value Y of the difference between the objective total score and the subjective score of the sample image A is 1. The server can determine that the objective total score and the subjective score of the sample image A are matched, and the sample image A is retained in the training sample set.

[0063] In this embodiment, the overall score of the sample image can also be calculated according to the weight value of the objective total score and the weight value of the subjective score of the sample image. It is assumed that the weight value of the objective total score in the training sample set is 60%, and the weight value of the subjective score is 40%. Of course, the weight value can be pre-set by the user or can also be obtained by analyzing the data of each sample image in the training sample set, and the present specification does not limit this. It is assumed that in the case where the absolute value of the difference between the objective total score and the overall score of the sample image is less than or equal to 1 and the absolute value of the difference between the subjective score and the overall score is less than or equal to 1, it can be considered that the objective total score and the overall score, and the subjective score and the overall score are matched, and the sample image can be retained in the training sample set. It is assumed that in the case where the absolute value of the difference between the objective total score and the overall score of the sample image is greater than 1, it can be considered that the objective total score and the overall score are not matched, and the sample image can be deleted from the training sample set. It is assumed that in the case where the absolute value of the difference between the subjective score and the overall score of the sample image is greater than 1, it can be considered that the subjective score and the overall score are not matched, and the sample image can be deleted from the training sample set.

[0064] Then the server can calculate that the score sum value of the sample image A in the objective dimension is "0", and the objective total score corresponding to the sample image A is "3". The subjective score of the sample image A is 2, and the absolute value Y of the difference between the objective total score and the subjective score of the sample image A is 1. The server can determine that the objective total score and the subjective score of the sample image A are matched, and the sample image A is retained in the training sample set. The server can screen the sample image A according to any of the above methods, and the present specification does not limit this.

[0065] In this embodiment, the sample images in the training sample set are screened according to the difference between the objective total score and the subjective score of the sample image in the objective dimension. This can avoid the training sample set containing sample images with inaccurate labeling by the labeler, can ensure the accuracy of the actual matching degree labeled on the sample image, can improve the quality of the sample images in the training sample set, and can improve the accuracy of the image analysis model trained subsequently according to the screened sample images.

[0066] Step 308: Extract the image features of the screened sample image, and train to obtain an image analysis model.

[0067] In this embodiment, the server can extract the sample features of the screened sample image in the preset dimensions of picture type, subjective score, product appearance, shooting effect, image definition, subject prominence, layout balance, and color harmony, respectively. The server can input the extracted sample features into the pre-trained image analysis model, and obtain the predicted matching degree of the sample image in the preset dimensions output by the image analysis model. The server can compare the actual matching degree labeled on the sample image with the predicted matching degree, and adjust the corresponding analysis parameters in the image analysis model according to the difference information between the actual matching degree and the predicted matching degree. The image analysis model can be a convolutional neural network (CNN) model, a recurrent neural network (RNN) model, or a generative adversarial network (GAN) model, or other types of neural network models, which are not limited in the specification. As shown below Figure 6 The predicted matching degree of the sample image A after inputting into the image analysis model is shown. Then the server can obtain the predicted matching degree of the sample image A in the subjective score dimension as "1", which is different from the actual matching degree "2"; the predicted matching degree of the sample image A in the shooting effect dimension is "uncertain", which is different from the actual matching degree "oppose"; and the predicted matching degree of the sample image A in the color harmony dimension is "oppose", which is different from the actual matching degree "agree". The server adjusts the corresponding analysis parameters in the image analysis model according to the difference information between the actual matching degree and the predicted matching degree of each sample image in the training sample set. For example, the difference value of 1 between the predicted matching degree and the actual matching degree of the subjective score of the sample image A is used to adjust the corresponding analysis parameters in the image analysis model, the difference value of 2 between "-1" corresponding to the predicted matching degree "oppose" of the sample image A in the color harmony and "+1" corresponding to the actual matching degree "agree" is used to adjust the corresponding analysis parameters in the image analysis model, and so on. After adjustment, the optimized image analysis model can be obtained, which can make the trained image analysis model more accurately predict the matching degree of the image in the preset dimensions.

[0068] Step 310, obtaining a target image B.

[0069] Step 312, determining the matching degree of the target image B in the preset dimensions.

[0070] In this embodiment, the server can obtain a target image B that needs to be analyzed, and the server can extract image features of the target image B in the eight dimensions of the preset dimension picture type, subjective score, product appearance, shooting effect, image definition, subject prominence, layout balance, and color harmony in response to an analysis instruction for the target image B. The server can input the extracted image features into the trained image analysis model to obtain the matching degrees of the target image in each preset dimension output by the image analysis model, as shown in Figure 7 Then the server can determine that the subjective score of the target image B is "3 points", the score of the objective dimension is "1 point", and the corresponding objective total score is "3 points".

[0071] Step 314, output image adjustment suggestions.

[0072] In this embodiment, it is assumed that the preset value of the target image B in each preset dimension is lower than 0 points, and image adjustment suggestions for the target image need to be output. Of course, the preset value can be set according to actual needs, and the present specification does not limit this. As shown in Figure 7 The server can determine that the preset value of the target image B in the preset dimensions, i.e., the shooting effect and the color harmony, is lower than 0 points. Then the server B can determine the image adjustment suggestions corresponding to each matching degree of the output target image according to the mapping relationship between the value in each dimension and the image adjustment suggestions, for example, the output image adjustment suggestions are "your product is more prominent, but the image shooting effect is general, the color harmony is poor, use more matched background color, or add some detail modification", and the like. Of course, the server can also determine the image adjustment suggestions corresponding to each matching degree of the target image according to the natural language dialogue system and corpus commonly used in intelligent design, and the present specification does not limit this. Then the user can quickly understand the corresponding image adjustment suggestions and improve the image.

[0073] Step 316, determine the analysis score calculation formula of the target image.

[0074] Step 318, obtain the analysis score of the target image B.

[0075] In this embodiment, the server can obtain the analysis score calculation formula of the target image according to the preset weight value of the objective total score of 60% and the weight value of the subjective score of 40%. Of course, the weight value can be preset by the user or can also be obtained by data analysis on the data of each sample image in the training sample set, and the present specification does not limit this. If the objective total score of the target image B is 3 points and the subjective score is 3 points, then the server can calculate the analysis score of the target image B to be 3 points.

[0076] In step 320, the image analysis result of the target image B is output.

[0077] In this embodiment, the server can send the image analysis result of the target image B to the electronic device for display, and the display page of the corresponding image analysis result can be as shown in Figure 8 The image analysis result can include a first display area 801 and a second display area 802. The first display area 801 is used to display the target image B input by the user, and the second display area 802 is used to display the analysis result of the target image B. The analysis result can include the matching degree of the target image B in the preset dimensions of picture type, subjective score, product appearance, shooting effect, image clarity, subject prominence, layout balance, and color harmony, as well as the corresponding image adjustment suggestion and the analysis score of the target image B. Of course, the image analysis result display page can also set the content to be displayed according to actual needs, for example, only display the page as shown in Figure 7 The present specification does not limit this.

[0078] From the above technical solutions, it can be seen that the server in the present specification can determine the matching degree of the target image with the preset analysis standard in the preset dimensions in response to the analysis instruction of the target image, so as to determine the analysis result and the image adjustment suggestion of the corresponding target image according to the matching degree. Each image determines the matching degree with the preset analysis standard in the same preset dimensions, which can ensure the objectivity of the matching degree of each image, is conducive to improving the objectivity and accuracy of the obtained image analysis result, and at the same time, the user can improve the target image according to the image adjustment suggestion, which can improve the efficiency of improving the image and significantly improve the user experience.

[0079] Figure 9 is a schematic structural diagram of a device provided by an exemplary embodiment. Please refer to Figure 9At the hardware level, the device comprises a processor 902, an internal bus 904, a network interface 906, a memory 908, and a non-volatile memory 910, and can also comprise other hardware required by the business. The processor 902 reads the corresponding computer program from the non-volatile memory 910 into the memory 908 and then runs, and forms an image analysis apparatus at the logical level. Of course, in addition to the software implementation, one or more embodiments of the present specification do not exclude other implementations, such as logic devices or a combination of software and hardware, and the like, that is, the execution subject of the following processing flow is not limited to each logical unit, but can also be hardware or logic devices.

[0080] Please refer to Figure 10 In the software implementation, the image analysis apparatus can comprise a determination unit 1002 and an output unit 1004. Wherein:

[0081] The determination unit 1002 is configured to determine the matching degree of the target image with the preset analysis standard in the preset dimension in response to an analysis instruction for the target image.

[0082] The output unit 1004 is configured to output an image adjustment suggestion in the case that the matching degree is lower than a preset value, and the image adjustment suggestion is used to indicate an adjustment processing in the preset dimension for the target image.

[0083] Optionally, the determination unit 1002 is specifically configured to:

[0084] extract image features of the target image in each preset dimension, input the extracted image features into a pre-trained image analysis model, and obtain the matching degree of the target image in the preset dimension output by the image analysis model; or

[0085] receive an analysis result provided by an external analysis object for the target image, and the analysis result comprises the matching degree of the target image in the preset dimension.

[0086] Optionally, the determination unit 1002 is specifically configured to:

[0087] extract sample features of each sample image in the preset dimension in a training sample set, wherein the sample image is labeled with an actual matching degree of the preset analysis standard in the preset dimension;

[0088] obtain the predicted matching degree of the sample image with the preset analysis standard in the preset dimension output by the image analysis model by inputting the sample features into the image analysis model;

[0089] Adjust an analysis parameter in the image analysis model according to difference information between the actual matching degree and the predicted matching degree, to obtain a trained image analysis model.

[0090] Optionally, the sample image is labeled with a first actual matching degree in a first preset dimension and a second actual matching degree in a second preset dimension associated with the first preset dimension, and the first actual matching degree matches the second actual matching degree.

[0091] Optionally, the first preset dimension is an overall score, and the second preset dimension is a local score.

[0092] Optionally, the sample image includes:

[0093] a user-input original image; and / or,

[0094] a color-adjusted image generated by color adjustment on the original image.

[0095] Optionally, the determining unit 1002 is configured to include at least one of the following:

[0096] color adjustment on a background part in the original image, color adjustment on a subject part in the original image, and color adjustment on a pre-set part in the original image.

[0097] Optionally, the output unit 1004 is specifically configured to:

[0098] output an image adjustment suggestion corresponding to the matching degree of the target image according to a mapping relationship between the matching degree on the predefined preset dimension and the image adjustment suggestion.

[0099] Optionally, the image analysis model training apparatus further includes:

[0100] a formula determining unit 1006 configured to determine an analysis score calculation formula of a target image according to a weight value corresponding to each preset dimension;

[0101] a calculation unit 1008 configured to input the matching degree of the target image on each preset dimension into the analysis score calculation formula, to calculate and output an analysis score of the target image.

[0102] Optionally, the preset dimension includes at least one of the following: an overall score, an image aesthetic degree, an image clarity, a subject prominence degree, and a layout balance condition.

[0103] Please refer to Figure 11 In a software implementation, the image analysis model training apparatus can include an expanding unit 1102 and a training unit 1104. Wherein:

[0104] The expansion unit 1102 is configured to obtain an original sample image and perform color adjustment on the original sample image to generate an expanded sample image.

[0105] The training unit 1104 is configured to use the original sample image and the expanded sample image as training samples to train an image analysis model, the image analysis model being configured to analyze a matching degree of a target image in a preset dimension with a preset analysis standard.

[0106] Optionally, the training unit 1104 is specifically configured to:

[0107] extract a sample feature of the training sample in the preset dimension, the training sample being labeled with an actual matching degree of the training sample in the preset dimension with the preset analysis standard;

[0108] obtain a predicted matching degree of the training sample in the preset dimension with the preset analysis standard output by the image analysis model by inputting the sample feature into the image analysis model;

[0109] adjust an analysis parameter in the image analysis model according to difference information between the actual matching degree and the predicted matching degree to obtain a trained image analysis model.

[0110] Optionally, the expansion unit 1102 is configured to perform at least one of the following expansion operations:

[0111] performing color adjustment on a background part of the original sample image, performing color adjustment on a subject part of the original sample image, or performing color adjustment on a preset part of the original sample image.

[0112] The system, device, module or unit illustrated in the above embodiments can be specifically implemented by a computer chip or entity, or by a product with certain functions. A typical implementation device is a computer, and the computer can be specifically a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an e-mail device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0113] In a typical configuration, a computer includes one or more processors (CPU), input / output interface, network interface, and memory.

[0114] The memory can include a non-persistent memory in a computer readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash memory (flash RAM). The memory is an example of a computer readable medium.

[0115] Computer-readable media includes permanent and non-permanent, movable and non-movable media that can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette, magnetic disk storage, quantum memory, graphene-based storage medium or other magnetic storage device, or any other non-transmission medium that can be used to store information accessible by a computing device. According to the definition herein, computer-readable media does not include transitory media such as modulated data signals and carriers.

[0116] It should also be noted that the terms "comprising", "containing", or any other variant thereof are intended to cover non-exclusive inclusion, such that processes, methods, articles or devices that include a series of elements not only include those elements, but also include other elements not explicitly listed or inherent to such processes, methods, articles or devices. Without more limitations, the element defined by the statement "comprising a" does not exclude the presence of additional identical elements in the process, method, article or device comprising the element.

[0117] The above describes specific embodiments of the present specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in an order different from the order in the embodiments and still achieve the desired result. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are possible or can be advantageous.

[0118] The terms used in one or more embodiments of the present specification are only for the purpose of describing specific embodiments and are not intended to limit one or more embodiments of the present specification. The singular forms "a", "an" and "the" used in one or more embodiments of the present specification and the appended claims are also intended to include plural forms, unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used herein means and includes any or all possible combinations of one or more associated listed items.

[0119] It should be understood that, although the terms first, second, third, etc. can be used herein to describe various information, the information should not be limited to these terms. These terms are only used to differentiate one piece of information from another. For example, a first information can also be termed a second information, and, similarly, a second information can also be termed a first information, without departing from the scope of the one or more embodiments. The word "if' can be interpreted to mean "upon" or "when" or "in response to the determination" depending on the context.

[0120] The foregoing description of the one or more embodiments of the present disclosure is by way of example only, and other embodiments of the present disclosure are within the scope of the present disclosure as delineated by the claims, with equivalents of the claims to be included therein.

Claims

1. An image analysis method characterized by, The method comprises: in response to an analysis instruction for a target image, determining a matching degree of the target image with a preset analysis standard in a preset dimension; outputting an image adjustment suggestion in a case where the matching degree is lower than a preset value, the image adjustment suggestion being used for indicating an adjustment processing in the preset dimension for the target image; the determining of the matching degree of the target image with the preset analysis standard in the preset dimension comprises: extracting image features of the target image in each preset dimension, and inputting the extracted image features into a pre-trained image analysis model to obtain a matching degree of the target image in the preset dimension output by the image analysis model.

2. The method of claim 1, wherein, The image analysis model is obtained by the following training method: respectively extracting sample features of each sample image in a preset dimension in a training sample set, wherein the sample image is labeled with an actual matching degree with the preset analysis standard in the preset dimension; obtaining a predicted matching degree of the sample image with the preset analysis standard in the preset dimension output by the image analysis model by inputting the sample features into the image analysis model; adjusting analysis parameters in the image analysis model according to difference information between the actual matching degree and the predicted matching degree to obtain a trained image analysis model.

3. The method of claim 2, wherein, The sample image is labeled with a first actual matching degree in a first preset dimension and a second actual matching degree in a second preset dimension associated with the first preset dimension, and the first actual matching degree matches the second actual matching degree.

4. The method of claim 3, wherein, The first preset dimension is an overall score, and the second preset dimension is a local score.

5. The method of claim 2, wherein, The sample image comprises: a user-input original image; and / or, a color-adjusted image generated by color adjustment on the original image.

6. The method of claim 5, wherein, The color adjustment on the original image comprises at least one of the following: color adjustment on a background part of the original image, color adjustment on a subject part of the original image, and color adjustment on a pre-set part of the original image.

7. The method of claim 1, wherein, The outputting of the image adjustment suggestion comprises: outputting an image adjustment suggestion corresponding to the matching degree of the target image according to a mapping relationship between the matching degree in the predefined preset dimension and the image adjustment suggestion.

8. The method of claim 1, wherein, Further comprising: determining an analysis score calculation formula of the target image according to a weight value corresponding to each preset dimension; inputting the matching degree of the target image in each preset dimension into the analysis score calculation formula to calculate and output an analysis score of the target image.

9. The method of claim 1, wherein, The preset dimension comprises at least one of the following: overall score, image aesthetic degree, image clarity, subject prominence degree, and layout balance condition. 10.A method for training an image analysis model, characterized in that, The method comprises: obtaining an original sample image and performing color adjustment on the original sample image to generate an expanded sample image; using the original sample image and the expanded sample image as training samples to train an image analysis model, the image analysis model being used for analyzing a matching degree of a target image with a preset analysis standard in a preset dimension; The matching degree of the analysis target image on the preset dimension with the preset analysis standard comprises: Extracting image features of the target image on each preset dimension, and inputting the extracted image features into a pre-trained image analysis model to obtain the matching degree of the target image on the preset dimension output by the image analysis model.

11. The method of claim 10, wherein, The training of the image analysis model comprises: Extracting sample features of the training sample on the preset dimension, the training sample being labeled with an actual matching degree of the training sample on the preset dimension with the preset analysis standard; Obtaining a predicted matching degree of the training sample on the preset dimension with the preset analysis standard output by the image analysis model by inputting the sample features into the image analysis model; Adjusting analysis parameters in the image analysis model according to difference information between the actual matching degree and the predicted matching degree to obtain a trained image analysis model.

12. The method of claim 10, wherein, The color adjustment of the original sample image comprises at least one of the following: Color adjustment of a background part of the original sample image, color adjustment of a subject part of the original sample image, and color adjustment of a pre-set part of the original sample image.

13. An image analysis result presentation page, characterized by, Comprise: A first display area for displaying a target image input by a user; A second display area for displaying an analysis result of the target image, the analysis result comprising at least one of the following: a matching degree of the target image on a preset dimension, an image adjustment suggestion, and an analysis score; the matching degree on the preset dimension being obtained by extracting image features of the target image on each preset dimension, inputting the extracted image features into a pre-trained image analysis model, and outputting by the image analysis model.

14. An image analysis apparatus characterized by comprising: Comprise: A determination unit configured to determine a matching degree of a target image on a preset dimension with a preset analysis standard in response to an analysis instruction of the target image; An output unit configured to output an image adjustment suggestion in a case where the matching degree is lower than a preset value, the image adjustment suggestion being used to indicate an adjustment processing of the target image on the preset dimension; The determination unit is specifically configured to: Extract image features of the target image on each preset dimension, and input the extracted image features into a pre-trained image analysis model to obtain the matching degree of the target image on the preset dimension output by the image analysis model.

15. A training device for an image analysis model, characterized in that, Comprise: An expansion unit configured to obtain an original sample image and perform color adjustment on the original sample image to generate an expanded sample image; A training unit configured to use the original sample image and the expanded sample image as training samples to train an image analysis model, the image analysis model being used to analyze a matching degree of a target image on a preset dimension with a preset analysis standard; The training unit is specifically configured to: The matching degree on the preset dimension is obtained by extracting image features of the target image on each preset dimension, inputting the extracted image features into a pre-trained image analysis model, and outputting by the image analysis model.

16. An electronic device, comprising: Comprise: A processor; a memory for storing processor-executable instructions; wherein the processor implements the method of any of claims 1-12 by executing the executable instructions.

17. A computer readable storage medium having stored thereon computer instructions, wherein, The instructions, when executed by a processor, perform steps of the method of any of claims 1-12.

Images