Image effect evaluation method and device, storage medium and computer equipment

Abstract

The invention discloses an image effect evaluation method and device, a storage medium and computer equipment. Comprising the following steps: obtaining delivery attribute data of a target image in response to an effect evaluation instruction of the target image; extracting image visual features of the target image and delivery attribute features of the delivery attribute data, and performing fusion processing on the image visual features and the delivery attribute features to obtain effect evaluation fusion features; the effect evaluation fusion feature is input into a preset effect evaluation model for image effect evaluation, an effect evaluation result of the target image is obtained, the preset effect evaluation model is obtained through training based on a sample data set with an effect label in advance, and the preset effect evaluation model is obtained through training based on the sample data set with the effect label. And then the effect evaluation results of the multiple images can be utilized to guide the generation and optimization of the images to be put. The method is suitable for image effect evaluation and image scene generation.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular, to a method, device, storage medium, and computer device for evaluating the effect of an image. Background Art

[0002] Currently, on e-commerce platforms, product information is displayed through images. In order to understand the performance of images in conveying information, stimulating users' desire to explore, etc., and thus optimize image design and improve image effects, it is necessary to analyze the effects of images.

[0003] Currently, the effects of images are usually analyzed manually. However, this way of effect analysis is time-consuming and laborious, and due to the uneven technical levels of analysts, there will be cases of incorrect analysis of image effects. Summary of the Invention

[0004] The present invention provides a method, device, storage medium, and computer device for evaluating the effect of an image, mainly capable of improving the efficiency and accuracy of image effect evaluation.

[0005] According to a first aspect of the present invention, there is provided a method for evaluating the effect of an image, including:

[0006] In response to an effect evaluation instruction for a target image, obtaining the placement attribute data of the target image;

[0007] Extracting the image visual features of the target image and the placement attribute features of the placement attribute data, and performing a fusion process on the image visual features and the placement attribute features to obtain an effect evaluation fusion feature;

[0008] Inputting the effect evaluation fusion feature into a preset effect evaluation model for image effect evaluation to obtain an effect evaluation result of the target image, where the preset effect evaluation model is pre-trained based on a sample data set with effect labels.

[0009] Optionally, the performing a fusion process on the image visual features and the placement attribute features to obtain an effect evaluation fusion feature includes:

[0010] Inputting the target image into a preset aesthetic scoring model for aesthetic scoring, and determining an aesthetic feature vector corresponding to the aesthetic score;

[0011] Performing a fusion process on the image visual features, the aesthetic features, and the placement attribute features to obtain an effect evaluation fusion feature, where the method for performing a fusion process on the image visual features, the aesthetic features, and the placement attribute features to obtain an effect evaluation fusion feature includes:

[0012] Taking the visual features of the image as the visual query vector, taking the concatenated features of the aesthetic features and the placement attribute features as the first key-value vector, and interacting the visual query vector with the first key-value vector to obtain visual context features;

[0013] Taking the aesthetic features as the aesthetic query vector, taking the concatenated features of the image visual features and the placement attribute features as the second key-value vector, and interacting the aesthetic query vector with the second key-value vector to obtain aesthetic context features;

[0014] Taking the placement attribute features as the placement query vector, taking the concatenated features of the image visual features and the aesthetic features as the third key-value vector, and interacting the placement query vector with the third key-value vector to obtain placement context features;

[0015] Aggregating the visual context features, the aesthetic context features, and the placement context features to obtain the effect evaluation fusion features.

[0016] Optionally, inputting the effect evaluation fusion features into a preset effect evaluation model for image effect evaluation to obtain the effect evaluation result of the target image, including:

[0017] Inputting the effect evaluation fusion features into a preset effect evaluation model for image effect evaluation to obtain the effect evaluation result of the target image in the form of a visual chart display;

[0018] Determining the target effects with a contribution degree greater than a preset threshold in the effect evaluation result, determining the evaluation regions corresponding to the target effects in the target image, and marking the evaluation regions and displaying the contribution degree.

[0019] Optionally, after inputting the effect evaluation fusion features into a preset effect evaluation model for image effect evaluation to obtain the effect evaluation result of the target image, the method further includes:

[0020] Responding to an image generation instruction of a to-be-generated image, and obtaining the image generation requirement information of the to-be-generated image;

[0021] Based on the image generation requirement information, extracting image effect evaluation keywords in an effect evaluation database, and mapping the extracted image effect evaluation keywords to image generation guiding elements that meet the image generation guiding format by using a preset mapping rule, where the effect evaluation database is constructed based on the effect evaluation results of multiple images;

[0022] Determine the image generation prompt word template for the to-be-generated image, and fill the product theme information of the to-be-generated image and the image generation guiding elements into the image generation prompt word template to obtain an image generation prompt word;

[0023] Input the image generation prompt word into a preset image generation model for image generation to obtain a newly generated image, where the preset image generation model is pre-trained based on a sample data set with image labels.

[0024] Optionally, the extracting of image effect evaluation keywords in the effect evaluation database based on the image generation requirement information includes:

[0025] Extract initial image effect evaluation keywords whose image effect evaluation results meet preset conditions in the effect evaluation database based on the image generation requirement information;

[0026] Determine the style combination probability of the image style corresponding to each of the initial image effect evaluation keywords, and determine the initial image effect evaluation keywords with the style combination probability greater than a preset probability threshold among the initial image effect evaluation keywords as the image effect evaluation keywords.

[0027] Optionally, the determining the style combination probability of the image style corresponding to each of the initial image effect evaluation keywords includes:

[0028] Divide each of the initial image effect evaluation keywords into multiple evaluation keyword groups according to the images to which they belong, and sort each evaluation keyword in each evaluation keyword group according to a preset sorting rule to obtain an evaluation keyword sequence corresponding to each evaluation keyword group;

[0029] Statistically analyze the transition probability between adjacent evaluation keywords in each evaluation keyword sequence, and determine the average value of the respective transition probabilities corresponding to each evaluation keyword group as the style combination probability of the image style corresponding to each initial image effect evaluation keyword in each evaluation keyword group.

[0030] Optionally, the inputting the image generation prompt word into a preset image generation model for image generation to obtain a newly generated image includes:

[0031] Determine the reference image of the to-be-generated image, and determine the initial seed value in the process of the preset image generation model generating an image;

[0032] Input the image generation prompt word and the initial seed value into the preset image generation model for image generation to obtain an initial image;

[0033] Determine the fluctuation range of the initial seed value based on the similarity between the initial image and the reference image, determine multiple seed values within the fluctuation range, and input each of the seed values and the image generation prompt into the preset image generation model for image generation to obtain candidate images corresponding to each of the seed values;

[0034] Respectively determine the similarity between each candidate image and the reference image, and use the candidate image corresponding to the maximum similarity as the newly generated image.

[0035] According to the second aspect of the present invention, there is provided an image effect evaluation device, including:

[0036] An acquisition unit, configured to acquire the placement attribute data of the target image in response to an image effect evaluation instruction for the target image;

[0037] A fusion unit, configured to extract the image visual feature of the target image and the placement attribute feature of the placement attribute data, and perform a fusion process on the image visual feature and the placement attribute feature to obtain an effect evaluation fusion feature;

[0038] An effect evaluation unit, configured to input the effect evaluation fusion feature into a preset effect evaluation model for image effect evaluation to obtain an effect evaluation result of the target image, where the preset effect evaluation model is pre-trained based on a sample data set with effect labels.

[0039] According to the third aspect of the present invention, there is provided a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the above-mentioned image effect evaluation method is implemented.

[0040] According to the fourth aspect of the present invention, there is provided a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, and when the processor executes the program, the above-mentioned image effect evaluation method is implemented.

[0041] An image effect evaluation method, device, storage medium and computer device provided by the present invention, compared with the current method of manually analyzing image effects, the present invention obtains the placement attribute data of the target image in response to an image effect evaluation instruction for the target image; and extracts the image visual features of the target image and the placement attribute features of the placement attribute data, and performs a fusion process on the image visual features and the placement attribute features to obtain an effect evaluation fusion feature; finally, inputs the effect evaluation fusion feature into a preset effect evaluation model for image effect evaluation to obtain an effect evaluation result of the target image, wherein the preset effect evaluation model is pre-trained based on a sample data set with effect labels. Thus, by using the preset effect evaluation model to analyze the image visual features of the target image and the placement attribute features of the placement attribute data to determine the effect evaluation result of the target image, manual participation is avoided, thereby improving the efficiency and accuracy of image effect evaluation and ensuring the degree of automation of image effect evaluation. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The illustrative embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0043] Figure 1 shows a flowchart of an image effect evaluation method provided by an embodiment of the present invention;

[0044] Figure 2 shows a flowchart of another image effect evaluation method provided by an embodiment of the present invention;

[0045] Figure 3 shows a schematic structural diagram of an image effect evaluation device provided by an embodiment of the present invention;

[0046] Figure 4 shows a schematic structural diagram of another image effect evaluation device provided by an embodiment of the present invention;

[0047] Figure 5 shows a schematic physical structure diagram of a computer device provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0048] The present invention will be described in detail below with reference to the drawings and in conjunction with the embodiments. It should be noted that, without conflict, the embodiments in this application and the features in the embodiments can be combined with each other.

[0049] Currently, the method of manually analyzing image effects is time-consuming and laborious, and if the analyst is negligent, it may lead to incorrect analysis of image effects.

[0050] To solve the above problems, an embodiment of the present invention provides a method for evaluating the effect of an image, as Figure 1 shown, the method includes:

[0051] 101. In response to an effect evaluation instruction for a target image, obtain the placement attribute data of the target image.

[0052] Among them, the target image can be any image, such as an advertisement image of food, daily necessities, clothing, etc.; the target image can be a display image such as an image of a target commodity; the placement attribute data refers to the total cost, exposure rate, click-through rate, transaction order amount, etc. of the target image during the placement process.

[0053] For the embodiment of the present invention, when it is necessary to analyze the image effect of the target image, an effect evaluation instruction for the target image is triggered. When the effect evaluation instruction is received, the placement attribute data such as the exposure rate and click-through rate of the target image is obtained in the database.

[0054] 102. Extract the image visual features of the target image and the placement attribute features of the placement attribute data, and perform a fusion process on the image visual features and the placement attribute features to obtain an effect evaluation fusion feature.

[0055] Among them, the image visual features refer to the features of information such as the color, texture, and object layout of the target image. For the embodiment of the present invention, in order to capture the complementary information between different features and thus improve the accuracy of effect evaluation, first, a fusion process needs to be performed on the image visual features and the placement attribute features. Based on this, the specific fusion process includes: taking the image visual features as an example, constructing a feature matrix from each image visual feature; determining the element mean of each element in the feature matrix, and subtracting each element in the feature matrix from the element mean to obtain a centered feature matrix; determining the covariance matrix corresponding to the centered feature matrix; performing eigenvalue decomposition on the covariance matrix to obtain matrix eigenvalues and matrix eigenvectors; based on the magnitudes of the matrix eigenvalues, selecting a preset number of matrix eigenvectors from the matrix eigenvectors, and determining the preset number of matrix eigenvectors as the principal component features in the image visual features. Similarly, the principal component features in the placement attribute features can be determined in the above manner; performing a feature-level cross process on the principal component features in the image visual features and the principal component features in the placement attribute features to obtain an effect evaluation fusion feature.

[0056] Among them, the preset number is set according to actual needs. Specifically, if one feature in the image visual features is [1 1 2 4 2] and another feature is [1 3 3 4 4], the feature matrix composed of the above two features is as follows:

[0057]

[0058] Among them, T represents the feature matrix. Then, the element mean corresponding to the elements in the first row of the feature matrix is determined, and each element in the first row is subtracted from this mean to obtain the subtraction results corresponding to each element in the first row. At the same time, the element mean corresponding to each element in the second row is determined, and each element in the second row is subtracted from the corresponding element mean to obtain the subtraction results corresponding to each element in the second row. Finally, the subtraction results corresponding to each element in the first row and the subtraction results corresponding to each element in the second row are used to form the centralized feature matrix as follows:

[0059]

[0060] Among them, T z represents the centralized feature matrix. Then, the covariance between the elements in each row of the centralized feature matrix is calculated. For example, the covariance between the elements in the first row Cov(X1, X1) = [(-1) 2 + (-1) 2 + (0) 2 + (2) 2 + (0) 2 / (5 - 1) = 1.5, the covariance Cov(X1, X2) between the first row elements and the second row elements is [(-1)×(-2) + (-1)×0 + 0×0 + 2×1 + 0×1] / (5 - 1) = 1. Thus, the covariance between all row elements can be calculated. Then, a covariance matrix is formed by all the covariances, and eigenvalue decomposition is performed on this covariance matrix. The specific decomposition method is as follows: First, determine the eigenvector group of the covariance matrix, and calculate the corresponding eigenvalues according to the eigenvector group. For example, if the obtained eigenvector group is k groups, then the corresponding eigenvalues are k, and each eigenvalue has its corresponding eigenvector. Then, sort the k eigenvalues in descending order to obtain the sorted eigenvalues. Then, select the top n (predetermined quantity) eigenvalues from the sorted eigenvalues, and determine the eigenvectors corresponding to the top n eigenvalues as the principal component features. Thus, the principal component features in the delivery attribute features can be determined in the above manner. For example, if the principal component features in the image visual features are (a1, a2) and the principal component features in the delivery attribute features are (b1, b2), then the method for performing feature-level cross-processing on the two principal component features is: After performing the Hadamard product on all the elements between the two features, perform a convolution transformation under a certain weight w to obtain the effect evaluation fusion feature f(w*(a1*b1, a2*b2)). It should be noted that the above examples are only illustrative and do not limit the embodiments of the present application. Thus, the embodiments of the present invention first determine the principal component features in the image visual features and the delivery attribute features. Since the principal components capture the main variation information in the features and ignore the secondary information and noise, the generalization ability of the model can be improved, and thus the accuracy of the image effect evaluation can be improved. At the same time, the embodiments of the present invention can automatically or explicitly combine different features by performing fusion processing on multiple features, thereby generating new feature combinations. These combined features may contain complex non-linear relationships between the original features, enabling the model to capture more refined and rich information in the data, that is, to make full use of the relationships between various data, extract more implicit features, make more full use of the data, and make the subsequent obtained image effect evaluation results more accurate, meeting the requirements of the actual application scenario.

[0061] 103. Input the effect evaluation fusion feature into a preset effect evaluation model for image effect evaluation to obtain the effect evaluation result of the target image, where the preset effect evaluation model is pre-trained based on a sample data set with effect labels.

[0062] For the embodiments of the present invention, in order to improve the effect evaluation accuracy of the preset effect evaluation model, it is first necessary to train and construct the preset effect evaluation model. Based on this, the method includes: obtaining a preset initial effect evaluation model and obtaining a sample data set, where the sample data set includes the placement attribute data of sample images with effect evaluation labels; dividing the sample data set into training data and test data, training the preset initial effect evaluation model with the training data, and testing the trained preset initial effect evaluation model with the test data. Finally, the trained preset initial effect evaluation model that meets the test conditions is used as the preset effect evaluation model, where the test conditions are such as the number of training times reaching the requirement or the accuracy of the trained model reaching the requirement, etc. Further, input the effect evaluation fusion features into the preset effect evaluation model for image effect evaluation, and output the effect evaluation result of the target image through the preset effect evaluation model. Among them, the effect evaluation result includes the evaluation result of visual attraction, the product information evaluation result, etc. The visual attraction evaluation result is such as: vivid colors: the image uses bright red and yellow, these colors are very eye-catching and easy to stand out among many images; clear product display: the product packaging and the real object show in the image are clearly displayed, allowing customers to clearly see the appearance and form of the product at a glance; attractive close-up of the real object: a close-up of the plate of dried tofu, with a bright red color, may be seasoned and looks very attractive, which can stimulate the appetite. The product information evaluation result is such as: prominent brand logo: the brand logo is clearly visible, which helps with brand recognition and building trust; prominent product name and display point: the name A, which implies good luck, is easy to attract consumers. At the same time, "value-for-money assorted pack" also indicates a diverse product combination and high cost performance; nutritious and healthy: "nutritious and healthy, loved by the whole family" highlights the healthy attribute of the product, meeting the pursuit of modern consumers for a healthy diet. The embodiments of the present invention use a model to evaluate the effect of images, avoiding manual participation, thereby improving the evaluation efficiency and evaluation accuracy of image effects. Further, according to the effect evaluation results of multiple images, it is possible to guide the generation or optimization of the images to be placed next.

[0063] In another embodiment of the present invention, in order to improve the accuracy of image effect evaluation, the effect evaluation fusion feature can be first input into a preset evaluation index prediction model for evaluation index prediction, and then the evaluation index and the effect evaluation fusion feature are jointly input into a large model, and the large model is used to evaluate the image for the evaluation index to obtain the evaluation results under each evaluation index. Among them, the evaluation indexes include visual attractiveness evaluation index, product name evaluation index, promotion information evaluation index, product highlight evaluation index, etc. For example, the effect evaluation result under the product name evaluation index can be that the meaning of the product name is auspicious and it is easy to attract consumers; the effect evaluation result under the promotion information evaluation index can be that the image indicates "super value assorted package", indicating that the product combination is diverse and cost-effective; the effect evaluation result under the product highlight evaluation index can be that the image indicates "nutritious and healthy", highlighting the health attribute of the product and meeting the pursuit of modern consumers for healthy diet.

[0064] In another embodiment of the present invention, in terms of the engineering architecture, in order to meet the real-time requirement of image effect evaluation, Flask (a popular Python framework for building web services) or FastAPI (a popular Python framework for building web services) can be used to build the service, and the image and data analysis are processed asynchronously. The front end may need to upload images and input placement attribute data, and the back end calls the model for effect evaluation, and the effect evaluation results are displayed in visual charts.

[0065] According to an image effect evaluation method provided by the present invention, compared with the current method of manually analyzing the image effect, the present invention obtains the placement attribute data of the target image in response to an effect evaluation instruction of the target image; extracts the image visual feature of the target image and the placement attribute feature of the placement attribute data, and performs fusion processing on the image visual feature and the placement attribute feature to obtain an effect evaluation fusion feature; finally, inputs the effect evaluation fusion feature into a preset effect evaluation model for image effect evaluation to obtain the effect evaluation result of the target image, where the preset effect evaluation model is pre-trained based on a sample data set with effect labels. Thus, the preset effect evaluation model is used to analyze the image visual feature of the target image and the placement attribute feature of the placement attribute data to determine the effect evaluation result of the target image, avoiding manual participation, thereby improving the efficiency and accuracy of image effect evaluation and ensuring the automation degree of image effect evaluation.

[0066] Further, in order to better illustrate the above process of image effect evaluation, as a refinement and extension of the above embodiment, the embodiment of the present invention provides another image effect evaluation method, as Figure 2 shown, the method includes:

[0067] 201. In response to the effect evaluation instruction of the target image, obtain the placement attribute data of the target image.

[0068] Specifically, when receiving the effect evaluation instruction of the target image, collect the placement attribute data of the target image in the database.

[0069] 202. Extract the image visual features of the target image and the placement attribute features of the placement attribute data, and perform fusion processing on the image visual features and the placement attribute features to obtain the effect evaluation fusion features.

[0070] For the embodiments of the present invention, in order to improve the accuracy of image effect evaluation and save computing resources, first perform preprocessing such as standardization on the placement attribute data. For example, standard deviation standardization can be performed on placement attribute data such as exposure, click-through rate, and conversion rate, and natural logarithmic function processing can be used for placement attribute data such as cost. Further, extract the image visual features of the image. For example, use a multimodal model, such as the CLIP ViT-L / 14 model, to generate image visual features of a preset dimension, and use methods such as word embedding to determine the placement attribute features of the placement attribute data. Then, perform fusion processing on the image visual features and the placement attribute features. In order to improve the accuracy of image effect evaluation, on this basis, the aesthetic features of the target image can also be extracted, and then the image effect evaluation is performed based on the aesthetic features, image visual features, and placement attribute features. Based on this, step 202 specifically includes: inputting the target image into a preset aesthetic scoring model for aesthetic scoring, and determining the aesthetic feature vector corresponding to the aesthetic score; performing fusion processing on the image visual features, the aesthetic features, and the placement attribute features to obtain the effect evaluation fusion features, where the method of performing fusion processing on the image visual features, the aesthetic features, and the placement attribute features to obtain the effect evaluation fusion features includes: using the image visual features as the visual query vector, using the concatenated feature of the aesthetic features and the placement attribute features as the first key-value vector, and performing interaction between the visual query vector and the first key-value vector to obtain the visual context feature; using the aesthetic features as the aesthetic query vector, using the concatenated feature of the image visual features and the placement attribute features as the second key-value vector, and performing interaction between the aesthetic query vector and the second key-value vector to obtain the aesthetic context feature; using the placement attribute features as the placement query vector, using the concatenated feature of the image visual features and the aesthetic features as the third key-value vector, and performing interaction between the placement query vector and the third key-value vector to obtain the placement context feature; aggregating the visual context feature, the aesthetic context feature, and the placement context feature to obtain the effect evaluation fusion feature.

[0071] Specifically, a preset aesthetic scoring model is pre-trained and constructed. The method for training and constructing the preset aesthetic scoring model includes: obtaining a preset initial aesthetic scoring model and a sample data set, where the sample data set includes sample images with aesthetic scores; dividing the sample data set into training data and test data, training the preset initial aesthetic scoring model with the training data, and testing the trained preset initial aesthetic scoring model with the test data. Finally, the preset initial aesthetic scoring model that meets the test conditions is used as the preset aesthetic scoring model, where the test conditions can be that the number of training times reaches the requirement or the test accuracy of the trained preset initial aesthetic scoring model reaches the requirement. Further, the target image is input into the preset aesthetic scoring model for scoring, and the preset aesthetic scoring model can output the aesthetic score of the target image. For example, the aesthetic score can be on a scale of 1-10, and the higher the score, the more beautiful the image is.

[0072] Further, methods such as word embedding are used to determine the aesthetic features corresponding to the aesthetic scores. Then, a cross-modal attention mechanism is adopted, such as realizing the interactive fusion of image visual features, aesthetic features, and placement attribute features through 3 layers of Transformer. For example, the following formula can be used to fuse the image visual features, aesthetic features, and placement attribute features:

[0073]

[0074] where Q is any one of the visual query vector, aesthetic query vector, and placement query vector. When Q is the visual query vector, K is the key of the concatenated feature of the aesthetic feature and the placement attribute feature, and V is the value corresponding to the key of the concatenated feature of the aesthetic feature and the placement attribute feature. When Q is the aesthetic query vector, K is the key of the concatenated feature of the image visual feature and the placement attribute feature, and V is the value corresponding to the key of the concatenated feature of the image visual feature and the placement attribute feature. When Q is the placement query vector, K is the key of the concatenated feature of the image visual feature and the aesthetic feature, and V is the value corresponding to the key of the concatenated feature of the image visual feature and the aesthetic feature. d k are the vector dimensions of the first key-value vector, or the second key-value vector, or the third key-value vector respectively. Attention(Q, K, V) is the visual context feature, or the aesthetic context feature, or the placement context feature, and softmax() is the feature interaction function. Further, the visual context feature, the aesthetic context feature, and the placement context feature are aggregated, such as horizontally concatenating the visual context feature, the aesthetic context feature, and the placement context feature to obtain the effect evaluation fusion feature. The effect evaluation fusion feature can be a feature of a preset dimension, such as a 768-dimensional fusion feature.

[0075] In another embodiment of the present invention, the following formula can also be used to interactively fuse each feature:

[0076] a ij = softmax(V T ·tanh(W q q i + W k k j ))

[0077] where a ij is the effect evaluation fusion feature, q i is the concatenated feature of the image visual feature and the aesthetic feature, k j is the placement attribute feature, V, W q and W k are feature fusion coefficients for enhancing the feature fusion expression ability, tanh() is an activation function used to process the feature fusion result, and softmax() is a feature interaction function. In another embodiment of the present invention, the weights corresponding to the image visual feature, the aesthetic feature, and the placement attribute feature can also be set respectively according to actual needs, and then, based on the weights, the image visual feature, the aesthetic feature, and the placement attribute feature are weighted and fused. By fusing multiple features in the embodiments of the present invention, the hidden layer information in the features can be provided, that is, richer information can be extracted to help the model better capture the complex patterns in the data, thereby improving the evaluation accuracy of the image effect.

[0078] 203. Input the effect evaluation fusion feature into a preset effect evaluation model for image effect evaluation to obtain the effect evaluation result of the target image, where the preset effect evaluation model is pre-trained based on a sample data set with effect labels.

[0079] For the embodiments of the present invention, after fusing the image visual feature, the aesthetic feature, and the placement attribute feature to obtain the effect evaluation fusion feature, it is necessary to perform effect evaluation based on the effect evaluation fusion feature using the preset effect evaluation model. Based on this, step 203 specifically includes: inputting the effect evaluation fusion feature into the preset effect evaluation model for image effect evaluation to obtain the effect evaluation result of the target image in the form of a visualization chart display; determining the target effect with a contribution degree greater than a preset threshold in the effect evaluation result, determining the evaluation area corresponding to the target effect in the target image, and marking the evaluation area and displaying the contribution degree.

[0080] Among them, the preset threshold is set according to actual requirements. Specifically, in the embodiments of the present invention, a preset effect evaluation model can be trained and constructed in advance based on a sample training set by using the XGBoost (eXtreme Gradient Boosting) method. During the model training process, the preset effect evaluation model can be trained and constructed based on test conditions such as the deviation degree of the effect evaluation of composition rationality being less than the first preset threshold, the color opposition space of color contrast being greater than the second preset threshold, and the recognition rate of key elements of information transmission efficiency being greater than the third preset threshold. The preset effect evaluation model can be an evaluation index prediction model. Next, a fine-tuned large model can also be used as an evaluation effect analysis model under different evaluation indexes. To improve the real-time performance of effect evaluation, the model can be deployed in a high-memory and high-concurrency system. At the same time, an acceleration device, such as ONNX Runtime (an open-source cross-platform machine learning inference engine), can be deployed in the system to accelerate the speed of image effect evaluation. The fine-tuned large model can finally output an effect evaluation result containing a multi-segment structure, such as outputting an effect evaluation result containing a 3-segment structure: core advantages → data evidence → optimization suggestions. The effect evaluation result can be displayed in the form of a visual chart. Further, to enhance the readability of the effect evaluation result, the evaluation effect can also be annotated in the target image, that is, first determine the influence degree of each effect evaluation index on the delivery attribute data, that is, the contribution degree. For example, if the contribution degree of the evaluation index of color contrast to the click-through rate is greater than the preset threshold, then determine the evaluation area corresponding to color contrast in the target image, and highlight and mark the evaluation area, and annotate the contribution degree on the evaluation area.

[0081] 204. In response to an image generation instruction of the image to be generated, obtain the image generation requirement information of the image to be generated.

[0082] Among them, if the target image is a peanut image, the image to be generated can be other types of images such as instant cereals; the image generation requirement information includes basic commodity information (such as commodity name, brand, model, packaging information, etc.), visual style requirements (such as color matching, font style, image composition, etc.), target audience, usage scenario, etc.

[0083] 205. Based on the image generation requirement information, extract image effect evaluation keywords in the effect evaluation database, and map the extracted image effect evaluation keywords to image generation guiding elements that meet the image generation guiding format by using a preset mapping rule, where the effect evaluation database is constructed based on the effect evaluation results of multiple images.

[0084] Among them, the preset mapping rule is the mapping rule from the image effect evaluation keywords in the effect evaluation database set according to actual requirements to the Prompt (prompt word format or image generation guidance format). For example, through a rule engine or a machine learning model, the structured image effect evaluation keywords are mapped to Prompt elements that meet the requirements of the preset image generation model. For example, the keyword "golden ratio composition" corresponds to the image generation guidance element "balanced composition with golden ratio" in the Prompt, and "high contrast" is mapped to "vivid color contrast".

[0085] For the embodiments of the present invention, multiple images can be evaluated for effects in the manner of steps 201-203. When evaluating the effects, the good effects can be evaluated (that is, analyzing the reasons for the good image, such as the reasons for good composition, color, style, etc.). Then, the effect evaluation database is composed of the good effect evaluation results of each image, and based on the effect evaluation database and the image generation requirement information, images are generated for the images to be generated. Based on this, first, it is necessary to extract image effect evaluation keywords in the effect evaluation database according to the image generation requirement information. The specific extraction method includes: based on the image generation requirement information, extracting initial image effect evaluation keywords whose image effect evaluation results in the effect evaluation database meet the preset conditions; determining the style combination probability of the image styles corresponding to each of the initial image effect evaluation keywords, and determining the initial image effect evaluation keywords with the style combination probability greater than the preset probability threshold among the initial image effect evaluation keywords as the image effect evaluation keywords. Among them, the method for determining the style combination probability of the image styles corresponding to each initial image effect evaluation keyword includes: dividing each of the initial image effect evaluation keywords into multiple evaluation keyword groups according to the images to which they belong, and sorting each evaluation keyword in each evaluation keyword group according to a preset sorting rule to obtain an evaluation keyword sequence corresponding to each evaluation keyword group; counting the transition probabilities between adjacent evaluation keywords in each evaluation keyword sequence, and determining the average value of the respective transition probabilities corresponding to each evaluation keyword group as the style combination probability of the image styles corresponding to each initial image effect evaluation keyword in each evaluation keyword group.

[0086] Among them, the preset conditions are set according to actual requirements, and conditions that meet the image generation requirement information are set. For example, if the target audience in the image generation requirement information is the elderly, the initial image effect evaluation keywords can be descriptive information for health; the preset probability threshold is set according to actual requirements; the preset sorting rule is set according to actual requirements, such as rules for sorting according to the evaluation order of evaluation keywords, sorting according to the first letter of evaluation keywords, etc.

[0087] Specifically, if the data in the effect evaluation database is structured data, the keywords in it are directly used. Otherwise, technologies such as natural language are used to extract the initial image effect evaluation keywords that meet the image generation requirement information from the effect evaluation database. For example, if the image generation requirement information is a modern style, the initial image effect evaluation keywords can be the respective initial image effect evaluation keywords corresponding to image A in the effect evaluation database (minimalist, black and white, bright), the respective initial image effect evaluation keywords corresponding to image B (monochromatic, single item prominent, modern elements), the respective initial image effect evaluation keywords corresponding to image C (black and white, bright, modern elements), etc. It should be noted that if the data in the effect evaluation database is structured, such as the effect analysis results of each picture having clear fields, such as color score, composition score, etc., then the Prompt can be directly generated according to these fields. If it is unstructured text, text mining technologies may be needed to extract keywords, such as using TF-IDF or BERT embedding clustering. Further, in order to select the evaluation keywords finally used for image generation from the respective initial image effect evaluation keywords, it is necessary to determine the style combination probability of the image styles corresponding to the respective initial image effect evaluation keywords. Specifically, based on the above example, the respective initial image effect evaluation keywords corresponding to image A (minimalist, black and white, bright) are determined as evaluation keyword group A, the respective initial image effect evaluation keywords corresponding to image B (monochromatic, single item prominent, modern elements) are determined as evaluation keyword group B, and the respective initial image effect evaluation keywords corresponding to image C (black and white, bright, modern elements) are determined as evaluation keyword group C. Taking the current keyword order as the sorting rule, the evaluation keyword sequence corresponding to evaluation keyword group A is (minimalist, black and white, bright), the evaluation keyword sequence corresponding to evaluation keyword group B is (monochromatic, single item prominent, modern elements), and the evaluation keyword sequence corresponding to evaluation keyword group C is (black and white, bright, modern elements). Further, if the evaluation keyword with a prior sort among two adjacent evaluation keywords in each of the said evaluation keyword sequences is k i , and the evaluation keyword with a posterior sort is k j , then the transition probability p(k i k j ) between the two adjacent evaluation keywords k j and k i is calculated according to the following formula:

[0088]

[0089] where k is the total number of initial image effect evaluation keywords, and count(k i → k j ) is the number of times the two evaluation keywords k i and k jThe number of co-occurrences in each evaluation keyword group, ∑ k count(k i →k k ) is the evaluation keyword k in each evaluation keyword group i The total number of occurrences. Thus, according to the above method, the transition probability between every two adjacent evaluation keywords in each evaluation keyword group can be determined, and then for each evaluation keyword group, the mean of all transition probabilities in the evaluation keyword group is determined as the style combination probability of the image style corresponding to each initial image effect evaluation keyword in the evaluation keyword group, and then the initial image effect evaluation keyword with a style combination probability greater than a preset probability threshold is determined as the image effect evaluation keyword. The embodiment of the present invention expresses the conversion law between different style combinations through the transition probability, so as to select the optimal effect evaluation keyword combination according to the transition law, which can improve the image generation quality and generation accuracy.

[0090] 206 . Determine an image generation prompt word template for the image to be generated, and fill the commodity theme information and image generation guide elements of the image to be generated into the image generation prompt word template to obtain the image generation prompt word.

[0091] The image generation prompt word template includes various title items, such as theme description title item, style setting title item, composition rule title item, lighting effect title item, artistic style title item, etc., which automatically links the product theme information and image effect evaluation keywords to the corresponding title items, so as to obtain the image generation prompt word. For example, if the image effect evaluation keyword is "high color contrast", relevant adjectives need to be added to the image generation prompt word template, such as "bright red and white form a strong contrast.

[0092] Specifically, the image effect evaluation keywords are converted into a format accepted by the image generation prompt word template to obtain image generation guiding elements, and finally the commodity theme information (such as commodity name, commodity description information, etc.) of the image to be generated and the image effect evaluation keywords are filled into the image generation prompt word template to obtain image generation prompt words. For example, if the image to be generated is glutinous rice balls, that is, the commodity theme information is glutinous rice balls, and the image effect evaluation keywords are "soft lighting" and "close-up", then the generated image generation prompt words are "a bowl of hot glutinous rice balls, soft side lighting, close-up, high details, 8k resolution, and warm family atmosphere". The embodiment of the present invention generates image generation prompt words through the image generation prompt word template, which can ensure the standardization of image generation prompt words, thereby improving the generation quality and generation accuracy of the image.

[0093] 207. Input the image generation prompt into a preset image generation model to generate an image, where the preset image generation model is pre-trained based on a sample data set with image labels.

[0094] For the embodiments of the present invention, in order to improve the image generation accuracy of the preset image generation model, it is necessary to train and construct the preset image generation model. Based on this, the method includes: obtaining a preset initial image generation model and a sample data set, where the sample data set includes sample image generation prompts with image labels; dividing the sample data set into training data and test data, training the preset initial image generation model with the training data, and testing the trained preset initial image generation model with the test data, and finally using the trained preset initial image generation model that meets the test conditions as the preset image generation model, where meeting the test conditions means that the training test meets the requirements or the test accuracy of the trained preset initial image generation model reaches the requirements. Further, after constructing the preset image generation model, it is necessary to use the preset image generation model to generate an image. Based on this, step 207 specifically includes: determining the reference image of the image to be generated and determining the initial seed value during the image generation process of the preset image generation model; inputting the image generation prompt and the initial seed value into the preset image generation model to generate an initial image; determining the fluctuation range of the initial seed value based on the similarity between the initial image and the reference image, and determining multiple seed values within the fluctuation range, and inputting each of the seed values and the image generation prompt into the preset image generation model to generate an image, obtaining candidate images corresponding to each of the seed values; respectively determining the similarity between each candidate image and the reference image, and using the candidate image corresponding to the maximum similarity as the newly generated image.

[0095] Among them, the reference image refers to the original image of the image to be generated, that is, an image that only contains the image to be generated, or a rough design image of the brand party with a general brand style; the initial seed value is the initialization state of the random noise that controls the preset image generation model, and the initial seed value can be generated based on the feature hash value of the reference image. For example, the initial seed value seed = 42. Then, the initial seed value and the above-mentioned image generation prompt words are jointly input into the preset image generation model. Through the preset image generation model, an initial image can be output, and the feature vectors of the reference image and the initial image are extracted respectively. Based on the feature vectors, the similarity between the reference image and the initial image is calculated, such as the cosine similarity. If the similarity is greater than the preset similarity threshold (the preset similarity threshold is set according to actual needs), a smaller fluctuation range of the initial seed value is set, such as seed ± 100. If the similarity is less than or equal to the preset similarity threshold, a larger fluctuation range of the initial seed value is set, such as seed ± 1000, and N seed values are randomly selected within the fluctuation range, such as N = 10. Then, the N seed values and the image generation prompt words are respectively input into the preset image generation model for image generation to obtain candidate images corresponding to the N seed values. Then, the similarity between each candidate image and the reference image is calculated respectively. Finally, the candidate image with the maximum similarity is used as the newly generated image. When generating an image in the embodiment of the present invention, introducing the seed value can ensure that the generated image not only conforms to the brand visual consistency but also has diversity, while reducing the manual screening cost.

[0096] According to another image effect evaluation method provided by the present invention, compared with the current method of manually analyzing the image effect, the present invention obtains the placement attribute data of the target image in response to an image effect evaluation instruction for the target image; extracts the image visual features of the target image and the placement attribute features of the placement attribute data, and performs a fusion process on the image visual features and the placement attribute features to obtain an effect evaluation fusion feature; finally, inputs the effect evaluation fusion feature into a preset effect evaluation model for image effect evaluation to obtain the effect evaluation result of the target image, where the preset effect evaluation model is pre-trained based on a sample data set with effect labels. Thus, by using the preset effect evaluation model to analyze the image visual features of the target image and the placement attribute features of the placement attribute data to determine the effect evaluation result of the target image, manual participation is avoided, thereby improving the image effect evaluation efficiency and evaluation accuracy, and ensuring the automation degree of the image effect evaluation.

[0097] Furthermore, as Figure 1 a specific implementation, the embodiment of the present invention provides an image effect evaluation device, as Figure 3 shown, the device includes: an acquisition unit 31, a fusion unit 32, and an effect evaluation unit 33.

[0098] The obtaining unit 31 can be configured to obtain the placement attribute data of the target image in response to an effect evaluation instruction of the target image.

[0099] The fusion unit 32 can be configured to extract the image visual feature of the target image and the placement attribute feature of the placement attribute data, and perform a fusion process on the image visual feature and the placement attribute feature to obtain an effect evaluation fusion feature.

[0100] The effect evaluation unit 33 can be configured to input the effect evaluation fusion feature into a preset effect evaluation model to perform image effect evaluation, and obtain an effect evaluation result of the target image, where the preset effect evaluation model is pre-trained based on a sample data set with effect labels.

[0101] In a specific application scenario, in order to perform a fusion process on the image visual feature and the placement attribute feature, as Figure 4 shown, the fusion unit 32 includes an aesthetic scoring module 321 and a fusion module 322.

[0102] The aesthetic scoring module 321 can be configured to input the target image into a preset aesthetic scoring model to perform aesthetic scoring, and determine an aesthetic feature vector corresponding to the aesthetic score.

[0103] The fusion module 322 can be configured to perform a fusion process on the image visual feature, the aesthetic feature, and the placement attribute feature to obtain an effect evaluation fusion feature. The method for performing a fusion process on the image visual feature, the aesthetic feature, and the placement attribute feature to obtain an effect evaluation fusion feature includes: using the image visual feature as a visual query vector, using the concatenated feature of the aesthetic feature and the placement attribute feature as a first key-value vector, and performing an interaction between the visual query vector and the first key-value vector to obtain a visual context feature; using the aesthetic feature as an aesthetic query vector, using the concatenated feature of the image visual feature and the placement attribute feature as a second key-value vector, and performing an interaction between the aesthetic query vector and the second key-value vector to obtain an aesthetic context feature; using the placement attribute feature as a placement query vector, using the concatenated feature of the image visual feature and the aesthetic feature as a third key-value vector, and performing an interaction between the placement query vector and the third key-value vector to obtain a placement context feature; and aggregating the visual context feature, the aesthetic context feature, and the placement context feature to obtain the effect evaluation fusion feature.

[0104] In a specific application scenario, in order to perform effect evaluation on the target image, the effect evaluation unit 33 includes an effect evaluation module 331 and an effect display module 332.

[0105] The effect evaluation module 331 can be used to input the effect evaluation fusion feature into a preset effect evaluation model for image effect evaluation, and obtain an effect evaluation result of the target image in the form of a visualized chart display.

[0106] The effect display module 332 can be used to determine a target effect with a contribution degree greater than a preset threshold in the effect evaluation result, determine an evaluation area corresponding to the target effect in the target image, and mark the evaluation area and display the contribution degree.

[0107] In a specific application scenario, in order to generate another image, the device further includes: an extraction unit 34, a filling unit 35, and an image generation unit 36.

[0108] The acquisition unit 31 can also be used to respond to an image generation instruction of an image to be generated, and acquire image generation requirement information of the image to be generated.

[0109] The extraction unit 34 can be used to extract keywords for image effect evaluation in an effect evaluation database based on the image generation requirement information, and map the extracted keywords for image effect evaluation to image generation guiding elements that meet the image generation guiding format by using a preset mapping rule, where the effect evaluation database is constructed based on effect evaluation results of multiple images.

[0110] The filling unit 35 can be used to determine an image generation prompt word template of the image to be generated, and fill the product theme information of the image to be generated and the image generation guiding elements into the image generation prompt word template to obtain an image generation prompt word.

[0111] The image generation unit 36 can be used to input the image generation prompt word into a preset image generation model for image generation to obtain a newly generated image, where the preset image generation model is pre-trained based on a sample data set with image labels. Then, based on the newly generated image, the generation or optimization of images for this or the next round of placement can be directly performed.

[0112] In a specific application scenario, in order to extract keywords for image effect evaluation in an effect evaluation database, the extraction unit 34 includes an extraction module 341 and a first determination module 342.

[0113] The extraction module 341 can be used to extract initial keywords for image effect evaluation whose image effect evaluation results meet preset conditions in the effect evaluation database based on the image generation requirement information.

[0114] The first determination module 342 may be configured to determine the style combination probability of the image style corresponding to each of the initial image effect evaluation keywords, and determine, among the initial image effect evaluation keywords, the initial image effect evaluation keywords with the style combination probability greater than a preset probability threshold as the image effect evaluation keywords.

[0115] In a specific application scenario, to determine the style combination probability of the image style corresponding to each initial image effect evaluation keyword, the first determination module 342 may specifically be configured to divide each of the initial image effect evaluation keywords into multiple evaluation keyword groups according to the images to which they belong, and sort each evaluation keyword in each evaluation keyword group according to a preset sorting rule to obtain an evaluation keyword sequence corresponding to each evaluation keyword group; count the transition probabilities between adjacent evaluation keywords in each evaluation keyword sequence, and determine the average value of the respective transition probabilities corresponding to each evaluation keyword group as the style combination probability of the image style corresponding to each initial image effect evaluation keyword in each evaluation keyword group.

[0116] In a specific application scenario, to generate an image, the image generation unit 36 includes a second determination module 361 and an image generation module 362.

[0117] The second determination module 361 may be configured to determine a reference image of the image to be generated and determine an initial seed value during the process of the preset image generation model generating an image.

[0118] The image generation module 362 may be configured to input the image generation prompt word and the initial seed value into the preset image generation model to generate an image, thereby obtaining an initial image.

[0119] The second determination module 361 may further be configured to determine the fluctuation range of the initial seed value based on the similarity between the initial image and the reference image, determine multiple seed values within the fluctuation range, and input each of the seed values and the image generation prompt word into the preset image generation model to generate an image, thereby obtaining candidate images corresponding to each of the seed values.

[0120] The second determination module 361 may specifically be configured to determine the similarity between each of the candidate images and the reference image, and use the candidate image corresponding to the maximum similarity as the newly generated image.

[0121] It should be noted that for other corresponding descriptions of each functional module involved in the image effect evaluation device provided in the embodiments of the present invention, reference may be made to Figure 1 the corresponding description of the method shown, which will not be elaborated herein.

[0122] Based on the above, such asFigure 1 For the method shown above, correspondingly, an embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, the following steps are implemented: in response to an effect evaluation instruction of a target image, obtain the placement attribute data of the target image; extract the image visual features of the target image and the placement attribute features of the placement attribute data, and perform a fusion process on the image visual features and the placement attribute features to obtain an effect evaluation fusion feature; input the effect evaluation fusion feature into a preset effect evaluation model for image effect evaluation to obtain an effect evaluation result of the target image, where the preset effect evaluation model is pre-trained based on a sample data set with effect labels.

[0123] Based on the above as Figure 1 shown in the method and as Figure 3 shown in the embodiment of the device, an embodiment of the present invention further provides an entity structure diagram of a computer device, as Figure 5 shown. The computer device includes: a processor 41, a memory 42, and a computer program stored on the memory 42 and executable on the processor. The memory 42 and the processor 41 are both arranged on a bus 43. When the processor 41 executes the program, the following steps are implemented: in response to an effect evaluation instruction of a target image, obtain the placement attribute data of the target image; extract the image visual features of the target image and the placement attribute features of the placement attribute data, and perform a fusion process on the image visual features and the placement attribute features to obtain an effect evaluation fusion feature; input the effect evaluation fusion feature into a preset effect evaluation model for image effect evaluation to obtain an effect evaluation result of the target image, where the preset effect evaluation model is pre-trained based on a sample data set with effect labels.

[0124] Through the technical solution of the present invention, the present invention obtains the placement attribute data of the target image in response to an effect evaluation instruction of the target image; extracts the image visual features of the target image and the placement attribute features of the placement attribute data, and performs a fusion process on the image visual features and the placement attribute features to obtain an effect evaluation fusion feature; finally, inputs the effect evaluation fusion feature into a preset effect evaluation model for image effect evaluation to obtain the effect evaluation result of the target image, wherein the preset effect evaluation model is pre-trained based on a sample data set with effect labels. Subsequently, new images can be generated based on the effect evaluation results of multiple images, and the image optimization for this round or the next placement can be directly continued based on the new images. Thus, the preset effect evaluation model is used to analyze the image visual features of the target image and the placement attribute features of the placement attribute data to determine the effect evaluation result of the target image, avoiding manual participation, thereby improving the efficiency and accuracy of image effect evaluation, ensuring the automation of image effect evaluation, and guiding the generation or optimization of the images to be placed according to the image effect evaluation results.

[0125] Obviously, those skilled in the art should understand that the above-mentioned modules or steps of the present invention can be implemented by a general-purpose computing device. They can be concentrated on a single computing device or distributed on a network composed of multiple computing devices. Optionally, they can be implemented by program codes executable by the computing device. Thus, they can be stored in a storage device and executed by the computing device. And in some cases, the steps shown or described can be executed in a different order than here, or they can be separately fabricated into individual integrated circuit modules, or multiple of them can be fabricated into a single integrated circuit module to implement. In this way, the present invention is not limited to any specific combination of hardware and software.

[0126] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An image effect evaluation method, characterized in that, Including: In response to an effect evaluation instruction for a target image, obtaining the placement attribute data of the target image; Extracting the image visual features of the target image and the placement attribute features of the placement attribute data, and performing a fusion process on the image visual features and the placement attribute features to obtain an effect evaluation fusion feature; Inputting the effect evaluation fusion feature into a preset effect evaluation model for image effect evaluation to obtain an effect evaluation result of the target image, where the preset effect evaluation model is pre-trained based on a sample data set with effect labels.

2. The method according to claim 1, wherein The performing a fusion process on the image visual features and the placement attribute features to obtain an effect evaluation fusion feature includes: Inputting the target image into a preset aesthetic scoring model for aesthetic scoring, and determining an aesthetic feature vector corresponding to the aesthetic score; Performing a fusion process on the image visual features, the aesthetic features, and the placement attribute features to obtain an effect evaluation fusion feature, where the method for performing a fusion process on the image visual features, the aesthetic features, and the placement attribute features to obtain an effect evaluation fusion feature includes: Using the image visual features as a visual query vector, using the concatenated feature of the aesthetic features and the placement attribute features as a first key-value vector, and performing an interaction between the visual query vector and the first key-value vector to obtain a visual context feature; Using the aesthetic features as an aesthetic query vector, using the concatenated feature of the image visual features and the placement attribute features as a second key-value vector, and performing an interaction between the aesthetic query vector and the second key-value vector to obtain an aesthetic context feature; Using the placement attribute features as a placement query vector, using the concatenated feature of the image visual features and the aesthetic features as a third key-value vector, and performing an interaction between the placement query vector and the third key-value vector to obtain a placement context feature; Aggregating the visual context feature, the aesthetic context feature, and the placement context feature to obtain the effect evaluation fusion feature.

3. The method according to claim 1, wherein The inputting the effect evaluation fusion feature into a preset effect evaluation model for image effect evaluation to obtain an effect evaluation result of the target image includes: Inputting the effect evaluation fusion feature into a preset effect evaluation model for image effect evaluation to obtain an effect evaluation result of the target image in a form of visualization chart display; Determining a target effect with a contribution degree greater than a preset threshold in the effect evaluation result, determining an evaluation area corresponding to the target effect in the target image, and marking the evaluation area and displaying the contribution degree.

4. The method according to claim 1, wherein After the inputting the effect evaluation fusion feature into a preset effect evaluation model for image effect evaluation to obtain an effect evaluation result of the target image, the method further includes: In response to an image generation instruction for an image to be generated, obtaining the image generation requirement information of the image to be generated; Based on the image generation requirement information, extract keywords for image effect evaluation in the effect evaluation database, and map the extracted keywords for image effect evaluation to image generation guiding elements that meet the image generation guiding format by using a preset mapping rule, where the effect evaluation database is constructed based on the effect evaluation results of multiple images; Determine the image generation prompt word template for the image to be generated, and fill the product theme information of the image to be generated and the image generation guiding elements into the image generation prompt word template to obtain an image generation prompt word; Input the image generation prompt word into a preset image generation model for image generation to obtain a newly generated image, where the preset image generation model is pre-trained based on a sample data set with image labels.

5. The method according to claim 4, characterized in that, The extracting keywords for image effect evaluation in the effect evaluation database based on the image generation requirement information includes: Based on the image generation requirement information, extract initial keywords for image effect evaluation in the effect evaluation database whose image effect evaluation results meet preset conditions; Determine the style combination probability of the image style corresponding to each of the initial keywords for image effect evaluation, and determine the initial keywords for image effect evaluation with a style combination probability greater than a preset probability threshold among the initial keywords for image effect evaluation as the keywords for image effect evaluation.

6. The method according to claim 5, wherein The determining the style combination probability of the image style corresponding to each of the initial keywords for image effect evaluation includes: Classify each of the initial keywords for image effect evaluation into multiple groups of evaluation keywords according to the images to which they belong, and sort each evaluation keyword in each group of evaluation keywords according to a preset sorting rule to obtain an evaluation keyword sequence corresponding to each group of evaluation keywords; Statistically analyze the transition probability between adjacent evaluation keywords in each evaluation keyword sequence, and determine the average value of the respective transition probabilities corresponding to each group of evaluation keywords as the style combination probability of the image style corresponding to each of the initial keywords for image effect evaluation in each group of evaluation keywords.

7. The method according to claim 4, characterized in that, The inputting the image generation prompt word into a preset image generation model for image generation to obtain a newly generated image includes: Determine the reference image of the image to be generated, and determine the initial seed value during the image generation process of the preset image generation model; Input the image generation prompt word and the initial seed value into the preset image generation model for image generation to obtain an initial image; Based on the similarity between the initial image and the reference image, determine the fluctuation range of the initial seed value, and determine multiple seed values within the fluctuation range. Input each of the seed values and the image generation prompt word into the preset image generation model for image generation to obtain candidate images corresponding to each of the seed values; Respectively determine the similarity between each candidate image and the reference image, and use the candidate image corresponding to the maximum similarity as the newly generated image.

8. An image effect evaluation device, characterized in that, including: An acquisition unit for acquiring the placement attribute data of the target image in response to an effect evaluation instruction of the target image; A fusion unit, configured to extract the image visual features of the target image and the placement attribute features of the placement attribute data, and perform fusion processing on the image visual features and the placement attribute features to obtain an effect evaluation fusion feature; An effect evaluation unit, configured to input the effect evaluation fusion feature into a preset effect evaluation model for image effect evaluation to obtain an effect evaluation result of the target image, wherein the preset effect evaluation model is pre-trained based on a sample data set with effect labels.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images