User design evaluation method and device based on artificial intelligence assistance
Through the user design evaluation model based on deep learning, the problems of low efficiency and large error of the traditional manual scoring model are solved, and the automation, accurate and efficient scoring of students' artistic design works are achieved.
Patent Information
- Application Number
- CN202510609548.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-13
- Publication Date
- 2025-08-12
AI Technical Summary
The traditional manual scoring model cannot meet the repeated timely feedback needs of art students, which has artificial fatigue and subjective errors, and it is difficult to achieve professional scoring of students' daily paintings.
The user-designed evaluation model based on deep learning, including the backbone network and five scoring branch networks, is adopted to achieve automated scoring through image acquisition, processing, model construction, training and scoring.
It achieves accurate and rapid rating of students' artistic design works, reduces labor costs, improves correction efficiency, reduces subjective errors, and has high consistency in the scoring standards.
Smart Images
Figure CN120471888A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of intelligent evaluation technology, and in particular to an artificial intelligence-assisted user design evaluation method and device. Background Art
[0002] With the continuous development of artificial intelligence (AI), AI has been widely applied in many fields and has had a profound impact on the development of various fields. Among them, AI has shown outstanding performance in the field of education. The intervention of AI has greatly enriched educational approaches and optimized resource allocation.
[0003] Art students require extensive, "mechanized" daily training and practice. Grading for art design is particularly crucial, but the traditional "one-on-one" or "one-on-many" grading model, which relies on instructors to grade their work, cannot meet the needs of art students for repeated review and timely feedback, and is labor-intensive. Furthermore, fatigue and other factors can lead to distractions and sloppy grading, and there are occasional errors caused by subjective factors, making the process highly subjective. Furthermore, art students need professional grading for their own work, but traditional methods are subject to numerous inconveniences.
[0004] Therefore, it is an urgent problem for those skilled in the art to propose a user design evaluation method and device based on artificial intelligence assistance to solve the problems existing in the prior art. Summary of the Invention
[0005] In view of this, the present invention provides a user design evaluation method and device based on artificial intelligence assistance, which can accurately and quickly score students' art design works with strong pertinence and high efficiency.
[0006] In order to achieve the above object, the present invention adopts the following technical solutions:
[0007] A user design evaluation method based on artificial intelligence assistance, comprising:
[0008] S1. Obtain historical user design images, annotate and score the images, and obtain a historical user design image and score dataset;
[0009] S2, divide the historical user design image and rating dataset into a training set and a test set according to a certain ratio;
[0010] S3. Build a user design evaluation model based on deep learning;
[0011] S4. Using the training set to train the constructed user design evaluation model to obtain a trained user design evaluation model;
[0012] S5. Input the test set into the trained user design evaluation model for verification, and obtain a verified user design evaluation model;
[0013] S6. Obtain a user design image to be rated, and input the user design image to be rated into a verified user design evaluation model for rating;
[0014] S7. Store the scored user design images and corresponding scores, and update the user design evaluation model.
[0015] In the above method, optionally, obtaining historical user design images in S1 includes: obtaining artwork images from an open source dataset, obtaining student art design work images from student historical assignments;
[0016] The acquired images are annotated and scored based on five aspects: completeness, modeling ability, object structure, aspect relationship, and tonal contrast, to obtain a historical user-designed image and scoring dataset.
[0017] The above method, optionally, the deep learning-based user design evaluation model in S3 includes: a backbone network and five scoring branch networks; the five scoring branch networks are: integrity scoring branch network, modeling ability scoring branch network, picture object structure scoring branch network, decent relationship scoring branch network and tone contrast scoring branch network.
[0018] In the above method, optionally, the backbone network includes multiple convolutional layers, and features of the input image are extracted through the multiple convolutional layers to obtain an input feature map.
[0019] In the above method, optionally, the five scoring branches have the same network structure, including: an efficient channel attention module and a regression network;
[0020] The efficient channel attention module processes the input feature map to generate channel weights, adjusts the feature values on each channel based on the channel weights, and performs Sigmoid activation on the adjusted channel feature values to obtain activation features. The activation features are then fused with the input feature map to obtain fused features.
[0021] The regression network includes: a global average pooling layer and three linear layers. The fused features are input to the global average pooling layer for global average pooling processing to obtain reduced dimensionality features. The global average pooling layer flattens the reduced dimensionality features into a one-dimensional vector through the flatten operation; the one-dimensional vector is input to the three linear layers for linear mapping and nonlinear transformation, and finally a corresponding score value is obtained.
[0022] In the above method, optionally, the constructed user design evaluation model is trained in S4, specifically:
[0023] S401: Input the training set data into the user design evaluation model based on deep learning for training to obtain an initial backbone network scoring model;
[0024] S402, training the five scoring branch networks one by one. When training any scoring branch network, lock the initial backbone network scoring model parameters and other scoring branch network parameters;
[0025] S403: Each time a scoring branch network is trained, it includes the previously trained scoring branch network. After the training of the five scoring branch networks is completed, the optimal scoring model is obtained.
[0026] In the above method, optionally, in S6, the user design to be rated is scored through a user design evaluation model, and finally a total score, a completeness score, a modeling ability score, a picture object structure score, a decent relationship score and a color contrast score are obtained.
[0027] An artificial intelligence-assisted user design evaluation device, which executes any one of the artificial intelligence-assisted user design evaluation methods described above, comprising:
[0028] Image acquisition module, image processing module, model building module, training module, evaluation module, storage module and display module;
[0029] The image acquisition module, image processing module, model building module, training module, evaluation module and display module are connected in sequence, and the evaluation module is also connected to the storage module;
[0030] Image acquisition module, which uses a scanner with a high-definition camera to obtain user-designed images to be rated;
[0031] Image processing module, which designs image preprocessing for users to be rated;
[0032] The model building module, based on deep learning, builds a user design evaluation model consisting of a backbone network and five scoring branch networks;
[0033] Training module, which trains the user design evaluation model built based on deep learning;
[0034] Evaluation module, which scores user designs and outputs the scoring results;
[0035] Storage module, storing user images and corresponding rating results;
[0036] The display module displays the user-designed images and corresponding scoring results.
[0037] It can be seen from the above technical solution that, compared with the prior art, the present invention provides a user design evaluation method and device based on artificial intelligence assistance, which has the following beneficial effects: the present invention significantly improves the performance of feature extraction through a local cross-channel attention mechanism and an adaptive one-dimensional convolution kernel, and performs feature extraction by sharing main network parameters, thereby improving the efficiency and performance of the user design evaluation model; the present invention uses AI technology for the evaluation of student art design works, which is more convenient and more real-time. By utilizing the present invention, low-cost grading can be achieved and the efficiency of grading can be improved; the present invention is specifically used to score student art design works, is not affected by subjective factors, and maintains consistent grading standards for exercises, and will not be hastily graded due to fatigue. It is more targeted and has higher accuracy. BRIEF DESCRIPTION OF THE DRAWINGS
[0038] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.
[0039] Figure 1 A flowchart of an artificial intelligence-assisted user design evaluation method provided by the present invention;
[0040] Figure 2 A schematic diagram of the structure of an artificial intelligence-assisted user design evaluation device provided by the present invention;
[0041] Figure 3 This is a schematic diagram of the final scoring results in the specific embodiment provided by the present invention. DETAILED DESCRIPTION
[0042] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.
[0043] In this application, relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. The terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or apparatus comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article or apparatus. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article or apparatus comprising the element.
[0044] Reference Figure 1 As shown, a user design evaluation method based on artificial intelligence assistance includes:
[0045] S1. Obtain historical user design images, annotate and score the images, and obtain a historical user design image and score dataset;
[0046] S2, divide the historical user design image and rating dataset into a training set and a test set according to a certain ratio;
[0047] S3. Build a user design evaluation model based on deep learning;
[0048] S4. Using the training set to train the constructed user design evaluation model to obtain a trained user design evaluation model;
[0049] S5. Input the test set into the trained user design evaluation model for verification, and obtain a verified user design evaluation model;
[0050] S6. Obtain a user design image to be rated, and input the user design image to be rated into a verified user design evaluation model for rating;
[0051] S7. Store the scored user design images and corresponding scores, and update the user design evaluation model.
[0052] Furthermore, obtaining historical user design images in S1 includes: obtaining artwork images from open source datasets, and obtaining student art design work images from student historical assignments;
[0053] The acquired images were annotated and scored based on five aspects: completeness, modeling ability, object structure, aspect relationship, and tonal contrast, to obtain a historical user design image and scoring dataset.
[0054] Specifically, user-designed images are annotated and scored, and scoring criteria for different types of images are formulated based on completeness, modeling ability, object structure, decent relationship, and color contrast. Each member of the scoring team cyclically assigns different images for scoring until the number of times each image is assigned reaches a threshold. Each member scores the assigned image based on the scoring criteria for completeness, modeling ability, object structure, decent relationship, and color contrast, and each image receives multiple scoring data. Multiple scoring data are collected and the average score is calculated as the final score to complete the scoring and annotation.
[0055] Furthermore, based on the correspondence between the integrity, modeling ability, picture object structure, decent relationship and color contrast score information and the score proportion, the product of the integrity, modeling ability, picture object structure, decent relationship and color contrast score and the score proportion is calculated, and the product of each sub-score and the score proportion is added to obtain the total score corresponding to the image to be identified.
[0056] Furthermore, the deep learning-based user design evaluation model in S3 includes: a backbone network and five scoring branch networks; the five scoring branch networks are: integrity scoring branch network, modeling ability scoring branch network, picture object structure scoring branch network, decent relationship scoring branch network and color contrast scoring branch network. Different scoring branch networks can be added according to different scoring criteria.
[0057] Furthermore, the backbone network includes multiple convolutional layers, which extract features from the input image to obtain an input feature map;
[0058] Specifically, convolutional feature extraction is used to extract the inherent features of the image. The specific process of convolutional feature extraction in large image recognition is to randomly select a small local area from the image as a training sample, learn some features from the small sample, and then use these features as filters to perform convolution operations with the original entire image to obtain the activation values of different features at any position in the original image. Given a large-size image with a resolution of r×c, it is defined as x_{large}; first, extract a×b small-size image sample x_{small} from x_{large}, and obtain k features and activation values by training the sparse autoencoder f(w(1)x_{small}+b(1)), where w(1) and b(1) are the trained parameters; then for each a×b size x_s in x_{large}, calculate the corresponding activation value f_s(w(1)xsmall+b(1)), and further use the activation value of x_small to perform convolution operation with these activation values f_s, and you can get k×(r-a+1)×(c-b+1) convolved feature maps. The pooling feature is specifically trained by inputting the features extracted by the convolution layer into the classifier. The pooling feature is specifically used to judge the overall semantics of the image.
[0059] Furthermore, the five scoring branches have the same network structure, including: an efficient channel attention module and a regression network;
[0060] The efficient channel attention module processes the input feature map to generate channel weights, adjusts the feature values on each channel based on the channel weights, and performs Sigmoid activation on the adjusted channel feature values to obtain activation features. The activation features are then fused with the input feature map to obtain fused features.
[0061] The regression network includes: a global average pooling layer and three linear layers. The fused features are input to the global average pooling layer for global average pooling processing to obtain reduced dimensionality features. The global average pooling layer flattens the reduced dimensionality features into a one-dimensional vector through the flatten operation; the one-dimensional vector is input to the three linear layers for linear mapping and nonlinear transformation, and finally a corresponding score value is obtained.
[0062] Furthermore, the constructed user design evaluation model is trained in S4, specifically:
[0063] S401: Input the training set data into the user design evaluation model based on deep learning for training to obtain an initial backbone network scoring model;
[0064] S402, training the five scoring branch networks one by one. When training any scoring branch network, lock the initial backbone network scoring model parameters and other scoring branch network parameters;
[0065] S403: Each time a scoring branch network is trained, it includes the previously trained scoring branch network. After the training of the five scoring branch networks is completed, the optimal scoring model is obtained.
[0066] Furthermore, in S6, the user design to be rated is scored through the user design evaluation model, and finally the total score, completeness score, modeling ability score, screen object structure score, decent relationship score and color contrast score are obtained.
[0067] Reference Figure 2 As shown, an artificial intelligence-assisted user design evaluation device executes any one of the artificial intelligence-assisted user design evaluation methods described above, including:
[0068] Image acquisition module, image processing module, model building module, training module, evaluation module, storage module and display module;
[0069] The image acquisition module, image processing module, model building module, training module, evaluation module and display module are connected in sequence, and the evaluation module is also connected to the storage module;
[0070] Image acquisition module, which uses a scanner with a high-definition camera to obtain user-designed images to be rated;
[0071] Image processing module, which designs image preprocessing for users to be rated;
[0072] The model building module, based on deep learning, builds a user design evaluation model consisting of a backbone network and five scoring branch networks;
[0073] Training module, which trains the user design evaluation model built based on deep learning;
[0074] Evaluation module, which scores user designs and outputs the scoring results;
[0075] Storage module, storing user images and corresponding rating results;
[0076] The display module displays the user-designed images and corresponding scoring results.
[0077] In a specific embodiment, the scoring information and scoring proportions of integrity, modeling ability, screen object structure, decent relationship and color contrast are as follows: integrity score 30%, modeling ability score 30%, screen object structure score 15%, decent relationship score 15% and color contrast score 10%. In other embodiments, the scoring information and scoring proportions can be adjusted as needed. The final scoring result is as follows: Figure 3 shown.
[0078] Each embodiment in this specification is described in a progressive manner. The same or similar parts between the embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments. In particular, for system or system embodiments, since they are basically similar to method embodiments, the description is relatively simple. For relevant parts, refer to the partial description of the method embodiment. The system and system embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. A person of ordinary skill in the art can understand and implement it without expending creative work.
[0079] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A user design evaluation method based on artificial intelligence assistance, characterized in that: include: S1. Obtain historical user design images, annotate and score the images, and obtain a historical user design image and score dataset; S2, divide the historical user design image and rating dataset into a training set and a test set according to a certain ratio; S3. Build a user design evaluation model based on deep learning; S4. Using the training set to train the constructed user design evaluation model to obtain a trained user design evaluation model; S5. Input the test set into the trained user design evaluation model for verification, and obtain a verified user design evaluation model; S6. Obtain a user design image to be rated, and input the user design image to be rated into a verified user design evaluation model for rating; S7. Store the scored user design images and corresponding scores, and update the user design evaluation model.
2. The user design evaluation method based on artificial intelligence assistance according to claim 1 is characterized in that: Obtaining historical user design images in S1 includes: obtaining artwork images from open source datasets and obtaining student art design images from student historical assignments; The acquired images are annotated and scored based on five aspects: completeness, modeling ability, object structure, aspect relationship, and tonal contrast, to obtain a historical user-designed image and scoring dataset.
3. The user design evaluation method based on artificial intelligence assistance according to claim 1 is characterized in that: The deep learning-based user design evaluation model in S3 includes: a backbone network and five scoring branch networks; the five scoring branch networks are: integrity scoring branch network, modeling ability scoring branch network, picture object structure scoring branch network, decent relationship scoring branch network and color contrast scoring branch network.
4. The user design evaluation method based on artificial intelligence assistance according to claim 3 is characterized in that: The backbone network includes multiple convolutional layers, which extract features of the input image through multiple convolutional layers to obtain the input feature map.
5. The user design evaluation method based on artificial intelligence assistance according to claim 4 is characterized in that: The five scoring branches have the same network structure, including: an efficient channel attention module and a regression network; The efficient channel attention module processes the input feature map to generate channel weights, adjusts the feature values on each channel based on the channel weights, and performs Sigmoid activation on the adjusted channel feature values to obtain activation features. The activation features are then fused with the input feature map to obtain fused features. The regression network includes: a global average pooling layer and three linear layers. The fused features are input to the global average pooling layer for global average pooling processing to obtain reduced dimensionality features. The global average pooling layer flattens the reduced dimensionality features into a one-dimensional vector through the flatten operation; the one-dimensional vector is input to the three linear layers for linear mapping and nonlinear transformation, and finally a corresponding score value is obtained.
6. The user design evaluation method based on artificial intelligence assistance according to claim 1 or 3, characterized in that: In S4, the constructed user design evaluation model is trained, specifically: S401: Input the training set data into the user design evaluation model based on deep learning for training to obtain an initial backbone network scoring model; S402, training the five scoring branch networks one by one. When training any scoring branch network, lock the initial backbone network scoring model parameters and other scoring branch network parameters; S403: Each time a scoring branch network is trained, it includes the previously trained scoring branch network. After the training of the five scoring branch networks is completed, the optimal scoring model is obtained.
7. The user design evaluation method based on artificial intelligence assistance according to claim 1 or 3, characterized in that: In S6, the user design to be rated is scored through the user design evaluation model, and the final score is the total score, completeness score, modeling ability score, screen object structure score, decent relationship score and color contrast score.
8. An artificial intelligence-assisted user design evaluation device, which executes an artificial intelligence-assisted user design evaluation method according to any one of claims 1 to 7, characterized in that: include: Image acquisition module, image processing module, model building module, training module, evaluation module, storage module and display module; The image acquisition module, image processing module, model building module, training module, evaluation module and display module are connected in sequence, and the evaluation module is also connected to the storage module; Image acquisition module, which uses a scanner with a high-definition camera to obtain user-designed images to be rated; Image processing module, which designs image preprocessing for users to be rated; The model building module, based on deep learning, builds a user design evaluation model consisting of a backbone network and five scoring branch networks; Training module, which trains the user design evaluation model built based on deep learning; Evaluation module, which scores user designs and outputs the scoring results; Storage module, storing user images and corresponding rating results; The display module displays the user-designed images and corresponding scoring results.