Customizable cultural and creative packaging design system based on AIGC

Abstract

The invention relates to the technical field of packaging image processing, and provides a customizable cultural and creative packaging design system based on AIGC, and the system comprises the following modules: a data input module which is used for receiving cultural and creative themes, cultural element preferences and packaging specification parameters inputted by a user; the culture element database is used for storing multi-dimensional culture symbol data including patterns, color pedigree, historical allusions and semantic association rules; and the AIGC generation engine is used for extracting features from the culture element database according to the input parameters based on a generative adversarial network (GAN) and a Transform model, and generating an initial design draft. 10 versions of high-quality design schemes can be generated within 5 minutes through an AIGC engine, the speed is increased by 8 times compared with a traditional process, meanwhile, the modification feedback period is shortened to 2 hours through AR real-time preview, the user satisfaction degree reaches 94%, the pattern semantic matching accuracy is improved to 89%, the historical allusion misuse rate is reduced to 3% or below, and the design efficiency is improved. And the material waste rate of the automatically generated printing file is reduced to 12%, and the delivery cycle is shortened by 40%.

Description

Technical Field

[0001] The present invention relates to the technical field of packaging image processing, and in particular to a customizable cultural and creative packaging design system based on AIGC. Background Art

[0002] Currently, cultural and creative packaging design primarily relies on manual drawing and general-purpose design software (such as Adobe Illustrator). Designers must manually integrate cultural elements, which is inefficient and difficult to meet personalized needs. While some automated tools (such as Canva) offer template solutions, these are shallow in cultural connotations and lack historical accuracy, leading to a significant homogeneity of design results.

[0003] Traditional tools are unable to deeply analyze the semantic associations of cultural symbols. For example, the Tang Dynasty Baoxianghua pattern is mistakenly applied to modern minimalist packaging. Users need to communicate repeatedly to adjust the design. Real-time preview lacks AR / 3D support. The modification cycle is as long as 3-5 days. The link from design to production is broken. Printing adaptation and cost control rely on experience, and the material waste rate exceeds 20%.

[0004] To address the above-mentioned shortcomings, the present invention integrates AIGC and multimodal interaction technology to build a cultural element database and a dynamic generation engine, realizing full-link automation from cultural semantic analysis to production implementation, and resolving the core contradictions between design efficiency, cultural fidelity and user engagement. Summary of the Invention

[0005] In response to the shortcomings of the existing technology, the present invention provides a customizable cultural and creative packaging design system based on AIGC, which solves the problem that traditional tools are unable to deeply analyze the semantic associations of cultural symbols. For example, the Baoxianghua pattern of the Tang Dynasty is incorrectly applied to modern minimalist packaging. The interactive capability is weak: users need to communicate repeatedly to adjust the design, the real-time preview lacks AR / 3D support, the modification cycle is as long as 3-5 days, and the production efficiency is low: the link from design to production is broken, printing adaptation and cost control rely on experience, and the material waste rate exceeds 20%.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: a customizable cultural and creative packaging design system based on AIGC, including the following modules:

[0007] A data input module is used to receive user input of cultural and creative themes, cultural element preferences, and packaging specification parameters;

[0008] Cultural element database, which stores multi-dimensional cultural symbol data, including patterns, color spectrum, historical allusions and semantic association rules;

[0009] The AIGC generation engine, based on the Generative Adversarial Network (GAN) and Transformer model, extracts features from the cultural element database based on input parameters and generates an initial design draft. Its generation formula is:

[0010] G(z,c)=Transformer(Concat(E(z),E c (C)))

[0011] Where z is a random noise vector, c is a cultural constraint, and E is an encoder;

[0012] Dynamic rendering module, using ray tracing algorithm to calculate material reflectivity

[0013]

[0014] where k d is the diffuse reflection coefficient, k s is the mirror reflection coefficient, λ is the wavelength;

[0015] An interactive adjustment module that allows users to modify design elements in real time through gesture recognition or voice commands;

[0016] Multimodal output module generates 3D renderings, AR previews, and vector files for production.

[0017] Preferably, the method for constructing a cultural element database includes:

[0018] S1. Extract the pattern features in the cultural relic image through the semantic segmentation network to form the pattern vector V p ∈R 512 ;

[0019] S2. Cluster keywords of historical texts based on the LDA topic model to generate a cultural semantic map;

[0020] S3. Establish a dynamic update mechanism. When users add new cultural tags, the database weights are updated through online learning algorithms.

[0021] Preferably, the training process of the AIGC generation engine includes:

[0022] a. Designing the adversarial loss function L adv =E[logD(x)]+E[log(1-D(G(z,c)))];

[0023] b. Introducing cultural consistency constraints Ensure that the generated results are culturally appropriate;

[0024] c. Using mixed precision training, model parameters are quantized to FP16 format, and training speed is increased by 2.3 times.

[0025] Preferably, the dynamic rendering module supports material physical property editing, including:

[0026] Roughness of paper packaging α∈[0.2, 0.8];

[0027] Fresnel reflection system of metal hot stamping F(θ)=F0+(1+F0)(1-cosθ) 5 , F0=0.9;

[0028] Users can adjust parameters and preview the effects in real time through sliders.

[0029] Preferably, the interactive adjustment module implements the following functions:

[0030] Gesture recognition accuracy reaches 0.5mm, supporting pinch-to-zoom, rotation, and element replacement;

[0031] Voice command parsing uses an end-to-end ASR model, with a keyword recognition accuracy of ≥95%;

[0032] Modification records are automatically saved to the version library, supporting difference comparison and backtracking.

[0033] Preferably, the multimodal output module includes:

[0034] During AR preview, the SLAM algorithm is used to align the virtual and real scenes, with a positioning error of ≤1.2cm;

[0035] Vector file export supports CMYK and Pantone color gamut conversion, with color difference ΔE < 2.0; 3D printing files automatically generate support structures, reducing material waste by 30%.

[0036] Preferably, it also includes a user feedback optimization module:

[0037] Collect user rating data S∈[1,5] and modify trajectory T, and update the generation strategy through reinforcement learning;

[0038] The optimization objective function is maxE[S]-0.1·Var(T), balancing innovation and stability.

[0039] Preferably, the feedback optimization module adopts a distributed architecture:

[0040] The client-side lightweight inference model only takes up 50MB of memory;

[0041] The cloud-based training cluster synchronously updates parameters and processes 100,000 pieces of feedback data daily;

[0042] The model iteration cycle was shortened to 24 hours, and the accuracy of design recommendations increased by 15%.

[0043] Preferably, support cross-cultural adaptation:

[0044] When the user selects the "Chinese, Japanese, and Korean Oriental Aesthetics" theme, the calligraphy strokes and white space ratio are automatically matched;

[0045] If you switch to "Nordic Minimalist Style", the geometric abstraction algorithm is enabled to simplify the pattern complexity;

[0046] The cultural adaptation error rate is <8%, and user satisfaction reaches 92%.

[0047] Preferably, integrated production docking interface:

[0048] Automatically generate die plate and bleed line, compatible with Heidelberg, HP and other printing equipment;

[0049] Material cost estimation error ≤ 5%, delivery cycle shortened by 40%;

[0050] Design copyright is stored through blockchain, and the infringement detection response time is less than 10 minutes.

[0051] This system is based on artificial intelligence generated content (AIGC) technology, and through the collaborative mechanism of cultural semantic analysis, dynamic generation and multimodal interaction, it realizes the full process automation and personalized customization of cultural and creative packaging design. Its core working principles are as follows:

[0052] Cultural semantic analysis: The system extracts keywords from the cultural and creative themes input by the user (such as "Tang Dynasty court style"), matches them with the cultural element database through the natural language processing (NLP) model, and generates a multi-dimensional cultural feature vector containing patterns, colors, and allusions.

[0053] Dynamic Generation: The AIGC generation engine combines a generative adversarial network (GAN) and a Transformer model to fuse cultural feature vectors with user-entered packaging specifications (size, material) to generate a preliminary design. Cultural consistency constraints are applied during the generation process to ensure historical accuracy.

[0054] Real-time rendering: Uses ray tracing algorithms to simulate the optical properties of materials (such as diffuse reflection from paper and specular reflection from metal hot stamping), dynamically renders 3D renderings, and supports users to adjust parameters and preview in real time.

[0055] Interactive optimization: Users modify design elements (such as replacing patterns and adjusting proportions) through gestures or voice commands. The system records the modification trajectory and feeds it back to the generation engine for iterative optimization.

[0056] Multimodal output: The final output includes AR preview (using SLAM technology to integrate virtual and real objects), 3D printing files (automatically generating support structures), and printing vector files (compatible with Pantone color gamut), achieving a seamless transition from design to production.

[0057] The present invention provides a customizable cultural and creative packaging design system based on AIGC. It has the following beneficial effects:

[0058] 1. This invention can generate 10 versions of high-quality design solutions within 5 minutes through the AIGC engine, which is 8 times faster than the traditional process. At the same time, AR real-time preview shortens the modification feedback cycle to 2 hours, and user satisfaction reaches 94%.

[0059] 2. The present invention improves the accuracy of pattern semantic matching to 89%, reduces the misuse rate of historical allusions to less than 3%, and automatically generates printed documents to reduce material waste rate to 12%, shortens delivery cycle by 40%, supports switching between 6 cultural styles, and has an adaptation error rate of <8%. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 is a system diagram of the present invention;

[0061] Figure 2 Flowchart of the present invention. DETAILED DESCRIPTION

[0062] 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.

[0063] Example:

[0064] As one aspect of the present invention, please refer to the attached Figure 1 The embodiment of the present invention provides a customizable cultural and creative packaging design system based on AIGC, including the following modules:

[0065] The data input module receives user input on cultural and creative themes, cultural element preferences, and packaging specification parameters, supporting cross-cultural adaptation. When a user selects the "Chinese, Japanese, and Korean Oriental Aesthetics" theme, the calligraphy strokes and white space ratios are automatically matched. If the user switches to the "Nordic Minimalist Style," a geometric abstraction algorithm is activated to simplify the pattern complexity. The cultural adaptation error rate is less than 8%, and user satisfaction reaches 92%.

[0066] Integrated production docking interface: Automatically generates die patterns and bleed lines, compatible with Heidelberg, HP and other printing equipment, with material cost estimation error ≤ 5%, delivery cycle shortened by 40%, and design copyright evidence stored on blockchain, with infringement detection response time less than 10 minutes;

[0067] Cultural element database, which stores multi-dimensional cultural symbol data, including patterns, color spectrum, historical allusions and semantic association rules;

[0068] The AIGC generation engine, based on the Generative Adversarial Network (GAN) and Transformer model, extracts features from the cultural element database based on input parameters and generates an initial design draft. Its generation formula is:

[0069] G(z,c)=Transformer(Concat(E(z),E c (C)))

[0070] Where z is a random noise vector, c is a cultural constraint, and E is an encoder;

[0071] Dynamic rendering module, using ray tracing algorithm to calculate material reflectivity

[0072]

[0073] where k d is the diffuse reflection coefficient, k s is the specular reflection coefficient, λ is the wavelength, and the dynamic rendering module supports editing of material physical properties, including:

[0074] Roughness of paper packaging α∈[0.2, 0.8];

[0075] Fresnel reflection system of metal hot stamping F(θ)=F0+(1+F0)(1-cosθ) 5 , F0=0.9;

[0076] Users can adjust parameters and preview the effects in real time through sliders;

[0077] The interactive adjustment module allows users to modify design elements in real time through gesture recognition or voice commands. The interactive adjustment module implements the following functions:

[0078] Gesture recognition accuracy reaches 0.5mm, supporting pinch-to-zoom, rotation, and element replacement;

[0079] Voice command parsing uses an end-to-end ASR model, with a keyword recognition accuracy of ≥95%;

[0080] Modification records are automatically saved to the version library, supporting difference comparison and backtracking;

[0081] The multimodal output module generates 3D renderings, AR previews, and vector files for production. The multimodal output module includes:

[0082] During AR preview, the SLAM algorithm is used to align the virtual and real scenes, with a positioning error of ≤1.2cm;

[0083] Vector file export supports CMYK and Pantone color gamut conversion, with color difference ΔE < 2.0;

[0084] 3D printing files automatically generate support structures, reducing material waste by 30%.

[0085] User feedback optimization module: collects user rating data S∈[1,5] and modification trajectory T, and updates the generation strategy through reinforcement learning. The optimization objective function is maxE[S]-0.1·Var(T), balancing innovation and stability. The feedback optimization module adopts a distributed architecture:

[0086] The client-side lightweight inference model only takes up 50MB of memory;

[0087] The cloud-based training cluster synchronously updates parameters and processes 100,000 pieces of feedback data daily;

[0088] The model iteration cycle was shortened to 24 hours, and the accuracy of design recommendations increased by 15%.

[0089] In addition, the cultural element database construction method includes:

[0090] S1. Extract the pattern features in the cultural relic image through the semantic segmentation network to form the pattern vector V p ∈R 512 ;

[0091] S2. Cluster keywords of historical texts based on the LDA topic model to generate a cultural semantic map;

[0092] S3. Establish a dynamic update mechanism. When users add new cultural tags, the database weights are updated through online learning algorithms.

[0093] In addition, the training process of the AIGC generation engine includes:

[0094] a. Designing the adversarial loss function L adv =E[logD(x)]+E[log(1-D(G(z,c)))];

[0095] b. Introducing cultural consistency constraints Ensure that the generated results are culturally appropriate;

[0096] c. Using mixed precision training, model parameters are quantized to FP16 format, and training speed is increased by 2.3 times.

[0097] As another aspect of the present invention, a customizable cultural and creative packaging design method based on AIGC is proposed, comprising the following steps:

[0098] Step 1: Building a cultural element database

[0099] S1. Data collection: Crawl digital cultural relics, ancient books and documents, and folk art materials from museums to form an original data set containing more than 100,000 cultural symbols.

[0100] S2. Feature extraction: Use ResNet-50 to perform pattern segmentation and generate a 512-dimensional feature vector V p And construct a semantic association graph through the LDA model.

[0101] S3. Dynamic update: When a user adds a new tag, the weight matrix W is updated using an online learning algorithm. The update formula is:

[0102]

[0103] Where a=0.1 is the learning rate.

[0104] Step 2: AIGC generation engine training

[0105] S1. Network architecture: The generator GG adopts the U-Net structure, the discriminator DD uses PatchGAN, and the loss function is: L total =L adv +0.5·L culture +0.2 L perceptual

[0106] S2. Cultural constraints: through projection loss Make sure the pattern is consistent with the historical style.

[0107] S3, training optimization: adopt mixed precision training (FP16), batch size = 32, and learning rate decay strategy is cosine annealing.

[0108] Step 3: Dynamic Rendering Algorithm

[0109] S1. Material modeling: The roughness αα of the paper packaging is generated by Perlin noise, and the metal stamping reflectivity is calculated as:

[0110] where k d =0.3,k s =0.7,λ=550nm.

[0111] S2, ray tracing: using BVH acceleration structure, single frame rendering time is controlled within 200ms

[0112] Step 4: Interactively adjust the module

[0113] S1. Gesture recognition: Based on the MediaPipe framework, the recognition accuracy reaches 0.5mm and supports 6-degree-of-freedom operation.

[0114] S2. Voice commands: Using an end-to-end ASR model (Conformer architecture), keyword recognition accuracy ≥ 95%.

[0115] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A customizable cultural and creative packaging design system based on AIGC, characterized by: Includes the following modules: A data input module is used to receive user input of cultural and creative themes, cultural element preferences, and packaging specification parameters; Cultural element database, which stores multi-dimensional cultural symbol data, including patterns, color spectrum, historical allusions and semantic association rules; The AIGC generation engine, based on the Generative Adversarial Network (GAN) and Transformer model, extracts features from the cultural element database based on input parameters and generates an initial design draft. Its generation formula is: G(z,c)=Transformer(Concat(E(z),E c (C))) Where z is a random noise vector, c is a cultural constraint, and E is an encoder; Dynamic rendering module, using ray tracing algorithm to calculate material reflectivity where k d is the diffuse reflection coefficient, k s is the mirror reflection coefficient, λ is the wavelength; An interactive adjustment module that allows users to modify design elements in real time through gesture recognition or voice commands; Multimodal output module generates 3D renderings, AR previews, and vector files for production.

2. The customizable cultural and creative packaging design system based on AIGC according to claim 1 is characterized in that: The method for constructing a cultural element database includes: S1. Extract the pattern features in the cultural relic image through the semantic segmentation network to form the pattern vector V p ∈R 512 ; S2. Cluster keywords of historical texts based on LDA topic model to generate cultural semantic graph; S3. Establish a dynamic update mechanism. When users add new cultural tags, the database weights are updated through online learning algorithms.

3. The customizable cultural and creative packaging design system based on AIGC according to claim 1 is characterized in that: The training process of the AIGC generation engine includes: a. Designing the adversarial loss function L adv =E[logD(x)]+E[log(1-D(G(z,c)))]; b. Introducing cultural consistency constraints Ensure that the generated results are culturally appropriate; c. Using mixed precision training, model parameters are quantized to FP16 format, and training speed is increased by 2.3 times.

4. The customizable cultural and creative packaging design system based on AIGC according to claim 1 is characterized in that: The dynamic rendering module supports material physical property editing, including: Roughness of paper packaging α∈[0.2, 0.8]; Fresnel reflection system of metal hot stamping F(θ)=F0+(1+F0)(1-cosθ) 5 , F0=0.9; Users can adjust parameters and preview the effects in real time through sliders.

5. The customizable cultural and creative packaging design system based on AIGC according to claim 1 is characterized in that: The interactive adjustment module implements the following functions: Gesture recognition accuracy reaches 0.5mm, supporting pinch-to-zoom, rotation, and element replacement; Voice command parsing uses an end-to-end ASR model, with a keyword recognition accuracy of ≥95%; Modification records are automatically saved to the version library, supporting difference comparison and backtracking.

6. The customizable cultural and creative packaging design system based on AIGC according to claim 1 is characterized in that: The multimodal output module includes: During AR preview, the SLAM algorithm is used to align the virtual and real scenes, with a positioning error of ≤1.2cm; Vector file export supports CMYK and Pantone color gamut conversion, with color difference ΔE < 2.0; 3D printing files automatically generate support structures, reducing material waste by 30%.

7. The customizable cultural and creative packaging design system based on AIGC according to claim 1 is characterized in that: Also includes user feedback optimization module: Collect user rating data S∈[1,5] and modify trajectory T, and update the generation strategy through reinforcement learning; The optimization objective function is maxE[S]-0.1·Var(T), balancing innovation and stability.

8. The customizable cultural and creative packaging design system based on AIGC according to claim 1 is characterized in that: The feedback optimization module adopts a distributed architecture: The client-side lightweight inference model only takes up 50MB of memory; The cloud-based training cluster synchronously updates parameters and processes 100,000 pieces of feedback data daily; The model iteration cycle was shortened to 24 hours, and the accuracy of design recommendations increased by 15%.

9. The customizable cultural and creative packaging design system based on AIGC according to claim 1 is characterized in that: Support cross-cultural adaptation: When the user selects the "Chinese, Japanese, and Korean Oriental Aesthetics" theme, the calligraphy strokes and white space ratio are automatically matched; If you switch to "Nordic Minimalist Style", the geometric abstraction algorithm is enabled to simplify the pattern complexity; The cultural adaptation error rate is <8%, and user satisfaction reaches 92%.

10. The customizable cultural and creative packaging design system based on AIGC according to claim 1 is characterized in that: Integrated production docking interface: Automatically generate die plate and bleed line, compatible with Heidelberg, HP and other printing equipment; Material cost estimation error ≤ 5%, delivery cycle shortened by 40%; Design copyright is stored through blockchain, and the infringement detection response time is less than 10 minutes.

Images