Assisted design system based on artificial intelligence

Abstract

The invention relates to the technical field of gift design, and particularly discloses an aided design system based on artificial intelligence. The system comprises an interaction module which performs multiple rounds of interaction with a user based on an AI interaction engine, analyzes user input information and obtains a design demand; the data storage module stores a design library and a user historical design version library; the processing module is in communication connection with the interaction module and is used for receiving the design requirements acquired by the interaction module, generating a 2D design drawing and converting the 2D design drawing into a 3D model; the system can interact with a user, a 2D design drawing and a 3D model are automatically generated according to input information of the user, and the design feasibility is verified in real time; operation is easy and convenient, flexibility is high, and production is adapted.

Description

Technical Field

[0001] The present invention relates to the technical field of gift design, and in particular to an artificial intelligence-based auxiliary design system. Background Art

[0002] Currently, gift design solutions on the market are mainly divided into two categories: traditional design software solutions and template-based design tool solutions, both of which have significant technical bottlenecks.

[0003] Traditional design software, such as Adobe Illustrator and CorelDRAW, requires users to possess professional design skills. They must manually draw vector graphics and undergo multiple reviews by merchants to adjust production parameters, resulting in lengthy design cycles and high labor costs (accounting for over 30%). For example, ordinary users require months of training to master complex path editing and layer management operations, and the design drafts often fail to produce as many as 40% of the first editions due to violations of production processes (such as embroidery thread spacing <0.2mm). On the other hand, although templated design tools simplify operations through preset element libraries (icons, text, etc.), their number of templates is limited and the element combinations are rigid. Users can only mechanically collage fixed elements by dragging and dropping, which makes it difficult to meet personalized needs, resulting in serious design homogeneity (for example, 80% of users use the same gift template). Existing solutions lack production adaptability, and designers need to repeatedly manually verify parameters such as material thickness and opening and closing structure, making it impossible to achieve full process automation from design to production.

[0004] Therefore, it is necessary to provide a new technical solution to overcome the above-mentioned defects. Summary of the Invention

[0005] The purpose of the present invention is to provide an artificial intelligence-based auxiliary design system that can effectively solve the above-mentioned technical problems.

[0006] In order to achieve the purpose of the present invention, the following technical solutions are adopted:

[0007] AI-based auxiliary design system, including:

[0008] The interactive module conducts multiple rounds of interactions with users based on the AI ​​interaction engine, analyzes user input information and obtains design requirements;

[0009] Data storage module, which stores design library and user historical design version library;

[0010] The processing module is in communication with the interaction module, and is used to receive the design requirements obtained by the interaction module, generate a 2D design drawing, and convert the 2D design drawing into a 3D model; the processing module is embedded with process rules to verify the feasibility of the design in real time.

[0011] Furthermore, the input information includes: a text description input by the user in a text input box, a sketch hand-drawn by the user on a sketch drawing board, a design outline drawn by the user using a shape tool, and a reference sketch uploaded by the user.

[0012] Furthermore, the AI ​​interaction engine extracts keywords from text descriptions through the NLP model; or performs semantic segmentation on the sketches input by the user through the U-Net model, extracts the main structure, and separates the background graffiti; when the user inputs text and sketches at the same time, the semantic information is aligned through the cross-modal attention mechanism.

[0013] Furthermore, the analysis of the text description includes: identifying and unifying the language, converting the mixed Chinese and English text into a single language; removing special characters and stop words; and performing word segmentation on the pre-processed text description to extract keywords, phrases or industry terms.

[0014] Furthermore, the processing module includes: a requirement parsing unit for receiving the design requirements acquired by the interaction module;

[0015] a design generation unit, which retrieves similar designs from the data storage module based on a Faiss vector index and generates a 2D design drawing based on the design requirements;

[0016] 3D modeling unit: converting the 2D design drawing into a 3D model through the Blender engine;

[0017] The production constraint verification unit is used to embed category process rules and verify the feasibility of design production in real time to ensure that the vector diagram conforms to the production process.

[0018] Furthermore, converting a 2D design drawing into a 3D model through the Blender engine includes: receiving the 2D design drawing output by the processing module, and using the svgpathtools library to parse the Bezier curves and straight line segments of the 2D design drawing; converting the path into Blender's Bezier Curve object and setting geometric properties; identifying dotted lines or specific color marks in the 2D design drawing as folding lines; generating solid thickness along the folding lines through Blender's Solidify modifier; automatically filling closed paths with planes and generating tubular structures for non-closed paths; presetting material node groups and embedding process rules; and dynamically loading corresponding node groups according to the material type selected by the user.

[0019] Furthermore, a real-time rendering engine based on WebGL is integrated into the interactive module, and WebWorker is used to separate rendering threads to achieve parallel processing of geometric calculation and material shading.

[0020] Furthermore, it also includes: an output module: used to convert the generated 3D model into a standard format file.

[0021] Compared with the existing technology, the present invention has the following beneficial effects: the present invention can interact with the user, automatically generate 2D design drawings and 3D models according to the user's input information, and verify the feasibility of the design in real time; it is easy to operate, highly flexible, and adaptable to production. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0023] Figure 1 This is a flow chart of the artificial intelligence-based auxiliary design system of the present invention;

[0024] Figure 2 The figure is a flow chart of the artificial intelligence-based auxiliary design system of the present invention. DETAILED DESCRIPTION

[0025] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments.

[0026] It should be understood that, although the various steps in the flow chart of each embodiment of the present invention are shown in sequence according to the indication of the arrows, these steps are not necessarily performed in sequence according to the order indicated by the arrows. Unless otherwise specified herein, the execution of these steps is not strictly limited in order, and these steps can be performed in other orders. Moreover, at least a portion of the steps in each embodiment may include a plurality of sub-steps or a plurality of stages, and these sub-steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these sub-steps or stages is not necessarily performed in sequence, but can be performed in turn or alternately with at least a portion of other steps or sub-steps or stages of other steps.

[0027] like Figures 1 to 2 As shown, the artificial intelligence-based auxiliary design system of the present invention includes:

[0028] The interactive module conducts multiple rounds of interactions with users based on the AI ​​interaction engine, parsing user input and obtaining design requirements. Design requirements include: gift type (box, bag, bottle), design style (retro, minimalist), and production process constraints (material type, printing process).

[0029] Input information includes text descriptions entered by the user in the text input box, sketches drawn by the user on the sketch board, design outlines drawn by the user using shape tools, and reference sketches uploaded by the user;

[0030] Specifically, the AI interaction engine extracts keywords in the text description through the NLP model, such as materials, styles, elements, etc.; or segments the main body of the sketch through the U-Net model; when the user inputs text and sketch simultaneously, the AI interaction engine aligns the text and sketch information through the cross-modal attention mechanism; the user can use text description or simple sketches, which is easy to operate.

[0031] The interaction module analyzes the user input information and obtains the design requirements, including: parsing the text description, parsing the sketch, and generating design requirements.

[0032] Parsing the text description includes: obtaining the text description, identifying and unifying the language, such as converting the "gold dragon pattern medal, vintage" with mixed Chinese and English into "gold dragon pattern medal, vintage style";

[0033] Removing special characters and filtering meaningless stop words, such as "de", "then", etc.;

[0034] Performing word segmentation on the preprocessed text description, extracting keywords, phrases or industry terms, such as "vintage", "hot stamping", "leather", etc.; word segmentation processing includes: using a text processing tool to perform word segmentation on the text description; the text processing tool includes: jieba segmentation or THULAC; the keyword extraction method includes: the TextRank algorithm based on BERT semantic enhancement or the sequence annotation model based on BERT, and the model is implemented through the BERT keyword extraction library;

[0035] The extraction of phrases and industry terms includes performing词性 analysis on the word segmentation results through a词性标注 tool, combining different词性词汇 to extract effective phrases and terms, and the词性标注 tool includes jieba词性标注 or LTP;

[0036] Parsing the sketch includes: unifying the size to a预定 pixel (such as 512×512 pixels), centering for display to eliminate scale deviation; grayscale and binary processing to convert the color sketch into a black and white line drawing; repairing broken lines and closing incomplete contours through morphological operations (dilation / erosion); performing semantic segmentation on the user input sketch through the U-Net model to extract the main structure (such as box body, bag body, bottle body, etc.), and at the same time separating background graffiti (such as misdrawn lines, text annotations); the segmentation result is repaired for broken lines through morphological operations (dilation / erosion) and used as the input of the 3D modeling unit to generate a parametric model; OCR recognizes the handwritten text in the sketch, deletes the text layer after extracting the handwritten text, and only retains the graphic data. When the user uploads a sketch file, before preprocessing, use a super-resolution model to improve the resolution of low-resolution pictures, and correct the tilt deformation caused by mobile phone shooting through the perspective correction method. It should be noted that there are some unclear or incorrect expressions in the original Chinese text, such as "词性" which is not clear in meaning. I have translated it as is as much as possible while trying to make sense of the overall context. If there are specific corrections or clarifications needed for the original text, it will help to improve the accuracy of the translation.

[0037] The matching degree between text and sketch is calculated through the cross-modal attention mechanism, and the feature similarity is calculated using the CLIP model fine-tuned with the gift design dataset. When the similarity is lower than the threshold (such as <0.7), the error correction process is triggered. For example: the text emphasizes "circle" but the sketch shows a square; the system error correction process is triggered, prompting the user to confirm the design intention.

[0038] As can be seen from the above description, in this embodiment, user input is parsed through natural language processing (NLP), image segmentation (U-Net), and cross-modal alignment (CLIP). Ordinary users do not need professional design skills and can express their design intentions through text descriptions or simple sketches. It also supports mixed Chinese and English input and automatic error correction, which improves the interaction tolerance rate, and the design intention recognition accuracy is greater than 90%.

[0039] Data storage module: Communicates with the interaction module and the processing module respectively, and stores the design library and the user's historical design version library. The reuse rate of the user's historical design version library is improved, reducing duplication of work; the design library includes: industry terms, design library, and material library. Industry terms include: keywords, phrases, and industry terms. The design library includes: vector library and the user's historical design version library. The vector library contains at least 100,000 vector images, including various cross-border e-commerce gifts, such as: hardware medals, commemorative coins, brooches, etc. The vector images include: 3D modeling and 2D plane drawings. The material library includes: leather, metal, plastic, glass, enamel, etc. Optionally, the design templates can be stylized through image clustering and style transfer algorithms.

[0040] The processing module is connected to the interaction module for communication.

[0041] The processing module includes: a requirement parsing unit for receiving the design requirements obtained by the interaction module;

[0042] The design generation unit retrieves similar designs from the data storage module based on the Faiss vector index and generates 2D design drawings based on the design requirements. 4K resolution design drawings are preferred. 2D design drawing generation takes ≤ 10 seconds, while traditional manual drawing takes 2 to 4 hours.

[0043] 3D modeling unit: converts 2D design drawings into 3D models through the Blender engine; specifically, the processing module extracts the text description entered by the user in the interactive module and determines whether the 2D design drawing needs to be modified; if so, the design requirements are updated based on the feedback information and the 2D design drawing is generated again; if not, the 2D design drawing is converted into a 3D model through the Blender engine.

[0044] The production constraint verification unit is used to embed category process rules, verify the feasibility of design production in real time, and ensure that the vector diagram complies with the production process; dynamically adjust design parameters (such as medal thickness and hanging ring position); process rules come from the industry standard database and are dynamically updated to the material library; verification methods include geometric analysis and physical simulation; such as the prompt "embroidery thread spacing must be greater than 0.2mm"; through production constraint verification, process rules are used as models to generate hard constraints, and the finished product qualification rate is increased from 65% to 95%.

[0045] Converting a 2D design drawing into a 3D model through the Blender engine includes: receiving the 2D design drawing output by the processing module, using the svgpathtools library to parse the Bezier curves and straight line segments of the 2D design drawing; converting the path into Blender's Bezier Curve object and setting the geometric properties; identifying the dotted lines or specific color marks in the 2D design drawing as folding lines; generating solid thickness along the folding lines through Blender's Solidify modifier; automatically filling closed paths with planes, and generating tubular structures for non-closed paths; presetting material node groups, embedding process rules, including PBR maps and physical properties such as roughness and refractive index; dynamically loading the corresponding node group according to the material type selected by the user; detecting the minimum bending radius through the geometric analysis algorithm, and automatically adding reinforcement ribs to the thin areas; the Blender engine can be selected as the StyleGAN3 engine, which requires sacrificing some detail accuracy.

[0046] Optionally, integrate the Bullet physics engine into Blender to simulate the opening and closing of gift boxes and load-bearing tests, for example, triggering a warning when the opening and closing angle of the box lid exceeds 90 degrees.

[0047] The interactive module is also used to display the 2D design drawings and 3D renderings of the design plan, realizing real-time preview of the 2D design drawings and 3D renderings. The 3D real-time preview function enhances user participation and increases the conversion rate by 40%.

[0048] The interactive module integrates a real-time rendering engine based on WebGL, and implements user interaction through the following steps: receiving 3D model files generated by Blender through the WebSocket protocol and performing LOD grading on the 3D model files; dynamically loading model data of the corresponding level according to the view distance, loading high-precision meshes (number of faces ≥ 100,000) in the near view, and switching to simplified models (number of faces ≤ 10,000) in the distant view; when the user rotates, zooms or clicks on the functional area of ​​the model, the coordinates of the target area are identified through ray collision detection; encapsulate user operation events (such as size adjustment and material replacement) as JSON instructions and push them to the processing module in real time for compliance verification; use WebWorker to separate the rendering thread, This system achieves parallel processing of geometric calculations and material shading; material data and geometric data are stored separately, and PBR maps are dynamically loaded based on the user's perspective; when the user clicks on the gift box handle area, ray detection obtains vertex coordinates and triggers the parameter editing panel. After the user changes the handle thickness from 1.2mm to 2.0mm, the WebGL engine sends the modification instruction to the processing module to verify the minimum wall thickness rule; after the verification passes, the processing module generates a new version of the 2D design drawing and triggers Blender to rebuild the 3D model; the WebGL engine receives the updated model file and seamlessly refreshes the rendering view; through real-time rendering, it is possible to replace materials, material maps, lighting models, and dynamic parameter adjustments, such as replacing metal materials with enamel materials. WebGL and Blender synchronize real-time data, use a material map compression algorithm (reducing transmission bandwidth by 80%), and the Unity engine can be used as the real-time rendering engine for WebGL, which requires higher terminal computing power.

[0049] The output module is electrically connected to the processing module and is used to convert the 3D model generated by the 3D modeling unit into a standard format file and transmit it to the merchant's production system through the API. Standard format files include JPG, PNG, SVG, DXF, STP, PDF format files, etc. The production file is standardized and the error rate of manual conversion is reduced.

[0050] The design cycle of this invention is shortened from 7 days to 1 hour, the user self-completion rate exceeds 80%, and it can reduce the merchant's design manpower input, and the labor cost is reduced by 50%.

[0051] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0052] The technical features of the above-mentioned embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above-mentioned embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0053] It should be understood that those skilled in the art can make improvements or changes based on the above description, and all such improvements and changes should fall within the scope of protection of the appended claims of the present invention.

Claims

1. The artificial intelligence-based auxiliary design system is characterized by: include: The interactive module conducts multiple rounds of interactions with users based on the AI ​​interaction engine, analyzes user input information and obtains design requirements; Data storage module, which stores design library and user historical design version library; The processing module is in communication with the interaction module, and is used to receive the design requirements obtained by the interaction module, generate a 2D design drawing, and convert the 2D design drawing into a 3D model; the processing module is embedded with process rules to verify the feasibility of the design in real time.

2. The artificial intelligence-based auxiliary design system according to claim 1, characterized in that: The input information includes: a text description input by the user in a text input box, a sketch hand-drawn by the user on a sketch drawing board, a design outline drawn by the user using a shape tool, and a reference sketch uploaded by the user.

3. The artificial intelligence-based auxiliary design system according to claim 1, characterized in that: The AI ​​interaction engine extracts keywords from text descriptions through the NLP model; or performs semantic segmentation on user-input sketches through the U-Net model, extracts the main structure, and separates background graffiti; when the user inputs text and sketches at the same time, the semantic information is aligned through the cross-modal attention mechanism.

4. The artificial intelligence-based auxiliary design system according to claim 3, characterized in that: The analysis of the text description includes: identifying and unifying the language, converting the mixed Chinese and English text into a single language; removing special characters and stop words; performing word segmentation on the pre-processed text description, and extracting keywords, phrases or industry terms therein.

5. The artificial intelligence-based auxiliary design system according to claim 1, characterized in that: The processing module includes: a requirement parsing unit for receiving the design requirements acquired by the interaction module; a design generation unit, which retrieves similar designs from the data storage module based on a Faiss vector index and generates a 2D design drawing based on the design requirements; 3D modeling unit: converting the 2D design drawing into a 3D model through the Blender engine; The production constraint verification unit is used to embed category process rules and verify the feasibility of design production in real time to ensure that the vector diagram conforms to the production process.

6. The artificial intelligence-based auxiliary design system according to claim 5, characterized in that: Converting a 2D design drawing into a 3D model through the Blender engine includes: receiving the 2D design drawing output by the processing module, parsing the Bezier curves and straight line segments of the 2D design drawing using the svgpathtools library; converting the path into a Bezier Curve object of Blender and setting geometric properties; identifying dotted lines or specific color marks in the 2D design drawing as folding lines; generating solid thickness along the folding lines through Blender's Solidify modifier; automatically filling closed paths with planes and generating tubular structures for non-closed paths; presetting material node groups and embedding process rules; and dynamically loading corresponding node groups according to the material type selected by the user.

7. The artificial intelligence-based auxiliary design system according to claim 6, characterized in that: The interactive module integrates a real-time rendering engine based on WebGL, uses WebWorker to separate rendering threads, and realizes parallel processing of geometric calculation and material shading.

8. The artificial intelligence-based auxiliary design system according to claim 1, characterized in that: Also includes: Output module: used for converting the generated 3D model into a standard format file.

Images