Product packaging design diagram generation method, device, medium and program product

Abstract

The specification provides a product packaging design diagram generation method, device, medium and program product, which first receives a product category of a product to be packaged input by a custom party, and then automatically maps the product category to product packaging attributes based on a knowledge graph, solving the selection difficulty problem caused by the lack of professional knowledge of the custom party; then, the existing producible packaging is recommended to the custom party as a candidate product packaging, and the structure information of the target product packaging selected by the custom party is obtained, the structure information is taken as a physical realizability constraint, and the product packaging structure attribute and the structure information are jointly input into a generative artificial intelligence model, so that the design diagram generated by the model meets the applicability requirement of the product packaging to the product and the structure constraint, reducing the repeated communication and secondary modification links, thereby reducing the communication cost and time cost in the traditional packaging customization process, and improving the generation efficiency of the product packaging design diagram.

Description

Technical Field

[0001] This specification relates to one or more embodiments in the field of computer technology, and more particularly to a method, apparatus, medium, and program product for generating product packaging design drawings. Background Technology

[0002] In e-commerce transactions, customized packaging is a common demand. Currently, the packaging customization process typically requires the client and seller to communicate manually to describe product characteristics and packaging needs, followed by a designer creating packaging design drafts based on the client's description. However, clients often lack packaging expertise and struggle to independently select suitable materials and structural types based on product characteristics. This leads to selection difficulties, inefficiency due to reliance on manual consultation, and a high risk of errors, making it difficult to accurately articulate feasible packaging requirements. Furthermore, the design and production stages are disconnected; designers focus only on visual style, resulting in packaging designs lacking physical feasibility constraints. This can cause the actual product packaging to deviate from expectations, requiring repeated communication and revisions. In summary, the traditional process of generating product packaging designs suffers from high communication costs and low efficiency. Summary of the Invention

[0003] In view of the above, one or more embodiments of this specification provide the following technical solutions: According to a first aspect of one or more embodiments of this specification, a method for generating product packaging design drawings is proposed. The method includes: performing intent recognition on input information from a customer to determine the product category included in the input information; searching for product packaging structural attributes that match product attribute information of the product category from a pre-constructed product packaging knowledge graph, wherein the product packaging knowledge graph is used to establish the association between product category, product attribute information, and product packaging structural attributes; recommending several candidate product packagings to the customer; determining the target product packaging selected by the customer from the several candidate product packagings; and obtaining a digital model of the target product packaging; wherein the several candidate product packagings are all existing manufacturable product packagings, and their digital models are used to provide physical realizability constraints; and inputting the matched product packaging structural attributes and the obtained digital model into a generative artificial intelligence model, so that the generative artificial intelligence model generates a product packaging design drawing that has the product packaging structural attributes and satisfies the physical realizability constraints provided by the digital model.

[0004] According to a second aspect of one or more embodiments of this specification, an electronic device is provided, comprising: a processor; a memory for storing processor-executable instructions; wherein the processor performs the executable instructions to implement the steps of the method as described in the first aspect of one or more embodiments of this specification.

[0005] According to a third aspect of one or more embodiments of this specification, a computer-readable storage medium is provided that stores computer instructions thereon, which, when executed by a processor, implement the steps of the method as described in the first aspect of one or more embodiments of this specification.

[0006] According to a fourth aspect of one or more embodiments of this specification, a computer program product is provided, comprising a computer program / instructions that, when executed by a processor, implement the steps of the method as described in the first aspect of one or more embodiments of this specification.

[0007] As can be seen from the above embodiments, this specification first receives the product category of the product to be packaged input by the customer, and then performs retrieval and matching based on a pre-built product packaging knowledge graph, automatically mapping the product category to the corresponding product packaging attributes. This solves the problem of selection difficulties caused by the customer's lack of professional knowledge and avoids reliance on manual consultation and communication bias. Next, it recommends existing manufacturable packaging as candidate product packaging to the customer, and obtains the structural information of the target product packaging selected by the customer. This structural information is used as a physical feasibility constraint, and the product packaging structural attributes and structural information are jointly input into the generative artificial intelligence model. This allows the design drawings generated by the model to simultaneously meet the product packaging's applicability requirements and structural constraints, reducing repeated communication and secondary modifications. This reduces the communication and time costs in the traditional packaging customization process and improves the generation efficiency of product packaging design drawings. Attached Figure Description

[0008] Figure 1 This is a schematic diagram illustrating an application scenario provided by an exemplary embodiment.

[0009] Figure 2 This is a flowchart of a method for generating a product packaging design drawing, provided as an exemplary embodiment.

[0010] Figure 3A This is a timing diagram of the generation process of a product packaging design drawing provided in an exemplary embodiment.

[0011] Figure 3B This is an exemplary embodiment of a product packaging design process, and provides an overall flowchart of the process.

[0012] Figure 4 This is a schematic diagram of an interactive interface for a product packaging design process, provided as an exemplary embodiment.

[0013] Figure 5A This is an exemplary embodiment of a partial redrawing process, and its overall flowchart is provided.

[0014] Figure 5B and Figure 5C This is an exemplary embodiment that provides a specific process for different stages in a partial redrawing process.

[0015] Figure 6 This is a schematic diagram of the structure of a device provided in an exemplary embodiment. Detailed Implementation

[0016] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.

[0017] The information about the customizer (including but not limited to the customizer's equipment information, personal information, etc.) and data (including but not limited to data used for analysis, stored data, and displayed data) involved in this manual are all information and data authorized by the customizer or fully authorized by all parties. The collection, use and processing of the relevant data shall comply with relevant laws, regulations and standards, and corresponding operation portals shall be provided for the customizer to choose to authorize or refuse.

[0018] Figure 1 This is a schematic diagram illustrating an online packaging customization application scenario provided by an exemplary embodiment. It should be noted that this application scenario is merely illustrative and does not constitute a limitation on the embodiments described in this specification. Figure 1 As shown, the system includes an e-commerce platform server 102, a client for merchants providing packaging customization services (hereinafter referred to as merchant client) 104, a client for manufacturers providing packaging production services (hereinafter referred to as manufacturer client) 106, and a client for the customizer (hereinafter referred to as customizer client) 108. Each of the merchant client 104, manufacturer client 106, and customizer client 108 can be an application or web browser running on a terminal device such as a personal computer, tablet, or smartphone.

[0019] During the packaging customization process, the customizer (U) first browses the product page through the customizer client (108) and selects "Customization Service," entering information such as the type and quantity of the product to be packaged, the product packaging attribute parameters (such as size, material, and process), and / or the target usage scenario (such as e-commerce transportation or gift box display). After receiving the customization request, the server (102) forwards it to the merchant client (104). Based on the customizer U's requirements, the merchant's designer generates a producible customization plan including design drawings, attribute parameters, and a quote, and uploads preliminary packaging design files (such as original design drafts and die-cutting diagrams) to the server (102) through the merchant client (104). The server (102) presents the plan and preview images to the customizer U, who can submit modification suggestions on the client (108) and return these suggestions to the merchant client (104) through the server (102). The merchant client (104) can update the packaging design files based on the modification suggestions until both parties confirm the final plan.

[0020] After the packaging design is confirmed, server 102 generates a product packaging production agreement and locks the packaging design files and attribute parameters. Subsequently, the merchant sends the final locked packaging design files and attribute parameters to server 102 via merchant client 104, which then forwards them to manufacturer client 106. Manufacturer client 106 then initiates the production process (such as printing and die-cutting) and transmits production progress data back to server 102 in real time, allowing the client U to track the process via client 108 until the finished product is delivered.

[0021] In the aforementioned process, the client (U) needs to communicate with the seller manually to describe the product characteristics and packaging requirements. Then, a designer creates product packaging design drafts based on the client's description. Because clients typically lack packaging expertise, they struggle to independently select suitable materials and structural types based on product characteristics. This leads to selection difficulties, inefficiency due to reliance on manual consultation, and a high risk of errors, making it difficult to accurately articulate feasible packaging requirements. Furthermore, the design and production stages are disconnected. Designers focus only on 2D visual styles, neglecting three-dimensional display effects. The resulting product packaging designs lack physical feasibility constraints, potentially causing the actual product packaging to deviate from expectations, requiring repeated communication and revisions. In summary, the traditional product packaging design process suffers from high communication costs and low efficiency.

[0022] Based on this, embodiments of this specification provide a method for generating product packaging design drawings. By using a knowledge graph to establish the association between the product category, product attribute information, and product packaging structural attributes of the product to be packaged, it can automatically map the product category input by the customizer to the corresponding product packaging structural attributes, solving the selection difficulties caused by the customizer's lack of professional knowledge. Furthermore, when designing product packaging, existing manufacturable packaging structural information is input as a physical feasibility constraint into the artificial intelligence model to generate product packaging design drawings, preventing the actual product packaging effect from deviating from expectations. In summary, the method of embodiments of this specification reduces repeated communication and secondary modifications, thereby reducing communication and time costs in the traditional packaging customization process and improving the efficiency of product packaging design drawing generation. The specific implementation details of the embodiments of this specification are illustrated below with reference to the accompanying drawings.

[0023] like Figure 2 The diagram shown is a flowchart of a method for generating product packaging design drawings according to an embodiment of this specification. The method includes: Step S12: Perform intent recognition on the input information from the customer to determine the product categories included in the input information; Step S14: Find the product packaging structure attributes that match the product attribute information of the product category from the pre-built product packaging knowledge graph. The product packaging knowledge graph is used to establish the relationship between product category, product attribute information and product packaging structure attributes. Step S16: Recommend several candidate product packaging to the customer, determine the target product packaging selected by the customer from the several candidate product packaging, and obtain the digital model of the target product packaging; the recommended candidate product packaging are all existing manufacturable product packaging, and their digital models are used to provide physical realizability constraints. Step S18: Input the matched product packaging structure attributes and the acquired digital model into the generative artificial intelligence model so that the generative artificial intelligence model can generate a product packaging design drawing that has product packaging structure attributes and meets the physical realizability constraints provided by the digital model.

[0024] The method in this embodiment can be executed by server 102 in the aforementioned application scenario.

[0025] In step S12, the customer U can generate input information via the customer client 108 in the form of voice or text. This input information can be natural language. The input information may include product categories (such as electronic products, food gift boxes, etc.). The server 102 can parse and recognize the intent of this input information using a built-in language model to extract the product category information. For example, when the customer U inputs "I need to customize a batch of cosmetic gift box packaging," the server 102 can recognize the product category as cosmetic gift boxes. Furthermore, the server 102 can also convert the various unstructured information obtained through intent recognition into structured data for subsequent processing.

[0026] In step S14, server 102 can pre-build a product packaging knowledge graph. This knowledge graph establishes the relationships between product categories, product attribute information, and product packaging structural attributes. Product attribute information describes the inherent physical characteristics (such as size, weight, material, and shape) and / or functional requirements (such as fragility, moisture resistance, display requirements, and stacking stability) of the product, directly affecting the protective and practical functions of the packaging. Product packaging structural attributes describe the packaging's construction (such as box design, layer division, divider layout, and sealing method) and / or its functions (such as shock absorption, visual display, easy opening, and resealability). For example, the product category "desktop crystal ball" has the product attribute information of being fragile and requiring display; therefore, its associated product packaging structural attributes could be a cardboard box with a window (structural construction) and internal foam recesses (shock absorption).

[0027] In some embodiments, the product packaging knowledge graph may include product category nodes (such as "cosmetic gift boxes," "electronic products," "food packaging," etc.), which serve as core classification identifiers for determining the product's category; product attribute nodes (such as size, weight, fragility, moisture protection requirements, etc.), used to describe the physical characteristics and functional features of the product; and product packaging structure attribute nodes (such as box structure, partition design, sealing method, window display, etc.), used to define the form and functional components of the product packaging. Furthermore, the graph includes edges connecting these nodes, used to describe the relationships between different product categories, product categories and product attribute information, and product attribute information and product packaging structure attributes, such as "belongs to," "applies to," "includes," "must match," etc. For example, the "desktop crystal ball" node is associated with the "must display" and "fragile" product attribute nodes via the "belong to" edge. These two product attribute nodes then point to the "cardboard box with window design" and "internal foam groove" product packaging structure attribute nodes via the "must match" edge, thus forming a complete logical chain of "category-attribute-structure."

[0028] In some embodiments, server 102 can construct a product packaging knowledge graph based on historical product packaging customization orders (hereinafter referred to as historical orders) using data mining and semantic analysis techniques. Specifically, server 102 first parses the historical orders, extracting unstructured text information input by the customizer (such as packaging purpose, category description and feature description of applicable products, style preferences, etc.) and structured attribute parameters locked after the historical order was completed (such as box structure, window design, waterproof material, process type, etc.). Subsequently, natural language processing techniques are used to identify entities in the unstructured text information and map the entities to corresponding nodes. For example, "category description of applicable products" is mapped to a product category node, "feature description of applicable products" is mapped to a product attribute information node, and structured attribute parameters are mapped to product packaging structure attribute nodes. Connection edges are then established between the corresponding nodes to generate a product packaging knowledge graph.

[0029] After determining the product categories included in the input information from the customizer, server 102 can search the product packaging knowledge graph for product packaging structure attributes that match the product attribute information of the product category. Specifically, server 102 can first use the product category as the starting point for retrieval, retrieve the product attribute information nodes connected to the corresponding product category nodes in the product packaging knowledge graph, and then further retrieve the product packaging structure attribute nodes connected to the product attribute information nodes. The product packaging structure attributes described by the retrieved product packaging structure attribute nodes are then determined as the product packaging structure attributes that match the product attribute information of the product category obtained in step S12.

[0030] In step S16, server 102 can recommend several candidate product packaging packages to the customer. Each recommended candidate product packaging package is an existing, manufacturable product packaging package, and these candidate product packaging packages have corresponding digital models (such as box-type 3D models, can-type 3D models, or bag-type 3D models) in server 102's database. The digital model refers to a digital representation of the candidate product packaging generated through data acquisition, modeling technology, or design software. It can accurately describe the geometric shape, structural attributes, material characteristics, and size proportions of the candidate product packaging, thereby providing physical feasibility constraints and supporting storage, analysis, simulation, modification, and interactive operations in a computer environment. In some embodiments, the number of candidate product packaging packages is greater than one, and different candidate product packaging packages can have different shapes, such as cylindrical, cubic, or irregular shapes.

[0031] In some embodiments, the digital model includes two-dimensional unfolded structural information, three-dimensional geometric information, and a processing and assembly constraint dataset. The two-dimensional unfolded structural information defines the dimensions, shapes, and two-dimensional arrangement relationships of each component in the product packaging design drawing in its unfolded planar state, such as the outlines, relative positions, and assembly references of each facet on the product packaging design drawing. The three-dimensional geometric information describes the three-dimensional form, three-dimensional positional relationships, and folding paths of each component in space, such as the forming angle of the box, the flipping trajectory of the flap, and the spatial angle between adjacent panels. The processing and assembly constraint dataset defines the geometric boundaries of the product packaging design drawing, the connection topology between components, and dimensional tolerance matching requirements, such as cutting boundaries, fold line positions, the mating clearance between the tongue and the slot, and the allowance of the adhesive area.

[0032] Generative AI models can determine the planar layout of each component in a product packaging design drawing based on 2D unfolded structural information. This planar layout can include the dimensions, shape contours, and 2D arrangement order of each component in its unfolded state. It can also construct the 3D layout of each component in the product packaging design drawing based on 3D geometric information. This 3D layout includes the solid geometry of each component, its relative position in space, and the path during folding motion. Furthermore, it can determine the spatial configuration information of the product packaging design drawing based on processing and assembly constraint datasets. This includes geometric boundaries (such as cutting contours and fold lines), the connection topology between components (such as insertion order and overlap direction), and dimensional tolerance matching relationships (such as male-female mating clearance and overlap allowance). After obtaining the above information, the generative AI model can integrate the planar layout, 3D layout, and spatial configuration information to generate a product packaging design drawing that satisfies the aforementioned constraints. This ensures that the surfaces, creases, and connection structures in the generated product packaging design drawing match the 2D and 3D definitions and processing and assembly constraints of the digital model.

[0033] After identifying several candidate product packages, server 102 can extract model parameters (including but not limited to 3D geometric data, material texture mapping information, and lighting environment parameters) of the corresponding digital models of each candidate product package, and encapsulate these model parameters into a data stream and send it to the client 108. Upon receiving the data stream, client 108 can call its locally integrated graphics rendering engine (such as WebGL or a native 3D engine) to construct and render 3D visualization images of each candidate product package in real time on the user interface (GUI) based on the received model parameters. These visualization images support interactive operations (such as 360-degree rotation and zooming to view details), enabling the client to intuitively compare the differences in shape, structure, and material between different candidate packages. The client can click or select on the displayed visualization images using the input device (such as a mouse or touchscreen) on client 108, generating selection instructions and sending them back to server 102. Server 102 can then determine the selected target product package based on the selection instructions and obtain its digital model. This embodiment recommends candidate product packaging to the customer, enabling the customer to intuitively compare the characteristics of different packaging shapes, structures, and size ratios, and thus independently choose the most suitable packaging solution according to their own needs.

[0034] In addition to planar layout, three-dimensional layout, and spatial configuration information, digital models can also include topological connection relationships, physical processing indicator lines, pasting areas, surface mapping parameters, windowed areas, and other information, which will not be listed here.

[0035] In step S18, server 102 can input the product packaging structure attributes matched in step S14 and the digital model obtained in step S16 into a generative artificial intelligence model, such as a multimodal language model or a diffusion model based on the Transformer architecture (such as DiT). The generative artificial intelligence model can generate product packaging design drawings that possess the aforementioned product packaging structure attributes and meet the physical realizability constraints provided by the aforementioned digital model. The generated product packaging design drawings can include at least one format, such as a 2D planar design drawing, a 3D scene diagram, and / or a die-cutting diagram of the product packaging. Among them, 2D planar design drawings refer to the graphic representation of each side of the product packaging unfolded into a two-dimensional view, which may include graphic elements, color labels, text information, dimension lines, and layout positioning marks of each side of the product packaging; 3D scene drawings refer to the product packaging simulation images with a sense of three-dimensionality and ambient lighting effects generated by three-dimensional modeling technology, which may include the geometric mesh model of the product packaging, material textures, lighting environment parameters, and viewpoint transformation or rotation animation sequences; die-cutting drawings refer to the structural engineering drawings of the product packaging used for die-cutting and creasing processing, which may include die-cutting outlines, creasing lines, crease lines, bleed areas, die numbers, material thickness, and bite positioning data.

[0036] In the example where the product packaging design includes 2D flat design, 3D scene rendering, and die-cutting pattern, the above product packaging design is generated based on the following method: First, the digital model can be converted into a 2D planar unfolded drawing. Specifically, the digital model of the product packaging can include structural feature lines, such as cutting lines, crease lines, and / or fold lines. The 2D planar unfolded drawing refers to the virtual cutting and flattening of the digital model along specific structural feature lines, resulting in a two-dimensional planar graphic that reflects the actual material shape, size proportions, and folding relationships of the product packaging.

[0037] Then, based on the 2D unfolded planar diagram and the product packaging structural attributes, a 2D planar design drawing of the product packaging can be generated. This process introduces the 2D unfolded planar diagram as a physical realizability constraint into the generation process, preventing the production of 2D planar design drawings that are only visually appealing but cannot be actually manufactured. The generative artificial intelligence model can render the 2D planar design drawing based on the product packaging structural attributes within the physical framework provided by the 2D unfolded planar diagram.

[0038] Subsequently, the 2D planar design is mapped onto the surface of the digital model using a UV mapping engine to generate a 3D scene of the product packaging. Specifically, the 2D planar design can be used as a texture map, mapped onto the geometric surface of the digital model via a UV mapping engine to restore the material texture and visual details of the product packaging. Combined with 3D rendering technology, it simulates the lighting reflections and shadow layers of a real environment, thereby generating a 3D scene of the product packaging that combines three-dimensional form with realistic visual effects. Since the 2D planar design is based on the digital model and conforms to the physical realizability constraints provided by the digital model, when the 2D design is mapped back to the surface of the 3D digital model using UV mapping technology, the structural features in the 2D planar design and the physical form of the digital model (such as folded edges, material thickness, and connection relationships) highly overlap in space, thus ensuring that the generated 3D scene also conforms to the physical realizability constraints provided by the digital model.

[0039] Next, based on the topology of the digital model, an engineering die-cutting pattern containing die-cutting line information is generated. Specifically, the geometric topology of the digital model can be analyzed first to extract structural features such as the cutting path defining the product packaging outline and the indentation trajectory indicating the folding position. Then, these structural features are projected onto a two-dimensional plane, and the die-cutting width is automatically compensated according to the processing accuracy requirements to draw an engineering die-cutting pattern containing precise die-cutting line information.

[0040] The generation process of the aforementioned 2D planar design drawings, 3D scene drawings, and die-cutting drawings shares the same model parameters of the digital model. The designer, the customer, and the factory collaborate based on the same set of standardized data, which reduces communication costs and ensures that the 2D planar design drawings and 3D scene drawings previewed by the customer are highly consistent with the final physical product packaging produced based on the die-cutting drawings. This reduces the return rate and modification costs caused by expected deviations.

[0041] In some embodiments, the server 102 can also parse the customization requirements for product packaging from the client's interaction information, and input the customization requirements, the matched product packaging structural attributes, and the acquired digital model into the generative artificial intelligence model, so that the generative artificial intelligence model generates a product packaging design drawing that meets the customization requirements, possesses the product packaging structural attributes, and satisfies the physical realizability constraints provided by the digital model. The client's interaction information may include the input information from step S12, or information selected by the client from preset interactive interface options. Customization requirements include, but are not limited to, visual style (e.g., minimalist, retro), physical specifications (e.g., size, shape), environmental constraints (e.g., recyclable materials, low-energy production), and / or target audience profiles (e.g., children, high-end business groups). This embodiment also uses the customization requirements as one of the inputs to the generative artificial intelligence model when generating the product packaging design drawing, enabling the generated product packaging design drawing to better meet the client's customization requirements.

[0042] In some embodiments, the generative AI model can also add visual elements to the generated product packaging design, such as brand logos, decorative backgrounds, gradient color bars, text descriptions, barcodes, QR codes, or product renderings, to make the product packaging display more vivid and aesthetically pleasing. However, if the added visual elements are located in the folded or die-cutting areas of the product packaging, after actual molding, the visual elements may be divided into two discontinuous parts by the fold lines, or part of the pattern may be cut off by the cutting areas, resulting in damage to the integrity of the visual elements and the expected display effect. To address this, the generative AI model can obtain physical processing indicator lines from the digital model, which are used to define the processing boundaries in the product packaging design, including cutting lines and fold lines. Cutting lines are used to determine the cutting area, i.e., the area retained as the outer contour of the finished packaging after die-cutting; fold lines are used to determine the folding area, i.e., the area that needs to be bent and deformed during the packaging forming process. Then, the layout position of the visual elements on the product packaging design can be determined based on the folding and cutting areas, so that the layout position is within the cutting area and outside the folding area. By constraining the position of visual elements using these preset conditions, the final product packaging ensures that each visual element remains complete, continuous, and logically arranged, preventing incomplete patterns or display anomalies caused by die-cutting or folding. Generative AI models can add visual elements to product packaging designs based on the planar layout, 3D layout, and spatial configuration information included in the digital model, as well as the layout positions determined through the aforementioned methods. This ensures that the added visual elements are constrained by their position within the layout, and that the structure of the product packaging design after adding visual elements is constrained by the planar layout, 3D layout, and spatial configuration information, preventing changes in the form of the product packaging design before and after adding visual elements and maintaining consistency in the product packaging design.

[0043] Furthermore, process rule verification can be performed on the generated product packaging design drawing to check whether the visual elements in the product packaging design drawing meet the layout position constraints. If not, an image inpainting model is used to locally correct the visual elements so that the corrected visual elements meet the layout position constraints. Specifically, each visual element in the product packaging design drawing can be traversed, and the boundary contours or pixel regions of the visual elements can be extracted. Each boundary contour or pixel region can be judged to meet the layout position constraints. If any visual element is partially or entirely located inside the folded area, distributed across the fold line, or exceeds the boundary of the cropping area, it is determined that it does not meet the layout position constraints. For visual elements that do not meet the constraints, a pre-trained image inpainting model can be used for local correction. The image inpainting model can adopt an architecture based on generative adversarial networks or diffusion models, which can identify the violation areas of visual elements (such as parts that cross fold lines) and generate repair content based on the non-violation visual features (such as color, texture, and pattern continuity) around the violation area to replace the violation area while maintaining the overall semantic consistency of the visual elements. During the correction process, if the visual element is text or a barcode, or other graphic with specific encoding rules, it can be first shifted to a safe position that meets the constraints, and then the background can be repaired at the original position. If the visual element is an unstructured pattern, such as a texture or gradient color bar, an image restoration model can be used to redraw the broken parts of the pattern on the fold lines or cut lines, making them continuous and complete within the constraint area. After the correction is completed, the process rule verification is performed again until all visual elements meet the constraints of the layout position.

[0044] In some embodiments, the client may wish to modify a local area on the product packaging design generated by the generative artificial intelligence model; this process is called local redrawing. In related technologies, the client needs to send the local redrawing request to the merchant via a chat tool. The merchant understands the request and forwards it to the designer, who then creates a rendering using drawing tools and returns it to the client. If the client is not satisfied, this process may need to be repeated multiple times. The entire process is cumbersome and inefficient. Therefore, this embodiment provides an editing interface on the client's client 108. The client can annotate the area to be edited on the product packaging design on this interface and input customization information. This customization information can be information described in natural language (referred to as natural language customization information), used to describe the intention to modify the area to be edited, such as color changes, pattern additions or replacements, element deletions, and / or minor size adjustments. Customization information is usually in conversational or unstructured text form, and its accuracy is low. The server 102 can obtain the aforementioned area to be edited and customization information.

[0045] Furthermore, an attribute parameter knowledge graph can be pre-constructed to establish a mapping relationship between customized information and standardized attribute parameters. Standardized attribute parameters refer to structured key-value pairs with clearly defined data types, value ranges, and semantic labels. For example, the customized information "change the logo color to red and slightly enlarge it" can be mapped to the standardized attribute parameters: {element: logo; color: "#FF0000"; magnification ratio: 1.1}. The attribute parameter knowledge graph can include information description nodes and parameter nodes, corresponding to the description information and standardized attribute parameters, respectively. These two types of nodes are connected by edges, each edge representing the semantic mapping relationship between the connected information description node and parameter node. Server 102 can map customized information to standardized attribute parameters based on the aforementioned attribute parameter knowledge graph. This method eliminates ambiguity and vagueness in the description information, enabling generative artificial intelligence models to obtain precise and executable modification instructions. Moreover, parameterized representation facilitates structured storage, reuse, and adjustment by server 102, thereby improving the stability and consistency of local editing.

[0046] Then, the area to be edited, standardized attribute parameters, and the original product packaging design can be selected. Figure 1 It also inputs a generative artificial intelligence model, which then performs local editing of the area to be edited based on standardized attribute parameters. This allows the customized modification needs of the client to be quickly met while keeping the rest of the design unchanged.

[0047] In some embodiments, after locally editing the area to be edited based on standardized attribute parameters, the surface normal information of the edited area can be estimated, the light and shadow distribution of the edited area can be calculated based on the ambient lighting diagram of the product packaging design drawing, and the light intensity, color and reflectivity of the edited area can be adjusted based on the light and shadow distribution and surface normal information.

[0048] Specifically, surface normal information can be estimated in the following ways: by using the 3D geometric information corresponding to the product packaging design drawing to extract the spatial orientation of the facet containing the area to be edited, or by predicting the pixel-by-pixel normal direction from the 2D design drawing using a depth estimation network. The ambient lighting map can be represented using a high dynamic range image (HDRI) or spherical harmonics to describe the incident light distribution in the virtual environment in which the packaging is located. When calculating the light and shadow distribution, a physically based rendering (PBR) model can be used, combining the surface normal, ambient lighting map, and material properties of the area to be edited (such as roughness and metallicity), to solve for the emitted radiance of each pixel in the area using the integral reflection equation.

[0049] Building upon this foundation, when adjusting the lighting of the edited area, a relighting method can be employed: maintaining the original region's geometry and material unchanged, only replacing or recalculating its lighting contribution. Specifically, the estimated surface normals and ambient lighting map are input into a pre-trained neural network relighting model (e.g., based on U-Net or Transformer architectures), or spherical harmonic lighting coefficients are directly used for linear superposition to adjust the region's lighting intensity, color channels, and reflectivity. During the adjustment process, it is crucial to maintain continuity in brightness and color temperature between the edited region and adjacent non-edited regions to avoid noticeable seams or abrupt changes in lighting and shadow.

[0050] Through the above adjustments, the visual elements added after local editing can naturally blend into the light and shadow environment of the original packaging surface. Their light and dark transitions, highlight positions, and shadow directions are consistent with the surrounding materials, eliminating the texture-like or floating feeling caused by mismatched lighting. At the same time, since the adjustment process is strictly based on surface normals and actual ambient lighting, even on folded or curved packaging, the edited content can present the correct light and shadow changes with the surface orientation, thereby significantly improving the visual realism of the design and the overall rendering quality.

[0051] In some embodiments, the product packaging knowledge graph is also used to establish the association between product packaging structural attributes and product packaging suppliers. Specifically, the product packaging knowledge graph may also include supplier nodes representing product packaging suppliers, which are connected to product packaging structural attribute nodes via edges. After generating the product packaging design drawing, supplier nodes connected to the product packaging structural attribute nodes can be retrieved from the product packaging knowledge graph. Then, the server 102 can recommend the corresponding product packaging suppliers to the customization party based on the retrieved supplier nodes, so that the customization party can determine the target supplier from the recommended product packaging suppliers. Further, the server 102 can encapsulate the product packaging design drawing according to a predefined data structure, generate a structured customization request, and push it to the target supplier through an API interface. After receiving the request, the target supplier can perform automatic quotation or order generation operations. In some embodiments, the product packaging design drawing (such as 2D planar design drawing, 3D scene drawing, and die-cutting drawing) and its attribute parameters (such as box type parameters, material specification parameters, etc.) can be encapsulated into a structured production requirement data package, and the structured production requirement data package can be pushed to the target supplier through the API interface between the server and the target supplier. Specifically, structured production demand data can be carried in structured customization requests and sent to the target supplier through the aforementioned API interface.

[0052] The following is combined Figure 3A The interaction sequence diagram shown Figure 3B The overall flowchart shown and Figure 4The interactive interface shown illustrates a specific embodiment of this specification. In this embodiment, the e-commerce platform can deploy an intelligent assistant, a recommendation and generation engine, and a knowledge base. The intelligent assistant is a conversational AI agent embedded in the e-commerce platform's front-end page, used to understand the customization intent of the customer and guide them into the packaging customization process. The overall process of this embodiment includes category analysis and packaging recommendation, multimodal visual generation, and delivery and transaction conversion.

[0053] Category Analysis and Packaging Recommendations This stage begins with requirement input and parsing. The customer can input their natural language requirements into the intelligent assistant via the customer client 108, such as "I want to customize product packaging." The intelligent assistant can further inquire about the product category for which the packaging will be used, such as "What type of product do you want to customize the packaging for?". The customer can provide input information as prompted by the intelligent assistant. This input information includes product category information and may also include other information such as design style, such as "Christmas-themed crystal ball" (Step 1). The intelligent assistant can then use Natural Language Processing (NLP) to parse the product category and style information from the input information in real time (Step 2).

[0054] Next, intelligent packaging material recommendations are performed. The intelligent assistant transmits the parsed product category and style information to the recommendation and generation engine, which can query a knowledge base that stores a product packaging knowledge graph (step 3). The intelligent assistant retrieves the product attribute information (such as fragile characteristics, display product) for the "crystal ball" product category from the product packaging knowledge graph, and further queries the associated product packaging structure attributes (such as a cardboard box with a window), and returns the results to the intelligent assistant (step 4).

[0055] Next, packaging options are displayed and selected. The intelligent assistant presents the matched product packaging (i.e., candidate product packaging) to the customizer in a visual format for confirmation or adjustment (step 5). The customizer can select the target product packaging based on the display results and further generate customization information (such as adjusting the window position, material thickness, etc.), or adjust the product packaging structural attributes (step 6).

[0056] Multimodal visual generation At this stage, the multimodal generation task is triggered (step 7). After the client confirms the customization information, the intelligent assistant triggers the recommendation and generation engine to start the multimodal generation task (step 8). The engine simultaneously calls the Wensheng image model to generate a 2D planar design drawing (i.e., Figure 3B (Step 8.1) Call the 3D engine to perform UV mapping and render a 3D scene map (i.e., Figure 3BThe system generates a 3D model (step 8.2) and a die-cutting diagram (step 8.3). Furthermore, after generating the 2D planar design, the intelligent assistant can automatically determine whether adjustments to visual elements on the product packaging are necessary. For example, if a visual element is outside the cutting area or inside the folding area (folding lines may cut off the visual element), adjustment is required. In this case, ControlNet or a rule engine can be used to automatically correct the position of the visual element, ensuring that the visual element is not affected by the folding line, thereby guaranteeing the integrity and visual effect of the design in the packaging application. If the visual element is inside the cutting area but outside the folding area, no adjustment is required, and a 3D scene diagram is directly generated from the 2D planar design. After generation, the 2D planar design, 3D scene diagram, and die-cutting diagram are returned (step 8.4).

[0057] Delivery and transaction conversion First, the recommendation and generation engine integrates and displays the generated product packaging design drawings (including 2D flat design drawings, 3D scene drawings, and die-cutting drawings) (step 9) for the client to preview the final effect. After the client confirms, the recommendation and generation engine integrates the 2D flat design drawings, 3D scene drawings, and die-cutting drawings (which can be DXF files) and packages them into a complete packaging design file (step 10). If the client believes that the generated product packaging design drawings need to be partially redrawn, they can also generate structured prompts through partial redrawing commands, regenerate the 2D flat design drawings, and adjust visual elements, etc.

[0058] After the client confirms the packaging design documents, the smart assistant can also prompt the client to select a product packaging supplier. After the supplier is selected, the smart assistant can automatically package the packaging design documents and the customization request form into a standardized inquiry document and send it to the supplier's system, achieving seamless connection from the design end to the production end (step 11).

[0059] After receiving the inquiry document, the supplier system can analyze the technical parameters (such as size, material, and process requirements), combine them with its own production capacity and cost calculation, and provide a precise quotation to the customer (step 12). This completes the closed loop of the entire process from the customer's input of requirements to production docking.

[0060] The technical solution of this embodiment can achieve the following technical effects: (1) By using existing manufacturable product packaging (i.e. target product packaging) as a physical feasibility constraint, we ensure that the packaging design complies with production specifications, eliminate perspective errors and structural conflicts, improve the consistency between the physical effect of the produced product packaging and the visual design draft, reduce the number of communications, and improve the efficiency of packaging customization.

[0061] (2) Automatically analyze product characteristics and intelligently recommend suitable product packaging, reducing the risk of selection error and meeting the needs of differentiated display.

[0062] (3) It supports the customization of text images in various styles, breaks through the template limitation, and can quickly generate personalized designs, shortening the packaging customization cycle.

[0063] Figure 5A The overall flow of the local redrawing process is shown. Figure 5B and Figure 5C The specific workflow for each stage of the partial redrawing process is illustrated. This embodiment embeds a partial redrawing function into the smart assistant. After generating a product packaging design drawing through a conversation with the smart assistant, the user can partially redraw the product packaging design drawing without leaving the conversation page, achieving deep customization. The process is as follows: Phase 1: Intent Recognition and Prompt Generation In this stage, the user's interactive behavior on the session page, input natural language (describing their intention to modify the area to be edited during partial redrawing), and uploaded reference materials are first obtained. Then, an intent recognition model is invoked, which identifies the user's intent based on the aforementioned interactive behavior, natural language, and reference materials. Next, customized information matching the intent recognition result is determined from a pre-established attribute parameter knowledge graph. This process is achieved by calculating the similarity between the intent recognition result and the customized information in the attribute parameter knowledge graph. Then, standardized attribute parameters related to the determined customized information are searched from the attribute parameter knowledge graph, and the found attribute parameters are normalized and filtered for effects. Finally, the user can annotate the mask of the area to be edited, generate local editing instructions, and use a Large Language Model (LLM) combined with the intent recognition result and normalized attribute parameters to output prompt text that fully describes the user's customization needs.

[0064] Phase Two: Image Generation In this stage, the Mask of the editable area marked by the customizer and the generated prompt text are input into the local redrawing model (which can be a generative artificial intelligence model like the diffusion model in the aforementioned embodiments) to perform local redrawing, resulting in a locally redrawn product packaging design. The local redrawing model can be pre-trained using product packaging design images from e-commerce platforms as sample data for customized training. The training process can be optimized for structural consistency. Specifically, technologies such as ControlNet can be introduced, using the original product packaging design image as input, and establishing a loss function based on the consistency of the main features between the locally redrawn product packaging design image output by the local redrawing model and the original product packaging design image, so as to make the main features of the product packaging design images before and after redrawing more consistent.

[0065] Furthermore, to enhance the realism of visual elements after partial redrawing, geometric deformation and lighting blending algorithms can be applied during the partial redrawing process. This algorithm first estimates the surface normal based on the depth information of the product packaging design, then performs perspective distortion on the uploaded visual elements to ensure they naturally conform to the surface of the product packaging design. Simultaneously, it analyzes the ambient lighting map corresponding to the original product packaging design, automatically generating matching highlights, shadows, and material reflection effects for the visual elements, thus achieving seamless visual integration.

[0066] Phase 3: Quality Screening To ensure the quality of partial redrawing, an image-text relevance scoring model trained on offline data can be used to evaluate the quality of the redrawn packaging design. First, the model determines the consistency between the redrawn packaging design and the customization information input by the client, combining image integrity and aesthetic quality for a comprehensive evaluation. This automatically selects high-quality packaging designs for priority display, while lowering the weight of inferior designs or filtering them out. Subsequently, text and image risk control are applied to the selected designs, filtering out those that do not meet the risk control rules, ultimately outputting high-quality packaging designs.

[0067] Phase Four: Transaction Conversion The client can initiate an inquiry or contact manufacturers who can undertake the customization based on the packaging design currently displayed on the user interface with a single click. After clicking the inquiry button, prompts can be generated based on the generated packaging design and corresponding attribute parameters and sent back to the merchant, thus saving communication costs for both parties regarding design details.

[0068] Through the above process, this embodiment can achieve the following technical effects: (1) It automatically generates customized packaging design drawings and enables instant preview. At the same time, it eliminates the need for professional designers, reducing labor costs and enabling large-scale processing of massive customization needs.

[0069] (2) By utilizing light and shadow blending and perspective transformation techniques, the generated packaging design drawings approach professional rendering effects, with clearly visible material textures. It supports inputting arbitrary material and color commands in natural language, breaking through the limitations of preset templates and meeting personalized needs.

[0070] (3) By using regional labeling and structural consistency constraints, the main structure of the product packaging remains unchanged, the modification accuracy is high, the "product does not match the description" is avoided, and pixel-level precise editing is achieved, so that the generated result meets expectations.

[0071] (4) The one-click inquiry function shortens the decision-making process and improves the order conversion rate of the customization party. Manufacturers can directly obtain the production-ready attribute parameters, eliminating misunderstandings and reducing communication costs.

[0072] Figure 6 This is a schematic structural diagram of a device provided in an exemplary embodiment. For example... Figure 6 As shown, device 400 mainly consists of a communication interface 402, a user interface 404, a processor 406, and a data storage 408. These components are interconnected and communicate with each other via a system bus, network, or other connection mechanism 410. The communication interface 402 enables device 400 to communicate with other devices, access networks, and transmission networks via analog or digital modulation. For example, the communication interface 402 may include a chipset and antenna for wireless communication with a radio access network or access point. Furthermore, the communication interface 402 can be a wired interface such as Ethernet, Token Ring, or a USB port, or a wireless interface such as Wi-Fi, Bluetooth, Global Positioning System (GPS), or a wide-area wireless interface (e.g., WiMAX or LTE). Of course, the communication interface 402 can also support other forms of physical layer interfaces and standard or proprietary communication protocols. The communication interface 402 may also include multiple physical communication interfaces, such as Wi-Fi, Bluetooth, and wide-area wireless interfaces.

[0073] User interface 404 includes receiving user input and providing output to the user. Therefore, user interface 404 may include input components such as a keypad, keyboard, touch-sensitive or presence-sensitive panel, computer mouse, trackball, joystick, microphone, still camera, and video camera, and output components such as a display screen (which may be combined with a touch-sensitive panel), CRT, LCD, LED, display using DLP technology, printer, and other similar devices known or developed in the future. User interface 404 may also generate auditory output via speakers, speaker jacks, audio output ports, audio output devices, headphones, and other similar devices known or developed in the future. In some embodiments, user interface 404 may include software, circuitry, or other forms of logic capable of transmitting and receiving data from external user input / output devices. Additionally or alternatively, device 400 may support remote access from other devices via communication interface 402 or another physical interface (not shown). User interface 404 may be configured to receive user input, the position and movement of which may be indicated by an indicator or cursor described herein. User interface 404 may also be configured as a display device for rendering or displaying text fragments.

[0074] Processor 406 may contain one or more general-purpose processors and / or special-purpose processors.

[0075] Data storage 408 may include one or more volatile and / or non-volatile storage components and may be integrated wholly or partially with processor 406. Data storage 408 may include removable and non-removable components.

[0076] Processor 406 is capable of executing program instructions 418 (e.g., compiled or uncompiled program logic and / or machine code) stored in data storage 408 to perform the various functions described herein. Data storage 408 may comprise a non-transitory computer-readable medium on which program instructions are stored, which, when executed by device 400, enable device 400 to perform any methods, processes, or functions disclosed in this specification and / or the accompanying drawings. Processor 406 executing program instructions 418 may result in processor 406 using data 412.

[0077] For example, program instructions 418 may include an operating system 422 (e.g., an operating system kernel, device drivers, and / or other modules) installed on device 400 and one or more applications 420 (e.g., a browser, social application, or game application). Similarly, data 412 may include operating system data 416 and application data 414. Operating system data 416 is primarily accessible to the operating system 422, while application data 414 is primarily accessible to one or more applications 420. Application data 414 may reside in a file system visible or hidden from the user of device 400.

[0078] Application 420 can communicate with operating system 422 through one or more application programming interfaces (APIs). These APIs help application 420 read and / or write application data 414, transmit or receive information via communication interface 402, receive or display information on user interface 404, etc.

[0079] In some terminology, application 420 may be simply referred to as "app". Furthermore, application 420 can be downloaded to device 400 through one or more online app stores or app markets. However, applications can also be installed on device 400 in other ways, such as through a web browser or a physical interface on device 400 (e.g., a USB port).

[0080] Based on the same concept as the methods described above, this specification also provides an electronic device, including: a processor; a memory for storing processor-executable instructions; wherein the processor performs the steps of the method as described in any of the above embodiments by executing the executable instructions.

[0081] Based on the same concept as the methods described above, this specification also provides a computer-readable storage medium having computer instructions stored thereon that, when executed by a processor, implement the steps of the methods as described in any of the above embodiments.

[0082] Based on the same concept as the methods described above, this specification also provides a computer program product, including a computer program / instructions that, when executed by a processor, implement the steps of the methods as described in any of the above embodiments.

[0083] What those skilled in the art will understand is: In this specification, the terms "comprising," "including," or any other variations thereof are intended to cover a non-exclusive inclusion, such that a process, method, product, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, product, or apparatus. Without further limitation, the presence of additional identical or equivalent elements in a process, method, product, or apparatus that includes said elements is not excluded.

[0084] In this specification, “a,” “an,” and “the” do not specifically refer to the singular, but may also include the plural.

[0085] In this specification, ordinal numbers such as "first," "second," etc., do not necessarily indicate order; they are often used to distinguish between objects. For example, "first server" and "second server" usually refer to two servers. To differentiate between these two servers, they are described as "first server" and "second server." Of course, sometimes these two servers may be the same server.

[0086] In this specification, unless explicitly stated otherwise, "receiving and sending data" does not necessarily mean direct receiving and sending; it can also mean indirect receiving and sending. For example, A receiving data sent by B can be understood as A directly receiving the data sent by B, or it can be understood as A indirectly receiving the data sent by B through other entities such as C. Similarly, B sending data to A can be understood as B sending the data directly to A, or it can be understood as B indirectly sending the data to A through other entities such as C. Here, C can be one entity, or it can be two or more entities.

[0087] In this specification, unless explicitly stated otherwise, the relationships between structures can be direct or indirect. For example, when describing "A is connected to B," unless it is explicitly stated that A and B are directly connected, it should be understood that A can be directly connected to B or indirectly connected to B. Similarly, when describing "A is on top of B," unless it is explicitly stated that A is directly above B (AB is adjacent and A is above B), it should be understood that A can be directly above B or indirectly above B (AB is separated by other elements, and A is above B). And so on.

[0088] This specification uses specific terms to describe embodiments thereof. Terms such as "an embodiment," "one embodiment," and / or "some embodiments" refer to a particular feature, structure, or characteristic associated with at least one embodiment of this specification. Therefore, it should be emphasized and noted that references to "an embodiment," "one embodiment," or "an alternative embodiment" in different locations throughout this specification do not necessarily refer to the same embodiment. Furthermore, those skilled in the art can combine and integrate the different embodiments or examples described herein, as well as the features of those different embodiments or examples, without contradiction.

[0089] Although one or more embodiments of this specification provide method steps as described in the embodiments or flowcharts, it is understood that the order of steps listed in the embodiments or flowcharts is only one of many possible execution orders and does not represent the only execution order. Therefore, when the claims involve method steps, any changes or adjustments to the order of such steps, or the parallelism between steps, are also within the scope of protection of the claims.

Claims

1. A method for generating a product packaging design drawing, the method comprising: The intent of the customer's input information is recognized to determine the product categories included in the input information; The product packaging structure attributes that match the product attribute information of the product category are found from the pre-built product packaging knowledge graph. The product packaging knowledge graph is used to establish the association between product category, product attribute information and product packaging structure attributes. Several candidate product packaging packages are recommended to the customer, the target product packaging package selected by the customer from the candidate product packaging packages is determined, and the digital model of the target product packaging package is obtained; the candidate product packaging packages are all existing manufacturable product packaging packages, and their digital models are used to provide physical realizability constraints. The matched product packaging structure attributes and the acquired digital model are input into the generative artificial intelligence model, so that the generative artificial intelligence model generates a product packaging design drawing that has the product packaging structure attributes and meets the physical realizability constraints provided by the digital model.

2. The method according to claim 1, wherein the product packaging design drawing includes a 2D planar design drawing, a 3D scene drawing, and a die-cutting drawing of the product packaging; generating the product packaging design drawing having the structural attributes of the product packaging and satisfying the physical realizability constraints provided by the digital model includes: The digital model is converted into a 2D planar unfolded drawing, and a 2D planar design drawing of the product packaging is generated based on the 2D planar unfolded drawing and the product packaging structural attributes. The 2D planar design drawing is mapped onto the surface of the digital model using a UV mapping engine to generate a 3D scene image of the product packaging; Based on the topology of the digital model, an engineering die-cutting pattern containing die-cutting line information is generated.

3. The method according to claim 2, wherein the product packaging knowledge graph is further used to establish the association between product packaging structural attributes and product packaging suppliers; the method further includes: Identify product packaging suppliers that are associated with the structural attributes of the product packaging from the product packaging knowledge graph; We recommend the product packaging suppliers we found to the custom packaging client.

4. The method according to claim 3, further comprising: The target supplier selected by the customizer from the found product packaging suppliers is determined; The product packaging design drawings and their attribute parameters are encapsulated into a structured production requirement data package; The structured production requirement data package is pushed to the target supplier via the API interface between the supplier and the target supplier.

5. The method according to claim 1, wherein the step of inputting the matched product packaging structure attributes and the acquired digital model into a generative artificial intelligence model, so that the generative artificial intelligence model generates a product packaging design drawing having the product packaging structure attributes and satisfying the physical realizability constraints provided by the digital model, includes: The customization requirements for product packaging were analyzed from the interaction information of the customization party; The customized requirements, the matched product packaging structure attributes, and the acquired digital model are input into the generative artificial intelligence model, so that the generative artificial intelligence model generates a product packaging design drawing that meets the customized requirements, possesses the product packaging structure attributes, and satisfies the physical realizability constraints provided by the digital model.

6. The method according to claim 5, wherein the customization requirements include at least one of the following: visual style, physical specifications, environmental constraints, and target audience.

7. The method according to claim 1, wherein the digital model includes two-dimensional unfolded structural information, three-dimensional geometric information, and a processing and assembly constraint dataset; the generative artificial intelligence model is configured to generate product packaging design drawings based on the following method: Based on the two-dimensional unfolded structural information, the planar layout of each component in the product packaging design drawing is determined. The planar layout includes the size, shape, and 2D arrangement relationship in the unfolded state. The three-dimensional layout of each component in the product packaging design drawing is constructed based on the three-dimensional geometric information. The three-dimensional layout includes the three-dimensional shape, 3D positional relationship and folding path. The spatial configuration information of the product packaging design drawing is determined based on the processing and assembly constraint dataset. The spatial configuration information includes geometric boundaries, component connection topology, and dimensional tolerance matching relationships. The product packaging design drawing is generated based on the planar layout, three-dimensional layout, and spatial configuration information.

8. The method according to claim 7, wherein the digital model further includes physical processing indicator lines; and the generative artificial intelligence model is further configured to add visual elements to the product packaging design drawing based on the following: Based on the position information of the physical processing indicator line, the folding area and the cutting area in the product packaging design drawing are determined, and the layout position of the visual elements on the product packaging design drawing is determined based on the folding area and the cutting area, so that the layout position is outside the folding area and inside the cutting area; Based on the planar layout, the three-dimensional layout, the spatial configuration information, and the layout position, visual elements are added to the product packaging design drawing, such that the added visual elements are constrained in position by the layout position, and the structure of the product packaging design drawing after adding visual elements is constrained by the planar layout, the three-dimensional layout, and the spatial configuration information.

9. The method according to claim 8, further comprising: The generated product packaging design drawing is subjected to process rule verification to check whether the visual elements in the product packaging design drawing meet the layout position constraints; If the conditions are not met, the visual elements are locally modified using an image restoration model so that the modified visual elements satisfy the layout position constraints.

10. The method according to claim 1, further comprising: Obtain the area to be edited marked by the customizer on the product packaging design drawing and the customization information entered by the customizer; The customization information is used to describe the intention to modify the area to be edited; The customized information is mapped to standardized attribute parameters based on a pre-built attribute parameter knowledge graph; wherein, the attribute parameter knowledge graph is used to establish the mapping relationship between customized information and standardized attribute parameters. The region to be edited, the standardized attribute parameters, and the product packaging design drawing are input into the generative artificial intelligence model, so that the generative artificial intelligence model can perform local editing on the region to be edited based on the standardized attribute parameters.

11. The method according to claim 10, further comprising, after performing local editing on the region to be edited based on the standardized attribute parameters: Estimate the surface normal information of the edited region; Based on the ambient lighting diagram of the product packaging design drawing, calculate the light and shadow distribution of the area to be edited after editing; Based on the light and shadow distribution and the surface normal information, the light intensity, color, and reflectivity of the edited area are adjusted.

12. An electronic device, comprising: processor; A memory for storing processor-executable instructions; wherein the processor implements the steps of the method as claimed in any one of claims 1 to 11 by executing the executable instructions.

13. A computer-readable storage medium having stored thereon computer instructions that, when executed by a processor, implement the steps of the method as claimed in any one of claims 1 to 11.

14. A computer program product comprising a computer program / instructions that, when executed by a processor, implement the steps of the method as claimed in any one of claims 1 to 11.

Images