Method, device and computer equipment for generating two-dimensional clothing pattern

Abstract

The present application relates to a method, apparatus, and computer device for generating a two-dimensional clothing pattern drawing, wherein the method comprises: performing three-dimensional reconstruction on an input image of a target garment to obtain a three-dimensional model of the target garment; identifying the three-dimensional model to obtain multiple three-dimensional curved surface pieces in the three-dimensional model; mapping each three-dimensional curved surface piece to a corresponding two-dimensional flat piece, and determining the piece parameter information of each two-dimensional flat piece; and generating a two-dimensional pattern drawing of the target garment based on each two-dimensional flat piece and the piece parameter information of each two-dimensional flat piece. Through the present application, the problems of low efficiency and insufficient accuracy of pattern drawing generation are solved, thereby achieving improved efficiency and accuracy of pattern drawing generation.

Description

Technical Field

[0001] The present application relates to the technical field of clothing pattern making, and in particular to a method, device and computer equipment for generating a two-dimensional clothing pattern making drawing. Background Art

[0002] During the garment manufacturing process, clothing pattern drawings are created to express the style, structure, dimensions, and process requirements of a garment. They typically serve as a crucial technical basis for fabric cutting, sewing, and other processes. However, existing pattern-making techniques typically rely on manual input of pattern parameters and empirical design, resulting in low efficiency and insufficient accuracy.

[0003] There is currently no effective solution to the problems of low efficiency and insufficient accuracy of plate-making drawings in related technologies. Summary of the Invention

[0004] In this embodiment, a method, apparatus, and computer device for generating a two-dimensional clothing pattern are provided to solve the problems of low pattern generation efficiency and insufficient pattern accuracy in related technologies.

[0005] In a first aspect, this embodiment provides a method for generating a two-dimensional clothing pattern drawing, comprising:

[0006] Performing three-dimensional reconstruction on an input image of a target garment to obtain a three-dimensional model of the target garment;

[0007] Identifying the three-dimensional model to obtain a plurality of three-dimensional curved surface pieces in the three-dimensional model;

[0008] Mapping each of the three-dimensional curved surface patches to a corresponding two-dimensional tiled patch, and determining patch parameter information of each of the two-dimensional tiled patches;

[0009] A two-dimensional pattern drawing of the target garment is generated based on each of the two-dimensional flat panels and the panel parameter information of each of the two-dimensional flat panels.

[0010] In some embodiments, identifying the three-dimensional model to obtain a plurality of three-dimensional curved surface pieces in the three-dimensional model includes:

[0011] identifying seam lines in the three-dimensional model using a semantic segmentation network;

[0012] Based on the recognition result, the three-dimensional model is segmented into a plurality of the three-dimensional curved surface panels.

[0013] In some embodiments, mapping each of the three-dimensional curved surface patches to a corresponding two-dimensional tiled patch includes:

[0014] Each of the three-dimensional curved surface panels is flattened into the corresponding two-dimensional tiled panel using a surface parameterization algorithm.

[0015] In some embodiments, before generating the two-dimensional pattern drawing of the target garment, the method further includes:

[0016] Identify other structural areas in the three-dimensional model and map the other structural areas to the corresponding two-dimensional flat panels; the other structural areas include pleated areas, pocket areas, and patch areas.

[0017] In some embodiments, the other structural region is the wrinkle region; and identifying the other structural region in the three-dimensional model and mapping the other structural region to the corresponding two-dimensional tiled piece includes:

[0018] Performing curvature analysis on the three-dimensional model to identify the wrinkle region in the three-dimensional model; the curvature value of the wrinkle region reaches a preset curvature threshold;

[0019] Based on the curvature value of the wrinkle area, the wrinkle area is mapped to a corresponding shadow area in the two-dimensional tiled piece.

[0020] In some embodiments, the input image is a single-view image; and performing three-dimensional reconstruction on the input image of the target garment to obtain a three-dimensional model of the target garment includes:

[0021] performing feature extraction on the input image of the target garment to obtain a first image feature of the input image;

[0022] Mapping the extracted first image features to a neural implicit field to obtain a corresponding first implicit geometric representation;

[0023] Matching the first implicit geometric representation with a preset implicit geometric representation template;

[0024] The three-dimensional model of the target garment is generated based on the implicit geometric representation template that matches the first implicit geometric representation.

[0025] In some embodiments, the input image is a multi-view image; and performing three-dimensional reconstruction on the input image of the target garment to obtain a three-dimensional model of the target garment includes:

[0026] performing feature extraction on the input image of the target garment to obtain a second image feature of each input image;

[0027] Mapping the extracted second image features to a neural implicit field to obtain a corresponding second implicit geometric representation;

[0028] Optimizing the second implicit geometric representation based on a photometric consistency constraint of the multi-view images;

[0029] The three-dimensional model of the target garment is generated based on the optimized second implicit geometric representation.

[0030] In some embodiments, after generating the two-dimensional pattern drawing of the target garment, the method further includes:

[0031] Determine the corresponding user size parameters according to the user size information;

[0032] Adjusting the two-dimensional pattern drawing of the target garment according to the user size parameter to obtain a user two-dimensional pattern drawing of the user garment;

[0033] A three-dimensional fitting image of the user is generated based on the user size information and the user two-dimensional pattern drawing.

[0034] In some embodiments, before performing three-dimensional reconstruction on the input image of the target garment, the method further includes:

[0035] The fabric features of the target garment are identified through the input image, wherein the fabric features include at least one of the following: fabric material, fabric color, and fabric structure.

[0036] In some embodiments, after obtaining the user two-dimensional pattern drawing of the user's clothing, the method further includes:

[0037] Training a pattern drawing optimization model based on clothing optimization information, size optimization parameters and fabric optimization parameters corresponding to the clothing optimization information, to obtain a trained pattern drawing optimization model; the clothing optimization information includes the user size information, the fabric characteristics, the user's two-dimensional pattern drawing, and user demand information;

[0038] The user's two-dimensional plate-making drawing is adjusted and optimized by using the trained plate-making drawing optimization model to obtain the optimized user's two-dimensional plate-making drawing.

[0039] In some embodiments, the method further comprises:

[0040] Based on the three-dimensional model and fabric features of the target clothing, clothing feature information of the target clothing is obtained by training a complete style-property joint model; the clothing feature information includes clothing property features and clothing style features.

[0041] In some embodiments, after obtaining the user two-dimensional pattern drawing of the user's clothing, the method further includes:

[0042] Determine the area of ​​each user's two-dimensional tiled piece in the user's two-dimensional layout;

[0043] The fabric usage of the user's clothing is determined based on the area, the fabric characteristics, and the clothing characteristic information of each of the user's two-dimensional flat pieces.

[0044] In some embodiments, the method further comprises:

[0045] The preset curvature threshold is dynamically adjusted based on the clothing feature information, and the wrinkle area is identified based on the adjusted preset curvature threshold.

[0046] In a second aspect, this embodiment provides a two-dimensional clothing pattern generation device, comprising:

[0047] A reconstruction module, configured to perform three-dimensional reconstruction on an input image of a target garment to obtain a three-dimensional model of the target garment;

[0048] a segmentation module, configured to identify the three-dimensional model and obtain a plurality of three-dimensional curved surface pieces in the three-dimensional model;

[0049] a flattening module, configured to map each of the three-dimensional curved surface patches to a corresponding two-dimensional tiled patch, and determine patch parameter information of each of the two-dimensional tiled patches;

[0050] A generating module is configured to generate a two-dimensional pattern drawing of the target garment based on each of the two-dimensional flat panels and the panel parameter information of each of the two-dimensional flat panels.

[0051] In a third aspect, a computer device is provided in this embodiment, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the two-dimensional clothing pattern generating method described in the first aspect is implemented.

[0052] Compared with the related art, the two-dimensional clothing pattern generation method, device and computer equipment provided in this embodiment obtain a three-dimensional model of the target clothing by performing three-dimensional reconstruction on the input image of the target clothing; identify the three-dimensional model to obtain multiple three-dimensional curved surface patches in the three-dimensional model; map each three-dimensional curved surface patch to a corresponding two-dimensional flat patch, and determine the patch parameter information of each two-dimensional flat patch; generate a two-dimensional pattern drawing of the target clothing based on each two-dimensional flat patch and the patch parameter information of each two-dimensional flat patch, thereby solving the problems of low pattern drawing generation efficiency and insufficient pattern drawing accuracy, and achieving improved pattern drawing generation efficiency and pattern drawing accuracy.

[0053] The details of one or more embodiments of the present application are set forth in the following drawings and description to make other features, objects, and advantages of the present application more readily apparent. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0055] Figure 1 This is a hardware structure block diagram of a terminal device for a two-dimensional clothing pattern generation method provided in an embodiment of the present application;

[0056] Figure 2 This is a flow chart of a method for generating a two-dimensional clothing pattern drawing provided by an embodiment of the present application;

[0057] Figure 3 This is a flow chart of a three-dimensional curved surface piece recognition method provided in one embodiment of the present application;

[0058] Figure 4 is a flowchart of another structural region identification method provided by an embodiment of the present application;

[0059] Figure 5 This is a flowchart of a three-dimensional reconstruction method based on a single-view image provided in one embodiment of the present application;

[0060] Figure 6 This is a flowchart of a three-dimensional reconstruction method based on multi-view images provided in one embodiment of the present application;

[0061] Figure 7 This is a flow chart of a two-dimensional clothing pattern optimization method provided by an embodiment of the present application;

[0062] Figure 8 This is a flow chart of a method for generating a two-dimensional clothing pattern drawing provided by a preferred embodiment of the present application;

[0063] Figure 9 This is a structural block diagram of a two-dimensional clothing pattern generation device provided in one embodiment of the present application.

[0064] In the figure: 102, processor; 104, memory; 106, transmission device; 108, input and output device; 10, reconstruction module; 20, segmentation module; 30, flattening module; 40, generation module. DETAILED DESCRIPTION

[0065] In order to more clearly understand the purpose, technical solutions and advantages of the present application, the present application is described and illustrated below in conjunction with the accompanying drawings and embodiments.

[0066] Unless otherwise defined, technical or scientific terms used in this application shall have the ordinary meanings as understood by persons of ordinary skill in the art to which this application belongs. The terms "a," "an," "the," "these," and similar expressions in this application do not denote limitations on quantity and may be singular or plural. The terms "comprise," "include," "have," and any variations thereof, as used in this application, are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or device comprising a series of steps or modules (units) is not limited to the listed steps or modules (units) but may include unlisted steps or modules (units) or other steps or modules (units) inherent to the process, method, product, or device. The terms "connected," "connected," "coupled," and similar expressions used in this application are not limited to physical or mechanical connections but may include electrical connections, whether direct or indirect. As used in this application, "plurality" means two or more. "And / or" describes an association between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can mean: A exists alone; A and B exist simultaneously; or B exists alone. Generally, the character " / " indicates that the objects in the preceding and following relationship are in an "or" relationship. The terms "first", "second", "third", etc. involved in this application are only used to distinguish similar objects and do not represent a specific ordering of the objects.

[0067] The method embodiment provided in this embodiment can be executed in a terminal, a computer or a similar computing device. For example, running on a terminal, Figure 1 FIG. 1 is a block diagram of the hardware structure of the terminal of the two-dimensional clothing pattern generation method of this embodiment. Figure 1 As shown, the terminal may include one or more ( Figure 1 The processor 102 (only one is shown) and a memory 104 for storing data, wherein the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA. The terminal may also include a transmission device 106 for communication functions and an input / output device 108. It will be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the above terminal. Figure 1 More or fewer components than shown, or with Figure 1 Different configurations shown.

[0068] Memory 104 can be used to store computer programs, such as software programs and modules of application software, such as the computer program corresponding to the two-dimensional garment pattern generation method in this embodiment. Processor 102 executes the computer programs stored in memory 104 to execute various functional applications and data processing, thereby implementing the aforementioned method. Memory 104 may include high-speed random access memory (RAM) and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some embodiments, memory 104 may further include memory remotely located from processor 102, which can be connected to the terminal via a network. Examples of such networks include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0069] Transmission device 106 is used to receive or transmit data via a network. This network may include a wireless network provided by the terminal's communications provider. In one embodiment, transmission device 106 includes a network interface controller (NIC), which can be connected to other network devices via a base station to enable communication with the Internet. In another embodiment, transmission device 106 may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0070] In this embodiment, a method for generating a two-dimensional clothing pattern is provided. Figure 2 Flowchart of the method for generating a two-dimensional clothing pattern drawing according to this embodiment. Figure 2 As shown, the process includes the following steps:

[0071] Step S220, performing three-dimensional reconstruction on the input image of the target garment to obtain a three-dimensional model of the target garment;

[0072] Specifically, an input image of a target garment is obtained. The target garment can be a short-sleeved shirt, a jacket, a dress, or trousers. The input image is a single-view image or a multi-view image of the target garment, with the multi-view image including at least two images from different viewpoints. The input image is preprocessed, and the preprocessing operations are adapted to the input image type, including but not limited to image normalization (such as uniform resolution and illumination correction) and image alignment (such as image alignment based on structure from motion or key point matching). A multi-scale convolutional neural network (such as a feature pyramid network or a high-resolution network) or a graph neural network is then used to extract features from the preprocessed input image to obtain image features. Based on the extracted image features, a 3D reconstruction is performed to obtain a 3D model of the target garment. 3D reconstruction methods include geometry-based 3D reconstruction algorithms and deep learning-based 3D reconstruction algorithms, which are not specifically limited here.

[0073] Exemplarily, when the input image is a monoscopic image of a target garment, feature extraction is performed on the input image to obtain a first image feature of the input image. The extracted first image feature is then mapped to a neural implicit field to obtain a corresponding first implicit geometric representation. Based on the first implicit geometric representation, an isosurface is extracted from the implicit field to generate a textured three-dimensional mesh to obtain a three-dimensional model of the target garment. In other embodiments, the first implicit geometric representation can be matched with an implicit geometric representation template preset in a garment morphology prior library. Based on the implicit geometric representation template that matches the first implicit geometric representation, a three-dimensional model of the target garment is generated. This combines prior knowledge to optimize the three-dimensional reconstruction process of monoscopic images, improving reconstruction efficiency and accuracy.

[0074] For example, when the input image is a multi-view image of a target garment, feature extraction is performed on the input image of the target garment to obtain second image features for each input image. The extracted second image features are then mapped to a neural implicit field to obtain a corresponding second implicit geometric representation. Based on the second implicit geometric representation, an isosurface is extracted from the implicit field to generate a textured three-dimensional mesh to obtain a three-dimensional model of the target garment. In other embodiments, the second implicit geometric representation can be optimized based on photometric consistency constraints of the multi-view images, and a three-dimensional model of the target garment is generated based on the optimized second implicit geometric representation. This improves the robustness of the three-dimensional reconstruction, adapts to complex scenes, and helps improve reconstruction accuracy.

[0075] Furthermore, through physics engines such as Blender and Bullet, gravity and cloth flexibility constraints are applied to the 3D model mesh. This allows the garment model's form to be dynamically adjusted through physical simulation post-processing, thereby correcting unreasonable geometric distortions generated during the modeling process, such as unnatural sharp wrinkles and excessive stretching. It should be noted that applying gravity can make the 3D garment model produce natural drooping and swaying in virtual space, consistent with the influence of real gravity, while applying cloth flexibility constraints is used to simulate deformation behaviors such as cloth stretching, bending, and wrinkling.

[0076] In this embodiment, single-view images or multi-view images can be used to achieve three-dimensional reconstruction, supporting flexible input types, lowering the data collection threshold, and helping to improve operational convenience.

[0077] Step S240, identifying the three-dimensional model to obtain a plurality of three-dimensional curved surface pieces in the three-dimensional model;

[0078] Specifically, the 3D model of the target garment is identified through relevant algorithms to segment the 3D model to obtain multiple 3D surface pieces. Garment pieces are the basic units that make up a complete garment, such as the front piece, back piece, sleeves, skirt piece, front trouser piece and back trouser piece.

[0079] Segmentation algorithms based on geometric features (e.g., segmenting by using sudden changes in curvature on the model surface as the boundaries between panels) and semantic priors can be used. For example, a semantic segmentation network can be used to identify seams in a 3D model and, based on the locations of the identified seams, segment the 3D model into multiple 3D curved panels.

[0080] Step S260 , mapping each three-dimensional curved surface patch to a corresponding two-dimensional tiled patch, and determining the patch parameter information of each two-dimensional tiled patch;

[0081] Specifically, each 3D surface patch is mapped to a corresponding 2D tiled patch, and the vertex coordinates and boundary parameters of the 2D tiled patch are calculated and output. The boundary parameters are used to describe the geometric characteristics and constraints of the patch edge, such as boundary length, boundary curvature, and boundary curve equation, to determine the basic geometric shape and key size information of the 2D tiled patch.

[0082] In this embodiment, the mapping of 3D curved surface pieces can be achieved using a variety of methods. For example, a surface parameterization algorithm (such as a stretch-minimization parameterization algorithm or a conformal mapping-based parameterization algorithm) can be used to unfold each 3D curved surface piece into a corresponding 2D tiled piece. Alternatively, a physics engine can be used to simulate the relaxation process of garment fabric, unfolding each 3D curved surface piece into a corresponding 2D tiled piece while minimizing deformation such as stretching and wrinkling after unfolding. Alternatively, a geometric transformation method such as orthogonal projection or perspective projection can be used to project each 3D curved surface piece into a corresponding 2D tiled piece, and projection distortion can be corrected.

[0083] Step S280 : generating a two-dimensional pattern drawing of the target garment based on each two-dimensional tiled piece and the piece parameter information of each two-dimensional tiled piece.

[0084] Specifically, the cutting parameter information of each two-dimensional flat piece is verified and corrected to avoid missing or contradictory parameters. Based on the verified and corrected cutting parameter information, each two-dimensional flat piece is arranged and combined according to preset requirements to reasonably handle the splicing relationship and alignment identification between the pieces. On this basis, an appropriate seam allowance is generated for the edge portion of each two-dimensional flat piece that needs to be sewn (which can be dynamically adjusted according to fabric characteristics and process requirements), and each two-dimensional flat piece is marked, including the piece type mark, the positioning mark of the edge or interior of the piece (such as notches, drill holes), and the range mark of the seam allowance. Finally, a complete and production-standard two-dimensional clothing pattern is generated, providing an accurate technical basis for the subsequent cutting and sewing processes.

[0085] During the garment manufacturing process, clothing pattern drawings are created to express the style, structure, dimensions, and process requirements of a garment. They typically serve as a crucial technical basis for fabric cutting, sewing, and other processes. However, existing pattern-making techniques typically rely on manual input of pattern parameters and empirical design, resulting in low efficiency and insufficient accuracy.

[0086] Compared with the prior art, the present invention provides a method, device and computer equipment for generating a two-dimensional clothing pattern drawing. The method, device and computer equipment perform three-dimensional reconstruction on the input image of the target clothing to obtain a three-dimensional model of the target clothing; identify the three-dimensional model to obtain multiple three-dimensional curved surface pieces in the three-dimensional model; map each three-dimensional curved surface piece to a corresponding two-dimensional flat piece, determine the piece parameter information of each two-dimensional flat piece; and generate a two-dimensional pattern drawing of the target clothing based on each two-dimensional flat piece and the piece parameter information of each two-dimensional flat piece. Based on this, the three-dimensional clothing model can reveal three-dimensional details that cannot be presented by the two-dimensional clothing image. By mapping each three-dimensional piece in the clothing three-dimensional model to a corresponding two-dimensional flat piece, accurate mapping from three dimensions to two dimensions is achieved, significantly improving the degree of restoration of the piece structure, more in line with the physical rules of clothing, and automatically generating a two-dimensional pattern drawing corresponding to each two-dimensional flat piece, optimizing the pattern making ambiguity, and completing accurate pattern making, thereby solving the problems of low pattern drawing generation efficiency and insufficient pattern drawing accuracy, improving the pattern drawing generation efficiency and pattern drawing accuracy, and providing an accurate structural reference for subsequent production processes.

[0087] In some of these embodiments, Figure 3 As shown, the step S240 of identifying the three-dimensional model to obtain a plurality of three-dimensional curved surface pieces in the three-dimensional model includes the following steps:

[0088] Step S241, identifying seam lines in the 3D model through a semantic segmentation network;

[0089] Step S242 : based on the recognition result, segment the 3D model into a plurality of 3D curved surface panels.

[0090] Specifically, the semantic segmentation network is used to identify the seams of the surface mesh of the three-dimensional model, obtain the positions of each seam line in the three-dimensional model, and based on the positions of each seam line, the three-dimensional model is segmented into multiple three-dimensional surface pieces. Clothing pieces are the basic units that make up a complete garment, such as the front piece, back piece, sleeves, skirt piece, front trouser piece and back trouser piece.

[0091] Among them, semantic segmentation networks include but are not limited to Mask2Former, U-Net, and Mask R-CNN.

[0092] Through this embodiment, the seam lines in the three-dimensional model are identified through a semantic segmentation network, and based on the recognition results, the three-dimensional model is segmented into multiple three-dimensional curved surface pieces, thereby achieving accurate segmentation of independent clothing pieces.

[0093] In some embodiments, mapping each three-dimensional curved surface patch to a corresponding two-dimensional tiled patch in step S260 includes the following steps:

[0094] Each 3D surface patch is flattened into a corresponding 2D tiled patch using a surface parameterization algorithm.

[0095] In this embodiment, a surface parameterization algorithm such as a minimum stretching parameterization algorithm (such as the low-distortion parameterization method ABF++ based on angle optimization) and a parameterization algorithm based on conformal mapping is used to flatten each three-dimensional surface patch into a corresponding two-dimensional tiled patch.

[0096] Exemplarily, a minimum stretch parameterization algorithm is applied to each three-dimensional surface piece to flatten each three-dimensional surface piece into a two-dimensional polygon to obtain a corresponding two-dimensional tiled piece, ensuring that the error between the unfolded shape and the actual garment piece is less than a preset threshold, such as 2%, thereby reducing shape distortion when the piece is unfolded and ensuring a high degree of match between the two-dimensional paper pattern and the garment prototype.

[0097] Through this embodiment, each three-dimensional surface piece is flattened into a corresponding two-dimensional flat piece through a surface parameterization algorithm, achieving a fidelity mapping from a three-dimensional surface to a two-dimensional plane, which can effectively improve the standardization of clothing process planning.

[0098] In some embodiments, before generating the two-dimensional pattern drawing of the target garment, the two-dimensional garment pattern drawing generating method further includes the following steps:

[0099] Identify other structural areas in the 3D model and map them to corresponding 2D flat pieces; other structural areas include pleated areas, pocket areas, and patch areas.

[0100] Specifically, the three-dimensional model of the target garment is identified to obtain other structural regions in the three-dimensional model, including wrinkle regions, pocket regions, and patch regions, and the other structural regions are mapped to corresponding two-dimensional flat pieces.

[0101] Among them, when identifying the wrinkle area, by performing curvature analysis on the three-dimensional model, the area whose curvature value reaches the preset curvature threshold is identified as the wrinkle area in the three-dimensional model, and based on the curvature value of the wrinkle area, the wrinkle area is mapped to the corresponding two-dimensional tiled piece.

[0102] When identifying pocket and patch regions, a region growing algorithm can be used, using normalized Gaussian curvature and normal vector consistency as the growth basis. Diffusion growth is performed from a selected seed point to form a preliminary structural region range. Area filtering and convexity verification are then combined to eliminate interference areas, ensuring that the identified regions are pocket and patch regions. Furthermore, conformal mapping technology is used to establish a precise correspondence between 3D mesh vertices and 2D UV coordinates, enabling accurate mapping of structural regions to 2D tiled patches. In other embodiments, the location and identification of pocket regions can also be achieved by analyzing the Gaussian curvature of the 3D model surface mesh to locate isolated raised areas.

[0103] It should be noted that in a two-dimensional tiled pattern, various methods can be used to mark each other structural region to clearly distinguish different structural regions. For example, different other structural regions can be marked with different colors, or a unique outline style can be generated for each other structural region, or different other structural regions can be marked with different symbols, etc., without specific limitations here.

[0104] Through this embodiment, different types of other structural areas in the three-dimensional model are identified, and each other structural area is mapped to a corresponding two-dimensional tiled piece, so as to accurately identify and map different other structural areas, which helps to improve the accuracy of subsequent generation of two-dimensional plate making drawings.

[0105] In some of these embodiments, the other structural region is a corrugated region; e.g. Figure 4 As shown, identifying other structural regions in the three-dimensional model and mapping the other structural regions to corresponding two-dimensional tiled pieces includes the following steps:

[0106] Step S251: performing curvature analysis on the three-dimensional model to identify wrinkle regions in the three-dimensional model; the curvature value of the wrinkle region reaches a preset curvature threshold;

[0107] Step S252 : Mapping the wrinkle area to a shadow area in the corresponding two-dimensional tiled piece based on the curvature value of the wrinkle area.

[0108] Specifically, during the 2D pattern generation process, a curvature analysis is performed on the 3D model of the target garment to calculate curvature information for the 3D model's surface mesh, including Gaussian curvature and mean curvature. This curvature information is then compared with a preset curvature threshold, and areas where the curvature reaches the threshold are marked as wrinkle areas. The preset curvature threshold is determined based on actual requirements, such as the garment type and fabric properties.

[0109] It can be understood that points with larger absolute values ​​of Gaussian curvature usually appear at the tips or intersection areas of the wrinkles, while the absolute values ​​of the average curvature are larger at the ridges or valleys of the wrinkles. Combining the Gaussian curvature and the average curvature for judgment can accurately determine the wrinkle area.

[0110] Furthermore, based on the curvature value of the wrinkle area, the wrinkle area is mapped to the shadow area in the corresponding two-dimensional tiled piece. The curvature value corresponds to the pixel grayscale of the shadow area. For example, the raised area of ​​the wrinkle uses high brightness, and the sunken area of ​​the wrinkle uses low brightness, etc., to achieve visualization of the wrinkle area. At the same time, it supports user interactive adjustment of shadow sensitivity, that is, by adjusting the parameters of the mapping function to change the mapping rules between the curvature value and the grayscale value to meet the analysis requirements of different accuracies.

[0111] Through this embodiment, curvature analysis is performed on the three-dimensional model to identify the wrinkle area in the three-dimensional model; the curvature value of the wrinkle area reaches a preset curvature threshold, and based on the curvature value of the wrinkle area, the wrinkle area is mapped to a shadow area in the corresponding two-dimensional flat piece, thereby realizing wrinkle recognition and calibration, improving the structural accuracy of the two-dimensional flat piece, and providing an accurate structural reference for subsequent production processes.

[0112] In some embodiments, the input image is a single-view image; Figure 5 As shown, the three-dimensional reconstruction of the input image of the target garment in step S220 to obtain a three-dimensional model of the target garment includes the following steps:

[0113] Step S221, performing feature extraction on the input image of the target garment to obtain a first image feature of the input image;

[0114] Step S222, mapping the extracted first image feature to the neural implicit field to obtain a corresponding first implicit geometric representation;

[0115] Step S223, matching the first implicit geometric representation with a preset implicit geometric representation template;

[0116] Step S224 : generating a three-dimensional model of the target garment based on the implicit geometric representation template that matches the first implicit geometric representation.

[0117] Specifically, when the input image is a single-view image, multi-scale convolutional neural network, graph neural network and other algorithms are used to extract features of the input image to obtain the first image feature of the input image. The first image feature includes local detail features and global semantic features.

[0118] Furthermore, the extracted first image features are mapped to a neural implicit field, and the occupancy probability or signed distance function (SDF) of each point in the space is predicted to obtain the corresponding first implicit geometric representation. This first implicit geometric representation is matched with implicit geometric representation templates preset in a clothing morphology prior library, thereby optimizing the implicit field prediction by retrieving similar templates. The clothing morphology prior library includes templates for different clothing types, such as SDF templates for short-sleeved shirts and shirts.

[0119] Afterwards, based on the implicit geometric representation template that matches the first implicit geometric representation, a textured three-dimensional mesh is generated by extracting isosurfaces from the implicit field to obtain a three-dimensional model of the target garment.

[0120] Through this embodiment, feature extraction is performed on the input image of the target clothing to obtain the first image feature of the input image, the extracted first image feature is mapped to the neural implicit field to obtain the corresponding first implicit geometric representation, and the first implicit geometric representation is matched with a preset implicit geometric representation template. Based on the implicit geometric representation template that matches the first implicit geometric representation, a three-dimensional model of the target clothing is generated. In this way, the three-dimensional reconstruction process of the single-view image is optimized in combination with prior knowledge, thereby improving the reconstruction efficiency and accuracy.

[0121] In some embodiments, the input image is a multi-view image; Figure 6 As shown, the three-dimensional reconstruction of the input image of the target garment in step S220 to obtain a three-dimensional model of the target garment includes the following steps:

[0122] Step S225, performing feature extraction on the input image of the target garment to obtain a second image feature of each input image;

[0123] Step S226, mapping the extracted second image features to the neural implicit field to obtain a corresponding second implicit geometric representation;

[0124] Step S227, optimizing the second implicit geometric representation based on the photometric consistency constraint of the multi-view images;

[0125] Step S228: Generate a three-dimensional model of the target garment based on the optimized second implicit geometric representation.

[0126] Specifically, when the input image is multi-view, algorithms such as multi-scale convolutional neural networks and graph neural networks are used to extract features from the input image to obtain a second image feature. For example, the multi-view image is fed into an encoder with shared weights to extract multiple sets of image features. These features are then fused using a cross-view attention mechanism (such as the Transformer architecture) to output the fused second image feature.

[0127] Furthermore, the extracted second image features are mapped to a neural implicit field, and the occupancy probability or signed distance function of each point in space is predicted to obtain the corresponding second implicit geometric representation. This second implicit geometric representation is then optimized based on the photometric consistency constraint of the multi-view image. Differentiable rendering (such as PyTorch3D and TensorFlowGraphics) is used to project the implicit field to each viewpoint to obtain a rendered image. The rendered image is then compared pixel by pixel with the true multi-view image. The loss value is calculated using a photometric consistency loss function. Based on the calculated photometric consistency loss, the implicit field expression is adjusted to reduce the difference between the rendered image and the true image, ensuring lighting rationality and facilitating fine-tuning of 3D model details in multi-view scenarios.

[0128] Afterwards, based on the optimized second implicit geometric representation, a textured three-dimensional mesh is generated by extracting isosurfaces from the implicit field to obtain a three-dimensional model of the target garment.

[0129] Through this embodiment, feature extraction is performed on the input image of the target clothing to obtain the second image features of each input image, and the extracted second image features are mapped to the neural implicit field to obtain the corresponding second implicit geometric representation. Based on the photometric consistency constraint of the multi-view images, the second implicit geometric representation is optimized, and then based on the optimized second implicit geometric representation, a three-dimensional model of the target clothing is generated, thereby improving the robustness of the three-dimensional reconstruction, adapting to complex scenes, and helping to improve the reconstruction accuracy.

[0130] In some of these embodiments, Figure 7 As shown in FIG, after generating the two-dimensional pattern drawing of the target garment, the following steps are also included:

[0131] Step S291, determining corresponding user size parameters based on user size information;

[0132] Step S292, adjusting the two-dimensional pattern drawing of the target garment according to the user size parameter to obtain a user two-dimensional pattern drawing of the user garment;

[0133] Step S293: Generate a three-dimensional fitting image of the user based on the user size information and the user's two-dimensional pattern drawing.

[0134] Specifically, user size information is obtained. This information can be in the form of images (e.g., full-body or half-body images of the user) and / or text (e.g., key size parameters such as chest circumference, waist circumference, hip circumference, and shoulder width, or body descriptions such as thin or pear-shaped). When the user size information is in the form of user size parameters, the user size parameters are obtained directly. When the user size information is in the form of images or body descriptions, the user size information can be converted into corresponding user size parameters using a size parameter model. The size parameter model is trained using user body images, body descriptions, and user size parameters.

[0135] Based on the user's size parameters, a parametric mapping relationship is used to adjust the 2D pattern drawing to obtain a user-specific 2D pattern drawing for the user's garment, i.e., a user-specific 2D pattern drawing adapted to the user's body shape. The user's size is pre-associated with key control points of the pattern pieces in the 2D pattern drawing (such as the neckline curvature inflection point and the dart endpoints) through parametric mapping rules. A non-uniform scaling algorithm is used to calculate differential displacements of each control point to dynamically adjust the pattern piece's contours. A constraint solving algorithm is also used to maintain the pattern piece's topological structure, ultimately generating a user-specific 2D pattern drawing adapted to the user's body shape.

[0136] Furthermore, a user image model is generated based on the user's size information. For example, a human body model can be reconstructed through 3D scanning or multi-view photography, or a parameterized avatar template, such as the Skinned Multi-Person Linear Model (SMPL), can be driven by input size data to generate a 3D human body mesh that fits the user's size, or a standard avatar image can be directly used.

[0137] In this embodiment, virtual fitting can be performed based on the user image modeling results. Specifically, based on the user's two-dimensional pattern drawing, a corresponding three-dimensional image of the user's clothing is produced. Then, through the virtual fitting tool, the clothing model is matched with the user image model to achieve a virtual fitting effect display. Alternatively, the pieces in the user's two-dimensional pattern drawing are adsorbed to the corresponding positions of the three-dimensional human body according to the actual sewing logic (such as aligning the sleeves with the shoulders). Through key point projection or UV mapping, a precise spatial correspondence between the pieces and the human body surface is established. Then, based on the physics engine, the sewing constraints between the pieces are simulated, and the mechanical properties of the fabric (such as elasticity and drape) are applied to simulate the natural form of the clothing when worn. Finally, a three-dimensional fitting image is output. In this way, the spatial positioning of the pieces and the virtual sewing solution are combined to restore the real sewing logic and improve the credibility of the fitting.

[0138] Through this embodiment, the corresponding user size parameters are determined according to the user size information, and the two-dimensional pattern drawing of the target garment is adjusted according to the user size parameters to obtain the user two-dimensional pattern drawing of the user garment, thereby achieving accurate adaptation of the garment pattern to the human body shape. Based on the user size information and the user two-dimensional pattern drawing, a user three-dimensional fitting image is generated to achieve visual verification of the garment design. Then, the user two-dimensional pattern drawing is further adjusted based on the fitting effect to improve the accuracy of the garment pattern drawing.

[0139] In some embodiments, before performing three-dimensional reconstruction on the input image of the target garment, the above-mentioned method for generating a two-dimensional garment pattern drawing further includes the following steps:

[0140] The fabric features of the target garment are identified by inputting an image, wherein the fabric features include at least one of the following: fabric material, fabric color, and fabric structure.

[0141] Specifically, the input image of the target garment is recognized to obtain the fabric features of the target garment. Fabric features include fabric raw material, fabric color, and fabric weave structure. Fabric raw material refers to the basic material that makes up the fabric, including natural fibers (such as cotton and linen) and chemical fibers (such as rayon, polyester, and nylon). Fabric weave structure refers to the arrangement and combination of fibers or yarns that make up the fabric, including yarn structure (such as single yarn and ply yarn) and fabric structure (such as plain weave, twill, and satin).

[0142] It is understandable that the fabric feature recognition process can be implemented in a manner selected according to actual application requirements, including but not limited to image processing algorithms, machine learning models or a combination thereof, which are not specifically limited here.

[0143] For example, for fabric raw material identification, the contrast, energy, entropy and other features of the texture can be extracted through the grayscale co-occurrence matrix, and the microscopic details of the fiber texture in the input image can be analyzed to determine the fabric raw material used by the target garment. For example, cotton fabrics usually have a rough texture, silk has a smooth texture, etc.; or, a Gabor filter can be used to capture the fiber arrangement pattern (such as chemical fiber has a regular grid texture) through multi-directional and multi-scale filtering responses to determine the fabric raw material used by the target garment; or, high-dimensional texture features can be extracted through a pre-trained model (such as a residual network, VGG), and the fabric type classification can be realized by combining a fine-tuning classification layer. At the same time, the channel attention module can be used to focus on key areas of the fabric (such as fiber seams) to improve the accuracy of fine-grained classification and accurately determine the fabric raw material used by the target garment.

[0144] Exemplarily, for fabric color recognition, the input image is converted from RGB color space to HSV color space or Lab color space, and the brightness and chromaticity information are separated to reduce light interference, so that the fabric color can be accurately identified based on the chromaticity information; or, color histogram statistics are performed on the chromaticity channels (such as the H channel of the HSV color space, the a channel and the b channel of the Lab color space), and the fabric color is accurately determined by matching with preset color templates (such as Pantone color cards, natural color system color cards); or, an unsupervised clustering algorithm (such as K-means clustering algorithm, density-based clustering algorithm DBSCAN) is used to extract the main color tone, or a semantic segmentation model (such as U-Net, fully convolutional network) is used to separate the clothing area of ​​the input image and then count the color distribution to accurately determine the fabric color.

[0145] For example, for fabric structure recognition, local binary patterns are used to quantify the grayscale differences between pixels, and by capturing the interweaving patterns of warp and weft yarns, details such as the diagonal features of twill weave can be accurately identified; or, the input image is converted to the frequency domain through Fourier transform, and structures such as plain weave (high-frequency energy concentration) and satin weave (low-frequency energy dominance) are distinguished based on the energy distribution characteristics; or, target detection models such as Faster R-CNN and YOLO are used to locate fabric unit structures such as yarn intersections, to achieve rapid classification and recognition of weave structures; or, graph convolutional networks are used to model yarn topological connection relationships, to effectively identify complex weave structures such as jacquard patterns.

[0146] Through this embodiment, the fabric features of the target garment are identified by inputting an image, and the fabric features include at least one of the fabric raw material, fabric color, and fabric structure, so as to accurately identify the fabric features of the garment.

[0147] In some embodiments, after obtaining the user's two-dimensional pattern drawing of the user's clothing, the above-mentioned method for generating a two-dimensional clothing pattern drawing further includes the following steps:

[0148] Based on the clothing optimization information, the size optimization parameters and fabric optimization parameters corresponding to the clothing optimization information, the pattern making optimization model is trained to obtain a trained pattern making optimization model; the clothing optimization information includes user size information, fabric characteristics, user two-dimensional pattern making drawings and user demand information;

[0149] Through the trained layout optimization model, the user's two-dimensional layout is adjusted and optimized to obtain the optimized user's two-dimensional layout.

[0150] Specifically, a training data set is constructed, which includes clothing optimization information and its corresponding three-dimensional image optimization information and fabric feature optimization information, wherein the clothing optimization information includes user size information, fabric feature information, a user two-dimensional pattern drawing and user demand information, and the user demand information includes user size demand information and / or user fabric demand information. The user demand information can be text and / or image, and the user demand information can be obtained by the user with reference to the three-dimensional fitting picture. For example, the neckline in the three-dimensional fitting picture is slightly large, and the user demand information may include "neckline reduction"; the size optimization parameters and fabric optimization parameters are adjustment parameters for adjusting the user's two-dimensional pattern drawing. For example, if the user demand information is "waist curve fits better" and "red", the corresponding size optimization parameters are two-dimensional pattern drawing adjustment parameters that make the waistline part in the two-dimensional pattern drawing closer to the user's waist circumference, and the fabric optimization parameters are fabric feature adjustment parameters that meet the user's color requirements.

[0151] Furthermore, a pattern drawing optimization model is trained based on the garment optimization information and the corresponding size optimization parameters and fabric optimization parameters to obtain a trained pattern drawing optimization model. Subsequently, the actual user size information, fabric feature information, the user's two-dimensional pattern drawing, and user requirement information are input into the trained pattern drawing optimization model to obtain size optimization parameters and / or fabric optimization parameters. The user's two-dimensional pattern drawing is then adjusted based on the size optimization parameters and / or fabric optimization parameters to obtain an optimized user's two-dimensional pattern drawing.

[0152] Through this embodiment, the pattern making optimization model is trained based on the clothing optimization information and its corresponding size optimization parameters and fabric optimization parameters to obtain a trained pattern making optimization model, and the user's two-dimensional pattern making is adjusted and optimized through the trained pattern making optimization model to obtain an optimized user two-dimensional pattern making model, thereby improving the accuracy of the clothing pattern making pattern, realizing personalized generation of clothing pattern making patterns, and supporting customized production.

[0153] In some embodiments, the above-mentioned method for generating a two-dimensional garment pattern drawing further includes the following steps:

[0154] Based on the three-dimensional model and fabric features of the target clothing, the clothing feature information of the target clothing is obtained by training a complete style-property joint model; the clothing feature information includes clothing property features and clothing style features.

[0155] Specifically, a training data set is constructed, which includes three-dimensional images of clothing, fabric features and corresponding clothing feature information. The clothing feature information includes clothing physical property features and clothing style features. Clothing style features can be manually labeled, and clothing physical property features can be obtained through measurement or simulation. The pre-constructed style-physical property joint model is trained based on the training data set.

[0156] Furthermore, the three-dimensional model and fabric features of the target clothing are input into a well-trained style-property joint model, and multi-layer feature extraction and semantic analysis are performed on the input data to obtain the clothing feature information of the target clothing, which includes clothing physical property features and clothing style features.

[0157] Clothing physical properties include objectively measurable physical and chemical properties of fabrics or other objective properties, such as shrinkage and elasticity. Clothing style features include the target clothing's type characteristics (e.g., casual wear, professional wear), aesthetic characteristics (e.g., elegant, lively, fashionable), scene characteristics (e.g., outdoor, home, formal), appearance characteristics (e.g., drape, gloss), and tactile characteristics (e.g., softness, roughness). Glossiness includes mercerization and high gloss.

[0158] Through this embodiment, based on the three-dimensional model and fabric features of the target clothing, the clothing feature information of the target clothing is obtained by training a complete style-physical property joint model; the clothing feature information includes clothing physical property features and clothing style features, thereby combining physical attributes with semantic style labels to construct a multimodal feature space and realize cross-modal reasoning, so that functional and aesthetic features can be given at the same time to support customized production.

[0159] In some embodiments, after obtaining the user's two-dimensional pattern drawing of the user's clothing, the following steps are further included:

[0160] Determine the area of ​​each user's two-dimensional tiled piece in the user's two-dimensional layout;

[0161] The fabric usage of each user's clothing is determined based on the area, fabric characteristics, and clothing characteristic information of each user's two-dimensional flat piece.

[0162] Specifically, the area of ​​each 2D flattened piece is calculated based on the cutting parameters of each piece in the 2D layout of the user's garment, as well as the piece structure (such as pleated areas, pocket areas, and patch areas) within the 2D flattened piece. The fabric usage of the garment is estimated based on the fabric characteristics and garment feature information of the user's garment. For example, if the garment is made of pure cotton, given its high shrinkage rate, it is necessary to allow for shrinkage and expansion when estimating fabric usage to ensure accurate estimation and quality of the finished product.

[0163] Furthermore, intelligent nesting can be performed based on factors such as fabric width, grain direction, and pattern repetition period to optimize piece layout, which helps improve the accuracy of estimating fabric usage while reducing production losses.

[0164] Through this embodiment, the area of ​​each user's two-dimensional flat piece in the user's two-dimensional pattern drawing is determined, and based on the area, fabric characteristics and clothing characteristic information of each user's two-dimensional flat piece, the fabric usage of the user's clothing is determined, and fabric calculation and nesting optimization are realized to reduce raw material costs and losses, while improving production efficiency and clothing effects.

[0165] In some embodiments, the above-mentioned method for generating a two-dimensional garment pattern drawing further includes the following steps:

[0166] The preset curvature threshold is dynamically adjusted based on the clothing feature information, and the wrinkle area is identified based on the adjusted preset curvature threshold.

[0167] Specifically, the preset curvature threshold is used to analyze and determine the wrinkle area of ​​the target garment. The curvature information of the surface mesh of the three-dimensional model is compared with the preset curvature threshold, and the area whose curvature value reaches the preset curvature threshold is detected as the wrinkle area.

[0168] In this embodiment, the preset curvature threshold is dynamically adjusted based on the target garment's characteristic information. For example, if the target garment is made of highly elastic spandex, the preset curvature threshold is dynamically lowered to accommodate the target garment's fabric. Wrinkle areas are then identified based on the adjusted preset curvature threshold, preventing missed or overdetected wrinkles.

[0169] Through this embodiment, the preset curvature threshold is dynamically adjusted based on the characteristic information of the clothing, and the wrinkle area is identified based on the adjusted preset curvature threshold, so as to achieve accurate identification of wrinkles, improve the structural accuracy of the two-dimensional flat-laid cutting pieces, and facilitate the subsequent accurate calculation of the fabric usage of the clothing.

[0170] The present embodiment is described and illustrated below through preferred embodiments.

[0171] Figure 8 Flowchart of the method for generating a two-dimensional clothing pattern drawing according to the preferred embodiment. Figure 8 As shown, the method for generating a two-dimensional clothing pattern drawing includes the following steps:

[0172] Step S801, identifying fabric features of a target garment through an input image, where the fabric features include fabric material, fabric color, and fabric structure;

[0173] Step S802, performing three-dimensional reconstruction on the input image of the target garment to obtain a three-dimensional model of the target garment;

[0174] Step S803: identifying seam lines in the 3D model using a semantic segmentation network, and segmenting the 3D model into a plurality of 3D curved surface panels based on the identification results;

[0175] Step S804: flatten each three-dimensional curved surface piece into a corresponding two-dimensional tiled piece using a surface parameterization algorithm, and determine the piece parameter information of each two-dimensional tiled piece;

[0176] Step S805, identifying other structural regions in the three-dimensional model and mapping the other structural regions to corresponding two-dimensional flat panels; the other structural regions include pleated regions, pocket regions, and patch regions;

[0177] Step S806: generating a two-dimensional pattern drawing of the target garment based on each two-dimensional tiled piece and the piece parameter information of each two-dimensional tiled piece;

[0178] Step S807: determining corresponding user size parameters based on the user size information, and adjusting the two-dimensional pattern drawing of the target garment based on the user size parameters to obtain a user two-dimensional pattern drawing of the user garment;

[0179] Step S808, generating a three-dimensional fitting image of the user based on the user size information and the user's two-dimensional pattern drawing;

[0180] Step S809: Based on the garment optimization information and its corresponding size optimization parameters and fabric optimization parameters, the pattern making optimization model is trained to obtain a trained pattern making optimization model; the user's two-dimensional pattern making model is adjusted and optimized using the trained pattern making optimization model to obtain an optimized user's two-dimensional pattern making model;

[0181] Step S810: Based on the three-dimensional model and fabric features of the target garment, clothing feature information of the target garment is obtained by training a complete style-property joint model; the clothing feature information includes clothing property features and clothing style features;

[0182] Step S811: determining the amount of fabric used for the user's clothing based on the area, fabric characteristics, and clothing characteristic information of each user's two-dimensional flattened piece in the user's two-dimensional pattern drawing.

[0183] This embodiment uses an input image to identify the fabric features of a target garment, including its material, color, and structure. The input image is then reconstructed into a 3D model of the target garment. A semantic segmentation network is then used to identify seams within the 3D model. Based on the recognition results, the 3D model is segmented into multiple 3D curved panels.

[0184] Furthermore, using a surface parameterization algorithm, each 3D curved surface piece is flattened into a corresponding 2D flat piece. The piece parameter information for each 2D flat piece is determined, and other structural regions in the 3D model are identified and mapped to corresponding 2D flat pieces. These other structural regions include pleat regions, pocket regions, and patch regions. Based on each 2D flat piece and its piece parameter information, a 2D pattern drawing of the target garment is generated. Based on the user size information, the corresponding user size parameters are determined, and the 2D pattern drawing of the target garment is adjusted based on the user size parameters to obtain a user 2D pattern drawing of the user garment. Based on the user size information and the user 2D pattern drawing, a user 3D fitting image is generated.

[0185] Furthermore, based on the clothing optimization information and its corresponding size optimization parameters and fabric optimization parameters, the pattern making optimization model is trained to obtain a trained pattern making optimization model; through the trained pattern making optimization model, the user's two-dimensional pattern making is adjusted and optimized to obtain an optimized user's two-dimensional pattern making, thereby improving the accuracy of the clothing pattern making, realizing personalized generation of clothing pattern making, and supporting customized production.

[0186] Afterwards, based on the three-dimensional model and fabric features of the target clothing, the clothing feature information of the target clothing is obtained by training a complete style-property joint model. The clothing feature information includes clothing physical property features and clothing style features. Based on the area, fabric features and clothing feature information of each user's two-dimensional flat piece in the user's two-dimensional pattern drawing, the fabric usage of the user's clothing is determined, which solves the problems of low efficiency and insufficient accuracy of pattern drawing generation, and improves the efficiency and accuracy of pattern drawing generation. At the same time, it can accurately estimate the amount of fabric required for the target clothing, which helps to reduce raw material costs and losses and improve production efficiency.

[0187] It should be noted that the steps shown in the above process or the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0188] This embodiment also provides a two-dimensional garment pattern generation device, which is used to implement the above-mentioned embodiments and preferred embodiments. Details already described will not be repeated here. The terms "module," "unit," "subunit," etc. used below may refer to a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation using hardware, or a combination of software and hardware, is also possible and contemplated.

[0189] Figure 9This is a block diagram of the structure of the two-dimensional clothing pattern generating device of this embodiment. Figure 9 As shown, the device includes:

[0190] The reconstruction module 10 is used to perform three-dimensional reconstruction on the input image of the target garment to obtain a three-dimensional model of the target garment;

[0191] The segmentation module 20 is used to identify the three-dimensional model and obtain multiple three-dimensional curved surface pieces in the three-dimensional model;

[0192] A flattening module 30 is used to map each 3D curved surface piece to a corresponding 2D tiled piece and determine the piece parameter information of each 2D tiled piece;

[0193] The generating module 40 is configured to generate a two-dimensional pattern drawing of the target garment based on each two-dimensional tiled piece and the piece parameter information of each two-dimensional tiled piece.

[0194] By using the device provided in this embodiment, a three-dimensional model of the target garment is obtained by performing three-dimensional reconstruction on an input image of the target garment; the three-dimensional model is identified to obtain multiple three-dimensional curved surface patches in the three-dimensional model; each three-dimensional curved surface patch is mapped to a corresponding two-dimensional flat patch, and the patch parameter information of each two-dimensional flat patch is determined; and based on the two-dimensional flat patches and the patch parameter information of each two-dimensional flat patch, a two-dimensional pattern drawing of the target garment is generated, thereby solving the problems of low pattern drawing generation efficiency and insufficient pattern drawing accuracy, and achieving improved pattern drawing generation efficiency and pattern drawing accuracy.

[0195] In some embodiments, the segmentation module 20 is further configured to identify seam lines in the three-dimensional model through a semantic segmentation network; and based on the identification result, segment the three-dimensional model into a plurality of three-dimensional curved surface panels.

[0196] In some embodiments, the flattening module 30 is further configured to flatten each three-dimensional surface patch into a corresponding two-dimensional tiled patch using a surface parameterization algorithm.

[0197] In some embodiments, the flattening module 30 is further configured to identify other structural regions in the three-dimensional model and map the other structural regions to corresponding two-dimensional flat pieces; the other structural regions include wrinkle regions, pocket regions, and patch regions.

[0198] In some embodiments, the flattening module 30 is further used to perform curvature analysis on the three-dimensional model to identify wrinkle areas in the three-dimensional model; the curvature value of the wrinkle area reaches a preset curvature threshold; based on the curvature value of the wrinkle area, the wrinkle area is mapped to a shadow area in the corresponding two-dimensional tiled piece.

[0199] In some embodiments, the reconstruction module 10 is further used to extract features from an input image of a target garment to obtain a first image feature of the input image; map the extracted first image feature to a neural implicit field to obtain a corresponding first implicit geometric representation; match the first implicit geometric representation with a preset implicit geometric representation template; and generate a three-dimensional model of the target garment based on the implicit geometric representation template that matches the first implicit geometric representation.

[0200] In some embodiments, the reconstruction module 10 is further used to extract features from the input images of the target garment to obtain second image features of each input image; map the extracted second image features to the neural implicit field to obtain a corresponding second implicit geometric representation; optimize the second implicit geometric representation based on the photometric consistency constraint of the multi-view images; and generate a three-dimensional model of the target garment based on the optimized second implicit geometric representation.

[0201] In some of these embodiments, Figure 9 On the basis of the above, the device also includes an adaptation module for determining corresponding user size parameters according to the user size information; adjusting the two-dimensional pattern making drawing of the target clothing according to the user size parameters to obtain the user two-dimensional pattern making drawing of the user clothing; and generating a user three-dimensional fitting image based on the user size information and the user two-dimensional pattern making drawing.

[0202] In some of these embodiments, Figure 9 On the basis of, the device also includes a recognition module for identifying fabric features of the target clothing through the input image, wherein the fabric features include at least one of the following: fabric raw material, fabric color, and fabric structure.

[0203] In some of these embodiments, Figure 9 On the basis of, the device also includes an optimization module, which is used to train the pattern making optimization model based on clothing optimization information, size optimization parameters and fabric optimization parameters corresponding to the clothing optimization information to obtain a trained pattern making optimization model; the clothing optimization information includes user size information, fabric characteristics, user two-dimensional pattern making map and user demand information; through the trained pattern making optimization model, the user two-dimensional pattern making map is adjusted and optimized to obtain an optimized user two-dimensional pattern making map.

[0204] In some of these embodiments, Figure 9 On the basis of the above, the device also includes an extraction module for obtaining clothing feature information of the target clothing by training a complete style-property joint model based on the three-dimensional model and fabric features of the target clothing; the clothing feature information includes clothing physical property features and clothing style features.

[0205] In some of these embodiments, Figure 9On the basis of the above, the device also includes an analysis module for determining the area of ​​each user's two-dimensional flat piece in the user's two-dimensional pattern making drawing; based on the area, fabric characteristics and clothing characteristic information of each user's two-dimensional flat piece, the fabric usage of the user's clothing is determined.

[0206] In some of these embodiments, Figure 9 On the basis of this, the device also includes an adjustment module for dynamically adjusting a preset curvature threshold based on clothing feature information, and identifying a wrinkle area based on the adjusted preset curvature threshold.

[0207] It should be noted that the above modules can be functional modules or program modules, and can be implemented through software or hardware. For modules implemented through hardware, the above modules can be located in the same processor; or the above modules can be located in different processors in any combination.

[0208] This embodiment further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0209] Optionally, the computer device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0210] Optionally, in this embodiment, the processor may be configured to execute the following steps through a computer program:

[0211] S1, perform 3D reconstruction on the input image of the target clothing to obtain a 3D model of the target clothing;

[0212] S2, identifying the three-dimensional model to obtain multiple three-dimensional surface pieces in the three-dimensional model;

[0213] S3, mapping each 3D curved surface patch to a corresponding 2D tiled patch, and determining the patch parameter information of each 2D tiled patch;

[0214] S4: Generate a two-dimensional pattern drawing of the target garment based on each two-dimensional tiled piece and the piece parameter information of each two-dimensional tiled piece.

[0215] It should be noted that, for specific examples in this embodiment, reference may be made to the examples described in the above embodiments and optional implementation modes, and will not be repeated in this embodiment.

[0216] In addition, in conjunction with the two-dimensional garment pattern generation method provided in the above embodiments, this embodiment may also provide a storage medium for implementation. The storage medium stores a computer program; when the computer program is executed by a processor, any of the two-dimensional garment pattern generation methods in the above embodiments is implemented.

[0217] It should be understood that the specific embodiments described herein are only used to explain this application and are not used to limit it. Based on the embodiments provided in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0218] Obviously, the accompanying drawings are merely examples or embodiments of the present application. A person skilled in the art can also apply the present application to other similar situations based on these drawings without inventive effort. Furthermore, it is understandable that, although the work involved in this development process may be complex and lengthy, certain design, manufacturing, or production changes based on the technical content disclosed in this application are merely routine technical means for a person skilled in the art and should not be considered to constitute a deficiency in the disclosure of the present application.

[0219] The term "embodiment" as used in this application refers to specific features, structures, or characteristics described in conjunction with the embodiment that can be included in at least one embodiment of this application. The appearance of this phrase in various places in the specification does not necessarily mean that the embodiment is the same, nor does it mean that it is mutually exclusive with other embodiments and is independent or optional. It is understood, either explicitly or implicitly, by those skilled in the art that the embodiments described in this application can be combined with other embodiments when there is no conflict.

[0220] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of patent protection. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. A method for generating a two-dimensional clothing pattern, characterized in that: include: Performing three-dimensional reconstruction on an input image of a target garment to obtain a three-dimensional model of the target garment; Identifying the three-dimensional model to obtain a plurality of three-dimensional curved surface pieces in the three-dimensional model; Mapping each of the three-dimensional curved surface patches to a corresponding two-dimensional tiled patch, and determining patch parameter information of each of the two-dimensional tiled patches; generating a two-dimensional pattern drawing of the target garment based on each of the two-dimensional flat panels and the panel parameter information of each of the two-dimensional flat panels; Determining corresponding user size parameters based on user size information; the user size information includes user key size parameters, images, or body descriptions; Wherein, when the user size information is an image or a body description, determining the corresponding user size parameters according to the user size information includes: converting the user size information into the corresponding user size parameters through a size parameter model; Adjusting the two-dimensional pattern drawing of the target garment according to the user size parameter to obtain a user two-dimensional pattern drawing of the user garment; generating a three-dimensional fitting image of the user based on the user size information and the user's two-dimensional pattern drawing; Identifying fabric features of the target garment through the input image, wherein the fabric features include at least one of the following: fabric material, fabric color, and fabric structure; Training a pattern drawing optimization model based on clothing optimization information, size optimization parameters and fabric optimization parameters corresponding to the clothing optimization information to obtain a trained pattern drawing optimization model; the clothing optimization information includes the user size information, the fabric characteristics, the user's two-dimensional pattern drawing, and user demand information; the user demand information is obtained by the user referring to the user's three-dimensional fitting image; The user's two-dimensional plate-making drawing is adjusted and optimized by using the trained plate-making drawing optimization model to obtain the optimized user's two-dimensional plate-making drawing.

2. The method for generating a two-dimensional clothing pattern according to claim 1, wherein: The identifying the three-dimensional model to obtain a plurality of three-dimensional curved surface pieces in the three-dimensional model includes: identifying seam lines in the three-dimensional model using a semantic segmentation network; Based on the recognition result, the three-dimensional model is segmented into a plurality of the three-dimensional curved surface panels.

3. The method for generating a two-dimensional clothing pattern according to claim 1, wherein: Mapping each of the three-dimensional curved surface patches to a corresponding two-dimensional tiled patch includes: Each of the three-dimensional curved surface panels is flattened into the corresponding two-dimensional tiled panel using a surface parameterization algorithm.

4. The method for generating a two-dimensional clothing pattern according to claim 1, wherein: Before generating the two-dimensional pattern drawing of the target garment, the method further includes: Identify other structural areas in the three-dimensional model and map the other structural areas to the corresponding two-dimensional flat panels; the other structural areas include pleated areas, pocket areas, and patch areas.

5. The method for generating a two-dimensional clothing pattern according to claim 4, wherein: The other structural region is the fold region; The identifying other structural regions in the three-dimensional model and mapping the other structural regions to the corresponding two-dimensional tiled pieces includes: Performing curvature analysis on the three-dimensional model to identify the wrinkle region in the three-dimensional model; the curvature value of the wrinkle region reaches a preset curvature threshold; Based on the curvature value of the wrinkle area, the wrinkle area is mapped to a corresponding shadow area in the two-dimensional tiled piece.

6. The method for generating a two-dimensional clothing pattern according to claim 1, wherein: The input image is a single-view image; Performing three-dimensional reconstruction on an input image of a target garment to obtain a three-dimensional model of the target garment includes: performing feature extraction on the input image of the target garment to obtain a first image feature of the input image; Mapping the extracted first image features to a neural implicit field to obtain a corresponding first implicit geometric representation; Matching the first implicit geometric representation with a preset implicit geometric representation template; The three-dimensional model of the target garment is generated based on the implicit geometric representation template that matches the first implicit geometric representation.

7. The method for generating a two-dimensional clothing pattern according to claim 1, wherein: The input image is a multi-view image; and the three-dimensional reconstruction of the input image of the target garment to obtain a three-dimensional model of the target garment includes: performing feature extraction on the input image of the target garment to obtain a second image feature of each input image; Mapping the extracted second image features to a neural implicit field to obtain a corresponding second implicit geometric representation; Optimizing the second implicit geometric representation based on a photometric consistency constraint of the multi-view images; The three-dimensional model of the target garment is generated based on the optimized second implicit geometric representation.

8. The method for generating a two-dimensional clothing pattern according to claim 1, wherein: The method further comprises: Based on the three-dimensional model and fabric features of the target clothing, clothing feature information of the target clothing is obtained by training a complete style-property joint model; the clothing feature information includes clothing property features and clothing style features.

9. The method for generating a two-dimensional clothing pattern according to claim 8, wherein: After obtaining the user two-dimensional pattern drawing of the user's clothing, the method further includes: Determine the area of ​​each user's two-dimensional tiled piece in the user's two-dimensional layout; The fabric usage of the user's clothing is determined based on the area, the fabric characteristics, and the clothing characteristic information of each of the user's two-dimensional flat pieces.

10. The method for generating a two-dimensional clothing pattern according to claim 8, wherein: The method further comprises: The preset curvature threshold is dynamically adjusted based on the clothing feature information, and the wrinkle area is identified based on the adjusted preset curvature threshold.

11. A two-dimensional clothing pattern generation device, characterized in that: include: A reconstruction module, configured to perform three-dimensional reconstruction on an input image of a target garment to obtain a three-dimensional model of the target garment; a segmentation module, configured to identify the three-dimensional model and obtain a plurality of three-dimensional curved surface pieces in the three-dimensional model; a flattening module, configured to map each of the three-dimensional curved surface patches to a corresponding two-dimensional tiled patch, and determine patch parameter information of each of the two-dimensional tiled patches; a generating module, configured to generate a two-dimensional pattern drawing of the target garment based on each of the two-dimensional flat panels and the panel parameter information of each of the two-dimensional flat panels; An adaptation module, configured to determine corresponding user size parameters based on user size information; the user size information includes key user size parameters, an image, or a body description; The adaptation module is further configured to convert the user size information into the corresponding user size parameters through a size parameter model when the user size information is an image or a body description; The adaptation module is further configured to adjust the two-dimensional pattern drawing of the target garment according to the user size parameters to obtain a user two-dimensional pattern drawing of the user garment; The adaptation module is further configured to generate a three-dimensional fitting image of the user based on the user size information and the user two-dimensional pattern drawing; a recognition module, configured to recognize fabric features of the target garment through the input image, wherein the fabric features include at least one of the following: fabric material, fabric color, and fabric structure; an optimization module, configured to train a pattern drawing optimization model based on garment optimization information, size optimization parameters corresponding to the garment optimization information, and fabric optimization parameters, to obtain a trained pattern drawing optimization model; the garment optimization information includes the user size information, the fabric characteristics, the user's two-dimensional pattern drawing, and user demand information; the user demand information is obtained by the user by referring to the user's three-dimensional fitting image; The optimization module is further used to adjust and optimize the user's two-dimensional plate-making drawing through the trained plate-making drawing optimization model to obtain the optimized user's two-dimensional plate-making drawing.

12. A computer device comprising a memory and a processor, characterized in that: A computer program is stored in the memory, and the processor is configured to run the computer program to execute the steps of the method for generating a two-dimensional garment pattern drawing according to any one of claims 1 to 10.

Images