Method and apparatus for optimizing clothing simulation parameter
The neural network-based method for estimating physical property parameters addresses the limitations of current simulation methods by learning mesh correlations and updating parameters to accurately simulate fabric draping, achieving efficient and accurate virtual fabric simulations.
Patent Information
- Application Number
- PCT/KR2024/007046
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2024-05-23
- Filing Date
- 2024-05-24
- Publication Date
- 2024-11-28
AI Technical Summary
Current methods for estimating physical property parameters for 3D clothing simulation in the fashion design industry are limited in speed and flexibility, failing to accurately replicate the physical properties of fabrics in virtual simulations.
A method using a neural network to generate and update physical property parameters for virtual fabrics by learning correlations between mesh structures and target drape data, minimizing errors through optimizer-based updates, allowing for efficient and accurate simulation of fabric draping.
This approach enables rapid convergence of physical property parameters within seconds to minutes, generating realistic drape simulation results that match actual fabric behavior, improving the accuracy and speed of virtual fabric simulations.
Smart Images

Figure KR2024007046_28112024_PF_FP_ABST
Abstract
Description
Method and device for optimizing costume simulation parameters
[0001] The following examples relate to a method and device for optimizing clothing simulation parameters, and more specifically, to estimating material properties for three-dimensional simulation.
[0002] In the fashion design industry, virtual simulation technology is used to develop new clothing designs. When fashion designers utilize virtual simulation, it is essential to determine material parameters that represent the physical properties of actual fabrics or textiles. Current methods for obtaining material parameters include using mechanical devices to measure the characteristics of fabric samples, optimization methods that minimize the difference between simulation results and target results, and learning models. However, there is a growing need to develop methods to improve the accuracy of material parameter estimation.
[0003]
[0004] The following examples provide a method for estimating material parameters for 3D simulation that satisfies both the speed of learning-based neural networks and the flexibility of optimization methods.
[0005] However, technical challenges are not limited to the technical challenges described above, and other technical challenges may exist.
[0006] A method for estimating physical property parameters for draping a virtual fabric according to one embodiment includes the steps of: generating a mesh by applying physical property parameters corresponding to the virtual fabric to a neural network, wherein the neural network learns a correlation between the physical property parameters and a mesh of a state in which the virtual fabric is draped on an object; obtaining simulated drape data corresponding to a type of target data regarding a drape of the virtual fabric based on the mesh; and updating the physical property parameters based on an error between the obtained simulated drape data and the target data.
[0007] The step of updating the above property parameters may include a step of updating the property parameters based on an optimizer for determining a value of the property parameters that minimizes the error.
[0008] The step of generating the mesh may include the step of obtaining a latent vector corresponding to the material parameter from a regressor of the neural network; and the step of generating the mesh corresponding to the latent vector based on a decoder of the neural network.
[0009] The above error may include at least one of an error corresponding to a 3D shape difference of a boundary curve between the target data and the simulated drape data; an error corresponding to a 2D shape difference of a boundary curve between the target data and the simulated drape data; an error corresponding to a difference in a depth image matrix of the target data and the simulated drape data; and an error corresponding to a difference in a reference point of the target data and the simulated drape data.
[0010] The method for estimating physical property parameters for drape simulation of the virtual fabric may further include a step of obtaining the physical property parameters based on the target data input by the user.
[0011] The type of the target data may include at least one of a type of 3D scan data regarding the drape of the virtual fabric; a type of 2D image data regarding the drape of the virtual fabric; a type of sketch data regarding the drape of the virtual fabric; a type of depth image data regarding the drape of the virtual fabric; and a type of numerical data regarding the drape of the virtual fabric.
[0012] The above target data may include at least one of drape data in the form of CIR, drape data in the form of SQR, and drape data in the form of CAP.
[0013] The above material property parameters may include at least one of a parameter relating to stretch stiffness and a parameter relating to bending stiffness.
[0014] The parameters relating to the tensile stiffness include first parameters and second parameters relating to factors of a stretch coefficient function, and may include a weft direction parameter, a warp direction parameter, and a bias direction parameter of each of the first parameter and the second parameter.
[0015] The parameters relating to the bending stiffness include third and fourth parameters relating to factors of a bending coefficient function, and each of the third and fourth parameters may include a weft direction parameter, a warp direction parameter, and a bias direction parameter.
[0016] The method for estimating physical property parameters for drape simulation of the virtual fabric may further include a step of generating a drape simulation result of the virtual fabric based on the updated physical property parameters.
[0017]
[0018] An electronic device for estimating physical property parameters for simulating a drape of a virtual fabric according to one embodiment includes a processor configured to generate a mesh by applying physical property parameters corresponding to the virtual fabric to a neural network, the neural network having learned a correlation between the physical property parameters and a mesh of a state in which the virtual fabric is draped on an object, and to obtain simulated drape data corresponding to a type of target data regarding a drape of the virtual fabric based on the mesh, and to update the physical property parameters based on an error between the obtained simulated drape data and the target data.
[0019] FIG. 1 is a flowchart of a method for estimating material parameters for drape simulation of a virtual fabric according to one embodiment.
[0020] FIG. 2 is a diagram illustrating the structure of a system for estimating material property parameters according to one embodiment.
[0021] Figures 3a and 3b are diagrams illustrating the determined material parameter values as a result of optimization of a set of material parameter values corresponding to a similar target drape.
[0022] Figures 4a to 4c are drawings for explaining examples of different types of drape shapes.
[0023] Figure 5 is a drawing for explaining the specific structure and learning method of a simulation prediction model according to one embodiment.
[0024] Figures 6a and 6b are diagrams illustrating a graph of reconstruction error and a graph of regression error according to the number of learning data, respectively.
[0025] Figures 7a to 7c are drawings illustrating target data according to one embodiment.
[0026] Figure 8 is an exemplary diagram of the configuration of a device according to one embodiment.
[0027] Specific structural or functional descriptions of the embodiments are disclosed for illustrative purposes only and may be modified and implemented in various forms. Therefore, the actual implementation is not limited to the specific embodiments disclosed, and the scope of this specification includes modifications, equivalents, or alternatives within the technical concepts described in the embodiments.
[0028] In connection with the description of the drawings, similar reference numerals may be used for similar or related components. The singular form of the noun corresponding to an item may include one or more of said items, unless the context clearly indicates otherwise.
[0029] In this document, each of the phrases "A or B", "at least one of A and B", "at least one of A or B", "A, B, or C", "at least one of A, B, and C", and "at least one of A, B, or C" may include any one of the items listed together in that phrase, or all possible combinations thereof.
[0030] Terms such as "first," "second," or "first" or "second" may be used simply to distinguish one component from another and do not limit the components in any other respect (e.g., importance or order). For example, a first component may be referred to as a second component, and similarly, a second component may also be referred to as a first component.
[0031] When a component (e.g., a first component) is referred to as being “coupled” or “connected” to another component (e.g., a second component), with or without the terms “functionally” or “communicatively,” it means that the component can be connected to the other component directly (e.g., wired), wirelessly, or through a third component.
[0032] Singular expressions include plural expressions unless the context clearly dictates otherwise. In this specification, the terms "comprises" or "has" should be understood to indicate the presence of a described feature, number, step, operation, component, part, or combination thereof, but not to exclude the possibility of the presence or addition of one or more other features, numbers, steps, operations, components, parts, or combinations thereof.
[0033] Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by a person of ordinary skill in the art. Terms defined in commonly used dictionaries should be interpreted to have a meaning consistent with their meaning in the context of the relevant technology, and will not be interpreted in an idealized or overly formal sense unless explicitly defined herein.
[0034] Hereinafter, embodiments will be described in detail with reference to the attached drawings. In the description with reference to the attached drawings, identical components are assigned the same reference numerals regardless of the drawing numbers, and redundant descriptions thereof will be omitted.
[0035]
[0036] FIG. 1 is a flowchart of a method for estimating material parameters for drape simulation of a virtual fabric according to one embodiment.
[0037] Referring to FIG. 1, a method for estimating physical property parameters according to an embodiment may include a step (110) of generating a mesh by applying physical property parameters corresponding to a virtual fabric to a neural network. The virtual fabric may correspond to a virtual implementation of a fabric that exists in reality.
[0038] Material parameters are parameters related to physical properties that appear in or can be measured from a fabric (or cloth or textile). For example, the material parameters may include parameters used in virtual simulations to reproduce the physical characteristics of actual fabrics. For example, the material parameters may include at least one of a parameter related to tensile stiffness and a parameter related to bending stiffness. Tensile stiffness is a numerical expression of the strength with which a fabric resists a tensile load (or a force that tries to stretch the fabric). Bending stiffness is a numerical expression of the strength with which a fabric resists a bending load (or a force that tries to bend the fabric). Specific examples of material parameters are described in detail below.
[0039] A method for estimating material properties according to one embodiment may include a step of obtaining material properties based on target data regarding the drape of a virtual fabric input by a user. Initial values of the material properties may be determined based on the target data. For example, the initial values of the material properties may be obtained by a measuring device that measures the material properties from the target data.
[0040] The target data may be data including information about the fabric of the virtual fabric or information about the drape of the virtual fabric. For example, the target data may include at least one of 3D scan data about the drape of the virtual fabric, 2D image data about the drape of the virtual fabric, sketch data about the drape of the virtual fabric, depth image data about the drape of the virtual fabric, and numerical data about the drape of the virtual fabric. Specifically, the target data may be experimentally acquired data regarding various data corresponding to any fabric. More specifically, the target data may be data confirming the result of draping any fabric on an object.
[0041] For example, 3D scan data regarding a drape of a virtual fabric may include data acquired by 3D scanning the shape of an actual fabric placed on an object. For example, 2D image data regarding a drape of a virtual fabric may include 2D image data acquired by projecting the shape of the fabric placed on an object onto a specific plane (e.g., top view, front view). For example, sketch data regarding a drape of a virtual fabric may include data sketching the shape of the fabric placed on an object. For example, depth image data regarding a drape of a virtual fabric may include data including depth information of the shape of the fabric placed on an object acquired from a depth sensor. For example, numerical data regarding a drape of a virtual fabric is numerical data for representing the shape of the fabric placed on an object, and may include data regarding the size of the fabric, the size of the area where the fabric and the object come into contact, the distance from the floor to the vertex of the fabric, etc.
[0042] The processor can obtain one or more material property parameters matching the target data. The processor can obtain the material property parameters from a database, such as a lookup table, or by processing (e.g., interpolating) information stored in the database. Alternatively, the processor can be configured to infer material property parameters matching the target data using a pre-trained neural network.
[0043] For example, the drape simulation results of a virtual fabric with initial values of material properties may not be sufficiently similar to the drape results of an actual fabric. Therefore, the initial material properties may be updated through the following steps (120, 130). The updating of the material properties may be performed iteratively, that is, steps (110, 120, 130) may be repeatedly performed based on the updated material properties. Finally, based on the updated material properties, the processor may generate the drape simulation results of the virtual fabric.
[0044] The neural network may include a neural network that has learned correlations between material parameters and a mesh of a virtual fabric draped (or placed) on an object. The object may refer to an object used to simulate the drape of the fabric. As an example, but not a limitation, the object may be a three-dimensional cylindrical cylinder. In the present disclosure, the cylinder may have various shapes other than a cylinder. As another example, the object may be a three-dimensional avatar. The neural network may be trained to generate a mesh corresponding to the shape of the virtual fabric draped on the object. The specific structure and learning method of the neural network are described in detail below.
[0045]
[0046] A method for estimating material parameters according to an embodiment may include a step (120) of acquiring simulated drape data corresponding to a type of target data regarding a drape of a virtual fabric based on a mesh. The target data may be data corresponding to a drape shape of a virtual fabric that is a target of a drape simulation result of the virtual fabric generated based on the material parameters. The simulated drape data may be drape data simulated using a spring mass model based on a mesh. In other words, the simulated drape data for the virtual fabric may be drape data for the virtual fabric corresponding to drape data when an actual fabric is draped on an object.
[0047] For example, the type of target data may include at least one of a type of 3D scan data regarding a drape of a virtual fabric, a type of 2D image data regarding a drape of a virtual fabric, a type of sketch data regarding a drape of a virtual fabric, a type of depth image data regarding a drape of a virtual fabric, and a type of numerical data regarding a drape of a virtual fabric.
[0048] The simulated drape data corresponding to the type of target data may correspond to the type of target data extracted from the mesh generated from the neural network. For example, if the type of target data is a type of 3D scan data regarding the drape of a virtual fabric, 3D data representing the shape of the virtual fabric may be extracted from the mesh. For example, if the type of target data is a type of 2D image data regarding the drape of a virtual fabric or a type of sketch data regarding the drape of a virtual fabric, the shape of the virtual fabric acquired from the mesh may be projected onto a plane corresponding to the target data to acquire 2D data. For example, if the type of target data is a type of depth image data regarding the drape of a virtual fabric, the depth image data of the virtual fabric may be extracted from the mesh. For example, if the type of target data is a type of numerical data regarding the drape of a virtual fabric, the numerical data included in the target data may be extracted from the mesh.
[0049] For example, the target data may include at least one of CIR-type drape data, SQR-type drape data, and CAP-type drape data. The CIR-type, SQR-type, and CAP-type drape data are types of drape types, which are described in detail below.
[0050] A method for estimating material properties according to an embodiment may include a step (130) of updating material properties based on an error between acquired simulated drape data and target data. The step (130) of updating material properties may include a step of determining values of optimized material properties. The values of the optimized material properties may include values of material properties that match the shape of the fabric indicated by the target data.
[0051] The step (130) of updating the material property parameters according to one embodiment may include a step of updating the material property parameters based on an optimizer for determining values of the material property parameters that minimize an error.
[0052] The error may include the meaning of an error function for optimization. The error may be determined based on a difference in the shape of the target data and the simulated drape data. For example, the error may include at least one of an error corresponding to a difference in the 3D shape of the boundary curve between the target data and the simulated drape data, an error corresponding to a difference in the 2D shape of the boundary curve between the target data and the simulated drape data, an error corresponding to a difference in the depth image matrix of the target data and the simulated drape data, and an error corresponding to a difference in the reference point of the target data and the simulated drape data. An optimization operation based on the error is described in detail below.
[0053] According to one embodiment, a method for estimating material properties may include a step of generating a drape simulation result of a virtual fabric based on updated material properties. The processor may generate a drape simulation result of a virtual fabric corresponding to the material properties determined by optimization. That is, the processor may generate a drape simulation result of a virtual fabric corresponding to target data. The drape simulation result of the virtual fabric may include 3D data of a drape shape of the fabric indicated by the target data. The drape shape of the fabric may refer to a shape of the fabric worn on an object. For example, the drape simulation result of the virtual fabric may include 3D graphic data of a drape shape of the fabric. For example, the 3D graphic data may be output through an interface.
[0054]
[0055] FIG. 2 is a diagram illustrating the structure of a system for estimating material property parameters according to one embodiment.
[0056] Referring to FIG. 2, a system (200) according to an embodiment may be a system in which the method for estimating a material parameter (201) described above with reference to FIG. 1 is performed. The material parameter (201) input to the system (200) may include at least one material parameter. The system (200) may estimate a simulation result using a pre-trained neural network of a simulation prediction model (210) to optimize the material parameter (201), which is a simulation parameter, and may reconstruct drape data (211) of a virtual fabric. The drape data (211) of a virtual fabric reconstructed from the estimated simulation result may indicate the shape of the simulated drape. The drape data (211) of a virtual fabric reconstructed from the estimated simulation result may be referred to as simulated drape data or simulated drape. The drape data (211) of the reconstructed virtual fabric can be compared with target data based on various material property parameters (201), and the objective function can be freely designed according to the user's needs.
[0057] Target data (202) input into the system (200) according to one embodiment may include various types of data including at least one of 3D scan information of a drape shape of an actual fabric (e.g., 3D contour), a 2D silhouette image (e.g., 2D sketch), data on the height of a vertex of a worn fabric (e.g., tip to ground), and a depth image. For example, the 3D scan information of a drape shape of an actual fabric may correspond to 3D scan data regarding the drape of the aforementioned virtual fabric. For example, the 2D silhouette image may correspond to 2D image data regarding the drape of the aforementioned virtual fabric. For example, the height of a vertex of the worn fabric may correspond to numerical data regarding the drape of the aforementioned virtual fabric.
[0058] According to an embodiment, the material property parameters (201) can be converged to an optimization result within seconds or minutes through the system (200). In addition, by differently designating the range of the material property parameters (201) through the system (200), the system (200) can obtain multiple sets of material property parameters (201) that generate drape simulation results of the same or similar virtual fabric. For example, FIGS. 3A and 3B illustrate material property parameter values determined as a result of optimization of a material property parameter set corresponding to a similar target drape. Referring to the screen (312) displaying the range and value of the material property parameter set of FIG. 3A, the range and value of each material property parameter included in the material property parameter set can be designated as 0 or more and 99 or less. The processor can determine the material property parameter values within the designated range through optimization. The processor can be configured to output the drape simulation results of the virtual fabric corresponding to the material property parameters determined through the screen (311). Referring to the screen (322) displaying the range and values of the material property parameter set of FIG. 3B, the processor can set the range of some material property parameters (323) to 50 or more and 99 or less, and set the range of other material property parameters (324) to 0 or more and 50 or less. Accordingly, the material property parameter values can be determined by optimization within the specified range. The processor can be set to output a drape simulation result of a virtual fabric corresponding to the material property parameters determined through the screen (321). The processor can obtain a plurality of material property parameter sets that generate drape simulation results of the same or similar virtual fabric.
[0059] Referring back to FIG. 2, for estimating the material properties parameter (201) according to one embodiment, the processor may be configured to use a mesh of simulation results estimated from a neural network-based simulation prediction model (210). The processor may obtain a gradient of a change in the material properties parameter (201) according to a change in the simulation results by estimating the simulation results using the neural network. The gradient of the change in the material properties parameter (201) may be utilized to improve the speed and accuracy of local optimization of the material properties parameter (201). Estimating the simulation results using the neural network of the present disclosure may be faster in computational speed than performing an actual simulation. In addition, the material properties estimation method according to one embodiment may be applied regardless of the type of simulator. In other words, the system (200) may be applied to any type of simulator.
[0060] For example, the simulation prediction model (210) may be a model based on a spring mass model. BDF-2 implicit integration may be applied to the simulation prediction model (210). The simulation prediction model (210) may estimate the drape simulation results by considering the stretch stiffness and bending stiffness, which have a significant influence on the drape shape of the fabric. The stress-strain relationship may be defined as a nonlinear function.
[0061] For example, the stretch coefficient function () can be defined as an exponential function of the stretched length as shown in Equation 1 below.
[0062]
[0063] In mathematical equation 1, is the tensile base factor, is the tensile growth factor, which is a factor related to the slope of the function.
[0064]
[0065] For example, the bending coefficient function b( ) can be defined as in mathematical equation 2 below.
[0066]
[0067] In mathematical expression 2 is the bending stretch factor, is the bending scale factor. is the angle at which two adjacent mesh triangles bend, when fully unfolded. =0 ° is. When bent beyond a certain level (cos( )<=t), flexural strength factor is defined as a value greater than or equal to 0 and less than or equal to 1. can be reduced by. The density of the fabric can also affect the drape shape of the fabric. The simulation prediction model (210) can estimate the simulation result using the value obtained by dividing the tensile stiffness value by the density value. In this case, the density can be fixed as a constant. For example, the material property parameter (201) to be estimated is , , and Tensile stiffness and bending stiffness can be defined by an anisotropic model independently defined for each of the weft, warp, and bias directions. The total number of material parameters (201) can be 12. Weft refers to the yarn in the transverse direction of the fabric and can also be called 'weft yarn'. In addition, 'warp' refers to the yarn in the longitudinal direction of the fabric and can also be called 'warp yarn'. In other words, among the material parameters (201), the parameter related to tensile stiffness is the first parameter related to the argument of the tensile modulus function (e.g., ) and the second parameter (e.g. ) and may include a weft direction parameter, a warp direction parameter, and a bias direction parameter of each of the first parameter and the second parameter. The parameters related to tensile stiffness may include six parameters. Among the material parameters (201), the parameters related to bending stiffness include a third parameter related to the argument of the bending coefficient function (e.g., ) and the fourth parameter (e.g. ) and may include a weft direction parameter, a warp direction parameter, and a bias direction parameter of each of the third and fourth parameters. The parameters relating to bending stiffness may include six parameters.
[0068] The drape shape of the target data (202) according to one embodiment may include a CIR shape, an SQR shape, and a CAP shape. For example, referring to FIG. 4A, the CIR shape (413) may be a drape shape of a portion not supported by an object when a circular specimen (e.g., a circular specimen with a diameter of 30 cm) (411) is placed on a disk (412) of a cylinder, which is an object. In the case of the CIR shape (413), since the width of all unsupported areas of the specimen is the same, the drape shape may not be biased according to the mass distribution. For example, referring to FIG. 4b, the SQR shape (423) may be a drape shape of a portion not supported by an object when a square specimen (e.g., a square specimen with a side length of 30 cm) (421) is placed on a disk (422) of a cylinder, which is an object. The square specimen (421) may be advantageous in distinguishing the weft and warp directions compared to other shapes of specimens. Referring to FIG. 4c, the CAP shape (433) may be a drape shape when a capelet specimen (431), that is, a donut-shaped specimen (e.g., a sample with a circular hole with a diameter of 15 cm in the center of a circular specimen with a diameter of 60 cm), is placed around the neck and shoulders of an object, which is a body (432).
[0069] Referring back to FIG. 2, a sufficient amount of data on actual fabrics can be used to generate training data for the simulation prediction model (210). For example, optimized material parameters for 5,000 different actual fabrics can be acquired as training data from any garment simulator (e.g., CLO3D). The material parameter data can be classified into k clusters using a Gaussian mixture model (GMM), and then equal-sized sampling can be performed for each cluster. For example, k can be set to 5 and 60,000 data can be sampled for each cluster, resulting in material parameter data for a total of 300,000 fabrics. The results of performing simulations on samples of three types of drape shapes (e.g., CIR, SQR, CAP) for each material parameter sample can be stored as a mesh. The sample used in the simulation can be made of a randomly sampled triangular mesh so that the maximum distance between vertices is a predetermined value. For example, the sample can be made of a randomly sampled triangular mesh so that the distance between vertices is less than 5 mm for the CIR and SQR shapes, and less than 10 mm for the CAP shape. The number of vertices (or triangles) of the sample can be 5,148 (10,102) for the CIR shape, 6,563 (12,880) for the SQR shape, and 4,890 (37,930) for the CAP shape. The simulation time step can be set to 0.03 seconds.
[0070] A simulation prediction model (210) according to one embodiment may include a neural network that predicts a drape mesh, which is a drape simulation result for a specific material parameter (201).
[0071] Referring to Fig. 5, the specific structure and learning of the simulation prediction model (210) will be described. Referring to Fig. 5, the simulation prediction model may include a regressor (510) and a decoder (520). The regressor (510) of the simulation prediction model (210) may estimate a latent vector of a simulated mesh from input material property parameters (501). For example, the latent vector may be a 512-dimensional vector. The decoder (520) may decode the estimated latent vector to restore the drape mesh.
[0072] An autoencoder (530) may be used to extract latent vectors. The decoder (520) of the simulation prediction model may be a decoder of the autoencoder (530). The autoencoder (530) may include a mesh-structure-specific convolutional neural network (e.g., SpiralNet++) capable of encoding the mesh structure while maintaining the mesh structure. Latent vectors of a certain size (e.g., 512 dimensions) may be extracted for three types of drape shapes.
[0073] A simulation prediction model according to one embodiment may include a regressor (510) which is a regression model. For example, the regressor (510) may include a multilayer perceptron (MLP) regression model for estimating the relationship between a material parameter and a latent vector. For example, the regressor (510) may include nine hidden layers each having 64, 128, 256, 512, 1024, 512, 256, 128, and 64 hidden units, respectively. A ReLU activation function may be included between each layer.
[0074] An autoencoder (530) can be trained using the simulated drape mesh of the training data. Using the trained autoencoder (530), the simulated drape mesh of all training data can be encoded, and latent vectors can be extracted. Then, a regressor (510) can be trained using the latent vectors.
[0075] For example, some randomly selected training data (e.g., 40,000 training data) may be used for training the autoencoder (530). The entire training data (e.g., 300,000 training data) may be used for training the regressor (510). For example, the graphs of FIGS. 6A and 6B respectively represent the reconstruction error and the regression error according to the number of training data. The reconstruction error of the autoencoder was already sufficiently low when trained with 40,000 training data, and there was no significant change in the error when training with more data. Meanwhile, in training the regressor, the regression error continuously decreases as the number of training data increases. This confirms that the diversity of the material parameters is greater than the diversity of the drape shape, and that the relationship between the material parameters and the drape shape is many-to-one.
[0076] Referring again to FIG. 2, an optimizer (220) according to one embodiment determines optimal material property parameters (201) that minimize the error between the target drape and the simulated drape. It may be a model for finding the target drape. The target drape may refer to the drape included in the target data (202). The simulated drape may refer to the drape included in the drape data (211) extracted from the mesh obtained from the simulation prediction model (210). The simulated drape may refer to the simulated drape data described above. An optimization problem according to one embodiment may be defined as in Mathematical Formula 3 below.
[0077]
[0078] In mathematical expression 3, P is a material property parameter (201), is a target drape included in the target data (202), and Sim(·) is a simulation prediction model (210). E(·) is an error function that calculates the difference between the target drape and the estimated simulated drape for the material parameter (201) P. D is a domain condition defined by the user. The set of material parameters (201) that generate a single drape shape may be assumed to be non-unique. Bound-constrained global optimization may be used to find different results depending on the user-specified conditions. For example, a dual annealing algorithm may be used for global optimization. Dual annealing may be adopted as one of the metaheuristic optimization algorithms for solving a global optimization problem. The processor may utilize dual annealing to minimize a nonlinear and irregular objective function. In particular, dual annealing may be a combination of a metaheuristic search method and sampling, which starts from a high-energy state and gradually converges to a lower-energy state during the process of exploring the parameter space to optimize the objective function. Dual annealing effectively explores the parameter space through a metaheuristic algorithm and random sampling, and the processor may adaptively set multiple hyperparameters that may affect the convergence speed or optimization performance during the exploration process through dual annealing to optimize the model of the present disclosure. For example, the L-BFGS-B algorithm may be used for local optimization. Local optimization may be performed using the gradient obtained by backpropagation of an optimization neural network.
[0079] Four error functions can be defined that can be used independently or together to train an optimization neural network according to one embodiment.
[0080] The material properties (201) can be estimated from the 3D boundary curve of the SQR-shaped drape. The difference in the shape of the boundary curve between the target drape and the simulated drape is expressed as the first error function E as shown in Equation 4 below. B3 can be defined as
[0081]
[0082] In mathematical equation 4, and are the ith vertices of the target drape and the simulated drape, respectively. is the set of indices of all sampling points on the boundary curve. is a set of indices of turning points of the boundary curve of the target drape. For example, referring to FIG. 7A, the target data may include 3D scan data (701) regarding a drape of a virtual fabric indicating the boundary curve and turning points of the target drape.
[0083] In Equation 4, ω(t) is a weight function that decreases over time (t). For example, it can be initially set to ω(0)=0.9, and ω can be decreased by 0.99% for every 100 evaluations of the objective function. When optimization begins, a high weight can be set for the number and location of turning points on the boundary curves that match, and as optimization progresses, a high weight can be set for the shape of the entire drape that matches.
[0084] The shape difference between the 2D boundary curves of the target drape and the simulated drape can be defined as a second error function. The second error function E B2 By projecting 3D points onto a 2D plane, the first error function E B3can be defined from. For example, a 2D boundary curve of a drape observed from a top view can be recorded using a rotoscope. For example, the 2D boundary curve can be drawn by a user. For example, referring to FIG. 7B, the target data can include sketch data (702) of a boundary curve of a CAP-shaped drape projected in a front view. The sketch data (702) of FIG. 7B can be an example of sketch data regarding a drape of the virtual fabric described above.
[0085] Target data (202) of the sketch data type can be used when it is difficult to digitally scan or photograph an actual drape. Furthermore, when actual fabric is not available, the designer can use the target data (202) to estimate material property parameters (201) by drawing a drape shape known through experience.
[0086] Depth images captured from various directions can be used as target data for a SQR-shaped drape. The depth images can contain data about the overall shape of the target drape and can also be used when it is difficult to accurately record the boundary curve of the target drape. The third error function E for the depth images D can be defined as in mathematical formula 5 below.
[0087]
[0088] In mathematical equation 5, and are matrices of the k-th depth images of the target drape and the simulated drape, respectively. There may be more than one depth image. For example, four depth images with a resolution of 256×256 pixels captured from different directions may be used.
[0089] When manually adjusting material properties (201), there are reference points that professional designers consider important. For example, referring to Fig. 7c, for a drape of a square sample, such as an SQR shape, the heights of the four corners (e.g., h1, h2, h3) may correspond to reference points. For example, for a T-shirt, the degree to which the rounded lines at the front and back of the neckline bend downward may also correspond to reference points. The fourth error function regarding the reference points can be defined as in mathematical equation 6 below:
[0090]
[0091] In mathematical expression 6, and are the lengths in the target drape and the simulated drape of the ith reference point, respectively.
[0092] Figure 8 is an exemplary diagram of the configuration of a device according to one embodiment.
[0093] Referring to FIG. 8, an electronic device (800) includes a processor (801), a memory (803), and an input / output device (I / O) (805). According to one embodiment, the electronic device (800) may include a device that performs the material parameter estimation method described above through FIGS. 1 to 7.
[0094] A processor (801) according to an embodiment may perform at least one operation related to the material parameter estimation method described above through FIGS. 1 to 7. For example, the processor (801) may perform at least one of an operation of generating a mesh by applying a material parameter corresponding to a virtual fabric to a neural network, an operation of obtaining drape data corresponding to a type of target data regarding a drape of the virtual fabric based on the mesh, and an operation of updating the material parameter based on an error between the obtained drape data and the target data.
[0095] The memory (803) according to one embodiment may be a volatile memory or a non-volatile memory, and may store data related to the material parameter estimation method described above through FIGS. 1 to 7. For example, the memory (803) may store data generated during the process of performing the material parameter estimation method or data required to perform the material parameter estimation method. For example, the memory (803) may store updated material parameters, and may store weight(s) and parameter(s) between layers included in the neural network and / or optimizer of the learned simulation prediction model.
[0096] According to one embodiment, the memory (803) may store a program implementing the material property parameter estimation method described above through FIGS. 1 to 7. The processor (801) may execute the program stored in the memory (803) and control the electronic device (800). The code of the program executed by the processor (801) may be stored in the memory (803).
[0097] An electronic device (800) according to an embodiment can receive data from a user through an input / output device (I / O) (805) and output generated data. For example, the electronic device (800) can receive target data through the input / output device (805) and output the values of determined physical property parameters or the simulation results of a virtual fabric generated based on the determined physical property parameters. The electronic device (800) can be connected to an external device (e.g., a personal computer or a network) through the input / output device (805) and exchange data.
[0098] The electronic device (800) according to one embodiment may further include other components not shown. For example, the electronic device (800) may further include a communication module that provides a function for the electronic device (800) to communicate with other electronic devices or other servers via a network. In addition, for example, the electronic device (800) may further include other components such as a transceiver, various sensors, a database, etc.
[0099]
[0100] The embodiments described above may be implemented using hardware components, software components, and / or a combination of hardware components and software components. For example, the devices, methods, and components described in the embodiments may be implemented using a general-purpose computer or a special-purpose computer, such as, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate array (FPGA), a programmable logic unit (PLU), a microprocessor, or any other device capable of executing instructions and responding to them. The processing device may execute an operating system (OS) and software applications running on the operating system. Furthermore, the processing device may access, store, manipulate, process, and generate data in response to the execution of the software. For ease of understanding, the processing device is sometimes described as being used alone; however, one of ordinary skill in the art will recognize that the processing device may include multiple processing elements and / or multiple types of processing elements. For example, a processing unit may include multiple processors, or a processor and a controller. Other processing configurations, such as parallel processors, are also possible.
[0101] Software may include a computer program, code, instructions, or a combination of one or more of these, and may configure a processing device to perform a desired operation or may independently or collectively command the processing device. The software and / or data may be stored on any type of machine, component, physical device, virtual equipment, computer storage medium, or device for interpretation by the processing device or for providing instructions or data to the processing device. The software may also be distributed over networked computer systems and stored or executed in a distributed manner. The software and data may be stored on a computer-readable recording medium.
[0102] The method according to the embodiment may be implemented in the form of program commands that can be executed through various computer means and recorded on a computer-readable medium. The computer-readable medium may store program commands, data files, data structures, etc., alone or in combination, and the program commands recorded on the medium may be those specially designed and configured for the embodiment or may be known and available to those skilled in the art of computer software. Examples of the computer-readable recording medium include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical media such as CD-ROMs and DVDs, magneto-optical media such as floptical disks, and hardware devices specially configured to store and execute program commands, such as ROMs, RAMs, and flash memories. Examples of program commands include not only machine language codes such as those generated by a compiler, but also high-level language codes that can be executed by a computer using an interpreter, etc.
[0103] The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the embodiments, and vice versa.
[0104] Although the embodiments described above have been described with limited drawings, those skilled in the art will appreciate that various technical modifications and variations can be applied based on the described embodiments. For example, appropriate results can still be achieved even if the described techniques are performed in a different order than described, and / or components of the described systems, structures, devices, circuits, etc. are combined or combined in a different manner than described, or are replaced or substituted with other components or equivalents.
[0105] Therefore, other implementations, other embodiments, and equivalents to the claims also fall within the scope of the claims described below.
Claims
1. In a method for estimating physical property parameters for drape simulation of a virtual fabric, A step of generating the mesh by applying the material properties corresponding to the virtual fabric to a neural network, wherein the neural network has learned a correlation between the material properties and a mesh of the virtual fabric draped on an object; A step of obtaining simulated drape data corresponding to the type of target data regarding the drape of the virtual fabric based on the mesh; and A step of updating the material property parameters based on the error between the obtained simulated drape data and the target data. including, method.
2. In paragraph 1, The step of updating the above physical property parameters is A step of updating the material property parameters based on an optimizer for determining the values of the material property parameters that minimize the above error. including, method.
3. In paragraph 1, The steps to create the above mesh are A step of obtaining a latent vector corresponding to the material parameter from the regressor of the neural network; and A step of generating the mesh corresponding to the latent vector based on the decoder of the neural network. including, method.
4. In paragraph 1, The above error is Error corresponding to the difference in 3D shape of the boundary curve between the target data and the simulated drape data; Error corresponding to the difference in 2D shape of the boundary curve between the target data and the simulated drape data; An error corresponding to the difference between the depth image matrix of the target data and the simulated drape data; and Error corresponding to the difference between the reference points of the above target data and the above simulated drape data Containing at least one of, method.
5. In paragraph 1, A step of obtaining the physical property parameters based on the target data input from the user. including more, method.
6. In paragraph 1, The type of the above target data is Type of 3D scan data regarding the drape of the above virtual fabric; Type of 2D image data regarding the drape of the above virtual fabric; Type of sketch data regarding the drape of the above virtual fabric; The type of depth image data regarding the drape of the above virtual fabric; and Type of numerical data regarding the drape of the above virtual fabric Containing at least one of, method.
7. In paragraph 1, The target data includes at least one of CIR-type drape data, SQR-type drape data, and CAP-type drape data. method.
8. In paragraph 1, The above physical property parameters are comprising at least one parameter relating to stretch stiffness and one parameter relating to bending stiffness, method.
9. In paragraph 8, The parameters related to the above tensile stiffness are Includes first and second parameters relating to the arguments of the stretch coefficient function, Each of the first parameter and the second parameter includes a weft direction parameter, a warp direction parameter, and a bias direction parameter, method.
10. In paragraph 8, The parameters related to the above bending stiffness are Including third and fourth parameters relating to the arguments of the bending coefficient function, Each of the third parameter and the fourth parameter includes a weft direction parameter, a warp direction parameter, and a bias direction parameter. method.
11. In paragraph 1, A step of generating a drape simulation result of the virtual fabric based on the updated material property parameters. including more, method.
12. A computer program stored on a medium to execute the method of claim 1 in combination with hardware.
13. In an electronic device for estimating physical property parameters for simulating the drape of a virtual fabric, By applying the material properties corresponding to the virtual fabric to a neural network, the neural network learns the correlation between the material properties and the mesh of the virtual fabric draped on the object, thereby generating the mesh, Based on the above mesh, simulated drape data corresponding to the type of target data regarding the drape of the virtual fabric is obtained, Based on the error between the obtained simulated drape data and the target data, the material property parameters are updated. processor including, Electronic devices.
Citation Information
Patent Citations
Parameter update device, classification device, parameter update program, and parameter update method
JP2021128569A
Electronic device and method for text recognition
KR1020240131210A
Method and apparatus of simulating apparel reflecting binding
KR102130252B1
System and method for ai based prediction of wearing fit
KR102224056B1
Method and apparatus for generating texture image
WO2023282614A1
Cited By
Method and device for acquiring geometric uncertainty of fabric composite yarn
CN119939949A
