Clothing simulation method and system based on simulation clothing and data driving and application
By using a data-driven approach based on simulated clothing, the real-time performance and resource consumption issues of clothing simulation in existing technologies have been resolved. This approach enables high-quality, low-resource-consumption clothing simulation, improving the realism and smoothness of virtual clothing.
Patent Information
- Application Number
- CN202410528535.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-04-29
- Publication Date
- 2025-10-31
AI Technical Summary
Existing clothing simulation technologies suffer from bottlenecks in real-time performance and computational resource consumption, making it impossible to efficiently display clothing effects in virtual environments. Furthermore, virtual clothing lacks realism and cannot effectively simulate the characteristics of clothing in the real world.
This paper adopts a data-driven approach based on simulated clothing. By acquiring clothing model meshes, performing preprocessing and smoothing to generate clothing skeletons, using a predictive network model to predict the transformation information of the clothing skeletons, and adjusting the clothing model mesh accordingly, and combining sewing thread and accessory processes for optimization, a deep learning network model is used for training and inference to achieve a realistic display of clothing in a virtual environment.
It achieves high real-time performance and high quality clothing simulation, reduces computing resource consumption, improves the realism and smoothness of virtual clothing, and enables high-quality clothing effects on a wider range of devices.
Smart Images

Figure CN120876676A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of 3D modeling, and more specifically, to a clothing simulation method, system, and application based on simulated clothing and data-driven methods. Background Technology
[0002] In the design field, 3D modeling has become an important tool. Traditional design methods are often limited by time and space; designers need to draw sketches on a flat surface and then create prototypes, which is time-consuming, labor-intensive, and costly. 3D modeling technology utilizes computer graphics and virtual reality to create a realistic 3D model on a computer. Designers can then design on this model without creating prototypes, significantly shortening the design cycle and reducing costs. Furthermore, problems can be identified and corrected in a timely manner during the design process, thereby improving the accuracy and quality of the design.
[0003] In the context of the rapid development of technologies such as metaverse, digital humans, games, and computer graphics, clothing simulation has become a key factor in enhancing virtual reality experiences and the realism of digital characters. However, how to efficiently and in real-time display the calculated clothing effects has become a common challenge for the industry.
[0004] Existing clothing simulation technologies suffer from real-time limitations. Due to the high computational load and complexity of the simulation, clothing effects cannot quickly adapt to user actions and scene changes in the virtual environment. This results in virtual character clothing appearing unrealistic or choppy during interactions. Furthermore, traditional machine learning methods can only train ordinary virtual clothing, leading to a lack of realism and an inability to effectively simulate the characteristics of real-world clothing. In addition, existing technologies typically require substantial computational resources and time for clothing calculations, necessitating the use of high-performance computing equipment. Summary of the Invention
[0005] To address the problems existing in current technologies, this application provides a clothing simulation method, system, and application based on simulated clothing and data-driven approaches. The specific solution is as follows:
[0006] The first part of this application provides a clothing simulation method based on simulated clothing and data-driven methods, including:
[0007] Obtain the simulated clothing and the clothing model mesh of the simulated clothing;
[0008] Preprocessing is performed based on the clothing technology involved in the simulated clothing, and animation data of the preprocessed simulated clothing in different poses is generated.
[0009] The animation data is smoothed, and the smoothed animation data is used to generate clothing skeletons;
[0010] The current human posture data is input into a pre-trained prediction network model to predict the transformation information of each bone on the clothing skeleton.
[0011] The clothing skeleton is adjusted based on the transformation information, and then the clothing model mesh is driven by the adjusted clothing skeleton.
[0012] In some specific embodiments, the garment manufacturing process includes a sewing thread process; the pretreatment of the sewing thread process specifically includes:
[0013] Obtain the individual panels corresponding to the simulated clothing;
[0014] Obtain the sewing thread on each plate and the vertex through which the sewing thread passes;
[0015] Merge two vertices connected by the same sewing thread on the simulated garment into one vertex, and calculate the coordinates of the merged vertex.
[0016] In some specific embodiments, the garment process includes an accessory process; the pretreatment of the accessory process specifically includes:
[0017] Obtain all accessories from the simulated clothing;
[0018] For each attachment, select one or more vertices as positioning vertices, and determine the vertices on the simulated clothing that have corresponding mapping relationships with each positioning vertices as reference vertices, and bind the positioning vertices and the corresponding reference vertices.
[0019] In some specific embodiments, during smoothing, the position of the vertex is moved along the line containing the normal vector of the vertex to maintain the shape of the simulated clothing.
[0020] In some specific embodiments, the human posture data includes the position of the human skeleton and the human mesh corresponding to the position of the human skeleton;
[0021] The transformation information includes the position of the garment skeleton and the rotation angle of the garment skeleton.
[0022] In some specific embodiments, the position information of each reference vertex is determined according to the transformation information, and the positioning vertex corresponding to the reference vertex is determined based on the corresponding mapping relationship, so as to restore the merged vertices.
[0023] In some specific embodiments, the training of the prediction network model specifically includes:
[0024] Acquire animation data containing the positions of the human skeleton and the clothing skeleton, and perform data preprocessing on the animation data to obtain training data;
[0025] The training data is divided into multiple consecutive time series in units of frames and imported into a preset network model for training. The model data is updated according to the smoothness loss function and the output results are encouraged to transition smoothly between consecutive frames until a preset stopping condition is met, and the trained prediction network model is obtained.
[0026] The second part of this application provides a clothing simulation system based on simulated clothing and data-driven methods, including:
[0027] The simulated clothing unit is used to acquire the simulated clothing and the clothing model mesh of the simulated clothing;
[0028] An animation data unit is used to preprocess the clothing technology involved in the simulated clothing and generate animation data of the simulated clothing in different poses after preprocessing.
[0029] The clothing skeleton unit is used to smooth the animation data and generate the clothing skeleton using the smoothed animation data.
[0030] The prediction unit is used to input the current human posture data into a pre-trained prediction network model to predict the transformation information of each bone on the clothing skeleton.
[0031] The output unit is used to adjust the clothing skeleton based on the transformation information, and then drive the clothing model mesh through the adjusted clothing skeleton.
[0032] Thirdly, this application provides a computer program product including executable instructions that, when executed, implement the clothing simulation method based on simulated clothing and data-driven methods as described in any of the first parts.
[0033] In the fourth part, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the clothing simulation method based on simulated clothing and data-driven methods as described in any of the first parts.
[0034] Beneficial Effects: This application provides a clothing simulation method, system, and application based on simulated clothing and data-driven simulation. It achieves high real-time performance, high quality, and low resource consumption in clothing simulation, effectively utilizing computing resources and reducing the complexity of the calculation, thereby enabling high-quality clothing effects on a wider range of devices. Employing data-driven algorithms and machine learning models to calculate clothing effects in real time allows the clothing to adapt to the user's movements more quickly, resulting in a more realistic and fluid performance in the virtual environment. By learning the real-world garment-making techniques of the simulated clothing, the quality and realism of the virtual clothing are improved.
[0035] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description
[0036] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.
[0037] Figure 1 This is a schematic diagram of the clothing simulation method of this application;
[0038] Figure 2 This is a schematic diagram illustrating the curvature-based smoothing principle of this application;
[0039] Figure 3 This is a schematic diagram of the effect before smoothing in this application;
[0040] Figure 4 This is a schematic diagram illustrating the effect of smoothing in this application;
[0041] Figure 5 This is a schematic diagram of the clothing simulation system module of this application.
[0042] Figure labels: 1-Simulation clothing unit; 2-Animation data unit; 3-Clothing skeleton unit; 4-Prediction unit; 5-Output unit. Detailed Implementation
[0043] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.
[0044] The terminology used in the various embodiments disclosed in this application is for the purpose of describing particular embodiments only and is not intended to limit the various embodiments disclosed in this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which the various embodiments disclosed in this application pertain. The terms (such as those defined in a generally used dictionary) are to be interpreted as having the same meaning as in the context of the relevant technical field and are not to be interpreted as having an idealized or overly formal meaning, unless clearly defined in the various embodiments disclosed in this application.
[0045] Example 1
[0046] This embodiment provides a clothing simulation method based on simulated clothing and data-driven approaches, enabling high real-time performance, high quality, and low resource consumption in clothing simulation. The flowchart of the clothing simulation method is attached. Figure 1 As shown, the specific solution is as follows:
[0047] A clothing simulation method based on simulated clothing and data-driven methods includes the following steps:
[0048] 101. Obtain the simulated clothing and the clothing model mesh of the simulated clothing;
[0049] 102. Based on the clothing technology involved in the simulated clothing, preprocessing is performed, and animation data of the preprocessed simulated clothing in different poses is generated;
[0050] 103. Smooth the animation data and use the smoothed animation data to generate clothing skeletons;
[0051] 104. Input the current human posture data into the pre-trained prediction network model to predict the transformation information of each bone on the clothing skeleton.
[0052] 105. Adjust the clothing skeleton based on the transformation information, and then drive the clothing model mesh through the adjusted clothing skeleton.
[0053] This embodiment of the clothing simulation scheme introduces a technology called simulated clothing, which highly reflects the details of real clothing. Simulated clothing refers to virtual clothing created through simulation technology. It can not only simulate the appearance of real clothing, but also its material, texture, dynamics, and other characteristics to achieve a lifelike effect.
[0054] The production of simulated clothing requires knowledge from various technical fields, including clothing design, simulation technology, and graphics rendering, and involves multiple stages. For example, firstly, a 3D model of the garment needs to be created based on design drawings or actual clothing. This model is then optimized for details, texture mapping, and material quality to make it more realistic. In some specific embodiments, simulated clothing uses special fabric and dyeing techniques to mimic the texture and luster of materials such as silk and real fur, enhancing the simulation effect. In some preferred embodiments, simulated clothing may employ special embroidery, sequins, appliqués, and other techniques to achieve the desired appearance and texture.
[0055] The mesh of a simulated garment model consists of a series of vertices, edges, and faces; that is, the mesh is composed of multiple polygons, which can be triangles, quadrilaterals, etc., depending on the situation. These vertices, edges, and faces define the shape and structure of the garment and, through their association with subsequent skeletons, drive the deformation of the garment. By creating vertices, edges, and faces, a 3D model of the garment can be constructed. Each vertex can be associated with one or more bones, and a weight value determines the degree to which the vertex is affected by the movement of the bones. Through the movement of the bones, the deformation of the model mesh can be driven, thereby simulating the shape of the garment in different poses. By adjusting parameters such as the shape, size, and weights of the model mesh, different human body shapes and garment styles can be simulated for design, fitting, and display purposes.
[0056] After acquiring the simulated clothing, it is necessary to obtain its animation data in different poses. In some embodiments, animation data can be generated using simulation technology. Since simulated clothing involves various processes, it requires special processing for convenient calculation.
[0057] The simulated clothing in this application employs real-world garment-making techniques, thereby improving the quality and realism of the virtual clothing. These techniques include sewing threads, topstitching, buttonholes, and buttoning. The specific techniques and processes used must be selected and applied according to different fabrics, design requirements, and production procedures to ensure the final garment quality and effect are optimal.
[0058] This embodiment divides the simulated clothing into two parts: a panel and accessories on the panel. Real-world clothing is mostly made by sewing together multiple pieces of fabric. The panel can be understood as the fabric on the simulated clothing, and the panels are sewn together with thread, requiring special treatment of the sewing thread. Accessories on the panel refer to various small parts attached to it, such as zippers, buckles, and decorations. The accessories are relatively small in volume compared to the panel and are not easily deformed. Their positions on the panel are relatively fixed, so they are not included in the calculation process. The position information of the accessories is obtained after the panel calculation is completed, based on the mapping relationship between the accessories and the panel.
[0059] In some specific embodiments, the garment manufacturing process includes a sewing thread process; preprocessing the sewing thread process specifically includes: obtaining various panels on the simulated garment, with the panels connected by sewing threads; obtaining the sewing threads on each panel and the vertices pierced by the sewing threads; merging two vertices connected by the same sewing thread on the simulated garment into one vertex, and calculating the coordinates of the merged vertex using a preset merging algorithm. Preferably, the merging algorithm specifically includes: obtaining the coordinates of the merged vertex by calculating the average of the vertex coordinates before merging.
[0060] Two vertices connected by the same sewing thread are adjacent to each other and have a high similarity. The two vertices connected by each sewing thread can be considered as the same vertex in space. This can be used to optimize the model by merging similar vertices to simplify the image structure and reduce the number of vertices. For example, averaging the coordinates of two vertices yields the merged vertex coordinates. The three vertices before and after merging are extremely close, exhibiting a high degree of similarity. This reduces the number of vertices, effectively lowering the model's complexity, reducing computational load, and simplifying the model structure.
[0061] In some specific embodiments, the garment manufacturing process includes an accessory manufacturing process. Preprocessing the accessory manufacturing process specifically includes: acquiring all accessories on the simulated garment; selecting one or more vertices for each accessory as positioning vertices, and determining the vertices on the simulated garment that have corresponding mapping relationships with each positioning vertices as reference vertices, and binding the positioning vertices and their corresponding reference vertices. Because accessories are small, not easily deformed, and have relatively fixed positions, they do not need to participate in position calculations. By setting positioning vertices and reference vertices, the mapping relationship between the accessories and the simulated garment is obtained. When the garment's position changes, the accessory position can be quickly calculated based on the mapping information, reducing the computational load.
[0062] More preferably, when the mesh consists of multiple triangles, after determining the positioning vertex, the position of the nearest triangle is calculated based on the position of the positioning vertex. The three vertices of this triangle are the reference vertices corresponding to the positioning vertex. The normal direction of the triangle is obtained by performing a cross product using the coordinates of the three vertices of the triangle. Finally, the minimum distance is calculated by projecting the normal direction using a dot product, thus obtaining the mapping relationship between the positioning vertex and the reference vertex. The specific expression is as follows:
[0063] Tri = Nearest(P) i )
[0064] Normal = CrossProduct(Tri) c -Tri a Tri b -Tri a )
[0065] Height = DotProcduct(P i -Tri a (Normal)
[0066] Where Tri represents the distance from the local vertex P i The nearest triangle, Tri a Tri b Tri c Let P represent the three vertices of triangle Tri, Normal represent the direction of the normal to triangle Tri, and Height represent point P. i The minimum distance to triangle Tri.
[0067] After preprocessing the animation data, a linear skinning decomposition algorithm is used to automatically generate the clothing skeleton and weights. The clothing skeleton is the virtual framework that drives the deformation of the clothing, while the weights represent the degree of influence of the bones on the clothing vertices. By adjusting the weights of the bones, the shape and movement of the clothing can be controlled. For example, in a virtual garment, if a vertex is affected by multiple bones simultaneously, the weights can be adjusted to determine which bone has the greatest influence on that vertex. Each vertex is assigned to one or more bones and a corresponding weight value is set. Furthermore, for complex clothing models, multiple layers of skeletal structure may be needed to better control its deformation. The shape and movement of the clothing are controlled by adjusting the relationship between vertices and bones. Through reasonable settings and adjustments, realistic virtual clothing effects can be achieved.
[0068] In some specific embodiments, the clothing is first denoised using Laplacian smoothing, and then a skinning algorithm is used to generate the clothing skeleton, preparing it for machine learning. The preferred skinning algorithm is linear skinning decomposition. Before training with animation data, the clothing mesh for each frame of animation is first subjected to curvature-based Laplacian smoothing. During smoothing, the position of a vertex is moved along the line containing its normal vector to maintain the shape of the simulated clothing. Traditional Laplacian smoothing would move point P to a position similar to the center of gravity. Curvature-based Laplacian smoothing moves the position opposite to the direction of the normal vector, effectively smoothing the model surface while maintaining the curvature characteristics and shape of the model, and resulting in a more uniform distribution of the generated skeleton. The principle is illustrated in the appendix. Figure 2 As shown in the attached figures, the model effects before and after curvature-based Laplacian smoothing are respectively... Figure 3 and 4 As shown.
[0069] The expression for smoothing is as follows:
[0070]
[0071]
[0072] Among them, Adj i (P) represents a point near point P, and Laplacian(P) represents the proportion of the weight of the i-th point near point P to the sum of the weights of all points near point P. i The value represents the influence coefficient of points near point P on it, ranging from 0 to 1. α and β represent the two included angles opposite the edges formed by point P and its adjacent points.
[0073] Linear skinning decomposition (LSD) is an algorithm for generating clothing skeletons. It associates clothing model meshes with skeletons and uses linear transformation matrices to map the motion of the skeletons onto the model mesh, thereby driving clothing deformation. In some specific embodiments, the process of constructing the clothing skeleton includes: adding skeletons to the clothing model and setting the hierarchical relationship and constraints of the skeletons according to the characteristics of the clothing, enabling the skeletons to drive the deformation of the clothing. Skeletons are bound to key parts of the model, such as shoulders, waist, and knees. Vertices in the model are bound to the skeletons, and a weight is set for each vertex. The weights determine the degree to which the vertex is affected by the movement of the skeletons. By adjusting the weights, the degree and detail of the deformation of the clothing in different poses can be controlled. Based on the movement of the skeletons, the deformation of the clothing model mesh is calculated using the LSD algorithm. This step involves calculating the linear relationship between the model vertices and the skeletons and mapping the motion of the skeletons onto the clothing model mesh. Based on actual needs and effects, the design, weights, and other parameters of the skeletons are adjusted to achieve the best clothing deformation effect.
[0074] In some specific embodiments, human pose data includes the position of the human skeleton and the corresponding human mesh; transformation information includes the position of the clothing skeleton and the rotation angle of the clothing skeleton. The input human pose data can be in the form of quaternions, Euler angles, transformation matrices, etc.
[0075] Regarding the prediction model, a deep learning network model can be selected. Preferably, the prediction model is a neural network model such as the Transformer model, GRU model, or LSTM model. The Transformer model is a neural network model based on a self-attention mechanism, consisting of an Encoder-Decoder architecture. It can handle sequence-to-sequence tasks. Through multi-layer self-attention mechanisms and linear transformations, it maps each word in the input sequence to a fixed-dimensional vector representation and calculates the relevance score between all words, thereby generating the output sequence. GRU and LSTM are both variants of recurrent neural networks (RNNs). They enhance the capabilities of RNNs by introducing gating mechanisms, enabling them to better handle sequence data.
[0076] In some specific embodiments, training the network model includes: acquiring animation data containing the positions of the human skeleton and clothing skeleton. Complete animation data includes the positions of the human skeleton and the corresponding clothing skeleton, covering various movements and poses. Next, the animation data is preprocessed to obtain training data. Preprocessing includes data cleaning, data alignment, and data standardization. Data cleaning removes potential outliers or missing values; data alignment ensures the human skeleton and clothing skeleton positions are aligned in time; and data standardization standardizes the skeleton position coordinates to ensure the input data is within a suitable range. The training data is then segmented into multiple consecutive time series, with each time step representing one frame, ensuring the sequence length is appropriate. The data is imported into a pre-defined network model for training. SGD and Adam optimizers are used to update model parameters. The model data is updated according to a smoothness loss function, encouraging smooth transitions in the output between consecutive frames. Each training iteration updates the model data according to the loss function until a preset stopping condition is met, resulting in a trained predictive network model. Considering the dependency of the neural network model, a loss function that guarantees animation smoothness is used for training, allowing the model to produce more vivid animation effects. The model can be divided into training, validation, and test sets, each imported into a separate set for iterative training until a preset stopping condition is met. Preferably, the preset stopping condition includes reaching a preset number of training iterations or the error rate falling below a preset value. Model performance is evaluated on the test set, using appropriate metrics to measure the continuity and accuracy of the predicted sequences. Based on the performance on the validation set, the model's hyperparameters, such as the number of layers, hidden units, and attention heads, are adjusted to improve continuity and accuracy. Finally, an ONNX file can be generated to increase the applicability of the prediction network model.
[0077] The transformation information of each virtual bone in the clothing is obtained through model inference, and the clothing skeleton is adjusted. The clothing skeleton adjusts the simulated clothing by changing the model mesh, thus obtaining the skinning effect of the clothing in the current human posture. Data-driven machine learning algorithms can display the movement effect of the clothing in real time, especially the dynamic effect. During preprocessing, after merging vertices, the inferred skinned vertex positions are still the merged vertex positions. However, due to the requirements of the fabrication process of the simulated clothing, the vertices of the simulated clothing displayed in real time do not need to be merged based on the stitching information. Therefore, they need to be restored based on the merged mapping information. The position information of each reference vertex is determined based on the transformation information, and the positioning vertices corresponding to the reference vertices are determined based on the corresponding mapping relationship to restore the merged vertices. Specifically, in the vertex merging stage, a map is used to associate the original vertex indices with the merged vertex indices. Then, after obtaining the vertex position information through model inference, the position information of the original vertices is set using the mapping relationship. After setting the vertex position information, the vertex position information can be fine-tuned by starting the simulation post-processing process as needed.
[0078] This embodiment provides a clothing simulation method based on simulated clothing and data-driven approaches. It achieves high real-time performance, high quality, and low resource consumption in clothing simulation, effectively utilizing computing resources and reducing the complexity of the calculation, thereby enabling high-quality clothing effects on a wider range of devices. Employing data-driven algorithms and machine learning models to calculate clothing effects in real time allows the clothing to adapt to the user's movements more quickly, resulting in a more realistic and fluid performance in the virtual environment. By learning the real-world garment-making techniques employed in the simulated clothing, the quality and realism of the virtual clothing are improved.
[0079] Example 2
[0080] This embodiment provides a clothing simulation system based on simulated clothing and data-driven methods. A schematic diagram of the system's modules is attached. Figure 5 As shown. The clothing simulation system specifically includes:
[0081] Simulation Clothing Unit 1 is used to acquire simulation clothing and the clothing model mesh of simulation clothing;
[0082] Animation data unit 2 is used to preprocess the clothing based on the clothing technology involved in the simulated clothing and generate animation data of the simulated clothing in different poses after preprocessing.
[0083] Clothing skeleton unit 3 is used to smooth the animation data and generate clothing skeleton using the smoothed animation data;
[0084] Prediction unit 4 is used to input the current human posture data into the pre-trained prediction network model in order to predict the transformation information of each bone on the clothing skeleton.
[0085] Output unit 5 is used to adjust the clothing skeleton based on transformation information, and then drive the clothing model mesh through the adjusted clothing skeleton.
[0086] This embodiment provides a clothing simulation system based on simulated clothing and data-driven methods, which systematizes the aforementioned clothing simulation method based on simulated clothing and data-driven methods, making it more practical.
[0087] Example 3
[0088] This application provides a computer program product including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform a clothing simulation method based on simulated clothing and data-driven methods according to Embodiment 1 of this application. The method includes the following steps:
[0089] 101. Obtain the simulated clothing and the clothing model mesh of the simulated clothing;
[0090] 102. Based on the clothing technology involved in the simulated clothing, preprocessing is performed, and animation data of the preprocessed simulated clothing in different poses is generated;
[0091] 103. Smooth the animation data and use the smoothed animation data to generate clothing skeletons;
[0092] 104. Input the current human posture data into the pre-trained prediction network model to predict the transformation information of each bone on the clothing skeleton.
[0093] 105. Adjust the clothing skeleton based on the transformation information, and then drive the clothing model mesh through the adjusted clothing skeleton.
[0094] This embodiment applies a clothing simulation method based on simulated clothing and data-driven methods to a computer program product, making it easy to execute.
[0095] Example 4
[0096] This application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of a clothing simulation method based on simulated clothing and data-driven methods as provided in Embodiment 1 of this application. The method includes:
[0097] 101. Obtain the simulated clothing and the clothing model mesh of the simulated clothing;
[0098] 102. Based on the clothing technology involved in the simulated clothing, preprocessing is performed, and animation data of the preprocessed simulated clothing in different poses is generated;
[0099] 103. Smooth the animation data and use the smoothed animation data to generate clothing skeletons;
[0100] 104. Input the current human posture data into the pre-trained prediction network model to predict the transformation information of each bone on the clothing skeleton.
[0101] 105. Adjust the clothing skeleton based on the transformation information, and then drive the clothing model mesh through the adjusted clothing skeleton.
[0102] The computer storage medium in this embodiment can be any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. For example, a computer-readable storage medium can be, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. This embodiment applies a clothing simulation method based on simulated clothing and data-driven clothing simulation to a computer-readable storage medium storing a computer program. When executed by a processor, this program implements the steps of the clothing simulation method provided in this application, which is simple, fast, easy to store, and not easily lost.
[0103] This application provides a clothing simulation method, system, and application based on simulated clothing and data-driven approaches. It enables high real-time performance, high quality, and low resource consumption in clothing simulation, effectively utilizing computing resources and reducing computational complexity, thereby achieving high-quality clothing effects on a wider range of devices. Employing data-driven algorithms and machine learning models for real-time clothing effect calculation allows the clothing to adapt to user movements more quickly, resulting in a more realistic and fluid performance in the virtual environment. By learning the real-world garment-making techniques employed in the simulated clothing, the quality and realism of the virtual clothing are improved.
[0104] Those skilled in the art will understand that the modules described above can be implemented using general-purpose computing systems. They can be centralized on a single computing system or distributed across a network of multiple computing systems. Optionally, they can be implemented using computer-executable program code, allowing them to be stored in a storage system for execution by the computing system. Alternatively, they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, this application is not limited to any particular combination of hardware and software.
[0105] Note that the above description is merely a preferred embodiment and the technical principles employed in this application. Those skilled in the art will understand that this application is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions can be made without departing from the scope of protection of this application. Therefore, although this application has been described in detail through the above embodiments, this application is not limited to the above embodiments. Many other equivalent embodiments may be included without departing from the concept of this application, and the scope of this application is determined by the scope of the appended claims.
[0106] The above disclosures are only a few specific implementation scenarios of this application. However, this application is not limited to these. Any variations that can be conceived by those skilled in the art should fall within the protection scope of this application.
Claims
1. A clothing simulation method based on simulated clothing and data-driven methods, characterized in that, include: Obtain the simulated clothing and the clothing model mesh of the simulated clothing; Preprocessing is performed based on the clothing technology involved in the simulated clothing, and animation data of the preprocessed simulated clothing in different poses is generated. The animation data is smoothed, and the smoothed animation data is used to generate clothing skeletons; The current human posture data is input into a pre-trained prediction network model to predict the transformation information of each bone on the clothing skeleton. The clothing skeleton is adjusted based on the transformation information, and then the clothing model mesh is driven by the adjusted clothing skeleton.
2. The clothing simulation method according to claim 1, characterized in that, The garment manufacturing process includes sewing thread processing; the pretreatment of the sewing thread processing specifically includes: Obtain the individual panels corresponding to the simulated clothing; Obtain the sewing thread on each plate and the vertex through which the sewing thread passes; Merge two vertices connected by the same sewing thread on the simulated garment into one vertex, and calculate the coordinates of the merged vertex.
3. The clothing simulation method according to claim 1, characterized in that, The garment manufacturing process includes accessory manufacturing; the pre-processing of the accessory manufacturing process specifically includes: Obtain all accessories from the simulated clothing; For each attachment, select one or more vertices as positioning vertices, and determine the vertices on the simulated clothing that have corresponding mapping relationships with each positioning vertices as reference vertices, and bind the positioning vertices and the corresponding reference vertices.
4. The clothing simulation method according to claim 1, characterized in that, During smoothing, the position of the vertex is moved along the line containing the normal vector of the vertex to maintain the shape of the simulated clothing.
5. The clothing simulation method according to claim 1, characterized in that, The human posture data includes the position of the human skeleton and the human mesh corresponding to the position of the human skeleton; The transformation information includes the position of the garment skeleton and the rotation angle of the garment skeleton.
6. The clothing simulation method according to claim 2, characterized in that, The position information of each reference vertex is determined based on the transformation information, and the positioning vertex corresponding to the reference vertex is determined based on the corresponding mapping relationship, so as to restore the merged vertices.
7. The clothing simulation method according to claim 1, characterized in that, The training of the prediction network model specifically includes: Acquire animation data containing the positions of the human skeleton and the clothing skeleton, and perform data preprocessing on the animation data to obtain training data; The training data is divided into multiple consecutive time series in units of frames and imported into a preset network model for training. The model data is updated according to the smoothness loss function and the output results are encouraged to transition smoothly between consecutive frames until a preset stopping condition is met, and the trained prediction network model is obtained.
8. A clothing simulation system based on simulated clothing and data-driven operation, characterized in that, include: The simulated clothing unit is used to acquire the simulated clothing and the clothing model mesh of the simulated clothing; An animation data unit is used to preprocess the clothing technology involved in the simulated clothing and generate animation data of the simulated clothing in different poses after preprocessing. The clothing skeleton unit is used to smooth the animation data and generate the clothing skeleton using the smoothed animation data. The prediction unit is used to input the current human posture data into a pre-trained prediction network model to predict the transformation information of each bone on the clothing skeleton. The output unit is used to adjust the clothing skeleton based on the transformation information, and then drive the clothing model mesh through the adjusted clothing skeleton.
9. A computer program product, characterized in that, It includes executable instructions that, when executed, implement the clothing simulation method based on simulated clothing and data-driven methods as described in any one of claims 1-7.
10. A computer-readable storage medium, characterized in that, The device contains a computer program that, when executed by a processor, implements the clothing simulation method based on simulated clothing and data-driven methods as described in any one of claims 1-7.
