Computer-Aided Design最新文献_第4页

Topo-GenMeta: Generative design of metamaterials based on diffusion model with attention to topology Topo-GenMeta：基于关注拓扑的扩散模型的超材料生成设计

IF 3.1 3区计算机科学 Q2 COMPUTER SCIENCE, SOFTWARE ENGINEERING

Computer-Aided Design

Pub Date : 2025-10-10 DOI: 10.1016/j.cad.2025.103977

Liang Du , Jiangbei Hu , Shengfa Wang , Yu Jiang , Na Lei , Ying He , Zhongxuan Luo

Metamaterials are a family of artificial materials that achieve unique properties by designing the shape of unit cell structures. Expanding the metamaterial unit cell library is a key focus in this field, with the aim of enhancing the design flexibility to meet multifunctional requirements across diverse physical scenarios. Recent advancements in data-driven generative techniques using deep learning have significantly sped up innovations in metamaterial design. However, existing approaches mostly focus on the geometric characteristics of unit structures without considering their topological properties explicitly, which we believe are essential for enhancing design diversity and enriching material properties. In this study, we propose a novel data-driven metamaterial design methodology that combines the denoising diffusion probabilistic model with the persistent homology technique. Our model is capable of generating high-fidelity and functionally effective unit structures. Furthermore, by incorporating topological properties derived from persistent homology into the diffusion process, our method facilitates the generation of a diversity of metamaterial unit structures with richer shapes and properties. To the best of our knowledge, this is the first approach to explicitly consider topological properties in metamaterial design. In addition, our method also supports multi-scale design applications, enabling the generation of metamaterial units that align with the desired properties to achieve the optimization objectives.

超材料是一类人工材料，通过设计单细胞结构的形状来实现独特的性能。扩展超材料单元胞库是该领域的一个关键焦点，其目的是提高设计灵活性，以满足不同物理场景的多功能需求。使用深度学习的数据驱动生成技术的最新进展大大加快了超材料设计的创新。然而，现有的方法大多侧重于单元结构的几何特征，而没有明确考虑其拓扑特性，我们认为这对于增强设计多样性和丰富材料特性至关重要。在这项研究中，我们提出了一种新的数据驱动的超材料设计方法，该方法将去噪扩散概率模型与持续同源技术相结合。我们的模型能够生成高保真度和功能有效的单元结构。此外，通过将源自持续同源的拓扑性质结合到扩散过程中，我们的方法有助于生成具有更丰富形状和性质的多种超材料单元结构。据我们所知，这是第一个在超材料设计中明确考虑拓扑特性的方法。此外，我们的方法还支持多尺度设计应用，能够生成与所需属性一致的超材料单元，以实现优化目标。

{"title":"Topo-GenMeta: Generative design of metamaterials based on diffusion model with attention to topology","authors":"Liang Du , Jiangbei Hu , Shengfa Wang , Yu Jiang , Na Lei , Ying He , Zhongxuan Luo","doi":"10.1016/j.cad.2025.103977","DOIUrl":"10.1016/j.cad.2025.103977","url":null,"abstract":"<div><div>Metamaterials are a family of artificial materials that achieve unique properties by designing the shape of unit cell structures. Expanding the metamaterial unit cell library is a key focus in this field, with the aim of enhancing the design flexibility to meet multifunctional requirements across diverse physical scenarios. Recent advancements in data-driven generative techniques using deep learning have significantly sped up innovations in metamaterial design. However, existing approaches mostly focus on the geometric characteristics of unit structures without considering their topological properties explicitly, which we believe are essential for enhancing design diversity and enriching material properties. In this study, we propose a novel data-driven metamaterial design methodology that combines the denoising diffusion probabilistic model with the persistent homology technique. Our model is capable of generating high-fidelity and functionally effective unit structures. Furthermore, by incorporating topological properties derived from persistent homology into the diffusion process, our method facilitates the generation of a diversity of metamaterial unit structures with richer shapes and properties. To the best of our knowledge, this is the first approach to explicitly consider topological properties in metamaterial design. In addition, our method also supports multi-scale design applications, enabling the generation of metamaterial units that align with the desired properties to achieve the optimization objectives.</div></div>","PeriodicalId":50632,"journal":{"name":"Computer-Aided Design","volume":"190 ","pages":"Article 103977"},"PeriodicalIF":3.1,"publicationDate":"2025-10-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145269415","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Mechanics simulation with Implicit Neural Representations of complex geometries 复杂几何的隐式神经表征力学模拟

IF 3.1 3区计算机科学 Q2 COMPUTER SCIENCE, SOFTWARE ENGINEERING

Computer-Aided Design

Pub Date : 2025-10-06 DOI: 10.1016/j.cad.2025.103978

Samundra Karki, Ming-Chen Hsu, Adarsh Krishnamurthy, Baskar Ganapathysubramanian

Implicit Neural Representations (INRs), characterized by neural network-encoded signed distance fields, provide a powerful means to represent complex geometries continuously and efficiently. While successful in computer vision and generative modeling, integrating INRs into computational analysis workflows, such as finite element simulations, remains underdeveloped, primarily due to the necessity of explicit geometry representations (meshes). Conventional mesh-based finite element methods (FEM) introduce computational overhead, discretization errors, and manual effort, particularly for intricate or dynamically evolving geometries. Although immersed boundary methods partially address these issues, they are susceptible to numerical artifacts from explicit boundary treatments. In this work, we propose an innovative computational framework that seamlessly combines INRs with the Shifted Boundary Method (SBM) for performing high-fidelity linear elasticity simulations without explicit geometry transformations. By directly querying the neural implicit geometry, we obtain the surrogate boundaries and distance vectors essential for SBM, effectively eliminating the intermediate meshing step. We demonstrate the efficacy and robustness of our approach through elasticity simulations on complex geometries sourced from diverse representations, including triangle soup and point cloud reconstructions (Stanford Bunny, Eiffel Tower, gyroids). Our method showcases significant computational advantages and accuracy, underscoring its potential in biomedical, geophysical, and advanced manufacturing applications, thus offering a versatile tool for geometric and physical modeling aligned with contemporary design and analysis workflows.

隐式神经表示（INRs）以神经网络编码的符号距离域为特征，为复杂几何图形的连续高效表示提供了强有力的手段。虽然在计算机视觉和生成建模方面取得了成功，但将inr集成到计算分析工作流程（如有限元模拟）中仍然不发达，主要是由于需要显式几何表示（网格）。传统的基于网格的有限元方法（FEM）引入了计算开销、离散误差和人工工作量，特别是对于复杂或动态发展的几何形状。虽然浸入式边界方法部分解决了这些问题，但它们容易受到显式边界处理的数值伪影的影响。在这项工作中，我们提出了一种创新的计算框架，将INRs与移位边界法（SBM）无缝结合，在没有显式几何变换的情况下执行高保真线性弹性模拟。通过直接查询神经隐式几何，我们获得了SBM所需的代理边界和距离向量，有效地消除了中间网格划分步骤。我们通过对来自不同表示的复杂几何图形进行弹性模拟，包括三角汤和点云重建（斯坦福兔、埃菲尔铁塔、陀螺仪），证明了我们方法的有效性和鲁棒性。我们的方法展示了显著的计算优势和准确性，强调了其在生物医学，地球物理和先进制造应用中的潜力，从而提供了与当代设计和分析工作流程相一致的几何和物理建模的多功能工具。

{"title":"Mechanics simulation with Implicit Neural Representations of complex geometries","authors":"Samundra Karki, Ming-Chen Hsu, Adarsh Krishnamurthy, Baskar Ganapathysubramanian","doi":"10.1016/j.cad.2025.103978","DOIUrl":"10.1016/j.cad.2025.103978","url":null,"abstract":"<div><div>Implicit Neural Representations (INRs), characterized by neural network-encoded signed distance fields, provide a powerful means to represent complex geometries continuously and efficiently. While successful in computer vision and generative modeling, integrating INRs into computational analysis workflows, such as finite element simulations, remains underdeveloped, primarily due to the necessity of explicit geometry representations (meshes). Conventional mesh-based finite element methods (FEM) introduce computational overhead, discretization errors, and manual effort, particularly for intricate or dynamically evolving geometries. Although immersed boundary methods partially address these issues, they are susceptible to numerical artifacts from explicit boundary treatments. In this work, we propose an innovative computational framework that seamlessly combines INRs with the Shifted Boundary Method (SBM) for performing high-fidelity linear elasticity simulations without explicit geometry transformations. By directly querying the neural implicit geometry, we obtain the surrogate boundaries and distance vectors essential for SBM, effectively eliminating the intermediate meshing step. We demonstrate the efficacy and robustness of our approach through elasticity simulations on complex geometries sourced from diverse representations, including triangle soup and point cloud reconstructions (Stanford Bunny, Eiffel Tower, gyroids). Our method showcases significant computational advantages and accuracy, underscoring its potential in biomedical, geophysical, and advanced manufacturing applications, thus offering a versatile tool for geometric and physical modeling aligned with contemporary design and analysis workflows.</div></div>","PeriodicalId":50632,"journal":{"name":"Computer-Aided Design","volume":"190 ","pages":"Article 103978"},"PeriodicalIF":3.1,"publicationDate":"2025-10-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145321805","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Text-driven 3D human motion generation for pose estimation using dual-transformer architecture 文本驱动的三维人体运动生成的姿态估计使用双变压器架构

IF 3.1 3区计算机科学 Q2 COMPUTER SCIENCE, SOFTWARE ENGINEERING

Computer-Aided Design

Pub Date : 2025-10-03 DOI: 10.1016/j.cad.2025.103991

Rizwan Abbas , Hua Gao , Xi Li

Text-to-motion generation has made significant progress in recent years. However, existing approaches struggle to generate high-quality 3D human motions that effectively capture pose estimation. These limitations are due to weak pose estimation and limited skeletal modeling. To address these limitations, we propose DT3DPE (Dual-Transformer for 3D Pose Estimation), a framework that integrates pose estimation to generate text-aligned, realistic 3D human motions. The proposed approach introduces residual vector quantization with additional layers for encoding pose tokens, enabling the capture of fine-grained details in body dynamics. Furthermore, DT3DPE employs a dual-transformer architecture, consisting of a masked transformer for motion token prediction and a residual transformer for refining motion details. This dual-transformer architecture allows the model to generate high-fidelity 3D human poses with precise body joint positioning and smooth temporal transitions. The experimental results on HumanML3D and KIT-ML datasets demonstrate that DT3DPE outperforms existing state-of-the-art methods in text-driven 3D human motion generation. Our code is available at https://github.com/swerizwan/DT3DPE.

近年来，文本到动作生成技术取得了重大进展。然而，现有的方法很难产生高质量的3D人体运动，有效地捕捉姿势估计。这些限制是由于弱姿态估计和有限的骨骼建模。为了解决这些限制，我们提出了DT3DPE（用于3D姿态估计的双变压器），这是一个集成姿态估计以生成文本对齐的逼真的3D人体运动的框架。该方法引入残差矢量量化和附加层来编码姿态标记，从而能够捕获身体动力学中的细粒度细节。此外，DT3DPE采用双变压器架构，包括用于运动令牌预测的屏蔽变压器和用于细化运动细节的残余变压器。这种双变压器结构使模型能够生成高保真的3D人体姿势，具有精确的身体关节定位和平滑的时间过渡。在HumanML3D和KIT-ML数据集上的实验结果表明，DT3DPE在文本驱动的3D人体运动生成方面优于现有的最先进方法。我们的代码可在https://github.com/swerizwan/DT3DPE上获得。

{"title":"Text-driven 3D human motion generation for pose estimation using dual-transformer architecture","authors":"Rizwan Abbas , Hua Gao , Xi Li","doi":"10.1016/j.cad.2025.103991","DOIUrl":"10.1016/j.cad.2025.103991","url":null,"abstract":"<div><div>Text-to-motion generation has made significant progress in recent years. However, existing approaches struggle to generate high-quality 3D human motions that effectively capture pose estimation. These limitations are due to weak pose estimation and limited skeletal modeling. To address these limitations, we propose DT3DPE (Dual-Transformer for 3D Pose Estimation), a framework that integrates pose estimation to generate text-aligned, realistic 3D human motions. The proposed approach introduces residual vector quantization with additional layers for encoding pose tokens, enabling the capture of fine-grained details in body dynamics. Furthermore, DT3DPE employs a dual-transformer architecture, consisting of a masked transformer for motion token prediction and a residual transformer for refining motion details. This dual-transformer architecture allows the model to generate high-fidelity 3D human poses with precise body joint positioning and smooth temporal transitions. The experimental results on HumanML3D and KIT-ML datasets demonstrate that DT3DPE outperforms existing state-of-the-art methods in text-driven 3D human motion generation. Our code is available at <span><span>https://github.com/swerizwan/DT3DPE</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50632,"journal":{"name":"Computer-Aided Design","volume":"190 ","pages":"Article 103991"},"PeriodicalIF":3.1,"publicationDate":"2025-10-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145269413","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

LoGAvatar: Local Gaussian Splatting for human avatar modeling from monocular video LoGAvatar：局部高斯喷溅，用于单目视频的人类头像建模

IF 3.1 3区计算机科学 Q2 COMPUTER SCIENCE, SOFTWARE ENGINEERING

Computer-Aided Design

Pub Date : 2025-09-29 DOI: 10.1016/j.cad.2025.103973

Jinsong Zhang , Xiongzheng Li , Hailong Jia , Jin Li , Zhuo Su , Guidong Wang , Kun Li

Avatar reconstruction from monocular videos plays a pivotal role in various virtual and augmented reality applications. Recent methods have utilized 3D Gaussian Splatting (GS) to model human avatars, achieving fast rendering speeds with high visual quality. However, due to the independent nature of GS primitives, existing approaches often struggle to capture high-fidelity details and lack the ability to edit the reconstructed avatars effectively. To address these limitations, we propose Local Gaussian Splatting Avatar (LoGAvatar), a novel framework designed to enhance both geometry and texture modeling of human avatars. Specifically, we introduce a hierarchical Gaussian splatting framework, where local GS primitives are predicted based on sampled points from a human template model, such as SMPL. For texture modeling, we design a convolution-based texture atlas that preserves spatial continuity and enriches fine details. By aggregating local information for both geometry and texture, our approach reconstructs high-fidelity avatars while maintaining real-time rendering efficiency. Experimental results on two public datasets demonstrate the superior performance of our method in terms of avatar fidelity and rendering quality. Moreover, based on our LoGAvatar, we can edit the shape and texture of the reconstructed avatar, which inspires more customized avatar applications. The code is available at http://cic.tju.edu.cn/faculty/likun/projects/LoGAvatar.

单目视频的虚拟化身重建在各种虚拟现实和增强现实应用中起着至关重要的作用。最近的方法利用三维高斯飞溅（GS）来建模人类化身，实现快速渲染速度和高视觉质量。然而，由于GS原语的独立性，现有的方法往往难以捕获高保真的细节，并且缺乏有效编辑重建头像的能力。为了解决这些限制，我们提出了局部高斯飞溅头像（LoGAvatar），这是一个新的框架，旨在增强人类头像的几何和纹理建模。具体来说，我们引入了一个分层高斯飞溅框架，其中局部GS原语是基于人类模板模型（如SMPL）的采样点来预测的。在纹理建模方面，我们设计了一个基于卷积的纹理图谱，既保留了空间连续性，又丰富了精细细节。通过聚合几何和纹理的局部信息，我们的方法在保持实时渲染效率的同时重建了高保真的化身。在两个公开数据集上的实验结果表明，我们的方法在头像保真度和渲染质量方面表现优异。此外，基于我们的LoGAvatar，我们可以编辑重构头像的形状和纹理，从而激发更多的定制头像应用。代码可在http://cic.tju.edu.cn/faculty/likun/projects/LoGAvatar上获得。

{"title":"LoGAvatar: Local Gaussian Splatting for human avatar modeling from monocular video","authors":"Jinsong Zhang , Xiongzheng Li , Hailong Jia , Jin Li , Zhuo Su , Guidong Wang , Kun Li","doi":"10.1016/j.cad.2025.103973","DOIUrl":"10.1016/j.cad.2025.103973","url":null,"abstract":"<div><div>Avatar reconstruction from monocular videos plays a pivotal role in various virtual and augmented reality applications. Recent methods have utilized 3D Gaussian Splatting (GS) to model human avatars, achieving fast rendering speeds with high visual quality. However, due to the independent nature of GS primitives, existing approaches often struggle to capture high-fidelity details and lack the ability to edit the reconstructed avatars effectively. To address these limitations, we propose Local Gaussian Splatting Avatar (LoGAvatar), a novel framework designed to enhance both geometry and texture modeling of human avatars. Specifically, we introduce a hierarchical Gaussian splatting framework, where local GS primitives are predicted based on sampled points from a human template model, such as SMPL. For texture modeling, we design a convolution-based texture atlas that preserves spatial continuity and enriches fine details. By aggregating local information for both geometry and texture, our approach reconstructs high-fidelity avatars while maintaining real-time rendering efficiency. Experimental results on two public datasets demonstrate the superior performance of our method in terms of avatar fidelity and rendering quality. Moreover, based on our LoGAvatar, we can edit the shape and texture of the reconstructed avatar, which inspires more customized avatar applications. The code is available at <span><span>http://cic.tju.edu.cn/faculty/likun/projects/LoGAvatar</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50632,"journal":{"name":"Computer-Aided Design","volume":"190 ","pages":"Article 103973"},"PeriodicalIF":3.1,"publicationDate":"2025-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145269414","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Fully discrete subdivision-based IGA scheme with decoupled structure and unconditional energy stability for the phase-field crystal model on surfaces 具有解耦结构和无条件能量稳定的基于完全离散细分的表面相场晶体模型IGA方案

IF 3.1 3区计算机科学 Q2 COMPUTER SCIENCE, SOFTWARE ENGINEERING

Computer-Aided Design

Pub Date : 2025-09-29 DOI: 10.1016/j.cad.2025.103969

Qing Pan , Yunqing Huang , Chong Chen , Xiaofeng Yang , Yongjie Jessica Zhang

In this work, we aim to numerically solve the phase-field crystal (PFC) model to simulate atomic growth on manifolds. The geometric complexity, pronounced curvature variations, and nonlinearities inherent in the physical model pose significant challenges, necessitating the development of efficient and robust numerical schemes that can handle strong coupling and nonlinear terms while accurately accounting for curved geometries. To address these challenges, we first adopt a subdivision-based isogeometric analysis (IGA) for spatial discretization. This approach effectively resolves geometric complexities by offering hierarchical refinability, geometric exactness, and adaptability to arbitrary topologies, while eliminating geometric errors commonly encountered in traditional finite element methods. For temporal discretization, the highly nonlinear terms in the model are addressed using the Invariant Energy Quadratization (IEQ) method, which linearizes the nonlinear terms and guarantees strict unconditional energy stability. However, the introduction of auxiliary variables in the IEQ method results in a linearly coupled system. To overcome this limitation and further enhance computational efficiency, we incorporate the Zero-Energy-Coupling (ZEC) approach, ultimately constructing a scheme that achieves second-order accuracy, linearity, unconditional energy stability, and a fully decoupled structure. We rigorously prove the energy stability and solvability of the proposed scheme and validate its accuracy and robustness through extensive numerical experiments conducted on manifolds, demonstrating its capability to handle intricate geometric structures and nonlinear dynamics effectively.

在这项工作中，我们的目标是数值求解相场晶体（PFC）模型来模拟流形上的原子生长。几何复杂性、明显的曲率变化和物理模型固有的非线性构成了重大挑战，需要开发有效且稳健的数值方案，以处理强耦合和非线性项，同时准确地考虑弯曲几何。为了解决这些挑战，我们首先采用基于细分的等高分析（IGA）进行空间离散化。该方法通过提供层次精细化性、几何精确性和对任意拓扑的适应性，有效地解决了几何复杂性，同时消除了传统有限元方法中常见的几何误差。对于时间离散化，采用不变能量二次化（IEQ）方法处理模型中的高度非线性项，该方法将非线性项线性化并保证严格的无条件能量稳定性。然而，在IEQ方法中引入辅助变量会导致线性耦合系统。为了克服这一限制并进一步提高计算效率，我们引入了零能量耦合（ZEC）方法，最终构建了一个实现二阶精度、线性、无条件能量稳定性和完全解耦结构的方案。通过对流形进行的大量数值实验，我们严格地证明了该方案的能量稳定性和可解性，并验证了其准确性和鲁棒性，证明了该方案能够有效地处理复杂的几何结构和非线性动力学。

{"title":"Fully discrete subdivision-based IGA scheme with decoupled structure and unconditional energy stability for the phase-field crystal model on surfaces","authors":"Qing Pan , Yunqing Huang , Chong Chen , Xiaofeng Yang , Yongjie Jessica Zhang","doi":"10.1016/j.cad.2025.103969","DOIUrl":"10.1016/j.cad.2025.103969","url":null,"abstract":"<div><div>In this work, we aim to numerically solve the phase-field crystal (PFC) model to simulate atomic growth on manifolds. The geometric complexity, pronounced curvature variations, and nonlinearities inherent in the physical model pose significant challenges, necessitating the development of efficient and robust numerical schemes that can handle strong coupling and nonlinear terms while accurately accounting for curved geometries. To address these challenges, we first adopt a subdivision-based isogeometric analysis (IGA) for spatial discretization. This approach effectively resolves geometric complexities by offering hierarchical refinability, geometric exactness, and adaptability to arbitrary topologies, while eliminating geometric errors commonly encountered in traditional finite element methods. For temporal discretization, the highly nonlinear terms in the model are addressed using the Invariant Energy Quadratization (IEQ) method, which linearizes the nonlinear terms and guarantees strict unconditional energy stability. However, the introduction of auxiliary variables in the IEQ method results in a linearly coupled system. To overcome this limitation and further enhance computational efficiency, we incorporate the Zero-Energy-Coupling (ZEC) approach, ultimately constructing a scheme that achieves second-order accuracy, linearity, unconditional energy stability, and a fully decoupled structure. We rigorously prove the energy stability and solvability of the proposed scheme and validate its accuracy and robustness through extensive numerical experiments conducted on manifolds, demonstrating its capability to handle intricate geometric structures and nonlinear dynamics effectively.</div></div>","PeriodicalId":50632,"journal":{"name":"Computer-Aided Design","volume":"190 ","pages":"Article 103969"},"PeriodicalIF":3.1,"publicationDate":"2025-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145222449","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A new measure of fairness for curves 曲线公平的新标准

IF 3.1 3区计算机科学 Q2 COMPUTER SCIENCE, SOFTWARE ENGINEERING

Computer-Aided Design

Pub Date : 2025-09-26 DOI: 10.1016/j.cad.2025.103979

Shoichi Tsuchie

This paper proposes a novel measure based on curvature variation to evaluate the fairness of curves. It is demonstrated that, in the simplest case, controlling the curvature using the proposed measure results in the log-aesthetic curve (LAC). In other words, by utilizing the proposed measure as a novel shape parameter, a unified framework can be established for aesthetic curves that accommodates a broader range of curvature variations, encompassing the LAC as a special case. Several examples are presented to illustrate curve evaluation using the proposed measure, along with its application to the approximation of aesthetic curves. The findings of this study offer a new perspective for understanding and evaluating the geometric properties of curves, with potential applications in curve design, analysis, and fairing.

提出了一种基于曲率变化的曲线公平性评价方法。结果表明，在最简单的情况下，使用所提出的测量方法控制曲率可以得到对数美观曲线（LAC）。换句话说，通过利用所提出的测量作为一个新的形状参数，可以为美学曲线建立一个统一的框架，以适应更广泛的曲率变化，包括LAC作为一个特殊情况。给出了几个例子来说明使用所提出的测量曲线评估，以及它在美学曲线近似中的应用。本研究结果为理解和评价曲线的几何特性提供了一个新的视角，在曲线设计、分析和整流方面具有潜在的应用价值。

引用次数: 0

Prototype optimization and self-training for few-shot 3D point cloud semantic segmentation 少镜头三维点云语义分割的原型优化与自训练

IF 3.1 3区计算机科学 Q2 COMPUTER SCIENCE, SOFTWARE ENGINEERING

Computer-Aided Design

Pub Date : 2025-09-26 DOI: 10.1016/j.cad.2025.103976

Jie Zhou , Yong Zhao , Fan Zhong

Few-shot point cloud segmentation aims to accurately decompose 3D point clouds into different semantic classes with few samples, and is crucial for subsequent tasks, such as analysis, modeling and editing. Despite the popularity of prototype-based approaches, prototypes often fail to adequately capture class-specific information. Therefore, for each class, a few points may exhibit significant differences from their prototype. And the lack of sufficient distinction between foreground and background prototypes presents a great challenge for precise segmentation. To address these issues, we propose a prototype optimization module to mitigate the interference among support prototypes, thereby generating prototypes of superior quality. These refined prototypes are capable of capturing the key characteristics of the data, which can prominently improve the generalization capability of our model. Then, we devise a self-training strategy that leverages pseudo query prototypes generated from high-confidence predicted labels. These prototypes are applied to query features to produce pseudo query labels and formulate a reconstruction constraint during training. By harnessing the contextual information embedded within query features, this approach significantly elevates segmentation performance. Extensive results on two popular benchmark datasets validate the superiority of our model, especially in the challenging 1-shot settings. Under the classic experimental setup, our method surpasses existing state-of-the-arts by 2.64% in 2-way 1-shot setting on the S3DIS dataset. On the ScanNet dataset, the improvements are 7.58% in 2-way 1-shot setting and 6.44% in 3-way 1-shot setting, respectively.

少镜头点云分割旨在用较少的样本将三维点云准确分解为不同的语义类，对后续的分析、建模和编辑等任务至关重要。尽管基于原型的方法很流行，但原型常常不能充分捕获特定于类的信息。因此，对于每个类，可能会有一些点与其原型有显著差异。前景和背景原型之间缺乏足够的区分，这给精确分割带来了很大的挑战。为了解决这些问题，我们提出了一个原型优化模块，以减轻支持原型之间的干扰，从而产生高质量的原型。这些精细化的原型能够捕获数据的关键特征，这可以显著提高我们模型的泛化能力。然后，我们设计了一种自我训练策略，利用高置信度预测标签生成的伪查询原型。将这些原型应用于查询特征，生成伪查询标签，并在训练过程中形成重构约束。通过利用嵌入在查询特征中的上下文信息，这种方法显著提高了分割性能。在两个流行的基准数据集上的广泛结果验证了我们模型的优越性，特别是在具有挑战性的1次射击设置中。在经典的实验设置下，我们的方法在S3DIS数据集上的2-way 1-shot设置比现有的最先进的方法高出2.64%。在ScanNet数据集上，2-way 1-shot设置的改进率为7.58%，3-way 1-shot设置的改进率为6.44%。

{"title":"Prototype optimization and self-training for few-shot 3D point cloud semantic segmentation","authors":"Jie Zhou , Yong Zhao , Fan Zhong","doi":"10.1016/j.cad.2025.103976","DOIUrl":"10.1016/j.cad.2025.103976","url":null,"abstract":"<div><div>Few-shot point cloud segmentation aims to accurately decompose 3D point clouds into different semantic classes with few samples, and is crucial for subsequent tasks, such as analysis, modeling and editing. Despite the popularity of prototype-based approaches, prototypes often fail to adequately capture class-specific information. Therefore, for each class, a few points may exhibit significant differences from their prototype. And the lack of sufficient distinction between foreground and background prototypes presents a great challenge for precise segmentation. To address these issues, we propose a prototype optimization module to mitigate the interference among support prototypes, thereby generating prototypes of superior quality. These refined prototypes are capable of capturing the key characteristics of the data, which can prominently improve the generalization capability of our model. Then, we devise a self-training strategy that leverages pseudo query prototypes generated from high-confidence predicted labels. These prototypes are applied to query features to produce pseudo query labels and formulate a reconstruction constraint during training. By harnessing the contextual information embedded within query features, this approach significantly elevates segmentation performance. Extensive results on two popular benchmark datasets validate the superiority of our model, especially in the challenging 1-shot settings. Under the classic experimental setup, our method surpasses existing state-of-the-arts by 2.64% in 2-way 1-shot setting on the S3DIS dataset. On the ScanNet dataset, the improvements are 7.58% in 2-way 1-shot setting and 6.44% in 3-way 1-shot setting, respectively.</div></div>","PeriodicalId":50632,"journal":{"name":"Computer-Aided Design","volume":"190 ","pages":"Article 103976"},"PeriodicalIF":3.1,"publicationDate":"2025-09-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145269538","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Adaptive gap closing for complex triangular mesh repair using geometric and topological characteristics 基于几何和拓扑特征的复杂三角形网格自适应闭合修复

IF 3.1 3区计算机科学 Q2 COMPUTER SCIENCE, SOFTWARE ENGINEERING

Computer-Aided Design

Pub Date : 2025-09-25 DOI: 10.1016/j.cad.2025.103981

Shuwei Shen , Shuai Zhou , Zhoufang Xiao , Jingchen Gao , Chenhao Xu

Gaps are prevalent defects in triangular meshes, often arising from various sources such as surface scanning and CAD model generation. Despite their significance, the automatic repair of complex gaps has received limited attention compared to other mesh imperfections. This study presents a novel surface-based gap-closing method for triangular mesh repair, leveraging both local geometric and topological characteristics to robustly match and merge gap boundaries. The proposed approach first employs a global–local vertex merging procedure with adaptive tolerances to eliminate duplicate vertices and simplify complex gaps. Subsequently, gaps are identified and classified into connected and disconnected types based on their topological and geometric features. For each detected gap, a non-iterative closing procedure is applied, simultaneously matching and merging all boundary vertices. An adaptive scheme is introduced to determine the geometric tolerance for vertex matching, ensuring the effective preservation of the original geometric shape. Extensive numerical experiments on a large dataset of discrete models demonstrate the effectiveness and robustness of the proposed method in closing both connected and disconnected gaps.

在三角形网格中，由于曲面扫描和CAD模型生成等原因，产生了许多缺陷。尽管具有重要意义，但与其他网格缺陷相比，复杂间隙的自动修复受到的关注有限。本文提出了一种新的基于表面的三角形网格修补方法，利用局部几何和拓扑特征对缝隙边界进行鲁棒匹配和合并。该方法首先采用具有自适应容差的全局-局部顶点合并过程，消除重复顶点，简化复杂间隙。然后，根据其拓扑和几何特征对间隙进行识别，并将其分为连通型和非连通型。对于每个检测到的间隙，应用非迭代闭合过程，同时匹配和合并所有边界顶点。引入了一种自适应方案来确定顶点匹配的几何公差，保证了原始几何形状的有效保留。在大量离散模型数据集上进行的大量数值实验证明了该方法在关闭连接和不连接间隙方面的有效性和鲁棒性。

{"title":"Adaptive gap closing for complex triangular mesh repair using geometric and topological characteristics","authors":"Shuwei Shen , Shuai Zhou , Zhoufang Xiao , Jingchen Gao , Chenhao Xu","doi":"10.1016/j.cad.2025.103981","DOIUrl":"10.1016/j.cad.2025.103981","url":null,"abstract":"<div><div>Gaps are prevalent defects in triangular meshes, often arising from various sources such as surface scanning and CAD model generation. Despite their significance, the automatic repair of complex gaps has received limited attention compared to other mesh imperfections. This study presents a novel surface-based gap-closing method for triangular mesh repair, leveraging both local geometric and topological characteristics to robustly match and merge gap boundaries. The proposed approach first employs a global–local vertex merging procedure with adaptive tolerances to eliminate duplicate vertices and simplify complex gaps. Subsequently, gaps are identified and classified into connected and disconnected types based on their topological and geometric features. For each detected gap, a non-iterative closing procedure is applied, simultaneously matching and merging all boundary vertices. An adaptive scheme is introduced to determine the geometric tolerance for vertex matching, ensuring the effective preservation of the original geometric shape. Extensive numerical experiments on a large dataset of discrete models demonstrate the effectiveness and robustness of the proposed method in closing both connected and disconnected gaps.</div></div>","PeriodicalId":50632,"journal":{"name":"Computer-Aided Design","volume":"190 ","pages":"Article 103981"},"PeriodicalIF":3.1,"publicationDate":"2025-09-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145222450","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

CORNet: A Consistency-based Outlier Rejection Network for non-rigid registration CORNet：一种基于一致性的非刚性配准离群值拒绝网络

IF 3.1 3区计算机科学 Q2 COMPUTER SCIENCE, SOFTWARE ENGINEERING

Computer-Aided Design

Pub Date : 2025-09-25 DOI: 10.1016/j.cad.2025.103980

Chang Yu, Sanguo Zhang, Li-Yong Shen

Non-rigid point cloud registration is an important problem in computer vision and graphics, aiming to find the warping function between deformed point clouds. In this paper, we propose CORNet, a consistency-based outlier rejection network for non-rigid registration. By leveraging the local geometric structure and probability distribution of point clouds, we obtain local spatial consistency and Gaussian probabilistic consistency. We then employ the Transformer mechanism, combined with consistency information, to classify inliers and outliers in correspondences, ultimately obtaining high-quality correspondences for non-rigid registration. Ablation studies validate the effectiveness of our method, and extensive experiments demonstrate that our method achieves state-of-the-art performance.

非刚性点云配准是计算机视觉和图形学中的一个重要问题，其目的是寻找变形点云之间的翘曲函数。本文提出了一种基于一致性的非刚性配准离群点拒绝网络CORNet。利用点云的局部几何结构和概率分布，获得局部空间一致性和高斯概率一致性。然后，我们使用Transformer机制，结合一致性信息，对对应中的内线和离群值进行分类，最终获得非刚性注册的高质量对应。烧蚀研究验证了我们方法的有效性，大量的实验表明我们的方法达到了最先进的性能。

引用次数: 0

Research on dynamic simulation and optimization of garment wrinkles combining computer vision and image processing 结合计算机视觉与图像处理的服装褶皱动态仿真与优化研究

IF 3.1 3区计算机科学 Q2 COMPUTER SCIENCE, SOFTWARE ENGINEERING

Computer-Aided Design

Pub Date : 2025-09-22 DOI: 10.1016/j.cad.2025.103982

Zhong Tian, Xu Xubing

This study presents a novel framework for real-time cloth wrinkle detection and optimisation, combining physics-based modelling with LSTM-Reinforcement Learning (LSTM-RL) and advanced computer vision techniques. A curated dataset of 45,876 annotated static garment images was used, featuring wrinkle attributes such as location, depth, width, and geometry. CNNs were employed for feature extraction, enhanced by Mask R-CNN to handle occlusions and RGBD data for depth-aware wrinkle modelling. A mass-spring system simulated fabric dynamics under environmental forces, while LSTM networks predicted the temporal evolution of wrinkles. Reinforcement learning dynamically adjusted fabric parameters, improving adaptability. The proposed RGBD model achieved a wrinkle detection accuracy of 96.4 %, outperforming the 2D model by over 9 %. Key metrics include an MSE of 0.0246, drift of 0.0187, and a reward value of -0.13543, with low policy and value loss confirming the RL agent’s learning stability. These results demonstrate high accuracy, real-time performance, and robustness, with strong implications for virtual fashion, robotics, and AR/VR applications.

本研究提出了一种新的实时布料皱纹检测和优化框架，将基于物理的建模与lstm -强化学习（LSTM-RL）和先进的计算机视觉技术相结合。使用了一个由45876张带注释的静态服装图像组成的精心策划的数据集，该数据集具有褶皱属性，如位置、深度、宽度和几何形状。采用cnn进行特征提取，通过Mask R-CNN进行增强处理遮挡，并使用RGBD数据进行深度感知皱纹建模。质量-弹簧系统模拟了环境力作用下的织物动力学，而LSTM网络预测了皱褶的时间演变。强化学习动态调整织物参数，提高适应性。所提出的RGBD模型的皱纹检测准确率达到96.4%，比2D模型高出9%以上。关键指标包括MSE为0.0246，漂移为0.0187，奖励值为-0.13543，低策略和价值损失证实了RL代理的学习稳定性。这些结果显示出高精度、实时性和鲁棒性，对虚拟时尚、机器人和AR/VR应用具有重要意义。

{"title":"Research on dynamic simulation and optimization of garment wrinkles combining computer vision and image processing","authors":"Zhong Tian, Xu Xubing","doi":"10.1016/j.cad.2025.103982","DOIUrl":"10.1016/j.cad.2025.103982","url":null,"abstract":"<div><div>This study presents a novel framework for real-time cloth wrinkle detection and optimisation, combining physics-based modelling with LSTM-Reinforcement Learning (LSTM-RL) and advanced computer vision techniques. A curated dataset of 45,876 annotated static garment images was used, featuring wrinkle attributes such as location, depth, width, and geometry. CNNs were employed for feature extraction, enhanced by Mask R-CNN to handle occlusions and RGBD data for depth-aware wrinkle modelling. A mass-spring system simulated fabric dynamics under environmental forces, while LSTM networks predicted the temporal evolution of wrinkles. Reinforcement learning dynamically adjusted fabric parameters, improving adaptability. The proposed RGBD model achieved a wrinkle detection accuracy of 96.4 %, outperforming the 2D model by over 9 %. Key metrics include an MSE of 0.0246, drift of 0.0187, and a reward value of -0.13543, with low policy and value loss confirming the RL agent’s learning stability. These results demonstrate high accuracy, real-time performance, and robustness, with strong implications for virtual fashion, robotics, and AR/VR applications.</div></div>","PeriodicalId":50632,"journal":{"name":"Computer-Aided Design","volume":"190 ","pages":"Article 103982"},"PeriodicalIF":3.1,"publicationDate":"2025-09-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145222448","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0