Enya Shen
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Enya Shen has a total of 37 co-authors. Here are the top co-authors by number of shared publications:
- Jianmin Wang
Tsinghua University
9 shared publications - Anchang Bao
School of Software, BNRist, Tsinghua University, Beijing, China
3 shared publications Jianmin Wang
3 shared publications- Leixian Shen
The Hong Kong University of Science and Technology, Hong Kong Sar, China
3 shared publications Zhiwei Tai
3 shared publicationsXiaolong Yin
2 shared publications
Latest Publications
- Open Access
- 9 November 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Graphics
- Vol. 45 (6) , 1-21
Point Containment Queries on Trimmed Surfaces via Winding Numbers and Periodicity Lifting
Point containment queries on trimmed surfaces are fundamental to CAD modeling, solid geometry processing, and surface tessellation. Existing approaches such as ray casting suffer from robustness issues, while state-of-the-art winding number methods remain computationally expensive. We propose a fast and robust method for point containment queries on trimmed surfaces, including those with periodic parameterizations. Our approach introduces a recursive winding number computation scheme that replaces the control-points bound with an ellipse-based bound for Bézier segments, avoiding costly curve subdivision. For periodic surfaces, we lift trimming curves to the universal covering space, allowing accurate winding number computation even for non-contractible loops and for loops whose p-curves appear open in the parameter domain due to periodic identification. Experiments on large-scale 2D datasets and real-world CAD models show that our method achieves substantial speedups over existing winding number algorithms. For valid periodic trimming loops, the method is theoretically well-defined; its resilience to mild geometric noise and open-boundary defects is evaluated empirically. We further demonstrate its effectiveness in B-Rep model intersection and in robust tessellation of trimmed surfaces.
- 9 November 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Graphics
- Vol. 45 (6) , 1-21
Open AccessPoint Containment Queries on Trimmed Surfaces via Winding Numbers and Periodicity Lifting
Point containment queries on trimmed surfaces are fundamental to CAD modeling, solid geometry processing, and surface tessellation. Existing approaches such as ray casting suffer from robustness issues, while state-of-the-art winding number methods remain computationally expensive. We propose a fast and robust method for point containment queries on trimmed surfaces, including those with periodic parameterizations. Our approach introduces a recursive winding number computation scheme that replaces the control-points bound with an ellipse-based bound for Bézier segments, avoiding costly curve subdivision. For periodic surfaces, we lift trimming curves to the universal covering space, allowing accurate winding number computation even for non-contractible loops and for loops whose p-curves appear open in the parameter domain due to periodic identification. Experiments on large-scale 2D datasets and real-world CAD models show that our method achieves substantial speedups over existing winding number algorithms. For valid periodic trimming loops, the method is theoretically well-defined; its resilience to mild geometric noise and open-boundary defects is evaluated empirically. We further demonstrate its effectiveness in B-Rep model intersection and in robust tessellation of trimmed surfaces.
- Open Access
- 19 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-11
Img2CADSeq: Image-to-CAD Generation via Sequence-Based Diffusion
- 19 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-11
Open AccessImg2CADSeq: Image-to-CAD Generation via Sequence-Based Diffusion
- Open Access
- 3 July 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Graphics
- Vol. 45 (4) , 1-16
Monte Carlo PDE Solvers for Nonlinear Radiative Boundary Conditions
Monte Carlo PDE solvers have become increasingly popular for solving heat-related partial differential equations in geometry processing and computer graphics due to their robustness in handling complex geometries. While existing methods can handle Dirichlet, Neumann, and linear Robin boundary conditions, nonlinear boundary conditions arising from thermal radiation remain largely unexplored. In this paper, we introduce a Picard-style fixed-point iteration framework that enables Monte Carlo PDE solvers to handle nonlinear radiative boundary conditions. While strict theoretical convergence is not generally guaranteed, our method remains stable and empirically convergent with a properly chosen relaxation coefficient. Even with imprecise initial boundary estimates, it progressively approaches the correct solution. Compared to standard linearization strategies, the proposed approach achieves significantly higher accuracy. To further address the high variance inherent in Monte Carlo estimators, we propose a heteroscedastic regression-based denoising technique specifically designed for on-boundary solution estimates, filling a gap left by prior variance reduction methods that focus solely on interior points. We validate our approach through extensive evaluations on synthetic benchmarks and demonstrate its effectiveness on practical heat radiation simulations with complex geometries.
- 3 July 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Graphics
- Vol. 45 (4) , 1-16
Open AccessMonte Carlo PDE Solvers for Nonlinear Radiative Boundary Conditions
Monte Carlo PDE solvers have become increasingly popular for solving heat-related partial differential equations in geometry processing and computer graphics due to their robustness in handling complex geometries. While existing methods can handle Dirichlet, Neumann, and linear Robin boundary conditions, nonlinear boundary conditions arising from thermal radiation remain largely unexplored. In this paper, we introduce a Picard-style fixed-point iteration framework that enables Monte Carlo PDE solvers to handle nonlinear radiative boundary conditions. While strict theoretical convergence is not generally guaranteed, our method remains stable and empirically convergent with a properly chosen relaxation coefficient. Even with imprecise initial boundary estimates, it progressively approaches the correct solution. Compared to standard linearization strategies, the proposed approach achieves significantly higher accuracy. To further address the high variance inherent in Monte Carlo estimators, we propose a heteroscedastic regression-based denoising technique specifically designed for on-boundary solution estimates, filling a gap left by prior variance reduction methods that focus solely on interior points. We validate our approach through extensive evaluations on synthetic benchmarks and demonstrate its effectiveness on practical heat radiation simulations with complex geometries.
- 1 July 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Visualization and Computer Graphics
- Vol. 32 (7) , 7341-7360
Surface Offsetting: A Survey From Geometric Construction to Neural Implicit Representations
Surface offsetting is a fundamental geometric operation in computer-aided design, manufacturing, robotics, and computational physics. Despite its conceptual simplicity, generating offset surfaces robustly and efficiently for complex and irregular geometries remains a persistent challenge, hindered by self-intersections, topological inconsistencies, and feature degradation. This paper reviews recent advances in the field of surface offsetting, offering a structured overview of the evolution of techniques from classical constructive methods to contemporary neural implicit representations. Our survey addresses a gap in the literature, as prior foundational reviews over the past two decades in this field focused primarily on parametric methods. We introduce a taxonomy that organizes existing algorithms into five principal classes: Constructive, Spatial Discretization, Optimization-based, Field-based, and Learning-based approaches. An analysis of 46 representative algorithms reveals trade-offs in algorithm design: achieving both geometric accuracy and topological correctness proves difficult, and representation choices introduce inherent complexity constraints. We identify open problems including offset generation for surfaces with open boundaries, preservation of thin features, and resolution of self-intersections in concave regions. These challenges point toward promising research directions, including extending classical offset theory to non-manifold and open-boundary domains, developing scalable geometric predicates, and designing hybrid neural representations that disentangle distance-field smoothness from geometric sharpness.
- 1 July 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Visualization and Computer Graphics
- Vol. 32 (7) , 7341-7360
Surface Offsetting: A Survey From Geometric Construction to Neural Implicit Representations
Surface offsetting is a fundamental geometric operation in computer-aided design, manufacturing, robotics, and computational physics. Despite its conceptual simplicity, generating offset surfaces robustly and efficiently for complex and irregular geometries remains a persistent challenge, hindered by self-intersections, topological inconsistencies, and feature degradation. This paper reviews recent advances in the field of surface offsetting, offering a structured overview of the evolution of techniques from classical constructive methods to contemporary neural implicit representations. Our survey addresses a gap in the literature, as prior foundational reviews over the past two decades in this field focused primarily on parametric methods. We introduce a taxonomy that organizes existing algorithms into five principal classes: Constructive, Spatial Discretization, Optimization-based, Field-based, and Learning-based approaches. An analysis of 46 representative algorithms reveals trade-offs in algorithm design: achieving both geometric accuracy and topological correctness proves difficult, and representation choices introduce inherent complexity constraints. We identify open problems including offset generation for surfaces with open boundaries, preservation of thin features, and resolution of self-intersections in concave regions. These challenges point toward promising research directions, including extending classical offset theory to non-manifold and open-boundary domains, developing scalable geometric predicates, and designing hybrid neural representations that disentangle distance-field smoothness from geometric sharpness.
- Open Access
- 12 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 37-41
English or Chinese? Investigating the Impact of Prompt Language on Large Language Models for Code Summarization
- 12 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 37-41
Open AccessEnglish or Chinese? Investigating the Impact of Prompt Language on Large Language Models for Code Summarization
- Open Access
- 14 December 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-11
Off-Centered WoS-Type Solvers with Statistical Weighting
- 14 December 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-11
Open AccessOff-Centered WoS-Type Solvers with Statistical Weighting
- 1 November 2025
- Published by China Science Publishing & Media Ltd. in Journal of Computer-Aided Design & Computer Graphics
- Vol. 37 (11) , 1993-2005
Efficient Boolean Operations for B-Rep Models
- 1 November 2025
- Published by China Science Publishing & Media Ltd. in Journal of Computer-Aided Design & Computer Graphics
- Vol. 37 (11) , 1993-2005
Efficient Boolean Operations for B-Rep Models
- Open Access
Fast and Robust Point Containment Queries on Trimmed Surface
Point containment queries on trimmed surfaces are fundamental to CAD modeling, solid geometry processing, and surface tessellation. Existing approaches such as ray casting and generalized winding numbers often face limitations in robustness and computational efficiency. We propose a fast and numerically stable method for performing containment queries on trimmed surfaces, including those with periodic parameterizations. Our approach introduces a recursive winding number computation scheme that replaces costly curve subdivision with an ellipse-based bound for Bezier segments, enabling linear-time evaluation. For periodic surfaces, we lift trimming curves to the universal covering space, allowing accurate and consistent winding number computation even for non-contractible or discontinuous loops in parameter domain. Experiments show that our method achieves substantial speedups over existing winding-number algorithms while maintaining high robustness in the presence of geometric noise, open boundaries, and periodic topologies. We further demonstrate its effectiveness in processing real B-Rep models and in robust tessellation of trimmed surfaces.
Open AccessFast and Robust Point Containment Queries on Trimmed Surface
Point containment queries on trimmed surfaces are fundamental to CAD modeling, solid geometry processing, and surface tessellation. Existing approaches such as ray casting and generalized winding numbers often face limitations in robustness and computational efficiency. We propose a fast and numerically stable method for performing containment queries on trimmed surfaces, including those with periodic parameterizations. Our approach introduces a recursive winding number computation scheme that replaces costly curve subdivision with an ellipse-based bound for Bezier segments, enabling linear-time evaluation. For periodic surfaces, we lift trimming curves to the universal covering space, allowing accurate and consistent winding number computation even for non-contractible or discontinuous loops in parameter domain. Experiments show that our method achieves substantial speedups over existing winding-number algorithms while maintaining high robustness in the presence of geometric noise, open boundaries, and periodic topologies. We further demonstrate its effectiveness in processing real B-Rep models and in robust tessellation of trimmed surfaces.
- Open Access
- 4 August 2025
- Published by China Science Publishing & Media Ltd. in Journal of Computer-Aided Design & Computer Graphics
- Vol. 37 (10) , 1666-1680
A Review of Conversion Techniques for Typical 3D Model Representation Methods
This paper systematically reviews common 3D model representation methods, including point clouds, voxels, meshes, boundary representations (B-Rep), and implicit representations such as signed distance fields (SDF) and neural radiance fields (NeRF), and introduces their differences in geometric accuracy, computational efficiency, and application scenarios. Furthermore, it explores conversion techniques between point clouds and meshes, point clouds and B-Rep, voxels and meshes, meshes and B-Rep, as well as meshes and implicit representations, covering both traditional geometric algorithms and deep learning-based intelligent approaches. The analysis evaluates the strengths and weaknesses of existing methods in terms of computational precision, topological consistency, and efficiency, highlighting key challenges such as low processing efficiency for high-resolution models and difficulties in preserving complex topological structures. The analysis indicates that future research should focus on developing multimodal fusion representation methods, optimizing high-precision conversion algorithms, developing dedicated conversion technologies for specific industries, and actively exploring AI-driven intelligent conversion paradigms to provide key technical support for the innovative development of fields such as intelligent manufacturing and digital twins.
- 4 August 2025
- Published by China Science Publishing & Media Ltd. in Journal of Computer-Aided Design & Computer Graphics
- Vol. 37 (10) , 1666-1680
Open AccessA Review of Conversion Techniques for Typical 3D Model Representation Methods
This paper systematically reviews common 3D model representation methods, including point clouds, voxels, meshes, boundary representations (B-Rep), and implicit representations such as signed distance fields (SDF) and neural radiance fields (NeRF), and introduces their differences in geometric accuracy, computational efficiency, and application scenarios. Furthermore, it explores conversion techniques between point clouds and meshes, point clouds and B-Rep, voxels and meshes, meshes and B-Rep, as well as meshes and implicit representations, covering both traditional geometric algorithms and deep learning-based intelligent approaches. The analysis evaluates the strengths and weaknesses of existing methods in terms of computational precision, topological consistency, and efficiency, highlighting key challenges such as low processing efficiency for high-resolution models and difficulties in preserving complex topological structures. The analysis indicates that future research should focus on developing multimodal fusion representation methods, optimizing high-precision conversion algorithms, developing dedicated conversion technologies for specific industries, and actively exploring AI-driven intelligent conversion paradigms to provide key technical support for the innovative development of fields such as intelligent manufacturing and digital twins.
- Open Access
- 1 December 2023
- Published by China Science Publishing & Media Ltd. in Journal of Computer-Aided Design & Computer Graphics
- Vol. 35 (12) , 1851-1862
A Method for Transforming B-Rep of 3D Ship Models into CSG
Model data representation is the foundation of the CAD geometry engine. The Boundary Representation (B-Rep) and Constructive Solid Geometry (CSG) representations are two commonly used model data representation methods. With the development of hybrid modeling, B-Rep and CSG hybrid modeling has been applied to large-scale models such as ships and the conversion between B-Rep and CSG is a key function. For the B-Rep to CSG conversion of primitives, a process was designed to parse the geometry structures in a B-Rep entity and combine them into a CSG primitive based on the topological structures. For conversion of complex entities, faces in a B-Rep entity are divided into different regions based on its convexity, and the boundary of each region is a cutting loop. Then a cutting loop recognition algorithm is proposed. Based on the cutting loop recognition results, a binary decomposition tree is designed for entity decomposition and CSG tree generation process. Fifteen types of basic entities and four complex models are used to make several B-Rep entities, and the conversion errors and time costs are measured. Experimental results show that the proposed method can achieve B-Rep to CSG conversion within an error range of 0.05% of the normalized model size, and has supported the model light-weighting function of the component sample modeling platform of the Ship Product Design System (SPD).
- 1 December 2023
- Published by China Science Publishing & Media Ltd. in Journal of Computer-Aided Design & Computer Graphics
- Vol. 35 (12) , 1851-1862
Open AccessA Method for Transforming B-Rep of 3D Ship Models into CSG
Model data representation is the foundation of the CAD geometry engine. The Boundary Representation (B-Rep) and Constructive Solid Geometry (CSG) representations are two commonly used model data representation methods. With the development of hybrid modeling, B-Rep and CSG hybrid modeling has been applied to large-scale models such as ships and the conversion between B-Rep and CSG is a key function. For the B-Rep to CSG conversion of primitives, a process was designed to parse the geometry structures in a B-Rep entity and combine them into a CSG primitive based on the topological structures. For conversion of complex entities, faces in a B-Rep entity are divided into different regions based on its convexity, and the boundary of each region is a cutting loop. Then a cutting loop recognition algorithm is proposed. Based on the cutting loop recognition results, a binary decomposition tree is designed for entity decomposition and CSG tree generation process. Fifteen types of basic entities and four complex models are used to make several B-Rep entities, and the conversion errors and time costs are measured. Experimental results show that the proposed method can achieve B-Rep to CSG conversion within an error range of 0.05% of the normalized model size, and has supported the model light-weighting function of the component sample modeling platform of the Ship Product Design System (SPD).