Neural signed distance functions (SDFs) are emerging as an effective representation for 3D shapes. State-of-the-art methods typically encode the SDF with a large, fixed-size neural network to approximate complex shapes implicit surfaces. Rendering these large networks is, however, computationally expensive since it requires many forward passes through every pixel, making representations impractical real-time graphics. We introduce efficient that, first time, enables rendering of...

Object deformation with linear blending dominates practical use as the fastest approach for transforming raster images, vector graphics, geometric models and animated characters. Unfortunately, schemes skeletons or cages are not always easy to because they may require manual weight painting modeling closed polyhedral envelopes around objects. Our goal is make design control of deformations simpler by allowing user work freely most convenient combination handle types. We develop weights that...

Object deformation with linear blending dominates practical use as the fastest approach for transforming raster images, vector graphics, geometric models and animated characters. Unfortunately, schemes skeletons or cages are not always easy to because they may require manual weight painting modeling closed polyhedral envelopes around objects. Our goal is make design control of deformations simpler by allowing user work freely most convenient combination handle types. We develop weights that...

Modern geometry processsing algorithms depend on an ever-growing toolbox of fundamental sub-routines and data structures. Prototyping from scratch requires much time building basic tools rather than focusing the novel research idea. Many existing code libraries have unsatisfactory APIs spent implementing is often replaced with learning complex, templated object hierarchies or memory layouts.

We propose a novel tetrahedral meshing technique that is unconditionally robust, requires no user interaction, and can directly convert triangle soup into an analysis-ready volumetric mesh. The approach based on several core principles: (1) initial mesh construction fully yet efficient, filtered exact computation (2) explicit (automatic or user-defined) tolerancing of the relative to surface input (3) iterative improvement with guarantees, at every step, output validity. quality resulting...

Empirically validating new 3D-printing related algorithms and implementations requires testing data representative of inputs encountered \emph{in the wild}. An ideal benchmarking dataset should not only draw from same distribution shapes people print in terms class (e.g., toys, mechanisms, jewelry), representation type triangle soup meshes) complexity number facets), but also capture problems artifacts endemic to 3D printing models self-intersections, non-manifoldness). We observe that...

Many machine learning models operate on images, but ignore the fact that images are 2D projections formed by 3D geometry interacting with light, in a process called rendering. Enabling ML to understand image formation might be key for generalization. However, due an essential rasterization step involving discrete assignment operations, rendering pipelines non-differentiable and thus largely inaccessible gradient-based techniques. In this paper, we present {\emph DIB-R}, differentiable...

Neural approximations of scalar- and vector fields, such as signed distance functions radiance have emerged accurate, high-quality representations. State-of-the-art results are obtained by conditioning a neural approximation with lookup from trainable feature grids [Liu et al. 2020; Martel 2021; Müller 2022; Takikawa 2021] that take on part the learning task allow for smaller, more efficient networks. Unfortunately, these usually come at cost significantly increased memory consumption...

Solid shapes in computer graphics are often represented with boundary descriptions, e.g. triangle meshes, but animation, physically-based simulation, and geometry processing more realistic accurate when explicit volume representations available. Tetrahedral meshes which exactly contain (interpolate) the input description desirable difficult to construct for a large class of meshes. Character CAD models composed many connected components numerous self-intersections, non-manifold pieces, open...

Skinning transformations are a popular way to articulate shapes and characters. However, traditional animation interfaces require all of the skinning be specified explicitly, typically using control structure (a rig). We propose system where user specifies only subset degrees freedom rest automatically inferred nonlinear, rigidity energies. By utilizing low-order model reformulating our energy functions accordingly, algorithm runs orders magnitude faster than previous methods without...

Many high-level geometry processing tasks rely on low-level constructive solid operations. Though trivial for implicit representations, boolean operations are notoriously difficult to execute robustly explicit boundary representations. Existing methods 3D triangle meshes fall short in one way or another. Some fast but fail produce closed, self-intersection free output. Other robust place prohibitively strict assumptions the input, e.g., no hollow cavities, non-manifold edges...

Inside-outside determination is a basic building block for higher-level geometry processing operations. Generalized winding numbers provide robust answer triangle meshes, regardless of defects such as self-intersections, holes or degeneracies. In this paper, we further generalize the number to point clouds. Previous methods evaluating are slow completely disconnected surfaces, soups or-in extreme case- We propose tree-based algorithm reduce asymptotic complexity generalized computation,...

We propose a method to design linear deformation subspaces, unifying blend skinning and generalized barycentric coordinates. Deformation subspaces cut down the time complexity of variational shape methods physics-based animation (reduced-order physics). Our feature many desirable properties: interpolation, smoothness, shape-awareness, locality, both constant precision. achieve these by minimizing quadratic energy, built via discrete Laplacian inducing precision on domain boundary. main...

Inverse reconstruction from images is a central problem in many scientific and engineering disciplines. Recent progress on differentiable rendering has led to methods that can efficiently differentiate the full process of image formation with respect millions parameters solve such problems via gradient-based optimization. At same time, availability cheap derivatives does not necessarily make an inverse easy solve. Mesh-based representations remain particular source irritation: adverse...

Neural implicit fields have recently emerged as a useful representation for 3D shapes. These are commonly represented neural networks which map latent descriptors and coordinates to function values. The descriptor of field acts deformation handle the shape it represents. Thus, smoothness with respect this is paramount performing shape-editing operations. In work, we introduce novel regularization designed encourage smooth spaces in by penalizing upper bound on field's Lipschitz constant....

No abstract available.

We present a new approach for generating global illumination renderings of hand-drawn characters using only small set simple annotations. Our system exploits the concept bas-relief sculptures, making it possible to generate 3D proxies suitable rendering without requiring side-views or extensive user input. formulate an optimization process that automatically constructs approximate geometry sufficient evoke impression consistent shape. The resulting renders provide richer stylization...

Abstract We propose the first reduced model simulation framework for deformable solid dynamics using autoencoder neural networks. provide a data‐driven approach to generating nonlinear spaces deformation dynamics. In contrast previous methods machine learning which accelerate by approximating time‐stepping function, we solve true equations of motion in latent‐space variational formulation implicit integration. Our produces drastically smaller than conventional linear reduction, improving...

Articulation of 3D characters requires control over many degrees freedom: a difficult task with standard 2D interfaces. We present tangible input device composed interchangeable, hot-pluggable parts. Embedded sensors measure the device's pose at rates suitable for real-time editing and animation. Splitter parts allow branching to accommodate any skeletal tree. During assembly, recognizes topological changes as individual or pre-assembled subtrees are plugged unplugged. A novel semi-automatic...

Many machine learning image classifiers are vulnerable to adversarial attacks, inputs with perturbations designed intentionally trigger misclassification. Current methods directly alter pixel colors and evaluate against norm-balls: smaller than a specified magnitude, according measurement norm. This evaluation, however, has limited practical utility since in the space do not correspond underlying real-world phenomena of formation that lead them no security motivation attached. Pixels natural...

This paper introduces a novel geometric multigrid solver for unstructured curved surfaces. Multigrid methods are highly efficient iterative solving systems of linear equations. Despite the success in problems defined on structured domains, generalizing to domains remains challenging problem. The critical missing ingredient is prolongation operator transfer functions across different levels. We propose method computing triangulated surfaces based intrinsic geometry, enabling an Our surface...

Many tasks in geometry processing and physical simulation benefit from multiresolution hierarchies. One important characteristic across a variety of applications is that coarser layers strictly encage finer layers, nesting one another. Existing techniques such as surface mesh decimation, voxelization, or contouring distance level sets do not provide sufficient control over the quality output surfaces while maintaining strict nesting. We propose solution enables use application-specific...