BEGIN:VCALENDAR VERSION:2.0 PRODID:Linklings LLC BEGIN:VTIMEZONE TZID:Asia/Seoul X-LIC-LOCATION:Asia/Seoul BEGIN:STANDARD TZOFFSETFROM:+0900 TZOFFSETTO:+0900 TZNAME:KST DTSTART:18871231T000000 DTSTART:19881009T020000 END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20230103T035346Z LOCATION:Auditorium\, Level 5\, West Wing DTSTART;TZID=Asia/Seoul:20221206T100000 DTEND;TZID=Asia/Seoul:20221206T120000 UID:siggraphasia_SIGGRAPH Asia 2022_sess153@linklings.com SUMMARY:Technical Papers Fast-Forward DESCRIPTION:Technical Papers\n\nA 2-hour preview session of all Technical Papers will also be held on the first day of the event where author(s) of each paper get less than a minute to wow the attendees with a brief overvi ew of their work. \n\nThe Technical Papers program is the premier internat ional forum for disseminating new scholarly work in computer graphics and interactive techniques. Technical Papers are published as a special issue of ACM Transactions on Graphics. In addition to papers selected by the SIG GRAPH Asia 2022 Technical Papers Jury, the conference presents papers that have been published in ACM Transactions on Graphics during the past year. Accepted papers adhere to the highest scientific standards.\n\nWoven Fabr ic Capture from a Single Photo\n\nJin, Wang, Hasan, Guo, Marschner...\n\nD igitally reproducing the appearance of woven fabrics is important in many applications of realistic rendering, from interior scenes to virtual chara cters. However, designing realistic shading models and capturing real fabr ic samples are both challenging tasks. Previous work ranges from applying gen...\n\n---------------------\nSurfaceVoronoi: Efficiently Computing Vor onoi Diagrams Over Mesh Surfaces with Arbitrary Distance Solvers\n\nXin, W ang, Xu, Yan, Chen...\n\nIn this paper, we propose to compute Voronoi diag rams over mesh surfaces driven by an arbitrary geodesic distance solver, a ssuming that the input is a triangle mesh as well as a collection of sites mathbf{P}={p_i}_{i=1}^m on the surface. We propose two key techniques to solve this problem. First, as...\n\n---------------------\nMotion In-betwe ening via Two-stage Transformers\n\nQin, Zheng, Zhou\n\nWe present a deep learning-based framework to synthesize motion in-betweening in a two-stage manner. Given some context frames and a target frame, the system can gene rate plausible transitions with variable lengths in a non-autoregressive f ashion. The framework consists of two Transformer Encoder-ba...\n\n------- --------------\nReference Based Sketch Extraction via Attention Mechanism\ n\nAshtari, Seo, Kang, Cha, Noh\n\nWe propose a model that extracts a sket ch from a colorized image in such a way that the extracted sketch has a li ne style similar to a given reference sketch while preserving the visual c ontent identically to the colorized image. Authentic sketches drawn by art ists have various sketch styles to add ...\n\n---------------------\nMeshT aichi: A Compiler for Efficient Mesh-based Operations\n\nYu, Xu, Kuang, Hu , Liu\n\nMeshes are an indispensable representation in many graphics appli cations because they provide conformal spatial discretizations. However, m esh-based operations are often slow due to unstructured memory access patt erns. We propose MeshTaichi, a novel mesh compiler that provides an intuit ive programmi...\n\n---------------------\nComputing Medial Axis Transform with Feature Preservation via Restricted Power Diagram\n\nWang, Wang, Wan g, Guo\n\nWe propose a novel framework for computing the medial axis trans form of 3D shapes while preserving their medial features via restricted po wer diagram (RPD). Medial features, including external features such as th e sharp edges and corners of the input mesh surface and internal features such as the se...\n\n---------------------\nHydrophobic and Hydrophilic So lid-Fluid Interaction\n\nLiu, Wang, Feng, Tang, Le...\n\nWe propose a nove l solid-fluid coupling method to capture the subtle hydrophobic and hydrop hilic interactions between liquid, solid, and air at their multi-phase jun ctions. The key component of our approach is a Lagrangian model that tackl es the coupling, evolution, and equilibrium of dynamic contac...\n\n------ ---------------\nCapturing and Animation of Body and Clothing from Monocul ar Video\n\nFeng, Yang, Pollefeys, Black, Bolkart\n\nWhile recent work has shown progress on extracting clothed 3D human avatars from a single image , video, or a set of 3D scans, several limitations remain. Most methods us e a holistic representation to jointly model the body and clothing, which means that the clothing and body cannot be separated for ...\n\n---------- -----------\nDepth of Field Aware Differentiable Rendering\n\nPidhorskyi, Bagautdinov, Ma, Saragih, Schwartz...\n\nCameras with a finite aperture di ameter exhibit defocus for scene elements that are not at the focus distan ce and have only a limited depth of field within which objects appear acce ptably sharp. In this work, we address the problem of applying inverse ren dering techniques to input data that exhibits...\n\n---------------------\ nImpact of correct and simulated focus cues on perceived realism\n\nMarch, Krishnan, Watt, Wernikowski, Gao...\n\nThe natural accommodation of the h uman eye to different distances results in focus cues, which contribute to depth perception and appearance. Since focus cues are very difficult to r eproduce in an electronic display, it is desirable to know how much they c ontribute to realistic image appearance. In ...\n\n---------------------\n DeepJoin: Learning a Joint Occupancy, Signed Distance, and Normal Field Fu nction for Shape Repair\n\nLamb, Banerjee, Banerjee\n\nWe introduce DeepJo in, an automated approach to generate high-resolution repairs for fracture d shapes using deep neural networks. Existing approaches to perform automa ted shape repair operate exclusively on symmetric objects, require a compl ete proxy shape, or predict restoration shapes using low-re...\n\n-------- -------------\nProgressive Simulation for Cloth Quasistatics\n\nZhang, Dum as, Fei, Jacobson, James...\n\nThe trade-off between speed and fidelity in cloth simulation is a fundamental computational problem in computer graph ics and computational design. Coarse cloth models provide the interactive performance required by designers, but they can not be simulated at higher resolutions (``up-resed'') without...\n\n---------------------\nMarginal Multiple Importance Sampling\n\nWest, Georgiev, Hachisuka\n\nMultiple impo rtance sampling (MIS) is a powerful tool to combine different sampling tec hniques in a provably good manner. MIS requires that the techniques' proba bility density functions (PDFs) are readily evaluable point-wise. However, this requirement may not be satisfied when (some of) those PDFs ...\n\n-- -------------------\nNeuralRoom: Geometry-Constrained Neural Implicit Surf aces for Indoor Scene Reconstruction\n\nWang, Li, Jiang, Zhou, Cao...\n\nW e present a novel neural surface reconstruction method called NeuralRoom f or reconstructing room-sized indoor scenes directly from a set of 2D image s. Recently, implicit neural representations have become a promising way t o reconstruct surfaces from multiview images due to their high-quality res ult...\n\n---------------------\nTransformer Inertial Poser: Real-time Hum an Motion Reconstruction from Sparse IMUs with Simultaneous Terrain Genera tion\n\nJiang, Ye, Gopinath, Won, Winkler...\n\nReal-time human motion rec onstruction from a sparse set of (e.g. six) wearable IMUs provides a non-i ntrusive and economic approach to motion capture. Without the ability to a cquire position information directly from IMUs, recent works took data-dri ven approaches that utilize large human motion datas...\n\n--------------- ------\nPupil-aware Holography\n\nChakravarthula, Baek, Schiffers, Tseng, Kuo...\n\nHolographic displays promise to deliver unprecedented display ca pabilities in augmented reality applications, featuring wide field of view , wide color gamut, spatial resolution and depth cues all in a compact for m factor. While emerging holographic display approaches have been successf ul in achievin...\n\n---------------------\nNeural Wavelet-domain Diffusio n for 3D Shape Generation\n\nHui, Li, Hu, Fu\n\nThis paper presents a new approach for 3D shape generation, enabling a direct generative modeling on a continuous implicit representation in wavelet frequency domain. Specifi cally, we propose a compact wavelet representation with a pair of coarse a nd detail coefficient volumes to implicitly represen...\n\n--------------- ------\nScene Synthesis from Human Motion\n\nYe, Wang, Li, Park, Liu...\n\ nLarge-scale capture of human motion with diverse, complex scenes, while i mmensely useful, is often considered prohibitively costly. Meanwhile, huma n motion alone contains rich information about the scene they reside in an d interact with. For example, a sitting human suggests the existence of a chair...\n\n---------------------\nAffordable Spectral Measurements of Tra nslucent Materials\n\nIser, Rittig, Nogué, Nindel, Wilkie\n\nWe present a spectral measurement approach for the bulk optical properties of transluce nt materials using only low-cost components. We focus on the translucent i nks used in full-color 3D printing, and develop a technique with a high sp ectral resolution, which is important for accurate color reproduc...\n\n-- -------------------\nShape from Release: Inverse Design and Fabrication of Controlled Release Structures\n\nPanetta, Mohammadian, Luci, Babaei\n\nOb jects with different shapes can dissolve in significantly different ways i nside a solution. Predicting different shapes’ dissolution dynamics is an important problem especially in pharmaceutics. More important and challeng ing, however, is controlling the dissolution via shape, i.e., designin...\ n\n---------------------\nDeclarative Specification for Unstructured Mesh Editing Algorithms\n\nJiang, Dai, Hu, Zhou, Dumas...\n\nWe introduce a nov el approach to describe mesh generation, mesh adaptation, and geometric mo deling algorithms relying on changing mesh connectivity using a high-level abstraction. The main motivation is to enable easy customization and deve lopment of these algorithms via a declarative specification ...\n\n------- --------------\nText2Light: Zero-shot Text-driven HDR Panorama Generation\ n\nChen, Wang, Liu\n\nHigh-quality HDRIs (High Dynamic Range Images), typi cally HDR panoramas, are one of the most popular ways to create photoreali stic lighting and 360-degree reflections of 3D scenes in graphics. Given t he difficulty of capturing HDRIs, a versatile and controllable generative model is highly desired, w...\n\n---------------------\nMotion Guided Deep Dynamic 3D Garments\n\nZhang, Ceylan, Mitra\n\nRealistic dynamic garments on animated characters have many AR/VR applications. While authoring such dynamic garment geometry is still a challenging task, data-driven simula tion provides an attractive alternative, especially if it can be controlle d simply using the motion of the underlying characte...\n\n--------------- ------\nDirect acquisition of volumetric scattering phase function using s peckle correlations\n\nAlterman, Saiko, Levin\n\nIn material acquisition w e want to infer the internal properties of materials from the way they sc atter light. In particular, we are interested in measuring the phase func tion of the material, governing the amount of energy scattered towards dif ferent directions. This phase function has been show...\n\n-------------- -------\nCurl-Flow: Boundary-Respecting Pointwise Incompressible Velocity Interpolation for Grid-Based Fluids\n\nChang, Partono, Azevedo, Batty\n\nW e propose to augment standard grid-based fluid solvers with pointwise dive rgence-free velocity interpolation, thereby ensuring exact incompressibili ty down to the sub-cell level. Our method takes as input a discretely dive rgence-free velocity field generated by a staggered grid pressure projecti on,...\n\n---------------------\nLearning Reconstructability for Drone Aer ial Path Planning\n\nLiu, Lin, Hu, Xie, Fu...\n\nWe introduce the first le arning-based reconstructability predictor to improve view and path plannin g for large-scale 3D urban scene acquisition using unmanned drones. In con trast to previous heuristic approaches, our method learns a model that exp licitly predicts how well a 3D urban scene will be re...\n\n-------------- -------\nMasked Lip-Sync Prediction by Audio-Visual Contextual Exploitatio n in Transformers\n\nSun, Zhou, Wang, Wu, Hong...\n\nPrevious studies have explored generating accurately lip-synced talking faces for arbitrary tar gets given audio conditions. However, most of them deform or generate the whole facial area, leading to non-realistic results. In this work, we delv e into the formulation of altering only the mouth shapes ...\n\n---------- -----------\nSHRED: 3D Shape Region Decomposition with Learned Local Opera tions\n\nJones, Habib, Ritchie\n\nWe present SHRED, a method for 3D SHape REgion Decomposition. SHRED takes a 3D point cloud as input and uses learn ed local operations to produce a segmentation that approximates fine-grain ed part instances. We endow SHRED with three decomposition operations: spl itting regions, fixing the boundaries ...\n\n---------------------\nBreaki ng Good: Fracture Modes for Realtime Destruction\n\nSellán, Luong, Mattos Da Silva, Ramakrishnan, Yang...\n\nDrawing a direct analogy with the well- studied vibration or elastic modes, we introduce an object’s fracture mode s, which constitute its preferred or most natural ways of breaking. We for mulate a sparsified eigenvalue problem, which we solve iteratively to obta in the n lowest-energy modes. The...\n\n---------------------\nNeural Clot h Simulation\n\nBertiche, Madadi, Escalera\n\nWe present a general framewo rk for the garment animation problem through unsupervised deep learning in spired in physically based simulation. Existing trends in the literature a lready explore this possibility. Nonetheless, these approaches do not hand le cloth dynamics. Here, we propose the first meth...\n\n----------------- ----\nA Monte Carlo Method for Fluid Simulation\n\nRioux-Lavoie, Sugimoto, Özdemir, Shimada, Batty...\n\nWe present a novel Monte Carlo-based fluid simulation approach capable of pointwise and stochastic estimation of flui d motion. Drawing on the Feynman-Kac representation of the vorticity trans port equation, we propose a recursive Monte Carlo estimator of the Biot-Sa vart law and extend it with a strea...\n\n---------------------\n$S^3$-Sli cer: A General Slicing Framework for Multi-Axis 3D Printing\n\nZhang, Fang , Huang, Dutta, Lefebvre...\n\nMulti-axis motion introduces more degrees o f freedom into the process of 3D printing to enable different objectives o f fabrication by accumulating materials layers upon curved layers. An exis ting challenge is how to effectively generate the curved layers satisfying multiple objectives simultaneously...\n\n---------------------\nDifferSke tching: How Differently Do People Sketch 3D Objects?\n\nXiao, Su, Liao, Li an, Song...\n\nMultiple sketch datasets have been proposed to understand h ow people draw 3D objects. However, such datasets are often of small scale and cover a small set of objects or categories. In addition, these datase ts contain freehand sketches mostly from expert users, making it difficult to compare the dra...\n\n---------------------\nRealistic Luminance in VR \n\nMatsuda, Chapiro, Zhao, Smith, Bachy...\n\nAs virtual reality (VR) hea dsets continue to achieve ever more immersive visuals along the axes of re solution, field of view, focal cues, distortion mitigation, and so on, the luminance and dynamic range of these devices falls far short of widely av ailable consumer televisions. While work remains to...\n\n---------------- -----\nPosition-Based Surface Tension Flow\n\nXing, Ruan, Wang, Zhu, Chen\ n\nThis paper presents a novel approach to simulating surface tension flow within a position-based dynamics (PBD) framework. We enhance the conventi onal PBD fluid method in terms of its surface representation and constrain t enforcement to furnish support for the simulation of interfacial phenome na driv...\n\n---------------------\nReconstructing editable prismatic CAD from rounded voxel models\n\nLambourne, Willis, Jayaraman, Zhang, Sanghi. ..\n\nReverse Engineering a CAD shape from other representations is an imp ortant geometric processing step for many downstream applications. In this work, we introduce a novel neural network architecture to solve this chal lenging task and approximate a smoothed signed distance function with an e ditable, c...\n\n---------------------\nPADL: Language-Directed Physics-Ba sed Character Control\n\nJuravsky, Guo, Fidler, Peng\n\nDeveloping systems that can synthesize natural and life-like motions for simulated character s has long been a focus for computer animation. But in order for these sys tems to be useful for downstream applications, they need not only produce high-quality motions, but must also provide an accessible an...\n\n------- --------------\nVideoReTalking: Audio-based Lip Synchronization for Talkin g Head Video Editing In the Wild\n\nCheng, Cun, Zhang, Xia, Yin...\n\nWe p resent VideoReTalking, a new system to edit the faces of a real-world talk ing head video according to an input audio, producing a high-quality and l ip-syncing output video even with a different emotion.Our system disentang les this objective into three sequential tasks: (1) face video generation ...\n\n---------------------\nNeural Parameterization for Dynamic Human He ad Editing\n\nMa, Li, Liao, Wang, Zhang...\n\nImplicit radiance functions emerged as a powerful scene representation for reconstructing and renderin g photo-realistic views of a 3D scene. These representations, however, suf fer from poor editability. On the other hand, explicit representations suc h as polygonal meshes allow easy editing, but are ...\n\n----------------- ----\nVOCAL: Vowel and Consonant Layering for Expressive Animator-Centric Singing Animation\n\nPan, Landreth, Fiume, Singh\n\nSinging and speaking are two fundamental forms of human communication. From a modeling perspect ive however, speaking can be seen as a subset of singing. We present VOCAL , a system that automatically generates expressive, animator-centric lower face animation from singing audio input. Articulatory ph...\n\n---------- -----------\nQuestSim: Human Motion Tracking from Sparse Sensors with Simu lated Avatars\n\nWinkler, Won, Ye\n\nReal-time tracking of human body moti on is crucial for interactive and\nimmersive experiences in AR/VR. However , very limited sensor data about\nthe body is available from standalone we arable devices such as HMDs (Head\nMounted Devices) or AR glasses. In this work, we present a reinforcement\nlearning f...\n\n---------------------\ nLearning-Based Bending Stiffness Parameter Estimation by a Drape Tester\n \nFeng, Huang, Xu, Wang\n\nReal-world fabrics often possess complicated no nlinear, anisotropic bending stiffness properties. Measuring the physical parameters of such properties for physics-based simulation is difficult y et unnecessary, due to the persistent existence of numerical errors in sim ulation technology. In this wo...\n\n---------------------\nAgileAvatar: Stylized 3D Avatar Creation via Cascaded Domain Bridging\n\nSang, Zhi, Son g, Liu, Lai...\n\nStylized 3D avatars have become increasingly prominent i n our modern life. Creating these avatars manually usually involves labori ous selection and adjustment of continuous and discrete parameters and is time-consuming for average users. Self-supervised approaches to automatica lly create 3D avatars ...\n\n---------------------\nSCULPTOR: Skeleton-Con sistent Face Creation Using a Learned Parametric Generator\n\nQiu, Li, He, Zhang, Zhang...\n\nRecent years have seen growing interest in 3D human fa ces modelling due to its wide applications in digital human, character gen eration and animation. Existing approaches overwhelmingly emphasized on mo deling the exterior shapes, textures and skin properties of faces, ignorin g the inherent correlati...\n\n---------------------\nHierarchical Layout Blending with Recursive Optimal Correspondence\n\nXu, Li, Yang, Shi, Fu... \n\nWe present a novel method for blending hierarchical layouts with seman tic labels. The core of our method is a hierarchical structure corresponde nce algorithm, which recursively finds optimal substructure correspondence s, achieving a globally optimal correspondence between a pair of hierarchi cal layo...\n\n---------------------\nConstant Time Median Filter using 2D Wavelet Matrix\n\nMoroto, Umetani\n\nThe median filter is a simple yet po werful noise reduction technique that is extensively applied in image, sig nal, and speech processing. It can effectively remove impulsive noise whil e preserving the content of the image by taking the median of neighboring pixels; thus, it has various applications,...\n\n---------------------\nSt yleBin: Stylizing Video by Example in Stereo\n\nKučera, Mould, Sýkora\n\nI n this paper we present StyleBin---an approach to example-based stylizatio n of videos that can produce consistent binocular depiction of stylized co ntent on stereoscopic displays. Given the target sequence and a set of sty lized keyframes accompanied by information about depth in the scene, we fo rmu...\n\n---------------------\nFluidic Topology Optimization with an Ani sotropic Mixture Model\n\nLi, Du, Grama Srinivasan, Wu, Zhu...\n\nFluidic devices are crucial components in many industrial applications involving f luid mechanics. Computational design of a high-performance fluidic system faces multifaceted challenges regarding its geometric representation and p hysical accuracy. We present a novel topology optimization method to d...\ n\n---------------------\nA Biologically Inspired Hair Aging Model\n\nBalb ão, Walter\n\nHair rendering has been a focal point of attention in comput er graphics for the last couple of decades. However, there have been few c ontributions to the modeling and rendering of the natural hair aging pheno menon. We present a new technique that simulates the process of hair grayi ng and hair thinni...\n\n---------------------\nVIINTER: View Interpolatio n With Implicit Neural Representations of Images\n\nFeng, Jabbireddy, Vars hney\n\nWe present VIINTER, a method for view interpolation by interpolati ng the implicit neural representation (INR) of the captured images. We lev erage the learned code vector associated with each image and interpolate b etween these codes to achieve viewpoint transitions. We propose several te chniques tha...\n\n---------------------\nForce-Aware Interface via Electr omyography for Natural VR/AR Interaction\n\nZhang, Liang, Chen, Torrens, A tashzar...\n\nWhile tremendous advances in visual and auditory realism hav e been made for virtual and augmented reality (VR/AR), introducing a plaus ible sense of physicality into the virtual world remains challenging. Clos ing the gap between real-world physicality and immersive virtual experienc e requires a close...\n\n---------------------\nAnimatomy: an Animator-cen tric, Anatomically Inspired System for 3D Facial Modeling, Animation and T ransfer\n\nChoi, Eom, Mouscadet, Cullingford, Ma...\n\nWe present Animatom y, a novel anatomic+animator centric representation of the human face. Pre sent FACS-based systems are plagued with problems of face muscle separatio n, coverage, opposition, and redundancy. We, therefore, propose a collecti on of muscle fiber curves as an anatomic basis, whose contr...\n\n-------- -------------\nMetappearance: Meta-Learning for Visual Appearance Reproduc tion\n\nFischer, Ritschel\n\nThere currently exist two main approaches to reproducing visual appearance using Machine Learning (ML): The first is tr aining models that generalize over different instances of a problem, e.g., different images of a dataset. As one-shot approaches, these offer fast i nference, but often fall short in...\n\n---------------------\nMake Your O wn Sprites: Aliasing-Aware and Cell-Controllable Pixelization\n\nWu, Chai, Zhao, Deng, Liu...\n\nPixel art is a unique art style with the appearance of low resolution images. In this paper, we propose a data-driven pixeliz ation method that can produce sharp and crisp cell effects with controllab le cell size. Our approach overcomes the limitation of existing learning-b ased methods in cell size c...\n\n---------------------\nMoRig: Motion-Awa re Rigging of Character Meshes from Point Clouds\n\nXu, Zhou, Yi, Kalogera kis\n\nWe present MoRig, a method that automatically rigs character meshes driven by single-view point cloud streams capturing the motion of perform ing characters. Our method is also able to animate the 3D meshes according to the captured point cloud motion. At the heart of our approach lies a d eep neural ...\n\n---------------------\nGeo-metric: A Perceptual Dataset of Distortions on Faces\n\nWolski, Trutoiu, Dong, Shen, MacKenzie...\n\nIn this work we take a novel perception-centered approach to quantify distor tions on 3D geometry of faces, to which humans are particularly sensitive. We generated a dataset, composed of 100 high-quality and demographically- balanced face scans. We then subjected these meshes to distortions that co ve...\n\n---------------------\nFast Dynamic Radiance Fields with Time-Awa re Neural Voxels\n\nFang, Yi, Wang, Xie, Zhang...\n\nNeural radiance field s (NeRF) have shown great success in modeling 3D scenes and synthesizing n ovel-view images. However, most previous NeRF methods take much time to op timize one single scene. Explicit data structures, e.g. voxel features, sh ow great potential to accelerate the training process. Ho...\n\n---------- -----------\nBoolSurf : Boolean Operations on Surfaces\n\nRiso, Nazzaro, P uppo, Jacobson, Zhou...\n\nWe port Boolean set operations between 2D shape s to surfaces of any genus, with any number of open boundaries. We combin e shapes bounded by sets of freely intersecting loops, consisting of geode sic lines and cubic B\'ezier splines lying on a surface. We compute the ar rangement of shapes directly on...\n\n---------------------\nMyStyle: A Pe rsonalized Generative Prior\n\nNitzan, Aberman, Liba, He, Yarom...\n\nWe i ntroduce MyStyle, a personalized deep generative prior trained with a few shots of an individual. MyStyle allows to reconstruct, enhance and edit im ages of a specific person, such that the output is faithful to the person’ s key facial characteristics. Given a small reference set of portrai...\n\ n---------------------\nDifferentiable rendering using RGBXY derivatives a nd optimal transport\n\nXing, Luan, Yan, Hu, Qian...\n\nWe present a novel differentiable rendering framework that is reformulated from the Lagrangi an view. Inspired by fluid simulation, traditional differentiable renderin g approach can be viewed as the Eulerian method due to the fact that image derivatives are computed locally on a fixed grid of screen-...\n\n------- --------------\nGreen Coordinates for Triquad Cages in 3D\n\nThiery, Boube keur\n\nWe introduce Green coordinates for triquad cages in 3D.\nBased on Green's third identity, they allow defining the harmonic deformation of a 3D point inside a cage as a linear combination of its vertices and face no rmals.\nUsing appropriate Neumann boundary conditions, the resulting defor mations are qu...\n\n---------------------\nUmeTrack: Unified multi-view e nd-to-end hand tracking for VR\n\nHan, Wu, Zhang, Liu, Zhang...\n\nReal-ti me tracking of 3D hand pose in world space is a challenging problem\nand p lays an important role in VR interaction. Existing work in this space are\ nlimited to either producing root-relative (versus world space) 3D pose or rely\non multiple stages such as generating heatmaps and kinematic optimi ...\n\n---------------------\nInteractive exploration of tension-compressi on mixed shells\n\nMiki, Mitchell\n\nAchieving a pure-compression stress s tate is considered central to the form-finding of shell structures. Howeve r, the pure-compression assumption restricts the geometry of the structure 's plan in that any free boundary edges cannot bulge outward. Allowing bot h tension and compression is essential so...\n\n---------------------\nPiv otal Tuning for Latent-based editing of Real Images\n\nRoich, Mokady, Berm ano, Cohen-Or\n\nWe bridge the distortion-editability tradeoff in StyleGAN inversion. By slightly altering the generator, an out-of-domain image is faithfully mapped into an in-domain latent code. The key idea is Pivotal T uning — a brief training process that preserves the editing quality of an in-domain laten...\n\n---------------------\nReconstructing Hand-Held Obje cts from Monocular Video\n\nHuang, Ji, He, Sun, He...\n\nThis paper presen ts an approach that reconstructs a hand-held object from a monocular video . In contrast to many recent methods that directly predict object geometry by a trained network, the proposed approach does not require any learned prior about the object and is able to recover more accurate a...\n\n------ ---------------\nPopStage: The Generation of Stage Cross-Editing Video bas ed on Spatio-Temporal Matching\n\nLee, Yoo, Cho, Kim, Im...\n\nStageMix is a mixed video that is created by concatenating the segments from various performance videos of an identical song in a visually smooth manner by mat ching the main subject's silhouette presented in the frame. We introduce P opStage, which allows users to generate a StageMix automatically. P...\n\n ---------------------\nReconstructing Personalized Semantic Facial NeRF Mo dels From Monocular Video\n\nGao, Zhong, Xiang, Hong, Guo...\n\nWe present a novel semantic model for human head defined with neural radiance field. The 3D-consistent head model consist of a set of disentangled and interpr etable bases, and can be driven by low-dimensional expression coefficients . Thanks to the powerful representation ability of neural radiance f...\n\ n---------------------\nICARUS: A Specialized Architecture for Neural Radi ance Fields Rendering\n\nRao, Yu, Wan, Zhou, Zheng...\n\nThe practical dep loyment of Neural Radiance Fields (NeRF) in rendering applications faces s everal challenges, with the most critical one being low rendering speed on even high-end graphic processing units (GPUs). In this paper, we present ICARUS, a specialized accelerator architecture tailored for N...\n\n------ ---------------\nLuisaRender: A High-Performance Rendering Framework with Layered and Unified Interfaces on Stream Architectures\n\nZheng, Zhou, Che n, Yan, Zhang...\n\nThe advancements in hardware have drawn more attention than ever to high-quality offline rendering with modern stream processors , both in the industry and in research fields. However, the graphics APIs are fragmented and existing shading languages lack high-level constructs s uch as polymorphism, whi...\n\n---------------------\nColor-Perception-Gui ded Display Power Reduction for Virtual Reality\n\nDuinkharjav, Chen, Tyag i, He, Zhu...\n\nBattery life is an increasingly urgent challenge for toda y's untethered VR and AR devices. However, the power efficiency of head-mo unted displays is naturally at odds with growing computational requirement s driven by better resolution, refresh rate, and dynamic ranges, all of wh ich reduce the sustai...\n\n---------------------\nSkinMixer: Blending 3D Animated Models\n\nNuvoli, Pietroni, Scateni, Cignoni, Tarini\n\nWe propos e a novel technique to compose new 3D animated models, such as videogame c haracters, by combining pieces from existing ones. Our method works on pro duction-ready rigged, skinned, and animated 3D models to reassemble new on es. We exploit Mix-and-match operations on the skeletons to trigger t...\n \n---------------------\nLearning to Generate 3D Shapes from a Single Exam ple\n\nWu, Zheng\n\nExisting generative models for 3D shapes are typically trained on a large 3D dataset, often of a specific object category. In t his paper, we investigate the deep generative model that learns from only a single reference 3D shape. Specifically, we present a multi-scale GAN-ba sed model designed to ca...\n\n---------------------\nStochastic Poisson S urface Reconstruction\n\nSellán, Jacobson\n\nWe introduce a novel statisti cal derivation of the classic Poisson Surface Reconstruction algorithm for recovering shapes from 3D point clouds. Instead of thorugh an implicit fu nction, we represent the reconstructed shape as a modified Gaussian Proces s, which allows us to respond to statistical quer...\n\n------------------ ---\nHuman Performance Modeling and Rendering via Neural Animated Mesh\n\n Zhao, Jiang, Yao, Zhang, Wang...\n\nWe have recently seen tremendous progr ess in the neural advances for photo-real human modeling and rendering. Bu t it's still challenging to integrate them into an existing mesh-based pip eline for downstream applications. In this paper, we present a comprehensi ve neural approach for high-quality reco...\n\n---------------------\nH❤️r tDown: Document Processor for Executable Linear Algebra Papers\n\nLi, Kami l, Jacobson, Gingold\n\nScientific documents describe a topic in a mix of prose and mathematical expressions. The prose refers to those expressions, which themselves must be encoded in, e.g., LaTeX. The resulting documents are static, even though most documents are now read digitally. Moreover, formulas must be implemented...\n\n---------------------\nLearning Virtual Chimeras by Dynamic Motion Reassembly\n\nLEE, Lee, Lee\n\nThe Chimera is a mythological hybrid creature composed of different animal parts. The chi mera’s movements are highly dependent on the spatial and temporal alignmen ts of its composing parts. In this paper, we present a novel algorithm tha t creates and animates chimeras by dynamically reassembli...\n\n---------- -----------\nEfficient Differentiation of Pixel Reconstruction Filters for Path-Space Differentiable Rendering\n\nYu, Zhang, Dong, Nowrouzezahrai, Z hao\n\nPixel reconstruction filters play an important role in physics-base d rendering and have been thoroughly studied. In physics-based differentia ble rendering, however, the proper treatment of pixel reconstruction filte rs has remained largely under-explored. We present a new technique to effi ciently dif...\n\n---------------------\nHidden Degrees of Freedom in Impl icit Vortex Filaments\n\nIshida, Wojtan, Chern\n\nThis paper presents a ne w representation of curve dynamics, with applications to vortex filaments in fluid dynamics. Instead of representing these filaments with explicit c urve geometry and Lagrangian equations of motion, we represent curves impl icitly with a new co-dimensional 2 level set descripti...\n\n------------- --------\nMixed Variational Finite Elements for Implicit Simulation of Def ormables\n\nTrusty, Kaufman, Levin\n\nWe propose and explore a new method for the implicit time integration of elastica. Key to our approach is the use of a mixed variational principle. In turn, its finite element discreti zation leads to an efficient and accurate sequential quadratic programming solver with a superset of the desirable p...\n\n---------------------\nCL IP-Mesh: Generating textured meshes from text using pretrained image-text models\n\nMohammad Khalid, Xie, Belilovsky, Popa\n\nWe present a technique for zero-shot generation of a 3D model using only a target text prompt. W ithout any 3D supervision our method deforms the control shape of a limit subdivided surface along with its texture map and normal map to obtain a 3 D asset that corresponds to the input text prompt and ca...\n\n----------- ----------\nRhythmic Gesticulator: Rhythm-Aware Co-Speech Gesture Synthesi s with Hierarchical Neural Embeddings\n\nAo, Gao, Lou, Chen, Liu\n\nAutoma tic synthesis of realistic co-speech gestures is an increasingly important yet challenging task in artificial embodied agent creation. Previous syst ems mainly focus on generating gestures in an end-to-end manner, which lea ds to difficulties in mining the clear rhythm and semantics due to the c.. .\n\n---------------------\nLaplacianFusion: Detailed 3D Clothed-Human Bod y Reconstruction\n\nKim, Nam, Kim, Park, Lee\n\nWe propose LaplacianFusion , a novel approach that reconstructs detailed and controllable 3D clothed- human body shapes from an input depth or 3D point cloud sequence.\nThe key idea of our approach is to use Laplacian coordinates, well-known differen tial coordinates that have been used for mesh editing...\n\n-------------- -------\nSimulation of Hand Anatomy Using Medical Imaging\n\nZheng, Wang, Huang, Barbic\n\nPrecision modeling of the hand internal musculoskeletal a natomy has been largely limited to individual poses, and has not been conn ected into continuous volumetric motion of the hand anatomy actuating acro ss the hand's entire range of motion. This is for a good reason, as hand a natomy and its motion...\n\n---------------------\nUnbiased Caustics Rende ring Guided by Representative Specular Paths\n\nLi, Wang, Tu, Xu, Holzschu ch...\n\nCaustics are interesting patterns caused by the light being focus ed when reflecting off glossy materials. Rendering them in Computer Graphi cs is still challenging: they correspond to high luminous intensity focuse d over a small area. Finding the paths that contribute to this small area is difficult,...\n\n---------------------\nQuadStream: A Quad-Based Scene Streaming Architecture for Novel Viewpoint Reconstruction\n\nHladky, Steng el, Vining, Kerbl, Seidel...\n\nCloud rendering is attractive when targeti ng thin client devices such as phones or VR/AR headsets, or any situation where a high-end GPU is not available due to thermal or power constraints. However, it introduces the challenge of streaming rendered data over a ne twork in a manner that is robust to ...\n\n---------------------\nGaussian Blue Noise\n\nAhmed, Ren, Wonka\n\nAmong the various approaches for produ cing point distributions with blue noise spectrum, we argue for an optimiz ation framework using Gaussian kernels.\nWe show that with a wise selectio n of optimization parameters, this approach attains unprecedented quality, provably surpassing the current state of...\n\n---------------------\nLea rning-based Inverse Rendering of Complex Indoor Scenes with Differentiable Monte Carlo Raytracing\n\nZhu, Luan, Huo, Lin, Zhong...\n\nIndoor scenes typically exhibit complex, spatially-varying appearance from global illumi nation, making the inverse rendering a challenging ill-posed problem. This work presents an end-to-end, learning-based inverse rendering framework i ncorporating differentiable Monte Carlo raytracing with importan...\n\n--- ------------------\nRFEPS: Reconstructing Feature-line Equipped Polygonal Surface\n\nXu, Wang, Dou, Zong, Xin...\n\nFeature lines are important geom etric cues in characterizing the structure of a CAD model. Despite great p rogress in both explicit reconstruction and implicit reconstruction, it re mains a challenging task to reconstruct a polygonal surface equipped with feature lines, especially when the input point...\n\n--------------------- \nCAD2Sketch: Generating Concept Sketches from CAD Sequences\n\nHähnlein, Li, Mitra, Bousseau\n\nConcept sketches are ubiquitous in industrial desig n, as they allow designers to quickly depict imaginary 3D objects.To const ruct their sketches with accurate perspective, designers rely on longstand ing drawing techniques, including the use of auxiliary construction lines to identify midpoints of pe...\n\n---------------------\nCompressing Geode sic Information for Fast Point-to-Point Geodesic Distance Queries\n\nGotsm an, Hormann\n\nGeodesic distances between pairs of points on a 3D mesh sur face are a crucial ingredient of many geometry processing tasks, but are n otoriously difficult to compute efficiently on demand. We propose a novel method for the compact storage of geodesic distance information, which ena bles answering poin...\n\n---------------------\nDynaGAN: Dynamic Few-shot Adaptation of GANs to Multiple Domains\n\nKim, Kang, Kim, Baek, Cho\n\nFe w-shot domain adaptation to multiple domains aims to learn a complex image distribution across multiple domains from a few training images. A naive solution here is to train a separate model for each domain using few-shot domain adaptation methods. Unfortunately, this approach mandates linearly- sc...\n\n---------------------\nGlobally Injective Flattening via a Reduce d Harmonic Subspace\n\nFargion, Weber\n\nWe present a highly efficient-and -robust method for free-boundary flattening of disk-like triangle meshes i n a globally injective manner. We show that by restricting the solution to a low-dimensional subspace of harmonic maps, we can dramatically accelera te the process while obtaining a low-distorti...\n\n---------------------\ nVersatile Control of Fluid-Directed Solid Objects Using Multi-Task Reinfo rcement Learning\n\nRen, Ye, Pan, Zhang\n\nWe propose a new fluid control method combining meta reinforcement learning and multi task training, whic h realize a versatile fluid-directed solid objects controller that can qui ckly transfer between fluid simulators with different parameters to accomp lish series of sub-tasks by encoding the simula...\n\n-------------------- -\nFast Octree Neighborhood Search for SPH Simulations\n\nFernández-Fernán dez, Westhofen, Löschner, Jeske, Longva...\n\nWe present a new octree-base d neighborhood search method for SPH simulation. A speedup of up to 1.9x i s observed in comparison to state-of-the-art methods which rely on uniform grids. While our method focuses on maximizing performance in fixed-radius SPH simulations, we show that it can also be used...\n\n----------------- ----\nIsometric Energies for Recovering Injectivity in Constrained Mapping \n\nDu, Kaufman, Zhou, Kovalsky, Yan...\n\nComputing injective maps with l ow distortions is a long-standing problem in computer graphics. Such maps are particularly challenging to obtain in the presence of positional const raints, because an injective initial map is often not available. Recently, several energies were proposed and shown to be...\n\n-------------------- -\nPlanar panels and planar supporting beams in architectural structures\n \nWang, Jiang, Tellier, Wallner, Pottmann\n\nWe investigate geometric prop erties and modeling capabilities of quad meshes with planar faces and plan ar parameter lines. These properties are beneficial from the structural vi ewpoint. We enable geometric design with planarity side conditions by work ing in the space of planes and with the appropria...\n\n------------------ ---\nFloRen: Real-time High-quality Human Performance Rendering via Appear ance Flow Using Sparse RGB Cameras\n\nShao, Chen, Zheng, Zhang, Zhang...\n \nWe propose FloRen, a novel system for real-time, high-resolution free-vi ew human synthesis. Our system runs at 15fps in 1K resolution with very sp arse RGB cameras. In FloRen, a coarse-level implicit geometry is recovered at first as initialization, and then processed by a neural rendering fram ework ...\n\n---------------------\nSprite-from-Sprite: Cartoon Animation Decomposition with Self-supervised Sprite Estimation\n\nZhang, Wong, Liu\n \nWe present an approach to decompose cartoon animation videos into a set of ``sprites'' --- the basic units of digital cartoons that depict the con tents and transforms of each animated objects. The sprites in real-world c artoons are unique: artists may draw arbitrary sprite animations for expre ssiven...\n\n---------------------\nEfficient Neural Radiance Fields for I nteractive Free-viewpoint Video\n\nLin, Peng, Xu, Yan, Shuai...\n\nThis pa per aims to tackle the challenge of efficiently producing interactive free -viewpoint videos. \nSome recent works equip neural radiance fields with i mage encoders, enabling them to generalize across scenes. When processing dynamic scenes, they can simply treat each video frame as an individual .. .\n\n---------------------\nDifferentiable Simulation of Inertial Musculot endons\n\nWang, Verheul, Yeo, Kalantari, Sueda\n\nWe propose a simple and practical approach for incorporating the effects of muscle inertia, which has been ignored by previous musculoskeletal simulators in both graphics a nd biomechanics. We approximate the inertia of the muscle by assuming that muscle mass is distributed along the centerline of th...\n\n------------- --------\nInteractive and Robust Mesh Booleans\n\nCherchi, Pellacini, Atte ne, Livesu\n\nBoolean operations between objects are among the most used p aradigms to create and edit digital shapes. Despite being conceptually sim ple, the computation of mesh booleans is notoriously challenging. The main issues come mostly from the presence of numerical approximations that mak e the detection an...\n\n---------------------\nDr.3D: Adapting 3D GANs to Artistic Drawings\n\nJin, Ryu, Kim, Baek, Cho\n\nWhile 3D GANs have recen tly demonstrated the high-quality synthesis of multi-view consistent image s and 3D shapes, they are mainly restricted to photo-realistic human portr aits. This paper aims to extend 3D GANs to a different, but meaningful vis ual form: artistic portrait drawings. However, extendi...\n\n------------- --------\nIDE-3D: Interactive Disentangled Editing for High-Resolution 3D- aware Portrait Synthesis\n\nSun, Wang, Shi, Wang, Liu...\n\nExisting 3D-aw are facial generation methods face a dilemma in quality versus editability : they either generate editable results in low resolution, or high quality ones with no editing flexibility. In this work, we propose a new approach that brings the best of both worlds together. Our system consis...\n\n--- ------------------\nDisplay Size and Targeting Performance: Small Hurts, L arge May Help\n\nKim, Madhusudan, Watson, Boudaoud, Tarrazo...\n\nWhich di splay size helps gamers win? Recommendations from the research and PC gami ng communities are contradictory. We find that as display size grows, targ eting performance improves. When size increases from 13" to 26", targeting time drops by over 3%. Further size increases from 26" through 39", 5...\ n\n---------------------\nVToonify: Controllable High-Resolution Portrait Video Style Transfer\n\nYang, Jiang, Liu, Loy\n\nGenerating high-quality a rtistic portrait videos is an important and desirable task in computer gra phics and vision. Although a series of successful portrait image toonifica tion models built upon the powerful StyleGAN have been proposed, these ima ge-oriented methods have obvious limitations when app...\n\n-------------- -------\nDressing Avatars: Deep Photorealistic Appearance for Physically S imulated Clothing\n\nXiang, Bagautdinov, Stuyck, Prada, Romero...\n\nDespi te recent progress in developing animatable full-body avatars, realistic m odeling of clothing - one of the core aspects of human self-expression - r emains an open challenge. State-of-the-art physical simulation methods can generate realistically behaving clothing geometry at interactive rates. . ..\n\n---------------------\nPattern-Based Cloth Registration and Sparse-V iew Animation\n\nHalimi, Stuyck, Xiang, Bagautdinov, Wen...\n\nWe propose a novel multi-view camera pipeline for the reconstruction and registration of dynamic clothing.\nOur proposed method relies on a specifically design ed pattern that allows for precise video tracking in each camera view. \nW e triangulate the tracked points and register the cloth surface in a ...\n \n---------------------\nDisentangled Image Colorization via Global Anchor s\n\nXia, Hu, Wong, Wang\n\nColorization is multimodal by nature and chall enges existing frameworks to achieve colorful and structurally consistent results. Even the sophisticated autoregressive model struggles to maintain long-distance color consistency due to the fragility of sequential depend ence.\nTo overcome this challenge...\n\n---------------------\nLearning to Relight Portrait Images via a Virtual Light Stage and Synthetic-to-Real A daptation\n\nYeh, Nagano, Khamis, Kautz, Liu...\n\nGiven a portrait image of a person and an environment map of the target lighting, portrait religh ting aims to re-illuminate the person in the image as if the person appear ed in an environment with the target lighting. To achieve high-quality res ults, recent methods rely on deep learning. An effectiv...\n\n------------ ---------\nEfficient Light Probes for Real-time Global Illumination\n\nGuo , Zong, Song, Fu, Tao...\n\nReproducing physically-based global illuminati on (GI) effects has been a long-standing demand for many real-time graphic al applications. In pursuit of this goal, many recent engines resort to so me form of light probes baked in a precomputation stage. Unfortunately, th e GI effects stemming from the p...\n\n---------------------\nRapid Face A sset Acquisition with Recurrent Feature Alignment\n\nLiu, Cai, Chen, Zhou, Zhao\n\nWe present Recurrent Feature Alignment (ReFA), an end-to-end neur al network for the very rapid creation of production-grade face assets fro m multi-view images. ReFA is on par with the industrial pipelines in quali ty for producing accurate, complete, registered, and textured assets direc tly applicabl...\n\n---------------------\nNeural Brushstroke Engine: Lear ning a Latent Style Space of Interactive Drawing Tools\n\nShugrina, Li, Fi dler\n\nWe propose Neural Brushstroke Engine, the first method to apply de ep generative models\nto learn a distribution of interactive drawing tools . \nOur conditional GAN model learns the latent \nspace of drawing styles from a small set (about 200) of unlabeled images in different media.\nOnce trained, a sing...\n\n---------------------\nDeepMVSHair: Deep Hair Model ing from Sparse Views\n\nKuang, Chen, Fu, Zhou, Zheng\n\nWe present DeepMV SHair, the first deep learning-based method for multi-view hair strand rec onstruction. The key component of our pipeline is HairMVSNet, a differenti able neural architecture which represents a spatial hair structure as a co ntinuous 3D hair growing direction field implicitly. Specific...\n\n------ ---------------\nAsynchronous Collaborative Autoscanning with Mode Switchi ng for Multi-Robot Scene Reconstruction\n\nGuo, Li, Xia, Hu, Liu\n\nWhen c onducting autonomous scanning for the online reconstruction of unknown ind oor environments, robots have to be competent at exploring the scene struc ture and reconstructing objects with high quality. Our key observation is that different tasks demand specialized scanning properties of robots: r.. .\n\n---------------------\n3QNet: 3D Point Cloud Geometry Quantization Co mpression Network\n\nHuang, zhang, Chen, Ding, Tai...\n\nSince the develop ment of 3D applications, the point cloud, as a spatial description easily acquired by sensors, has been widely used in multiple areas such as SLAM a nd 3D reconstruction. Point Cloud Compression (PCC) has also attracted mor e attention as a primary step before point cloud transferring...\n\n------ ---------------\nProduction-Ready Face Re-Aging for Visual Effects\n\nZoss , Chandran, Sifakis, Gross, Gotardo...\n\nPhotorealistic digital re-aging of faces in video is becoming increasingly common in entertainment and adv ertising. But the predominant 2D painting workflow often requires frame-b y-frame manual work that can take days to accomplish, even by skilled arti sts. Although research on facial image re-agi...\n\n--------------------- \nA Neural Galerkin Solver for Accurate Surface Reconstruction\n\nHuang, C hen, Hu\n\nTo reconstruct meshes from the widely-available 3D point cloud data, implicit shape representation is among the primary choices as an int ermediate form due to its superior representation power and robustness in topological optimizations. Although different parameterizations of the imp licit fields ha...\n\n---------------------\nGloss management for consiste nt reproduction of real and virtual objects\n\nChen, Piovarci, Wang, Seide l, Didyk...\n\nA good match of material appearance between real-world obje cts and their digital on-screen representations is critical for many appli cations such as fabrication, design, and e-commerce. However, faithful app earance reproduction is challenging, especially for complex phenomena, suc h as gloss. In most...\n\n---------------------\nPhysical Interaction: Rec onstructing Hand-object Interactions with Physics\n\nHu, Yi, Zhang, Yong, Xu\n\nSingle view-based reconstruction of hand-object interaction is chall enging due to the severe observation missing caused by occlusions. This pa per proposes a physics-based method to better solve the ambiguities in the reconstruction. It first proposes a force-based dynamic model of the in-h and object...\n\n---------------------\nImplicit Conversion of Manifold B- Rep Solids by Neural Halfspace Representation\n\nGuo, Liu, Pan, Guo\n\nWe present a novel implicit representation --- neural halfspace representatio n (NH-Rep), to convert manifold B-Rep solids to implicit representations. NH-Rep is a Boolean tree built on a set of implicit functions represented by the neural network, and the composite Boolean function is capable of re p...\n\n---------------------\nScalable multi-class sampling via filtered sliced optimal transport\n\nSALAUN, Geogiev, Seidel, Singh\n\nWe propose a continuous domain formulation of Wasserstein barycenters\nfor multi-class (-purpose) point set optimization. Our formulation is sys-\ntematically d erived to handle hundreds to thousands of classes for different\nsampling applications. We develop a practical optimization scheme that is\nclos...\ n\n---------------------\nShape Completion with Points in the Shadow\n\nZh ang, Zhao, Wang, Hu\n\nSingle-view point cloud completion aims to recover the full geometry of an object based on only limited observation, which is extremely hard due to the data sparsity and occlusion. The core challenge is to generate plausible geometries to fill the unobserved part of the ob ject based on a partial sca...\n\n---------------------\nNeural Point Cata caustics for Novel-View Synthesis of Reflections\n\nKopanas, Leimkühler, R ainer, Jambon, Drettakis\n\nView-dependent effects such as reflections pos e a substantial challenge for image-based and neural rendering algorithms. Above all, curved reflectors are particularly hard, as they lead to highl y non-linear reflection flows as the camera moves. We introduce Neural Poi nt Catacaustics, a new point-bas...\n\n---------------------\nVideo-driven Neural Physically-based Facial Asset for Production\n\nZhang, Zeng, Zhang , Lin, Cao...\n\nProduction-level workflows for producing convincing 3D dy namic human faces have long relied on an assortment of labor-intensive too ls for geometry and texture generation, motion capture and rigging, and ex pression synthesis. Recent neural approaches automate individual component s but the correspondi...\n\n---------------------\nIsotropic ARAP energy u sing Cauchy-Green invariants\n\nLin, Chitalu, Komura\n\nIsotropic As-Rigid -As-Possible (ARAP) energy has been popular for shape editing, mesh parame trisation and soft-body simulation for almost two decades. However, a form ulation using Cauchy-Green (CG) invariants has always been unclear, due to a rotation-polluted trace term that cannot be directly expr...\n\n------- --------------\nAssemble Them All: Physics-Based Planning for Generalizabl e Assembly by Disassembly\n\nTian, Xu, Li, Luo, Sueda...\n\nAssembly plann ing is the core of automating product assembly, maintenance, and recycling for modern industrial manufacturing. Despite its importance and long hist ory of research, planning for mechanical assemblies when given the final a ssembled state remains a challenging problem. This is due to the...\n\n--- ------------------\nThe Continuity of Locomotion: Rethinking Conventions f or Locomotion and Its Visualization in Shared Virtual Reality Spaces\n\nFr eiwald, Schmidt, Riecke, Steinicke\n\nNatural interaction between multiple users within a shared virtual environment (VE) relies on each others' awa reness of the current position of the interaction partners. This, however, cannot be warranted when users employ noncontinuous locomotion techniques , such as teleportation, which may cause c...\n\n---------------------\nLa youtEnhancer: Generating Good Indoor Layouts from Imperfect Data\n\nLeimer , Guerrero, Weiss, Musialski\n\nWe address the problem of indoor layout sy nthesis, which is a topic of continuing research interest in computer grap hics. The newest works made significant progress using data-driven generat ive methods; however, these approaches rely on suitable datasets. In pract ice, some desirable layout propertie...\n\n---------------------\nLook-Ahe ad Training with Learned Reflectance Loss for Single-Image SVBRDF Estimati on\n\nZhou, Kalantari\n\nIn this paper, we propose a novel optimization-ba sed method to estimate the reflectance properties of a near planar surface from a single input image. Specifically, we perform test-time optimizatio n by directly updating the parameters of a neural network to minimize the test error. Since single imag...\n\n---------------------\nTileGen: Tileab le, Controllable Material Generation and Capture\n\nZhou, Hasan, Deschaint re, Guerrero, Sunkavalli...\n\nRecent methods (e.g. MaterialGAN) have used unconditional GANs to generate per-pixel material maps, or as a prior to reconstruct materials from input photographs. These models can generate va ried random material appearance, but do not have any mechanism to constrai n the generated material to a speci...\n\n---------------------\nStitch it in Time: GAN-Based Facial Editing of Real Videos\n\nTzaban, Mokady, Gal, Bermano, Cohen-Or\n\nThe ability of Generative Adversarial Networks to enc ode rich semantics within their latent space has been widely adopted for f acial image editing. However, replicating their success with videos has pr oven challenging. Applying StyleGAN editing over real videos introduces tw o main challenges: (i) St...\n\n---------------------\nNeuralMarker: A Fra mework for Learning General Marker Correspondence\n\nHuang, Pan, Pan, Bian , Xu...\n\nWe tackle the problem of estimating correspondences from a gene ral marker, such as a movie poster, to an image that captures such a marke r. Conventionally, this problem is addressed by fitting a homography model based on sparse feature matching. However, they are only able to handle p lane-like marke...\n\n---------------------\nDifferentiable Point-Based Ra diance Fields for Efficient View Synthesis\n\nZhang, Baek, Rusinkiewicz, H eide\n\nWe propose a differentiable rendering algorithm for efficient nove l view synthesis. By departing from volume-based representations in favor of a learned point representation, we improve on existing methods more tha n an order of magnitude in memory and runtime, both in training and infere nce. The met...\n\n---------------------\nExact 3D Path Generation via 3D Cam-Linkage Mechanisms\n\nCheng, Song, Lu, Chew, Liu\n\nExact 3D path gene ration is a fundamental problem of designing a mechanism to make a point e xactly move along a prescribed 3D path, driven by a single actuator. Exist ing mechanisms are insufficient to address this problem. Planar linkages a nd their combinations with gears and/or plate cams can only ...\n\n------- --------------\nDifferentiable Rendering of Neural SDFs through Reparamete rization\n\nBangaru, Gharbi, Luan, Li, Sunkavalli...\n\nWe present a metho d to automatically compute correct gradients with respect to geometric sce ne parameters in neural SDF renderers. Recent physically-based differentia ble rendering techniques for meshes have used edge-sampling to handle disc ontinuities, particularly at object silhouettes, but SDFs do...\n\n------- --------------\nFast Stabilization of Inducible Magnet Simulation\n\nKim, Han\n\nThis paper presents a novel method for simulating inducible rigid m agnets efficiently and stably. In the proposed method, inducible magnets a re magnetized by a modified magnetization dynamics, so that the magnetic e quilibrium can be obtained in a computationally efficient manner. Furtherm ore, our mo...\n\n---------------------\nUniColor: A Unified Framework for Multi-Modal Colorization with Transformer\n\nHuang, Zhao, Liao\n\nWe prop ose the first unified framework UniColor to support colorization in multip le modalities, including both unconditional and conditional ones, such as stroke, exemplar, text, and even a mix of them. Rather than learning a sep arate model for each type of condition, we introduce a two-stage colori... \n\n---------------------\nDeep Adaptive Sampling and Reconstruction using Analytic Distributions\n\nSalehi, Manzi, Roethlin, Schroers, Weber...\n\n We propose an adaptive sampling and reconstruction method for offline Mont e Carlo rendering. Our method produces sampling maps constrained by a user -defined budget that minimize the expected future denoising error. Compare d to other state-of-the-art methods, which produce the necessary training data...\n\n---------------------\nLightweight Neural Basis Functions for A ll-Frequency Shading\n\nXu, Zeng, Wu, Wang, Yan\n\nBasis functions provide both the abilities for compact representation and the properties for effi cient computation. Therefore, they are pervasively used in rendering to pe rform all-frequency shading. However, common basis functions, including sp herical harmonics (SH), wavelets, and spherical Gaussia...\n\n------------ ---------\nHigh-Order Directional Fields\n\nBoksebeld, Vaxman\n\nWe introd uce a framework for representing face-based directional fields of an arbit rary piecewise-polynomial order. Our framework is based on a primal-dual d ecomposition of fields, where the exact component of a field is the gradie nt of piecewise-polynomial conforming function, and the coexact compo...\n \n---------------------\nFast Editing of Singularities in Field-Aligned St ripe Patterns\n\nNoma, Umetani, Kawahara\n\n\textit{Field-aligned parametr ization} is a method that maps a scalar function onto a surface, such that the gradient vector of the scalar function matches the input vector field . Using this idea, one can produce a stripe pattern that is convenient for various purposes such as remeshing, texture syn...\n\n------------------- --\nAugmented Paths and Reodesics for Topologically-Stable Matching\n\nSah illioğlu, Horsman\n\nWe present a fully-automatic 3D isometric shape corre spondence method that computes point-to-point dense correspondences in the presence of topological noise. In order to make this process work well, w e introduce two novel concepts: path augmentation with heat vectors and ro bust geodesics, reodesics...\n\n---------------------\nDifferentiable Hybr id Traffic Simulation\n\nSon, Qiao, Sewall, Lin\n\nWe introduce a novel di fferentiable hybrid traffic simulator, which simulates traffic using a hyb rid model of both macroscopic and microscopic models and can be directly i ntegrated into a neural network for traffic control and flow optimization. This is the first differentiable traffic simulator for...\n\n------------ ---------\nNeRFFaceEditing: Disentangled Face Editing in Neural Radiance F ields\n\nJiang, Chen, Liu, Fu, Gao\n\nRecent methods for synthesizing 3D-a ware face images have achieved rapid development thanks to neural radiance fields, allowing for high quality and fast inference speed. However, exis ting solutions for editing facial geometry and appearance independently us ually require retraining and are not optim...\n\n---------------------\nWa ter Simulation and Rendering from a Still Photograph\n\nSugimoto, He, Liao , Sander\n\nWe propose an approach to simulate and render realistic water animation from a single still input photograph. We first segment the water surface, estimate rendering parameters, and compute water reflection text ures with a combination of neural networks and traditional optimization te chniques. Then w...\n\n---------------------\nControl VAE: Model-Based Lea rning of Generative Controllers for Physics-Based Characters\n\nYao, Song, Chen, Liu\n\nIn this paper, we introduce Control VAE, a novel model-based framework for learning generative motion control policies based on variat ional autoencoders (VAE). Our framework can learn a rich and flexible late nt representation of skills and a skill-conditioned generative control pol icy from a diverse...\n\n---------------------\nElastoMonolith: A Monolith ic Optimization-based Liquid Solver for Contact-Aware Elastic-Solid Coupli ng\n\nTakahashi, Batty\n\nSimultaneous coupling of diverse physical system s poses significant computational challenges in terms of speed, quality, a nd stability. Rather than treating all components with a single discretiza tion methodology (e.g., smoothed particles, material point method, Euleria n grid, etc.) that is ill-suit...\n\n---------------------\nContinuous def ormation based panelization for design rationalization\n\nJadon, Thomaszew sky, Apolinarska, Poranne\n\nDesign rationalization is the process of simp lifying a 3D shape to enable cost-efficient manufacturing. A common approa ch is to approximate the input shape by a collection of simple units, suc h as flat or spherical panels, that are easy to manufacture and simple to assemble. This panelization proce...\n\n---------------------\nNeural Phot o-Finishing\n\nTseng, Zhang, Jebe, Zhang, Xia...\n\nImage processing pipel ines are ubiquitous and we rely on them either directly, by filtering or a djusting an image post-capture, or indirectly, as image signal processing pipeline (ISP) on broadly deployed camera systems. Used by artists, photog raphers, system engineers, and for downstream vision tas...\n\n----------- ----------\nNeural James-Stein Combiner for Unbiased and Biased Renderings \n\nGu, Iglesias-Guitian, Moon\n\nUnbiased rendering algorithms such as pa th tracing produce accurate images given an infinite number of samples, bu t in practice, the techniques often leave visually distracting artifacts ( i.e., noise) in their rendered images due to a limited time budget. A favo red approach for mitigating the noise ...\n\n---------------------\nMIPNet : Neural Normal-to-Anisotropic-Roughness MIP mapping\n\nGauthier, Faury, L evallois, Thonat, Thiery...\n\nWe present MIPNet, a novel approach for SVB RDF mipmapping which preserves material appearance under varying view dist ances and lighting conditions. As in classical mipmapping, our method expl icitly encodes the multiscale appearance of materials in a SVBRDF mipmap p yramid. To do so, we use a tensor-b...\n\n---------------------\nAn Implic it Parametric Morphable Dental Model\n\nZhang, Elgharib, Fox, Gu, Theobalt ...\n\n3D Morphable models of the human body capture variations among subj ects and are useful in reconstruction and editing applications. Current de ntal models use an explicit mesh scene representation and model only the t eeth, ignoring the gum. In this work, we present the first parametric 3D m orphable de...\n\n---------------------\nEfficient Neural Style Transfer f or Volumetric Simulations\n\nAurand, Ortiz, Nauer, C. Azevedo\n\nArtistica lly controlling fluids has always been a challenging task. Recently, volum etric Neural Style Transfer (NST) techniques have been used to artisticall y manipulate smoke simulation data with 2D images. In this work, we revisi t previous volumetric NST techniques for smoke, proposing a suite of ...\n \n---------------------\nFDNeRF: Few-shot Dynamic Neural Radiance Fields f or Face Reconstruction and Expression Editing\n\nZHANG, LI, WAN, WANG, LIA O\n\nWe propose a Few-shot Dynamic Neural Radiance Field (FDNeRF), the fir st NeRF-based method capable of reconstruction and expression editing of 3 D faces based on a small number of dynamic images. Unlike existing dynamic NeRFs that require dense images as input and can only be modeled for a si ngle iden...\n\n---------------------\nNeuLighting: Neural Lighting for Fr ee Viewpoint Outdoor Scene Relighting with Unconstrained Photo Collections \n\nLi, Guo, Fei, Li, Guo\n\nWe propose NeuLighting, a new framework for f ree viewpoint outdoor scene relighting from a sparse set of unconstrained in-the-wild photo collections. Our framework represents all the scene comp onents as continuous functions parameterized by MLPs that take a 3D locati on and the lighting condition as ...\n\n\nRegistration Category: FULL ACCE SS, EXPERIENCE PLUS ACCESS, EXPERIENCE ACCESS, TRADE EXHIBITOR\n\nLanguage : ENGLISH\n\nFormat: IN-PERSON END:VEVENT END:VCALENDAR