Language-guided 3D scene understanding is important for advancing applications in robotics, AR/VR, and human-computer interaction, enabling models to comprehend and interact with 3D environments through natural language. While 2D vision-language models (VLMs) have achieved remarkable success in 2D VQA tasks, progress in the 3D domain has been significantly slower due to the complexity of 3D data and the high cost of manual annotations. In this work, we introduce SplatTalk, a novel method that uses a generalizable 3D Gaussian Splatting (3DGS) framework to produce 3D tokens suitable for direct input into a pretrained LLM, enabling effective zero-shot 3D visual question answering (3D VQA) for scenes with only posed images. During experiments on multiple benchmarks, our approach outperforms both 3D models trained specifically for the task and previous 2D-LMM-based models utilizing only images (our setting), while achieving competitive performance with state-of-the-art 3D LMMs that additionally utilize 3D inputs. Project website: https://splat-talk.github.io/

We present a generative approach to forecast long-term future human behavior in 3D, requiring only weak supervision from readily available 2D human action data. This is a fundamental task enabling many downstream applications. The required ground-truth data is hard to capture in 3D (mocap suits, expensive setups) but easy to acquire in 2D (simple RGB cameras). Thus, we design our method to only require 2D RGB data at inference time while being able to generate 3D human motion sequences. We use a differentiable 2D projection scheme in an autoregressive manner for weak supervision, and an adversarial loss for 3D regularization. Our method predicts long and complex human behavior sequences (e.g., cooking, assembly) consisting of multiple sub-actions. We tackle this in a semantically hierarchical manner, jointly predicting high-level coarse action labels together with their low-level fine-grained real-izations as characteristic 3D human poses. We observe that these two action representations are coupled in nature, and joint prediction benefits both action and pose forecasting. Our experiments demonstrate the complementary nature of joint action and 3D pose prediction: our joint approach outperforms each task treated individually, enables robust longer-term sequence prediction, and improves over alter-native approaches to forecast actions and characteristic 3D poses.

Neural Radiance Fields (NeRFs) have demonstrated amazing ability to synthesize images of 3D scenes from novel views. However, they rely upon specialized volumetric rendering algorithms based on ray marching that are mismatched to the capabilities of widely deployed graphics hardware. This paper introduces a new NeRF representation based on textured polygons that can synthesize novel images efficiently with standard rendering pipelines. The NeRF is represented as a set of polygons with textures representing binary opacities and feature vectors. Traditional rendering of the polygons with a z-buffer yields an image with features at every pixel, which are interpreted by a small, view-dependent MLP running in a fragment shader to produce a final pixel color. This approach enables NeRFs to be rendered with the traditional polygon rasterization pipeline, which provides massive pixel-level parallelism, achieving interactive frame rates on a wide range of compute platforms, including mobile phones. Project page: https://mobile-nerf.github.io

Traditional 3D scene understanding approaches rely on labeled 3D datasets to train a model for a single task with supervision. We propose OpenScene, an alternative approach where a model predicts dense features for 3D scene points that are co-embedded with text and image pixels in CLIP feature space. This zero-shot approach enables task-agnostic training and open-vocabulary queries. For example, to perform SOTA zero-shot 3D semantic segmentation it first infers CLIP features for every 3D point and later classifies them based on similarities to embeddings of arbitrary class labels. More interestingly, it enables a suite of open-vocabulary scene understanding applications that have never been done before. For example, it allows a user to enter an arbitrary text query and then see a heat map indicating which parts of a scene match. Our approach is effective at identifying objects, materials, affordances, activities, and room types in complex 3D scenes, all using a single model trained without any labeled 3D data.

We address efficient and structure-aware 3D scene representation from images. Nerflets are our key contribution-a set of local neural radiance fields that together represent a scene. Each nerflet maintains its own spatial position, orientation, and extent, within which it contributes to panoptic, density, and radiance reconstructions. By leveraging only photometric and inferred panoptic image supervision, we can directly and jointly optimize the parameters of a set of nerflets so as to form a decomposed representation of the scene, where each object instance is represented by a group of nerflets. During experiments with indoor and outdoor environments, we find that nerflets: (1) fit and approximate the scene more efficiently than traditional global NeRFs, (2) allow the extraction of panoptic and photometric renderings from arbitrary views, and (3) enable tasks rare for NeRFs, such as 3D panoptic segmentation and interactive editing. Our project page.

Removing clutter from scenes is essential in many applications, ranging from privacy-concerned content filtering to data augmentation. In this work, we present an automatic system that removes clutter from 3D scenes and inpaints with coherent geometry and texture. We propose techniques for its two key components: 3D segmentation based on shared properties and 3D inpainting, both of which are important problems. The definition of 3D scene clutter (frequently-moving objects) is not well captured by commonly-studied object categories in computer vision. To tackle the lack of well-defined clutter annotations, we group noisy fine-grained labels, leverage virtual rendering, and impose an instance-level area-sensitive loss. Once clutter is removed, we inpaint geometry and texture in the resulting holes by merging inpainted RGB-D images. This requires novel voting and pruning strategies that guarantee multi-view consistency across individually inpainted images for mesh reconstruction. Experiments on ScanNet and Matterport3D dataset show that our method outperforms baselines for clutter segmentation and 3D in-painting, both visually and quantitatively. Project page: https://weify627.github.io/clutter/.

For a robot to personalize physical assistance effectively, it must learn user preferences that can be generally reapplied to future scenarios. In this work, we investigate personalization of household cleanup with robots that can tidy up rooms by picking up objects and putting them away. A key challenge is determining the proper place to put each object, as people's preferences can vary greatly depending on personal taste or cultural background. For instance, one person may prefer storing shirts in the drawer, while another may prefer them on the shelf. We aim to build systems that can learn such preferences from just a handful of examples via prior interactions with a particular person. We show that robots can combine language-based planning and perception with the few-shot summarization capabilities of large language models (LLMs) to infer generalized user preferences that are broadly applicable to future interactions. This approach enables fast adaptation and achieves 91.2% accuracy on unseen objects in our benchmark dataset. We also demonstrate our approach on a real-world mobile manipulator called TidyBot, which successfully puts away 85.0% of objects in real-world test scenarios.

We propose the task of forecasting characteristic 3d poses: from a short sequence observation of a person, predict a future 3d pose of that person in a likely action-defining, characteristic pose - for instance, from observing a person picking up an apple, predict the pose of the person eating the apple. Prior work on human motion prediction estimates future poses at fixed time intervals. Although easy to define, this frame-by-frame formulation confounds temporal and intentional aspects of human action. Instead, we define a semantically meaningful pose prediction task that decouples the predicted pose from time, taking inspiration from goal-directed behavior. To predict characteristic poses, we propose a probabilistic approach that models the possible multimodality in the distribution of likely characteristic poses. We then sample future pose hypotheses from the predicted distribution in an autoregressive fashion to model dependencies between joints. To evaluate our method, we construct a dataset of manually annotated characteristic 3d poses. Our experiments with this dataset suggest that our proposed probabilistic approach outperforms state-of-the-art methods by 26% on average.

We present Panoptic Neural Fields (PNF), an object-aware neural scene representation that decomposes a scene into a set of objects (things) and background (stuff). Each object is represented by an oriented 3D bounding box and a multi-layer perceptron (MLP) that takes position, direction, and time and outputs density and radiance. The background stuff is represented by a similar MLP that additionally outputs semantic labels. Each object MLPs are instance-specific and thus can be smaller and faster than previous object-aware approaches, while still leveraging category-specific priors incorporated via meta-learned initialization. Our model builds a panoptic radiance field representation of any scene from just color images. We use off-the-shelf algorithms to predict camera poses, object tracks, and 2D image semantic segmentations. Then we jointly optimize the MLP weights and bounding box parameters using analysis-by-synthesis with self-supervision from color images and pseudo-supervision from predicted semantic segmentations. During experiments with real-world dynamic scenes, we find that our model can be used effectively for several tasks like novel view synthesis, 2D panoptic segmentation, 3D scene editing, and multiview depth prediction.

The goal of this work is to perform 3D reconstruction and novel view synthesis from data captured by scanning platforms commonly deployed for world mapping in urban outdoor environments (e.g., Street View). Given a sequence of posed RGB images and lidar sweeps acquired by cameras and scanners moving through an outdoor scene, we produce a model from which 3D surfaces can be extracted and novel RGB images can be synthesized. Our approach extends Neural Radiance Fields, which has been demonstrated to synthesize realistic novel images for small scenes in controlled settings, with new methods for leveraging asynchronously captured lidar data, for addressing exposure variation between captured images, and for leveraging predicted image segmentations to supervise densities on rays pointing at the sky. Each of these three extensions provides significant performance improvements in experiments on Street View data. Our system produces state-of-the-art 3D surface reconstructions and synthesizes higher quality novel views in comparison to both traditional methods (e.g. COLMAP) and recent neural representations (e.g. Mip-NeRF).

A classical problem in computer vision is to infer a 3D scene representation from few images that can be used to render novel views at interactive rates. Previous work focuses on reconstructing pre-defined 3D representations, e.g. textured meshes, or implicit representations, e.g. radiance fields, and often requires input images with precise camera poses and long processing times for each novel scene. In this work, we propose the Scene Representation Transformer (SRT), a method which processes posed or unposed RGB images of a new area, infers a “set-latent scene representation ”, and synthesises novel views, all in a single feed-forward pass. To calculate the scene representation, we propose a generalization of the Vision Transformer to sets of images, enabling global information integration, and hence 3D reasoning. An efficient decoder transformer parameterizes the light field by attending into the scene representation to render novel views. Learning is supervised end-to-end by minimizing a novel-view reconstruction error. We show that this method outperforms recent baselines in terms of PSNR and speed on synthetic datasets, including a new dataset created for the paper. Further, we demonstrate that SRT scales to support interactive visualization and semantic segmentation of real-world outdoor environments using Street View imagery.

Neural fields have emerged as a new paradigm for representing signals, thanks to their ability to do it compactly while being easy to optimize. In most applications, however, neural fields are treated like a black box, which precludes many signal manipulation tasks. In this paper, we propose a new class of neural fields called basis-encoded polynomial neural fields (PNFs). The key advantage of a PNF is that it can represent a signal as a composition of a number of manipulable and interpretable components without losing the merits of neural fields representation. We develop a general theoretical framework to analyze and design PNFs. We use this framework to design Fourier PNFs, which match state-of-the-art performance in signal representation tasks that use neural fields. In addition, we empirically demonstrate that Fourier PNFs enable signal manipulation applications such as texture transfer and scale-space interpolation. Code is available at https://github.com/stevenygd/PNF.

We introduce neural dual contouring (NDC), a new data-driven approach to mesh reconstruction based on dual contouring (DC). Like traditional DC, it produces exactly one vertex per grid cell and one quad for each grid edge intersection, a natural and efficient structure for reproducing sharp features. However, rather than computing vertex locations and edge crossings with hand-crafted functions that depend directly on difficult-to-obtain surface gradients, NDC uses a neural network to predict them. As a result, NDC can be trained to produce meshes from signed or unsigned distance fields, binary voxel grids, or point clouds (with or without normals); and it can produce open surfaces in cases where the input represents a sheet or partial surface. During experiments with five prominent datasets, we find that NDC, when trained on one of the datasets, generalizes well to the others. Furthermore, NDC provides better surface reconstruction accuracy, feature preservation, output complexity, triangle quality, and inference time in comparison to previous learned (e.g., neural marching cubes, convolutional occupancy networks) and traditional (e.g., Poisson) methods. Code and data are available at https://github.com/czq142857/NDC.

With the recent growth of urban mapping and autonomous driving efforts, there has been an explosion of raw 3D data collected from terrestrial platforms with lidar scanners and color cameras. However, due to high labeling costs, ground-truth 3D semantic segmentation annotations are limited in both quantity and geographic diversity, while also being difficult to transfer across sensors. In contrast, large image collections with ground-truth semantic segmentations are readily available for diverse sets of scenes. In this paper, we investigate how to use only those labeled 2D image collections to supervise training 3D semantic segmentation models. Our approach is to train a 3D model from pseudo-labels derived from 2D semantic image segmentations using multi-view fusion. We address several novel issues with this approach, including how to select trusted pseudo-labels, how to sample 3D scenes with rare object categories, and how to decouple input features from 2D images from pseudo-labels during training. The proposed network architecture, 2D3DNet, achieves significantly better performance (+6.2-11.4 mIoU) than baselines during experiments on a new urban dataset with lidar and images captured in 20 cities across 5 continents.

Localizing the camera in a known indoor environment is a key building block for scene mapping, robot navigation, AR, etc. Recent advances estimate the camera pose via optimization over the 2D/3D-3D correspondences established between the coordinates in 2D/3D camera space and 3D world space. Such a mapping is estimated with either a convolution neural network or a decision tree using only the static input image sequence, which makes these approaches vulnerable to dynamic indoor environments that are quite common yet challenging in the real world. To address the aforementioned issues, in this paper, we propose a novel outlier-aware neural tree which bridges the two worlds, deep learning and decision tree approaches. It builds on three important blocks: (a) a hierarchical space partition over the indoor scene to construct the decision tree; (b) a neural routing function, implemented as a deep classification network, employed for better 3D scene understanding; and (c) an outlier rejection module used to filter out dynamic points during the hierarchical routing process. Our proposed algorithm is evaluated on the RIO-10 benchmark developed for camera relocalization in dynamic indoor environments. It achieves robust neural routing through space partitions and outperforms the state-of-the-art approaches by around 30% on camera pose accuracy, while running comparably fast for evaluation.

We present a method that synthesizes novel views of complex scenes by interpolating a sparse set of nearby views. The core of our method is a network architecture that includes a multilayer perceptron and a ray transformer that estimates radiance and volume density at continuous 5D locations (3D spatial locations and 2D viewing directions), drawing appearance information on the fly from multiple source views. By drawing on source views at render time, our method hearkens back to classic work on image-based rendering (IBR), and allows us to render high-resolution imagery. Unlike neural scene representation work that optimizes per-scene functions for rendering, we learn a generic view interpolation function that generalizes to novel scenes. We render images using classic volume rendering, which is fully differentiable and allows us to train using only multi-view posed images as supervision. Experiments show that our method outperforms recent novel view synthesis methods that also seek to generalize to novel scenes. Further, if fine-tuned on each scene, our method is competitive with state-of-the-art single-scene neural rendering methods.1

We study an unsupervised domain adaptation problem for the semantic labeling of 3D point clouds, with a particular focus on domain discrepancies induced by different LiDAR sensors. Based on the observation that sparse 3D point clouds are sampled from 3D surfaces, we take a Complete and Label approach to recover the underlying surfaces before passing them to a segmentation network. Specifically, we design a Sparse Voxel Completion Network (SVCN) to complete the 3D surfaces of a sparse point cloud. Unlike semantic labels, to obtain training pairs for SVCN requires no manual labeling. We also introduce local adversarial learning to model the surface prior. The recovered 3D surfaces serve as a canonical domain, from which semantic labels can transfer across different LiDAR sensors. Experiments and ablation studies with our new benchmark for cross-domain semantic labeling of LiDAR data show that the proposed approach provides 6.3-37.6% better performance than previous domain adaptation methods.

We introduce Multiresolution Deep Implicit Functions (MDIF), a hierarchical representation that can recover fine geometry detail, while being able to perform global operations such as shape completion. Our model represents a complex 3D shape with a hierarchy of latent grids, which can be decoded into different levels of detail and also achieve better accuracy. For shape completion, we propose latent grid dropout to simulate partial data in the latent space and therefore defer the completing functionality to the decoder side. This along with our multires design significantly improves the shape completion quality under decoder-only latent optimization. To the best of our knowledge, MDIF is the first deep implicit function model that can at the same time (1) represent different levels of detail and allow progressive decoding; (2) support both encoder-decoder inference and decoder-only latent optimization, and fulfill multiple applications; (3) perform detailed decoder-only shape completion. Experiments demonstrate its superior performance against prior art in various 3D reconstruction tasks.

This paper proposes a method for representation learning of multimodal data using contrastive losses. A traditional approach is to contrast different modalities to learn the information shared among them. However, that approach could fail to learn the complementary synergies between modalities that might be useful for downstream tasks. Another approach is to concatenate all the modalities into a tuple and then contrast positive and negative tuple correspondences. However, that approach could consider only the stronger modalities while ignoring the weaker ones. To address these issues, we propose a novel contrastive learning objective, TupleInfoNCE. It contrasts tuples based not only on positive and negative correspondences, but also by composing new negative tuples using modalities describing different scenes. Training with these additional negatives encourages the learning model to examine the correspondences among modalities in the same tuple, ensuring that weak modalities are not ignored. We provide a theoretical justification based on mutual-information for why this approach works, and we propose a sample optimization algorithm to generate positive and negative samples to maximize training efficacy. We find that TupleInfoNCE significantly outperforms previous state of the arts on three different downstream tasks.

The ability to communicate intention enables decentralized multi-agent robots to collaborate while performing physical tasks. In this work, we present spatial intention maps, a new intention representation for multi-agent vision-based deep reinforcement learning that improves coordination between decentralized mobile manipulators. In this representation, each agent’s intention is provided to other agents, and rendered into an overhead 2D map aligned with visual observations. This synergizes with the recently proposed spatial action maps framework, in which state and action representations are spatially aligned, providing inductive biases that encourage emergent cooperative behaviors requiring spatial coordination, such as passing objects to each other or avoiding collisions. Experiments across a variety of multi-agent environments, including heterogeneous robot teams with different abilities (lifting, pushing, or throwing), show that incorporating spatial intention maps improves performance for different mobile manipulation tasks while significantly enhancing cooperative behaviors.

3D object detection is a key module for safety-critical robotics applications such as autonomous driving. For these applications, we care most about how the detections affect the ego-agent’s behavior and safety (the egocentric perspective). Intuitively, we seek more accurate descriptions of object geometry when it’s more likely to interfere with the ego-agent’s motion trajectory. However, current detection metrics, based on box Intersection-over-Union (IoU), are object-centric and aren’t designed to capture the spatio-temporal relationship between objects and the ego-agent. To address this issue, we propose a new egocentric measure to evaluate 3D object detection, namely Support Distance Error (SDE). Our analysis based on SDE reveals that the egocentric detection quality is bounded by the coarse geometry of the bounding boxes. Given the insight that SDE would benefit from more accurate geometry descriptions, we propose to represent objects as amodal contours, specifically amodal star-shaped polygons, and devise a simple model, StarPoly, to predict such contours. Our experiments on the large-scale Waymo Open Dataset show that SDE better reflects the impact of detection quality on the ego-agent’s safety compared to IoU; and the estimated contours from StarPoly consistently improve the egocentric detection quality over recent 3D object detectors.

Many applications in 3D shape design and augmentation require the ability to make specific edits to an object's semantic parameters (e.g., the pose of a person's arm or the length of an airplane's wing) while preserving as much existing details as possible. We propose to learn a deep network that infers the semantic parameters of an input shape and then allows the user to manipulate those parameters. The network is trained jointly on shapes from an auxiliary synthetic template and unlabeled realistic models, ensuring robustness to shape variability while relieving the need to label realistic exemplars. At testing time, edits within the parameter space drive deformations to be applied to the original shape, which provides semantically-meaningful manipulation while preserving the details. This is in contrast to prior methods that either use autoencoders with a limited latent-space dimensionality, failing to preserve arbitrary detail, or drive deformations with purely-geometric controls, such as cages, losing the ability to update local part regions. Experiments with datasets of chairs, airplanes, and human bodies demonstrate that our method produces more natural edits than prior work.

The goal of this project is to learn a 3D shape representation that enables accurate surface reconstruction, compact storage, efficient computation, consistency for similar shapes, generalization across diverse shape categories, and inference from depth camera observations. Towards this end, we introduce Local Deep Implicit Functions (LDIF), a 3D shape representation that decomposes space into a structured set of learned implicit functions. We provide networks that infer the space decomposition and local deep implicit functions from a 3D mesh or posed depth image. During experiments, we find that it provides 10.3 points higher surface reconstruction accuracy (F-Score) than the state-of-the-art (OccNet), while requiring fewer than 1\% of the network parameters. Experiments on posed depth image completion and generalization to unseen classes show 15.8 and 17.8 point improvements over the state-of-the-art, while producing a structured 3D representation for each input with consistency across diverse shape collections.

Realistic color texture generation is an important step in RGB-D surface reconstruction, but remains challenging in practice due to inaccuracies in reconstructed geometry, misaligned camera poses, and view-dependent imaging artifacts. In this work, we present a novel approach for color texture generation using a conditional adversarial loss obtained from weakly-supervised views. Specifically, we propose an approach to produce photorealistic textures for approximate surfaces, even from misaligned images, by learning an objective function that is robust to these errors. The key idea of our approach is to learn a patch-based conditional discriminator which guides the texture optimization to be tolerant to misalignments. Our discriminator takes a synthesized view and a real image, and evaluates whether the synthesized one is realistic, under a broadened definition of realism. We train the discriminator by providing as `real' examples pairs of input views and their misaligned versions - so that the learned adversarial loss will tolerate errors from the scans. Experiments on synthetic and real data under quantitative or qualitative evaluation demonstrate the advantage of our approach in comparison to state of the art (see Figure 1, right). Our code is publicly available1with video demonstration2.

Shape priors learned from data are commonly used to reconstruct 3D objects from partial or noisy data. Yet no such shape priors are available for indoor scenes, since typical 3D autoencoders cannot handle their scale, complexity, or diversity. In this paper, we introduce Local Implicit Grid Representations, a new 3D shape representation designed for scalability and generality. The motivating idea is that most 3D surfaces share geometric details at some scale - i.e., at a scale smaller than an entire object and larger than a small patch. We train an autoencoder to learn an embedding of local crops of 3D shapes at that size. Then, we use the decoder as a component in a shape optimization that solves for a set of latent codes on a regular grid of overlapping crops such that an interpolation of the decoded local shapes matches a partial or noisy observation. We demonstrate the value of this proposed approach for 3D surface reconstruction from sparse point observations, showing significantly better results than alternative approaches.

We propose DOPS, a fast single-stage 3D object detection method for LIDAR data. Previous methods often make domain-specific design decisions, for example projecting points into a bird-eye view image in autonomous driving scenarios. In contrast, we propose a general-purpose method that works on both indoor and outdoor scenes. The core novelty of our method is a fast, single-pass architecture that both detects objects in 3D and estimates their shapes. 3D bounding box parameters are estimated in one pass for every point, aggregated through graph convolutions, and fed into a branch of the network that predicts latent codes representing the shape of each detected object. The latent shape space and shape decoder are learned on a synthetic dataset and then used as supervision for the end-to-end training of the 3D object detection pipeline. Thus our model is able to extract shapes without access to ground-truth shape information in the target dataset. During experiments, we find that our proposed method achieves state-of-the-art results by~5% on object detection in ScanNet scenes, and it gets top results by 3.4% in the Waymo Open Dataset, while reproducing the shapes of detected cars.

We investigate whether a robot arm can learn to pick and throw arbitrary rigid objects into selected boxes quickly and accurately. Throwing has the potential to increase the physical reachability and picking speed of a robot arm. However, precisely throwing arbitrary objects in unstructured settings presents many challenges: from acquiring objects in grasps suitable for reliable throwing, to handling varying object-centric properties (e.g., mass distribution, friction, shape) and complex aerodynamics. In this work, we propose an end-to-end formulation that jointly learns to infer control parameters for grasping and throwing motion primitives from visual observations (RGB-D images of arbitrary objects in a bin) through trial and error. Within this formulation, we investigate the synergies between grasping and throwing (i.e., learning grasps that enable more accurate throws) and between simulation and deep learning (i.e., using deep networks to predict residuals on top of control parameters predicted by a physics simulator). The resulting system, TossingBot, is able to grasp and successfully throw arbitrary objects into boxes located outside its maximum reach range at 500+ mean picks per hour (600+ grasps per hour with 85% throwing accuracy); and generalizes to new objects and target locations.

We introduce, TextureNet, a neural network architecture designed to extract features from high-resolution signals associated with 3D surface meshes (e.g., color texture maps). The key idea is to utilize a 4-rotational symmetric(4-RoSy) field to define a domain for convolution on a surface. Though 4-RoSy fields have several properties favor-able for convolution on surfaces (low distortion, few singularities, consistent parameterization, etc.), orientations are ambiguous up to 4-fold rotation at any sample point. So, we introduce a new convolutional operator invariant to the4-RoSy ambiguity and use it in a network to extract features from high-resolution signals on geodesic neighborhoods of a surface. In comparison to alternatives, such as PointNet-based methods which lack a notion of orientation, the coherent structure given by these neighborhoods results in significantly stronger features. As an example application, we demonstrate the benefits of our architecture for 3D semantic segmentation of textured 3D meshes. The results show that our method outperforms all existing methods on the basis of mean IoU by a significant margin in both geometry-only(6.4%) and RGB+Geometry (6.9-8.2%) settings.

This paper addresses the task of estimating the light arriving from all directions to a 3D point observed at a selected pixel in an RGB image. This task is challenging because it requires predicting a mapping from a partial RGB observation by a camera to a complete illumination map for a different 3D point, which depends on the 3D location of the selected pixel, the distribution of unobserved light sources, the occlusions by scene geometry, etc. Previous methods attempt to learn this complex mapping directly using a single black-box neural network which often fails to estimate high-frequency lighting details for scenes with complicated 3D geometry. Instead, we propose "Neural Illumination," a new approach that decomposes illumination prediction into several simpler differentiable sub-tasks: 1) geometry estimation, 2) scene completion, and 3) LDR-to-HDR estimation. The advantage of this approach is that the sub-tasks are relatively easy to learn and can be trained with direct supervision, while the whole pipeline is fully differentiable and can be fine-tuned with end-to-end supervision. Experiments show that our approach performs significantly better quantitatively and qualitatively than prior work.

Template 3D shapes are useful for many tasks in graphics and vision, including fitting observation data, analyzing shape collections, and transferring shape attributes. Because of the variety of geometry and topology of real-world shapes, previous methods generally use a library of hand-made templates. In this paper, we investigate learning a general shape template from data. To allow for widely varying geometry and topology, we choose an implicit surface representation based on composition of local shape elements. While long known to computer graphics, this representation has not yet been explored in the context of machine learning for vision. We show that structured implicit functions are suitable for learning and allow a network to smoothly and simultaneously fit multiple classes of shapes. The learned shape template supports applications such as shape exploration, correspondence, abstraction, interpolation, and semantic segmentation from an RGB image.

In this work, we introduce the novel problem of identifying dense canonical 3D coordinate frames from a single RGB image. We observe that each pixel in an image corresponds to a surface in the underlying 3D geometry, where a canonical frame can be identified as represented by three orthogonal axes, one along its normal direction and two in its tangent plane. We propose an algorithm to predict these axes from RGB. Our first insight is that canonical frames computed automatically with recently introduced direction field synthesis methods can provide training data for the task. Our second insight is that networks designed for surface normal prediction provide better results when trained jointly to predict canonical frames, and even better when trained to also predict 2D projections of canonical frames. We conjecture this is because projections of canonical tangent directions often align with local gradients in images, and because those directions are tightly linked to 3Dcanonical frames through projective geometry and orthogonality constraints. In our experiments, we find that our method predicts 3D canonical frames that can be used in applications ranging from surface normal estimation, feature matching, and augmented reality.

Abstract Designing fonts and typefaces is a difficult process for both beginner and expert typographers. Existing workflows require the designer to create every glyph, while adhering to many loosely defined design suggestions to achieve an aesthetically appealing and coherent character set. This process can be significantly simplified by exploiting the similar structure character glyphs present across different fonts and the shared stylistic elements within the same font. To capture these correlations, we propose learning a stroke‐based font representation from a collection of existing typefaces. To enable this, we develop a stroke‐based geometric model for glyphs, a fitting procedure to reparametrize arbitrary fonts to our representation. We demonstrate the effectiveness of our model through a manifold learning technique that estimates a low‐dimensional font space. Our representation captures a wide range of everyday fonts with topological variations and naturally handles discrete and continuous variations, such as presence and absence of stylistic elements as well as slants and weights. We show that our learned representation can be used for iteratively improving fit quality, as well as exploratory style applications such as completing a font from a subset of observed glyphs, interpolating or adding and removing stylistic elements in existing fonts.

We propose a new procedure to guide training of a data-driven shape generative model using a structure-aware loss function. Complex 3D shapes often can be summarized using a coarsely defined structure which is consistent and robust across variety of observations. However, existing synthesis techniques do not account for structure during training, and thus often generate implausible and structurally unrealistic shapes. During training, we enforce structural constraints in order to enforce consistency and structure across the entire manifold. We propose a novel methodology for training 3D generative models that incorporates structural information into an end-to-end training pipeline.

We present Im2Pano3D, a convolutional neural network that generates a dense prediction of 3D structure and a probability distribution of semantic labels for a full 360° panoramic view of an indoor scene when given only a partial observation (= 50%) in the form of an RGB-D image. To make this possible, Im2Pano3D leverages strong contextual priors learned from large-scale synthetic and real-world indoor scenes. To ease the prediction of 3D structure, we propose to parameterize 3D surfaces with their plane equations and train the model to predict these parameters directly. To provide meaningful training supervision, we use multiple loss functions that consider both pixel level accuracy and global context consistency. Experiments demonstrate that Im2Pano3D is able to predict the semantics and 3D structure of the unobserved scene with more than 56% pixel accuracy and less than 0.52m average distance error, which is significantly better than alternative approaches.

The goal of our work is to complete the depth channel of an RGB-D image. Commodity-grade depth cameras often fail to sense depth for shiny, bright, transparent, and distant surfaces. To address this problem, we train a deep network that takes an RGB image as input and predicts dense surface normals and occlusion boundaries. Those predictions are then combined with raw depth observations provided by the RGB-D camera to solve for depths for all pixels, including those missing in the original observation. This method was chosen over others (e.g., inpainting depths directly) as the result of extensive experiments with a new depth completion benchmark dataset, where holes are filled in training data through the rendering of surface reconstructions created from multiview RGB-D scans. Experiments with different network inputs, depth representations, loss functions, optimization methods, inpainting methods, and deep depth estimation networks show that our proposed approach provides better depth completions than these alternatives.

This paper presents a robotic pick-and-place system that is capable of grasping and recognizing both known and novel objects in cluttered environments. The key new feature of the system is that it handles a wide range of object categories without needing any task-specific training data for novel objects. To achieve this, it first uses a category-agnostic affordance prediction algorithm to select and execute among four different grasping primitive behaviors. It then recognizes picked objects with a cross-domain image classification framework that matches observed images to product images. Since product images are readily available for a wide range of objects (e.g., from the web), the system works out-of-the-box for novel objects without requiring any additional training data. Exhaustive experimental results demonstrate that our multi-affordance grasping achieves high success rates for a wide variety of objects in clutter, and our recognition algorithm achieves high accuracy for both known and novel grasped objects. The approach was part of the MIT-Princeton Team system that took 1st place in the stowing task at the 2017 Amazon Robotics Challenge. All code, datasets, and pre-trained models are available online at http://arc.cs.princeton.edu.

Skilled robotic manipulation benefits from complex synergies between non-prehensile (e.g. pushing) and prehensile (e.g. grasping) actions: pushing can help rearrange cluttered objects to make space for arms and fingers; likewise, grasping can help displace objects to make pushing movements more precise and collision-free. In this work, we demonstrate that it is possible to discover and learn these synergies from scratch through model-free deep reinforcement learning. Our method involves training two fully convolutional networks that map from visual observations to actions: one infers the utility of pushes for a dense pixel-wise sampling of end-effector orientations and locations, while the other does the same for grasping. Both networks are trained jointly in a Q-learning framework and are entirely self-supervised by trial and error, where rewards are provided from successful grasps. In this way, our policy learns pushing motions that enable future grasps, while learning grasps that can leverage past pushes. During picking experiments in both simulation and real-world scenarios, we find that our system quickly learns complex behaviors even amid challenging cases of tightly packed clutter, and achieves better grasping success rates and picking efficiencies than baseline alternatives after a few hours of training. We further demonstrate that our method is capable of generalizing to novel objects. Qualitative results (videos), code, pre-trained models, and simulation environments are available at http://vpg.cs.princeton.edu/

Access to large, diverse RGB-D datasets is critical for training RGB-D scene understanding algorithms. However, existing datasets still cover only a limited number of views or a restricted scale of spaces. In this paper, we introduce Matterport3D, a large-scale RGB-D dataset containing 10,800 panoramic views from 194,400 RGB-D images of 90 building-scale scenes. Annotations are provided with surface reconstructions, camera poses, and 2D and 3D semantic segmentations. The precise global alignment and comprehensive, diverse panoramic set of views over entire buildings enable a variety of supervised and self-supervised computer vision tasks, including keypoint matching, view overlap prediction, normal prediction from color, semantic segmentation, and region classification.

We propose the idea of using a generative adversarial network (GAN) to assist users in designing real-world shapes with a simple interface. Users edit a voxel grid with a Minecraft-like interface. Yet they can execute a SNAP command at any time, which transforms their rough model into a desired shape that is both similar and realistic. They can edit and snap until they are satisfied with the result. The advantage of this approach is to assist novice users to create 3D models characteristic of the training data by only specifying rough edits. Our key contribution is to create a suitable projection operator around a 3D-GAN that maps an arbitrary 3D voxel input to a latent vector in the shape manifold of the generator that is both similar in shape to the input but also realistic. Experiments show our method is promising for computer-assisted interactive modeling.

A key requirement for leveraging supervised deep learning methods is the availability of large, labeled datasets. Unfortunately, in the context of RGB-D scene understanding, very little data is available - current datasets cover a small range of scene views and have limited semantic annotations. To address this issue, we introduce ScanNet, an RGB-D video dataset containing 2.5M views in 1513 scenes annotated with 3D camera poses, surface reconstructions, and semantic segmentations. To collect this data, we designed an easy-to-use and scalable RGB-D capture system that includes automated surface reconstruction and crowd-sourced semantic annotation.We show that using this data helps achieve state-of-the-art performance on several 3D scene understanding tasks, including 3D object classification, semantic voxel labeling, and CAD model retrieval.

RGB-D scanning of indoor environments is important for many applications, including real estate, interior design, and virtual reality. However, it is still challenging to register RGB-D images from a hand-held camera over a long video sequence into a globally consistent 3D model. Current methods often can lose tracking or drift and thus fail to reconstruct salient structures in large environments (e.g., parallel walls in different rooms). To address this problem, we propose a fine-to-coarse global registration algorithm that leverages robust registrations at finer scales to seed detection and enforcement of new correspondence and structural constraints at coarser scales. To test global registration algorithms, we provide a benchmark with 10,401 manually-clicked point correspondences in 25 scenes from the SUN3D dataset. During experiments with this benchmark, we find that our fine-to-coarse algorithm registers long RGB-D sequences better than previous methods.

This paper focuses on semantic scene completion, a task for producing a complete 3D voxel representation of volumetric occupancy and semantic labels for a scene from a single-view depth map observation. Previous work has considered scene completion and semantic labeling of depth maps separately. However, we observe that these two problems are tightly intertwined. To leverage the coupled nature of these two tasks, we introduce the semantic scene completion network (SSCNet), an end-to-end 3D convolutional network that takes a single depth image as input and simultaneously outputs occupancy and semantic labels for all voxels in the camera view frustum. Our network uses a dilation-based 3D context module to efficiently expand the receptive field and enable 3D context learning. To train our network, we construct SUNCG - a manually created largescale dataset of synthetic 3D scenes with dense volumetric annotations. Our experiments demonstrate that the joint model outperforms methods addressing each task in isolation and outperforms alternative approaches on the semantic scene completion task. The dataset and code is available at http://sscnet.cs.princeton.edu.

Convolutional networks for image classification progressively reduce resolution until the image is represented by tiny feature maps in which the spatial structure of the scene is no longer discernible. Such loss of spatial acuity can limit image classification accuracy and complicate the transfer of the model to downstream applications that require detailed scene understanding. These problems can be alleviated by dilation, which increases the resolution of output feature maps without reducing the receptive field of individual neurons. We show that dilated residual networks (DRNs) outperform their non-dilated counterparts in image classification without increasing the models depth or complexity. We then study gridding artifacts introduced by dilation, develop an approach to removing these artifacts (degridding), and show that this further increases the performance of DRNs. In addition, we show that the accuracy advantage of DRNs is further magnified in downstream applications such as object localization and semantic segmentation.

Matching local geometric features on real-world depth images is a challenging task due to the noisy, low-resolution, and incomplete nature of 3D scan data. These difficulties limit the performance of current state-of-art methods, which are typically based on histograms over geometric properties. In this paper, we present 3DMatch, a data-driven model that learns a local volumetric patch descriptor for establishing correspondences between partial 3D data. To amass training data for our model, we propose a self-supervised feature learning method that leverages the millions of correspondence labels found in existing RGB-D reconstructions. Experiments show that our descriptor is not only able to match local geometry in new scenes for reconstruction, but also generalize to different tasks and spatial scales (e.g. instance-level object model alignment for the Amazon Picking Challenge, and mesh surface correspondence). Results show that 3DMatch consistently outperforms other state-of-the-art approaches by a significant margin. Code, data, benchmarks, and pre-trained models are available online at http://3dmatch.cs.princeton.edu.

Indoor scene understanding is central to applications such as robot navigation and human companion assistance. Over the last years, data-driven deep neural networks have outperformed many traditional approaches thanks to their representation learning capabilities. One of the bottlenecks in training for better representations is the amount of available per-pixel ground truth data that is required for core scene understanding tasks such as semantic segmentation, normal prediction, and object boundary detection. To address this problem, a number of works proposed using synthetic data. However, a systematic study of how such synthetic data is generated is missing. In this work, we introduce a large-scale synthetic dataset with 500K physically-based rendered images from 45K realistic 3D indoor scenes. We study the effects of rendering methods and scene lighting on training for three computer vision tasks: surface normal prediction, semantic segmentation, and object boundary detection. This study provides insights into the best practices for training with synthetic data (more realistic rendering is worth it) and shows that pretraining with our new synthetic dataset can improve results beyond the current state of the art on all three tasks.

Computational manufacturing technologies such as 3D printing hold the potential for creating objects with previously undreamed-of combinations of functionality and physical properties. Human designers, however, typically cannot exploit the full geometric (and often material) complexity of which these devices are capable. This STAR examines recent systems developed by the computer graphics community in which designers specify higher-level goals ranging from structural integrity and deformation to appearance and aesthetics, with the final detailed shape and manufacturing instructions emerging as the result of computation. It summarizes frameworks for interaction, simulation, and optimization, as well as documents the range of general objectives and domain-specific goals that have been considered. An important unifying thread in this analysis is that different underlying geometric and physical representations are necessary for different tasks: we document over a dozen classes of representations that have been used for fabrication-aware design in the literature. We analyze how these classes possess obvious advantages for some needs, but have also been used in creative manners to facilitate unexpected problem solutions.

This paper investigates a method for semantic segmentation of small objects in terrestrial LIDAR scans in urban environments. The core research contribution is a hierarchical segmentation algorithm where potential merges between segments are prioritized by a learned affinity function and constrained to occur only if they achieve a significantly high object classification probability. This approach provides a way to integrate a learned shape-prior (the object classifier) into a search for the best semantic segmentation in a fast and practical algorithm. Experiments with LIDAR scans collected by Google Street View cars throughout ~100 city blocks of New York City show that the algorithm provides better segmentations and classifications than simple alternatives for cars, vans, traffic lights, and street lights.

This paper describes an automatic algorithm for global alignment of LiDAR data collected with Google Street View cars in urban environments. The problem is challenging because global pose estimation techniques (GPS) do not work well in city environments with tall buildings, and local tracking techniques (integration of inertial sensors, structure-from-motion, etc.) provide solutions that drift over long ranges, leading to solutions where data collected over wide ranges is warped and misaligned by many meters. Our approach to address this problem is to extract “semantic features” with object detectors (e.g., for facades, poles, cars, etc.) that can be matched robustly at different scales, and thus are selected for different iterations of an ICP algorithm. We have implemented an all-to-all, non-rigid, global alignment based on this idea that provides better results than alternatives during experiments with data from large regions of New York, San Francisco, Paris, and Rome.

Abstract Increasingly, companies are creating product advertisements and catalog images using computer renderings of 3D scenes. A common goal for these companies is to create aesthetically appealing compositions that highlight objects of interest within the context of a scene. Unfortunately, this goal is challenging, not only due to the need to balance the trade‐off among aesthetic principles and design constraints, but also because of the huge search space induced by possible camera parameters, object placement, material choices, etc. Previous methods have investigated only optimization of camera parameters. In this paper, we develop a tool that starts from an initial scene description and a set of high‐level constraints provided by a stylist and then automatically generates an optimized scene whose 2D composition is improved. It does so by locally adjusting the 3D object transformations, surface materials, and camera parameters. The value of this tool is demonstrated in a variety of applications motivated by product catalogs, including rough layout refinement, detail image creation, home planning, cultural customization, and text inlay placement. Results of a perceptual study indicate that our system produces images preferable for product advertisement compared to a more traditional camera‐only optimization.

This paper presents a method for learning to predict the stylistic compatibility between 3D furniture models from different object classes: e.g., how well does this chair go with that table? To do this, we collect relative assessments of style compatibility using crowdsourcing. We then compute geometric features for each 3D model and learn a mapping of them into a space where Euclidean distances represent style incompatibility. Motivated by the geometric subtleties of style, we introduce part-aware geometric feature vectors that characterize the shapes of different parts of an object separately. Motivated by the need to compute style compatibility between different object classes, we introduce a method to learn object class-specific mappings from geometric features to a shared feature space. During experiments with these methods, we find that they are effective at predicting style compatibility agreed upon by people. We find in user studies that the learned compatibility metric is useful for novel interactive tools that: 1) retrieve stylistically compatible models for a query, 2) suggest a piece of furniture for an existing scene, and 3) help guide an interactive 3D modeler towards scenes with compatible furniture.

This paper proposes a group annotation approach to interactive semantic labeling of data and demonstrates the idea in a system for labeling objects in 3D LiDAR scans of a city. In this approach, the system selects a group of objects, predicts a semantic label for it, and highlights it in an interactive display. In response, the user either confirms the predicted label, provides a different label, or indicates that no single label can be assigned to all objects in the group. This sequence of interactions repeats until a label has been confirmed for every object in the data set. The main advantage of this approach is that it provides faster interactive labeling rates than alternative approaches, especially in cases where all labels must be explicitly confirmed by a person. The main challenge is to provide an algorithm that selects groups with many objects all of the same label type arranged in patterns that are quick to recognize, which requires models for predicting object labels and for estimating times for people to recognize objects in groups. We address these challenges by defining an objective function that models the estimated time required to process all unlabeled objects and approximation algorithms to minimize it. Results of user studies suggest that group annotation can be used to label objects in LiDAR scans of cities significantly faster than one-by-one annotation with active learning.

As 3D acquisition devices and modeling tools become widely available there is a growing need for automatic algorithms that analyze the semantics and functionality of digitized shapes. Most recent research has focused on analyzing geometric structures of shapes. Our work is motivated by the observation that a majority of man-made shapes are designed to be used by people. Thus, in order to fully understand their semantics, one needs to answer a fundamental question: "how do people interact with these objects?" As an initial step towards this goal, we offer a novel algorithm for automatically predicting a static pose that a person would need to adopt in order to use an object. Specifically, given an input 3D shape, the goal of our analysis is to predict a corresponding human pose, including contact points and kinematic parameters. This is especially challenging for man-made objects that commonly exhibit a lot of variance in their geometric structure. We address this challenge by observing that contact points usually share consistent local geometric features related to the anthropometric properties of corresponding parts and that human body is subject to kinematic constraints and priors. Accordingly, our method effectively combines local region classification and global kinematically-constrained search to successfully predict poses for various objects. We also evaluate our algorithm on six diverse collections of 3D polygonal models (chairs, gym equipment, cockpits, carts, bicycles, and bipedal devices) containing a total of 147 models. Finally, we demonstrate that the poses predicted by our algorithm can be used in several shape analysis problems, such as establishing correspondences between objects, detecting salient regions, finding informative viewpoints, and retrieving functionally-similar shapes.

We present AttribIt, an approach for people to create visual content using relative semantic attributes expressed in linguistic terms. During an off-line processing step, AttribIt learns semantic attributes for design components that reflect the high-level intent people may have for creating content in a domain (e.g. adjectives such as "dangerous", "scary" or "strong") and ranks them according to the strength of each learned attribute. Then, during an interactive design session, a person can explore different combinations of visual components using commands based on relative attributes (e.g. "make this part more dangerous"). Novel designs are assembled in real-time as the strengths of selected attributes are varied, enabling rapid, in-situ exploration of candidate designs. We applied this approach to 3D modeling and web design. Experiments suggest this interface is an effective alternative for novices performing tasks with high-level design goals.

As large repositories of 3D shape collections continue to grow, understanding the data, especially encoding the inter-model similarity and their variations, is of central importance. For example, many data-driven approaches now rely on access to semantic segmentation information, accurate inter-model point-to-point correspondence, and deformation models that characterize the model collections. Existing approaches, however, are either supervised requiring manual labeling; or employ super-linear matching algorithms and thus are unsuited for analyzing large collections spanning many thousands of models. We propose an automatic algorithm that starts with an initial template model and then jointly optimizes for part segmentation, point-to-point surface correspondence, and a compact deformation model to best explain the input model collection. As output, the algorithm produces a set of probabilistic part-based templates that groups the original models into clusters of models capturing their styles and variations. We evaluate our algorithm on several standard datasets and demonstrate its scalability by analyzing much larger collections of up to thousands of shapes.

Abstract In this paper, we propose an automatic algorithm for finding a correspondence map between two 3D surfaces. The key insight is that global reflective symmetry axes are stable, recognizable, semantic features of most real‐world surfaces. Thus, it is possible to find a useful map between two surfaces by first extracting symmetry axis curves, aligning the extracted curves, and then extrapolating correspondences found on the curves to both surfaces. The main advantages of this approach are efficiency and robustness: the difficult problem of finding a surface map is reduced to three significantly easier problems: symmetry detection, curve alignment, and correspondence extrapolation, each of which has a robust, polynomial‐time solution (e.g., optimal alignment of 1D curves is possible with dynamic programming). We investigate of this approach on a wide range of examples, including both intrinsically symmetric surfaces and polygon soups, and find that it is superior to previous methods in cases where two surfaces have different overall shapes but similar reflective symmetry axes, a common case in computer graphics.

This paper investigates "Schelling points" on 3D meshes, feature points selected by people in a pure coordination game due to their salience. To collect data for this investigation, we designed an online experiment that asked people to select points on 3D surfaces that they expect will be selected by other people. We then analyzed properties of the selected points, finding that: 1) Schelling point sets are usually highly symmetric, and 2) local curvature properties (e.g., Gauss curvature) are most helpful for identifying obvious Schelling points (tips of protrusions), but 3) global properties (e.g., segment centeredness, proximity to a symmetry axis, etc.) are required to explain more subtle features. Based on these observations, we use regression analysis to combine multiple properties into an analytical model that predicts where Schelling points are likely to be on new meshes. We find that this model benefits from a variety of surface properties, particularly when training data comes from examples in the same object class.

We present a method for synthesizing 3D object arrangements from examples. Given a few user-provided examples, our system can synthesize a diverse set of plausible new scenes by learning from a larger scene database. We rely on three novel contributions. First, we introduce a probabilistic model for scenes based on Bayesian networks and Gaussian mixtures that can be trained from a small number of input examples. Second, we develop a clustering algorithm that groups objects occurring in a database of scenes according to their local scene neighborhoods. These contextual categories allow the synthesis process to treat a wider variety of objects as interchangeable. Third, we train our probabilistic model on a mix of user-provided examples and relevant scenes retrieved from the database. This mixed model learning process can be controlled to introduce additional variety into the synthesized scenes. We evaluate our algorithm through qualitative results and a perceptual study in which participants judged synthesized scenes to be highly plausible, as compared to hand-created scenes.

This article presents a framework for symmetry-guided texture synthesis and processing. It is motivated by the long-standing problem of how to optimize, transfer, and control the spatial patterns in textures. The key idea is that symmetry representations that measure autocorrelations with respect to all transformations of a group are a natural way to describe spatial patterns in many real-world textures. To leverage this idea, we provide methods to transfer symmetry representations from one texture to another, process the symmetries of a texture, and optimize textures with respect to properties of their symmetry representations. These methods are automatic and robust, as they don't require explicit detection of discrete symmetries. Applications are investigated for optimizing, processing, and transferring symmetries and textures.

Large collections of 3D models from the same object class (e.g., chairs, cars, animals) are now commonly available via many public repositories, but exploring the range of shape variations across such collections remains a challenging task. In this work, we present a new exploration interface that allows users to browse collections based on similarities and differences between shapes in user-specified regions of interest (ROIs). To support this interactive system, we introduce a novel analysis method for computing similarity relationships between points on 3D shapes across a collection. We encode the inherent ambiguity in these relationships using fuzzy point correspondences and propose a robust and efficient computational framework that estimates fuzzy correspondences using only a sparse set of pairwise model alignments. We evaluate our analysis method on a range of correspondence benchmarks and report substantial improvements in both speed and accuracy over existing alternatives. In addition, we demonstrate how fuzzy correspondences enable key features in our exploration tool, such as automated view alignment, ROI-based similarity search, and faceted browsing.

We introduce a simple formula for 4-point planar warping that produces provably good 2D deformations. In contrast to previous work, the new deformations minimize the maximum conformal distortion and spread the distortion equally across the domain. We derive closed-form formulas for computing the 4-point interpolant and analyze its properties. We further explore applications to 2D shape deformations by building local deformation operators that use thin-plate splines to further deform the 4-point interpolant to satisfy certain boundary conditions. Although this modification no longer has any theoretical guarantees, we demonstrate that, practically, these local operators can be used to create compound deformations with fewer control points and smaller worst-case distortions in comparisons to the state-of-the-art.

This paper describes a fully automatic pipeline for finding an intrinsic map between two non-isometric, genus zero surfaces. Our approach is based on the observation that efficient methods exist to search for nearly isometric maps (e.g., Möbius Voting or Heat Kernel Maps), but no single solution found with these methods provides low-distortion everywhere for pairs of surfaces differing by large deformations. To address this problem, we suggest using a weighted combination of these maps to produce a "blended map." This approach enables algorithms that leverage efficient search procedures, yet can provide the flexibility to handle large deformations. The main challenges of this approach lie in finding a set of candidate maps { m i } and their associated blending weights { b i ( p )} for every point p on the surface. We address these challenges specifically for conformal maps by making the following contributions. First, we provide a way to blend maps, defining the image of p as the weighted geodesic centroid of m i ( p ). Second, we provide a definition for smooth blending weights at every point p that are proportional to the area preservation of m i at p . Third, we solve a global optimization problem that selects candidate maps based both on their area preservation and consistency with other selected maps. During experiments with these methods, we find that our algorithm produces blended maps that align semantic features better than alternative approaches over a variety of data sets.

Abstract The goal of our work is to develop an algorithm for automatic and robust detection of global intrinsic symmetries in 3D surface meshes. Our approach is based on two core observations. First, symmetry invariant point sets can be detected robustly using critical points of the Average Geodesic Distance (AGD) function. Second, intrinsic symmetries are self‐isometries of surfaces and as such are contained in the low dimensional group of Möbius transformations. Based on these observations, we propose an algorithm that: 1) generates a set of symmetric points by detecting critical points of the AGD function, 2) enumerates small subsets of those feature points to generate candidate Möbius transformations, and 3) selects among those candidate Möbius transformations the one(s) that best map the surface onto itself. The main advantages of this algorithm stem from the stability of the AGD in predicting potential symmetric point features and the low dimensionality of the Möbius group for enumerating potential self‐mappings. During experiments with a benchmark set of meshes augmented with human‐specified symmetric correspondences, we find that the algorithm is able to find intrinsic symmetries for a wide variety of object types with moderate deviations from perfect symmetry.

Abstract A geodesic is a parameterized curve on a Riemannian manifold governed by a second order partial differential equation. Geodesics are notoriously unstable: small perturbations of the underlying manifold may lead to dramatic changes of the course of a geodesic. Such instability makes it difficult to use geodesics in many applications, in particular in the world of discrete geometry. In this paper, we consider a geodesic as the indicator function of the set of the points on the geodesic. From this perspective, we present a new concept called fuzzy geodesics and show that fuzzy geodesics are stable with respect to the Gromov‐Hausdorff distance. Based on fuzzy geodesics, we propose a new object called the intersection configuration for a set of points on a shape and demonstrate its effectiveness in the application of finding consistent correspondences between sparse sets of points on shapes differing by extreme deformations.

We introduce the Symmetry Factored Embedding (SFE) and the Symmetry Factored Distance (SFD) as new tools to analyze and represent symmetries in a point set. The SFE provides new coordinates in which symmetry is "factored out," and the SFD is the Euclidean distance in that space. These constructions characterize the space of symmetric correspondences between points -- i.e., orbits. A key observation is that a set of points in the same orbit appears as a clique in a correspondence graph induced by pairwise similarities. As a result, the problem of finding approximate and partial symmetries in a point set reduces to the problem of measuring connectedness in the correspondence graph, a well-studied problem for which spectral methods provide a robust solution. We provide methods for computing the SFE and SFD for extrinsic global symmetries and then extend them to consider partial extrinsic and intrinsic cases. During experiments with difficult examples, we find that the proposed methods can characterize symmetries in inputs with noise, missing data, non-rigid deformations, and complex symmetries, without a priori knowledge of the symmetry group. As such, we believe that it provides a useful tool for automatic shape analysis in applications such as segmentation and stationary point detection.

Measuring distances between pairs of points on a 3D surface is a fundamental problem in computer graphics and geometric processing. For most applications, the important properties of a distance are that it is a metric, smooth, locally isotropic, globally “shape-aware,” isometry-invariant, insensitive to noise and small topology changes, parameter-free, and practical to compute on a discrete mesh. However, the basic methods currently popular in computer graphics (e.g., geodesic and diffusion distances) do not have these basic properties. In this article, we propose a new distance measure based on the biharmonic differential operator that has all the desired properties. This new surface distance is related to the diffusion and commute-time distances, but applies different (inverse squared) weighting to the eigenvalues of the Laplace-Beltrami operator, which provides a nice trade-off between nearly geodesic distances for small distances and global shape-awareness for large distances. The article provides theoretical and empirical analysis for a large number of meshes.

We present a multiple-feature approach for determining matches between small fragments of archaeological artifacts such as Bronze-Age and Roman frescoes. In contrast with traditional 2D and 3D shape matching approaches, we introduce a set of feature descriptors that are based on not only color and shape, but also normal maps. These are easy to acquire and combine high data quality with discriminability and robustness to some types of deterioration. Our feature descriptors range from general-purpose to domain-specific, and are quick to compute and match. We have tested our system on three datasets of fresco fragments, demonstrating that multi-cue matching using different subsets of features leads to different tradeoffs between efficiency and effectiveness. In particular, we show that normal-based features are more effective than color-based ones at similar computational complexity, and that 3D features are more discriminative than ones based on 2D or normals, but at higher computational cost. We also demonstrate how machine learning techniques can be used to effectively combine our new features with traditional ones. Our results show good retrieval performance, significantly improving upon the match prediction rate of state-of-the-art 3D matching algorithms, and are expected to extend to general matching problems in applications such as texture synthesis and forensics.

This paper investigates the design of a system for recognizing objects in 3D point clouds of urban environments. The system is decomposed into four steps: locating, segmenting, characterizing, and classifying clusters of 3D points. Specifically, we first cluster nearby points to form a set of potential object locations (with hierarchical clustering). Then, we segment points near those locations into foreground and background sets (with a graph-cut algorithm). Next, we build a feature vector for each point cluster (based on both its shape and its context). Finally, we label the feature vectors using a classifier trained on a set of manually labeled objects. The paper presents several alternative methods for each step. We quantitatively evaluate the system and tradeoffs of different alternatives in a truthed part of a scan of Ottawa that contains approximately 100 million points and 1000 objects of interest. Then, we use this truth data as a training set to recognize objects amidst approximately 1 billion points of the remainder of the Ottawa scan.

Abstract This paper introduces a framework for defining a shape‐aware distance measure between any two points in the interior of a surface mesh. Our framework is based on embedding the surface mesh into a high‐dimensional space in a way that best preserves boundary distances between vertices of the mesh, performing a mapping of the mesh volume into this high‐dimensional space using barycentric coordinates, and defining the interior distance between any two points simply as their Euclidean distance in the embedding space. We investigate the theoretical properties of the interior distance in relation to properties of the chosen boundary distances and barycentric coordinates, and we investigate empirical properties of the interior distance using diffusion distance as the prescribed boundary distance and mean value coordinates. We prove theoretically that the interior distance is a metric, smooth, interpolating the boundary distances, and reproducing Euclidean distances, and we show empirically that it is insensitive to boundary noise and deformation and quick to compute. In case the barycentric coordinates are non‐negative we also show a maximum principle exists. Finally, we use it to define a new geometric property, barycentroid of shape, and show that it captures the notion of semantic center of the shape.

Identifying a protein's functional sites is an important step towards characterizing its molecular function. Numerous structure- and sequence-based methods have been developed for this problem. Here we introduce ConCavity, a small molecule binding site prediction algorithm that integrates evolutionary sequence conservation estimates with structure-based methods for identifying protein surface cavities. In large-scale testing on a diverse set of single- and multi-chain protein structures, we show that ConCavity substantially outperforms existing methods for identifying both 3D ligand binding pockets and individual ligand binding residues. As part of our testing, we perform one of the first direct comparisons of conservation-based and structure-based methods. We find that the two approaches provide largely complementary information, which can be combined to improve upon either approach alone. We also demonstrate that ConCavity has state-of-the-art performance in predicting catalytic sites and drug binding pockets. Overall, the algorithms and analysis presented here significantly improve our ability to identify ligand binding sites and further advance our understanding of the relationship between evolutionary sequence conservation and structural and functional attributes of proteins. Data, source code, and prediction visualizations are available on the ConCavity web site (http://compbio.cs.princeton.edu/concavity/).

This paper describes a benchmark for evaluation of 3D mesh segmentation salgorithms. The benchmark comprises a data set with 4,300 manually generated segmentations for 380 surface meshes of 19 different object categories, and it includes software for analyzing 11 geometric properties of segmentations and producing 4 quantitative metrics for comparison of segmentations. The paper investigates the design decisions made in building the benchmark, analyzes properties of human-generated and computer-generated segmentations, and provides quantitative comparisons of 7 recently published mesh segmentation algorithms. Our results suggest that people are remarkably consistent in the way that they segment most 3D surface meshes, that no one automatic segmentation algorithm is better than the others for all types of objects, and that algorithms based on non-local shape features seem to produce segmentations that most closely resemble ones made by humans.

This paper investigates the ability of sparse line drawings to depict 3D shape. We perform a study in which people are shown an image of one of twelve 3D objects depicted with one of six styles and asked to orient a gauge to coincide with the surface normal at many positions on the object's surface. The normal estimates are compared with each other and with ground truth data provided by a registered 3D surface model to analyze accuracy and precision. The paper describes the design decisions made in collecting a large data set (275,000 gauge measurements) and provides analysis to answer questions about how well people interpret shapes from drawings. Our findings suggest that people interpret certain shapes almost as well from a line drawing as from a shaded image, that current computer graphics line drawing techniques can effectively depict shape and even match the effectiveness of artist's drawings, and that errors in depiction are often localized and can be traced to particular properties of the lines used. The data collected for this study will become a publicly available resource for further studies of this type.

The goal of our work is to develop an efficient, automatic algorithm for discovering point correspondences between surfaces that are approximately and/or partially isometric. Our approach is based on three observations. First, isometries are a subset of the Möbius group, which has low-dimensionality -- six degrees of freedom for topological spheres, and three for topological discs. Second, computing the Möbius transformation that interpolates any three points can be computed in closed-form after a mid-edge flattening to the complex plane. Third, deviations from isometry can be modeled by a transportation-type distance between corresponding points in that plane. Motivated by these observations, we have developed a Möbius Voting algorithm that iteratively: 1) samples a triplet of three random points from each of two point sets, 2) uses the Möbius transformations defined by those triplets to map both point sets into a canonical coordinate frame on the complex plane, and 3) produces "votes" for predicted correspondences between the mutually closest points with magnitude representing their estimated deviation from isometry. The result of this process is a fuzzy correspondence matrix, which is converted to a permutation matrix with simple matrix operations and output as a discrete set of point correspondences with confidence values. The main advantage of this algorithm is that it can find intrinsic point correspondences in cases of extreme deformation. During experiments with a variety of data sets, we find that it is able to find dozens of point correspondences between different object types in different poses fully automatically.

This paper presents the results of a study in which artists made line drawings intended to convey specific 3D shapes. The study was designed so that drawings could be registered with rendered images of 3D models, supporting an analysis of how well the locations of the artists' lines correlate with other artists', with current computer graphics line definitions, and with the underlying differential properties of the 3D surface. Lines drawn by artists in this study largely overlapped one another (75% are within 1mm of another line), particularly along the occluding contours of the object. Most lines that do not overlap contours overlap large gradients of the image intensity, and correlate strongly with predictions made by recent line drawing algorithms in computer graphics. 14% were not well described by any of the local properties considered in this study. The result of our work is a publicly available data set of aligned drawings, an analysis of where lines appear in that data set based on local properties of 3D models, and algorithms to predict where artists will draw lines for new scenes.

The goal of this paper is to investigate a new shape analysis method based on randomized cuts of 3D surface meshes. The general strategy is to generate a random set of mesh segmentations and then to measure how often each edge of the mesh lies on a segmentation boundary in the randomized set. The resulting "partition function" defined on edges provides a continuous measure of where natural part boundaries occur in a mesh, and the set of "most consistent cuts" provides a stable list of global shape features. The paper describes methods for generating random distributions of mesh segmentations, studies sensitivity of the resulting partition functions to noise, tessellation, pose, and intra-class shape variations, and investigates applications in mesh visualization, segmentation, deformation, and registration.

Selecting the most important regions of a surface is useful for shape matching and a variety of applications in computer graphics and geometric modeling. While previous research has analyzed geometric properties of meshes in isolation, we select regions that distinguish a shape from objects of a different type. Our approach to analyzing distinctive regions is based on performing a shape-based search using each region as a query into a database. Distinctive regions of a surface have shape consistent with objects of the same type and different from objects of other types. We demonstrate the utility of detecting distinctive surface regions for shape matching and other graphics applications including mesh visualization, icon generation, and mesh simplification.

Databases of 3D shapes have become widespread for a variety of applications, and a key research problem is searching these databases for similar shapes. This paper introduces a method for finding distinctive features of a shape that are useful for determining shape similarity. Although global shape descriptors have been developed to facilitate retrieval, they fail when local shape properties are the distinctive features of a class. Alternatively, local shape descriptors can be generated over the surface of shapes, but then storage and search of the descriptors becomes unnecessarily expensive, as perhaps only a few descriptors are sufficient to distinguish classes. The challenge is to select local descriptors from a query shape that are most distinctive for retrieval. Our approach is to define distinction as the retrieval performance of a local shape descriptor. During a training phase, we estimate descriptor likelihood using a multivariate Gaussian distribution of real-valued shape descriptors, evaluate the retrieval performance of each descriptor from a training set, and average these performance values at every likelihood value. For each query, we evaluate the likelihood of local shape descriptors on its surface and lookup the expected retrieval values learned from the training set to determine their predicted distinction values. We show that querying with the most distinctive shape descriptors provides favorable retrieval performance during tests with a database of common graphics objects.

Detailed surface geometry contributes greatly to the visual realism of 3D face models. However, acquiring high-resolution face geometry is often tedious and expensive. Consequently, most face models used in games, virtual reality, or computer vision look unrealistically smooth. In this paper, we introduce a new statistical technique for the analysis and synthesis of small three-dimensional facial features, such as wrinkles and pores. We acquire high-resolution face geometry for people across a wide range of ages, genders, and races. For each scan, we separate the skin surface details from a smooth base mesh using displaced subdivision surfaces. Then, we analyze the resulting displacement maps using the texture analysis/synthesis framework of Heeger and Bergen, adapted to capture statistics that vary spatially across a face. Finally, we use the extracted statistics to synthesize plausible detail on face meshes of arbitrary subjects. We demonstrate the effectiveness of this method in several applications, including analysis of facial texture in subjects with different ages and genders, interpolation between high-resolution face scans, adding detail to low-resolution face scans, and adjusting the apparent age of faces. In all cases, we are able to re-produce fine geometric details consistent with those observed in high resolution scans.

Symmetry is an important cue for many applications, including object alignment, recognition, and segmentation. In this paper, we describe a planar reflective symmetry transform (PRST) that captures a continuous measure of the reflectional symmetry of a shape with respect to all possible planes. This transform combines and extends previous work that has focused on global symmetries with respect to the center of mass in 3D meshes and local symmetries with respect to points in 2D images. We provide an efficient Monte Carlo sampling algorithm for computing the transform for surfaces and show that it is stable under common transformations. We also provide an iterative refinement algorithm to find local maxima of the transform precisely. We use the transform to define two new geometric properties, center of symmetry and principal symmetry axes, and show that they are useful for aligning objects in a canonical coordinate system. Finally, we demonstrate that the symmetry transform is useful for several applications in computer graphics, including shape matching, segmentation of meshes into parts, and automatic viewpoint selection.

In recent years, many shape representations and geometric algorithms have been proposed for matching 3D shapes. Usually, each algorithm is tested on a different (small) database of 3D models, and thus no direct comparison is available for competing methods. We describe the Princeton Shape Benchmark (PSB), a publicly available database of polygonal models collected from the World Wide Web and a suite of tools for comparing shape matching and classification algorithms. One feature of the benchmark is that it provides multiple semantic labels for each 3D model. For instance, it includes one classification of the 3D models based on function, another that considers function and form, and others based on how the object was constructed (e.g., man-made versus natural objects). We find that experiments with these classifications can expose different properties of shape-based retrieval algorithms. For example, out of 12 shape descriptors tested, extended Gaussian images by B. Horn (1984) performed best for distinguishing man-made from natural objects, while they performed among the worst for distinguishing specific object types. Based on experiments with several different shape descriptors, we conclude that no single descriptor is best for all classifications, and thus the main contribution of this paper is to provide a framework to determine the conditions under which each descriptor performs best.

In this paper, we investigate a data-driven synthesis approach to constructing 3D geometric surface models. We provide methods with which a user can search a large database of 3D meshes to find parts of interest, cut the desired parts out of the meshes with intelligent scissoring, and composite them together in different ways to form new objects. The main benefit of this approach is that it is both easy to learn and able to produce highly detailed geometric models -- the conceptual design for new models comes from the user, while the geometric details come from examples in the database. The focus of the paper is on the main research issues motivated by the proposed approach: (1) interactive segmentation of 3D surfaces, (2) shape-based search to find 3D models with parts matching a query, and (3) composition of parts to form new models. We provide new research contributions on all three topics and incorporate them into a prototype modeling system. Experience with our prototype system indicates that it allows untrained users to create interesting and detailed 3D models.

With recent improvements in methods for the acquisition and rendering of 3D models, the need for retrieval of models has gained prominence in the graphics and vision communities. A variety of methods have been proposed that enable the efficient querying of model repositories for a desired 3D shape. Many of these methods use a 3D model as a query and attempt to retrieve models from the database that have a similar shape.In this paper we consider the implications of anisotropy on the shape matching paradigm. In particular, we propose a novel method for matching 3D models that factors the shape matching equation as the disjoint outer product of anisotropy and geometric comparisons. We provide a general method for computing the factored similarity metric and show how this approach can be applied to improve the matching performance of many existing shape matching methods.

A long-standing challenge in geometric modeling is providing a natural, intuitive interface for making local deformations to 3D surfaces. Previous approaches have provided either interactive manipulation or physical simulation to control surface deformations. In this paper, we investigate combining these two approaches with a painting interface that gives the user direct, local control over a physical simulation. The paint a user applies to the model defines its instantaneous surface velocity, the user can effect surface deformations. We have found that this painting metaphor gives the user direct, local control over surface deformations for several applications: creating new models, removing noise from existing models, and adding geometric texture to an existing surface at multiple scales.

Image-based rendering (IBR) systems enable virtual walkthroughs of photorealistic environments by warping and combining reference images to novel viewpoints under interactive user control. A significant challenge in such systems is to automatically compute image correspondences that enable accurate image warping.In this paper, we describe a new algorithm for computing a globally consistent set of image feature correspondences across a wide range of viewpoints suitable for IBR walkthroughs. We first detect point features in a dense set of omnidirectional images captured on an eye-height plane. Then, we track these features from image to image, identifying potential correspondences when two features track to the same position in the same image. Among the potential correspondences, we select the maximal consistent set using a greedy graph-labeling algorithm.A key feature of our approach is that it exploits the multiple paths that can be followed between images in order to increase the number of feature correspondences between distant images. We demonstrate the benefits of this approach in a real-time IBR walkthrough system where novel images are reconstructed as the user moves interactively.

As the number of 3D models available on the Web grows, there is an increasing need for a search engine to help people find them. Unfortunately, traditional text-based search techniques are not always effective for 3D data. In this article, we investigate new shape-based search methods. The key challenges are to develop query methods simple enough for novice users and matching algorithms robust enough to work for arbitrary polygonal models. We present a Web-based search engine system that supports queries based on 3D sketches, 2D sketches, 3D models, and/or text keywords. For the shape-based queries, we have developed a new matching algorithm that uses spherical harmonics to compute discriminating similarity measures without requiring repair of model degeneracies or alignment of orientations. It provides 46 to 245% better performance than related shape-matching methods during precision--recall experiments, and it is fast enough to return query results from a repository of 20,000 models in under a second. The net result is a growing interactive index of 3D models available on the Web (i.e., a Google for 3D models).

A long-standing research problem in computer graphics is to reproduce the visual experience of walking through a large photorealistic environment interactively. On one hand, traditional geometry-based rendering systems fall short of simulating the visual realism of a complex environment. On the other hand, image-based rendering systems have to date been unable to capture and store a sampled representation of a large environment with complex lighting and visibility effects. In this paper, we present a "sea of images," a practical approach to dense sampling, storage, and reconstruction of the plenoptic function in large, complex indoor environments. We use a motorized cart to capture omnidirectional images every few inches on a eye-height plane throughout an environment. The captured images are compressed and stored in a multiresolution hierarchy suitable for real-time prefetching during an interactive walkthrough. Later, novel images are reconstructed for a simulated observer by resampling nearby captured images. Our system acquires 15,254 images over 1,050 square feet at an average image spacing of 1.5 inches. The average capture and processing time is 7 hours. We demonstrate realistic walkthroughs of real-world environments reproducing specular reflections and occlusion effects while rendering 15-25 frames per second.

Measuring the similarity between 3D shapes is a fundamental problem, with applications in computer graphics, computer vision, molecular biology, and a variety of other fields. A challenging aspect of this problem is to find a suitable shape signature that can be constructed and compared quickly, while still discriminating between similar and dissimilar shapes.In this paper, we propose and analyze a method for computing shape signatures for arbitrary (possibly degenerate) 3D polygonal models. The key idea is to represent the signature of an object as a shape distribution sampled from a shape function measuring global geometric properties of an object. The primary motivation for this approach is to reduce the shape matching problem to the comparison of probability distributions, which is simpler than traditional shape matching methods that require pose registration, feature correspondence, or model fitting.We find that the dissimilarities between sampled distributions of simple shape functions (e.g., the distance between two random points on a surface) provide a robust method for discriminating between classes of objects (e.g., cars versus airplanes) in a moderately sized database, despite the presence of arbitrary translations, rotations, scales, mirrors, tessellations, simplifications, and model degeneracies. They can be evaluated quickly, and thus the proposed method could be applied as a pre-classifier in a complete shape-based retrieval or analysis system concerned with finding similar whole objects. The paper describes our early experiences using shape distributions for object classification and for interactive web-based retrieval of 3D models.

Realistic modeling of reverberant sound in 3D virtual worlds provides users with important cues for localizing sound sources and understanding spatial properties of the environment. Unfortunately, current geometric acoustic modeling systems do not accurately simulate reverberant sound. Instead, they model only direct transmission and specular reflection, while diffraction is either ignored or modeled through statistical approximation. However, diffraction is important for correct interpretation of acoustic environments, especially when the direct path between sound source and receiver is occluded.

Measuring the similarity between 3D shapes is a fundamental problem, with applications in computer vision, molecular biology, computer graphics, and a variety of other fields. A challenging aspect of this problem is to find a suitable shape signature that can be constructed and compared quickly, while still discriminating between similar and dissimilar shapes. In this paper, we propose and analyze a method for computing shape signatures for arbitrary (possibly degenerate) 3D polygonal models. The key idea is to represent the signature of an object as a shape distribution sampled from a shape function measuring the global geometric properties of an object. The primary motivation for this approach is to reduce the shape matching problem to the comparison of probability distributions, which is simpler than traditional shape matching methods that require pose registration, feature correspondence or model fitting. We find that the dissimilarities between sampled distributions of simple shape functions (e.g. the distance between two random points on a surface) provide a robust method for discriminating between classes of objects (e.g. cars versus airplanes) in a moderately sized database, despite the presence of arbitrary translations, rotations, scales, reflections, tessellations, simplifications and model degeneracies. They can be evaluated quickly, and thus the proposed method could be applied as a pre-classifier in an object recognition system or in an interactive content-based retrieval application.

We describe a system for non-photorealistic rendering (NPR) of virtual environments. In real time, it synthesizes imagery of architectural interiors using stroke-based textures. We address the four main challenges of such a system — interactivity, visual detail, controlled stroke size, and frame-to-frame coherence — through image based rendering (IBR) methods. In a preprocessing stage, we capture photos of a real or synthetic environment, map the photos to a coarse model of the environment, and run a series of NPR filters to generate textures. At runtime, the system re-renders the NPR textures over the geometry of the coarse model, and it adds dark lines that emphasize creases and silhouettes. We provide a method for constructing non-photorealistic textures from photographs that largely avoids seams in the resulting imagery. We also offer a new construction, art-maps, to control stroke size across the images. Finally, we show a working system that provides an immersive experience rendered in a variety of NPR styles.

Geometric acoustic modeling systems spatialize sounds according to reverberation paths from a sound source to a receiver to give an auditory impression of a virtual 3D environment. These systems are useful for concert hall design, teleconferencing, training and simulation, and interactive virtual environments. In many cases, such as in an interactive walkthrough program, the reverberation paths must be updated within strict timing constraints ‐ e.g., as the sound receiver moves under interactive control by a user. In this paper, we describe a geometric acoustic modeling algorithm that uses a priority queue to trace polyhedral beams representing reverberation paths in best‐first order up to some termination criteria (e.g., expired time‐slice). The advantage of this algorithm is that it is more likely to find the highest priority reverberation paths within a fixed time‐slice, avoiding many geometric computations for lower‐priority beams. Yet, there is overhead in computing priorities and managing the priority queue. The focus of this paper is to study the trade‐offs of the priority‐driven beam tracing algorithm with different priority functions. During experiments computing reverberation paths between a source and a receiver in a 3D building environment, we find that priority functions incorporating more accurate estimates of source‐to‐receiver path length are more likely to find early reverberation paths useful for spatialization, especially in situations where the source and receiver cannot reach each other through trivial reverberation paths. However, when receivers are added to the environment such that it becomes more densely and evenly populated, this advantage diminishes.

Realistic acoustic modeling is essential for spatializing sound in distributed virtual environments where multiple networked users move around and interact visually and aurally in a shared vir-tual world. Unfortunately, current methods for computing accurate acoustical models are not fast enough for real-time auralization of sounds for simultaneously moving sources and receivers. In this pa-per, we present three new beam tracing algorithms that greatly ac-celerate computation of reverberation paths in a distributed virtual environment by taking advantage of the fact that sounds can only be generated or heard at the positions of “avatars ” representing the users. The priority-driven beam tracing algorithm performs a best-first search of a cell adjacency graph, and thus enables new termina-tion criteria with which all early reflection paths can be found very efficiently. The bidirectional beam tracing algorithm combines sets of beams traced from pairs of avatar locations to find reverbera-tion paths between them while requiring significantly less compu-tation than previous unidirectional algorithms. The amortized beam tracing algorithm computes beams emanating from box-shaped re-gions of space containing predicted avatar locations and re-uses those beams multiple times to compute reflections paths as each avatar moves inside the box. Cumulatively, these algorithms enable speedups of approximately two orders of magnitude over previous methods. They are incorporated into a time-critical multiprocess-ing system that allocates its computational resources dynamically in order to compute the highest priority reverberation paths between moving avatar locations in real-time with graceful degradation and adaptive refinement. Key Words: Virtual environment systems, virtual reality, acoustic modeling, auralization, beam tracing.

Virtual environment research has focused on interactive image generation and has largely ignored acoustic modeling for spatialization of sound.Yet, realistic auditory cues can complement and enhance visual cues to aid navigation, comprehension, and sense of presence in virtual environments.A primary challenge in acoustic modeling is computation of reverberation paths from sound sources fast enough for real-time auralization.We have developed a system that uses precomputed spatial subdivision and "beam tree" data structures to enable real-time acoustic modeling and auralization in interactive virtual environments.The spatial subdivision is a partition of 3D space into convex polyhedral regions (cells) represented as a cell adjacency graph.A beam tracing algorithm recursively traces pyramidal beams through the spatial subdivision to construct a beam tree data structure representing the regions of space reachable by each potential sequence of transmission and specular reflection events at cell boundaries.From these precomputed data structures, we can generate high-order specular reflection and transmission paths at interactive rates to spatialize fixed sound sources in real-time as the user moves through a virtual environment.Unlike previous acoustic modeling work, our beam tracing method: 1) supports evaluation of reverberation paths at interactive rates, 2) scales to compute highorder reflections and large environments, and 3) extends naturally to compute paths of diffraction and diffuse reflection efficiently.We are using this system to develop interactive applications in which a user experiences a virtual environment immersively via simultaneous auralization and visualization.

Consistent repreaentations of the boundary and interior of thredimensional solid objects are required by applications ramging from interactive visualization to finite element analysis. However, most commonly available models of solid objects contain errors and inconsistencies. We describe an algorithm that automatically constructs consistent representations of the solid objects modeled by an arbitrary set of polygons. The key feature of our algorithm is that it first partitions space into a set of polyhedral regions and then determines which regions are solid based on region adjacency relationships. Fromthe solid polyhedral regions, we are able to output umsistent boundary and solid representations in a variety of iile formats. Unlike previous approaches, our solid-based approach is effective even when the input polygons intersect, overlap, are wrongly-oriented, have T-junctions, or are unconnected.

Article Coarse-grained parallelism for hierarchical radiosity using group iterative methods Share on Author: Thomas A. Funkhouser Bell Laboratories, Murray Hill, NJ Bell Laboratories, Murray Hill, NJView Profile Authors Info & Claims SIGGRAPH '96: Proceedings of the 23rd annual conference on Computer graphics and interactive techniquesAugust 1996 Pages 343–352https://doi.org/10.1145/237170.237273Online:01 August 1996Publication History 30citation402DownloadsMetricsTotal Citations30Total Downloads402Last 12 Months2Last 6 weeks0 Get Citation AlertsNew Citation Alert added!This alert has been successfully added and will be sent to:You will be notified whenever a record that you have chosen has been cited.To manage your alert preferences, click on the button below.Manage my AlertsNew Citation Alert!Please log in to your account Save to BinderSave to BinderCreate a New BinderNameCancelCreateExport CitationPublisher SiteGet Access

This paper describes the client-server design, implementation and experimental results for a system that supports real-time visual interaction between a large number of users in a shared 3D virtual environment. The key feature of the system is that server-based visibility algorithms compute potential visual interactions between entities representing users in order to reduce the number of messages required to maintain consistent state among many workstations distributed across a wide-area network. When an entity changes state, update messages are sent only to workstations with entities that can potentially perceive the change—i.e., ones to which the update is visible. Initial experiments show a 40x decrease in the number of messages processed by client workstations during tests with 1024 entities interacting in a large densely occluded virtual environment.

We describe a system that computes radiosity solutions for polygonal environments much larger than can be stored in main memory. The solution is stored in and retrieved from a database as the computation proceeds. Our system is based on two ideas: the use of visibility oracles to find source and blocker surfaces potentially visible to a receiving surface; and the use of hierarchical techniques to represent interactions between large surfaces efficiently, and to represent the computed radiosity solution compactly. Visibility information allows the environment to be partitioned into subsets, each containing all the information necessary to transfer light to a cluster of receiving polygons. Since the largest subset needed for any particular cluster is much smaller than the total size of the environment, these subset computations can be performed in much less memory than can classical or hierarchical radiosity. The computation is then ordered for further efficiency. Careful ordering of energy transfers minimizes the number of database reads and writes. We report results from large solutions of unfurnished and furnished buildings, and show that our implementation's observed running time scales nearly linearly with both local and global model complexity.

We describe an adaptive display algorithm for interactive frame rates during visualization of very complex virtual environments.The algorithm relies upon a hierarchical model representation in which objects are described at multiple levels of detail and can be drawn with various rendering algorithms.The idea behind the algorithm is to adjust image quality adaptively to maintain a uniform, user-specified target frame rate.We perform a constrained optimization to choose a level of detail and rendering algorithm for each potentially visible object in order to generate the "best" image possible within the target frame time.Tests show that the algorithm generates more uniform frame rates than other previously described detail elision algorithms with little noticeable difference in image quality during visualization of complex models.

We describe techniques for managing large amounts of data during an interactive walkthrough of an architectural model.These techniques are based on a spatial subdivision, visibility analysis, and a display database containing objects described at multiple levels of detail.In each frame of the walkthrough, we compute a set of objects to render, i.e. those potentially visible from the observer's viewpoint, and a set of objects to swap into memory, i.e. those that might become visible in the near future.We choose an appropriate level of detail at which to store and to render each object, possibly using very simple representations for objects that appear small to the observer, thereby saving space and time.Using these techniques, we cull away large portions of the model that are irrelevant from the observer's viewpoint, and thereby achieve interactive frame rates.