VMV17

Permanent URI for this collection

https://diglib.eg.org/handle/10.2312/2631688

Browse

Now showing 1 - 20 of 21

Appearance Bending: A Perceptual Editing Paradigm for Data-Driven Material Models
(The Eurographics Association, 2017) Mylo, Marlon; Giesel, Martin; Zaidi, Qasim; Hullin, Matthias; Klein, Reinhard; Matthias Hullin and Reinhard Klein and Thomas Schultz and Angela Yao
Data-driven representations of material appearance play an important role in a wide range of applications. Unlike with analytical models, however, the intuitive and efficient editing of tabulated reflectance data is still an open problem. In this work, we introduce appearance bending, a set of image-based manipulation operators, such as thicken, inflate, and roughen, that implement recent insights from perceptual studies. In particular, we exploit a link between certain perceived visual properties of a material, and specific bands in its spectrum of spatial frequencies or octaves of a wavelet decomposition. The result is an editing interface that produces plausible results at interactive rates, even for drastic manipulations. We present the effectiveness of our method on a database of bidirectional texture functions (BTFs) for a variety of material samples.
C++ Compile Time Polymorphism for Ray Tracing
(The Eurographics Association, 2017) Zellmann, Stefan; Lang, Ulrich; Matthias Hullin and Reinhard Klein and Thomas Schultz and Angela Yao
Reducing the amount of conditional branching instructions in innermost loops is crucial for high performance code on contemporary hardware architectures. In the context of ray tracing algorithms, typical examples for branching in inner loops are the decisions what type of primitive a ray should be tested against for intersection, or which BRDF implementation should be evaluated at a point of intersection. Runtime polymorphism, which is often used in those cases, can lead to highly expressive but poorly performing code. Optimization strategies often involve reduced feature sets (e.g. by simply supporting only a single geometric primitive type), or an upstream sorting step followed by multiple ray tracing kernel executions, which effectively places the branching instruction outside the inner loop. In this paper we propose C++ compile time polymorphism as an alternative optimization strategy that does on its own not reduce branching, but that can be used to write highly expressive code without sacrificing optimization potential such as early binding or inlining of tiny functions. We present an implementation with modern C++ that we integrate into a ray tracing template library. We evaluate our approach on CPU and GPU architectures.
Compression of Non-Manifold Polygonal Meshes Revisited
(The Eurographics Association, 2017) Buelow, Max von; Guthe, Stefan; Goesele, Michael; Matthias Hullin and Reinhard Klein and Thomas Schultz and Angela Yao
Polygonal meshes are used in various fields ranging from CAD to gaming and web based applications. Reducing the size required for storing and transmitting these meshes by taking advantage of redundancies is an important aspect in all of these cases. In this paper, we present a connectivity based compression approach that predicts attributes and stores differences to the predictions together with minimal connectivity information. It is an extension to the Cut-Border Machine and applicable to arbitrary manifold and non-manifold polygonal meshes containing multiple attributes of different types. It compresses both the connectivity and attributes without loss outside of re-ordering vertices and polygons. In addition, an optional quantization step can be used to further reduce the data if a certain loss of accuracy is acceptable. Our method outperforms state-of-the-art compression techniques, including specialized triangle mesh compression approaches when applicable. Typical compression rates for our approach range from 2:1 to 6:1 for lossless compression and up to 25:1 when quantizing to 14 bit accuracy.
Data Driven Synthesis of Hand Grasps from 3-D Object Models
(The Eurographics Association, 2017) Majumder, Soumajit; Chen, Haojiong; Yao, Angela; Matthias Hullin and Reinhard Klein and Thomas Schultz and Angela Yao
Modeling and predicting human hand grasping interactions is an active area of research in robotics, computer vision and computer graphics. We tackle the problem of predicting plausible hand grasps and the contact points given an input 3-D object model. Such a prediction task can be difficult due to the variations in the 3-D structure of daily use objects as well as the different ways that similar objects can be manipulated. In this work, we formulate grasp synthesis as a constrained optimization problem which takes into account the anthropomorphic and kinematic limitations of a human hand as well as the local and global geometric properties of the interacting object. We evaluate our proposed algorithm on twelve 3-D object models of daily use and demonstrate that our algorithm can successfully predict plausible hand grasps and contact points on the object.
Dense and Scalable Reconstruction from Unstructured Videos with Occlusions
(The Eurographics Association, 2017) Wei, Jian; Resch, Benjamin; Lensch, Hendrik P. A.; Matthias Hullin and Reinhard Klein and Thomas Schultz and Angela Yao
Depth-map-based multi-view stereo algorithms typically recover textureless surfaces by assuming smoothness per view, so they require processing different views to solve occlusions. Moreover, the highly redundant viewpoints of videos make exhaustive calculation of depth maps unfeasible for large scenes. This paper achieves dense and scalable reconstruction from videos by adaptively selecting a minimum subset of views from the unstructured camera paths, that are most beneficial for incremental occlusion handling and coverage improvement. Furthermore, we simplify and optimize each set of locally consistent points as the points accumulated from a cluster of previously processed views. By combining content-aware view selection and clustering, as well as cluster-wise point merging, our approach can reduce both computational and memory costs while producing accurate, concise, and dense 3D points, even for homogeneous areas. The superior efficiency and point-level fashion of our operations facilitate 3D modeling at large scales.
Efficient Lifted Relaxations of the Quadratic Assignment Problem
(The Eurographics Association, 2017) Burghard, Oliver; Klein, Reinhard; Matthias Hullin and Reinhard Klein and Thomas Schultz and Angela Yao
Quadratic assignment problems (QAPs) and quadratic assignment matchings (QAMs) recently gained a lot of interest in computer graphics and vision, e.g. for shape and graph matching. Literature describes several convex relaxations to approximate solutions of the NP-hard QAPs in polynomial time. We compare the convex relaxations recently introduced in computer graphics and vision to established approaches in discrete optimization. Building upon a unified constraint formulation we theoretically analyze their solution spaces and their approximation quality. Experiments on a standard benchmark as well as on instances of the shape matching problems support our analysis. It turns out that often the bounds of a tight linear relaxation are competitive with the bounds of semidefinite programming (SDP) relaxations, while the linear relaxation is often much faster to calculate. Indeed, for many instances the bounds of the linear relaxation are only slightly worse than the SDP relaxation of Zhao [ZKRW98,PR09], which itself is at least as accurate as the relaxations currently used in computer graphics and vision. Solving the SDP relaxations can often be accelerated considerably from hours to minutes using the recently introduced approximation method for trace bound SDPs [WSvdHT16], but nonetheless calculating linear relaxations is faster in most cases. For the shape matching problem all relaxations generate the optimal solution, only that the linear relaxation does so faster. Our results generalize as well to QAMs for which we deliver new relaxations. Furthermore by interpreting the Product Manifold Filter [VLR∗17] in the context of QAPs we show how to automatically calculate correspondences between shapes of several hundred points.
Finite Time Steady Vector Field Topology - Theoretical Foundation and 3D Case
(The Eurographics Association, 2017) Friederici, Anke; Günther, Tobias; Rössl, Christian; Theisel, Holger; Matthias Hullin and Reinhard Klein and Thomas Schultz and Angela Yao
Vector Field Topology is the standard approach for the analysis of asymptotic particle behavior in a vector field flow: A topological skeleton is separating the flow into regions by the movement of massless particles for an integration time converging to infinity. In some use cases however only a finite integration time is feasible. To this end, the idea of a topological skeleton with an augmented finite-time separation measure was introduced for 2D vector fields. We lay the theoretical foundation for that method and extend it to 3D vector fields. From the observation of steady vector fields in a temporal context we show the Galilean invariance of Vector Field Topology. In addition, we present a set of possible visualizations for finite-time topology on 3D topological skeletons.
Improved Image Classification using Topological Persistence
(The Eurographics Association, 2017) Dey, Tamal Krishna; Mandal, Sayan; Varcho, William; Matthias Hullin and Reinhard Klein and Thomas Schultz and Angela Yao
Image classification has been a topic of interest for many years. With the advent of Deep Learning, impressive progress has been made on the task, resulting in quite accurate classification. Our work focuses on improving modern image classification techniques by considering topological features as well. We show that incorporating this information allows our models to improve the accuracy, precision and recall on test data, thus providing evidence that topological signatures can be leveraged for enhancing some of the state-of-the art applications in computer vision.
Improving Layout Quality by Mixing Treemap-Layouts Based on Data-Change Characteristics
(The Eurographics Association, 2017) Bethge, Joseph; Hahn, Sebastian; Döllner, Jürgen; Matthias Hullin and Reinhard Klein and Thomas Schultz and Angela Yao
This paper presents a hybrid treemap layout approach that optimizes layout-quality metrics by combining state-of-the-art treemap layout algorithms. It utilizes machine learning to predict those metrics based on data metrics describing the characteristics and changes of the dataset. For this, the proposed approach uses a neural network which is trained on artificially generated dataset,s containing a total of 15.8 million samples. The resulting model is integrated into an approach called Smart- Layouting. This approach is evaluated on real-world data from 100 publicly available software repositories. Compared to other state-of-the-art treemap algorithms it reaches an overall better result. Additionally, this approach can be customized by an end user's needs. The customization allows for specifying weights for the importance of each layout-quality metric. The results indicate, that the algorithm is able to adapt successfully towards a given set of weights.
Interactive Visualization of Gaps and Overlaps for Large and Dynamic Sphere Packings
(The Eurographics Association, 2017) Gu, Feng; Yang, Zhixing; Kolonko, Michael; Grosch, Thorsten; Matthias Hullin and Reinhard Klein and Thomas Schultz and Angela Yao
To gain insight into many properties of granular matter, a particle packing can be simulated. For a dry particle mixture, collective rearrangement is often used as an iterative process to place the particles. In this paper, we present a new visualization technique to judge the quality of a collective rearrangement simulation of many spheres with a given particle size distribution. In addition to a visualization of the spheres themselves, we directly visualize the gaps and overlaps of the spheres in each iteration. This allows to see the regions where the simulation is not yet converged as well as the free spaces where spheres can still move into. Our method supports millions of spheres at interactive to real-time frame rates, allowing the user to inspect the sphere packing during the simulation.We demonstrate that this type of visualization better shows the structure of the current sphere arrangement than standard techniques like 2D clipping planes and therefore serves as a visual feedback to support the development of the packing simulation.
LiteMaker: Interactive Luminaire Development using Progressive Photon Tracing and Multi-Resolution Upsampling
(The Eurographics Association, 2017) Krösl, Katharina; Luksch, Christian; Schwärzler, Michael; Wimmer, Michael; Matthias Hullin and Reinhard Klein and Thomas Schultz and Angela Yao
Industrial applications like luminaire development (the creation of a luminaire in terms of geometry and material) or lighting design (the efficient and aesthetic placement of luminaires in a virtual scene) rely heavily on high realism and physically correct simulations. Using typical approaches like CAD modeling and offline rendering, this requirement induces long processing times and therefore inflexible workflows. In this paper, we combine a GPU-based progressive photon-tracing algorithm to accurately simulate the light distribution of a luminaire with a novel multi-resolution image-filtering approach that produces visually meaningful intermediate results of the simulation process. By using this method in a 3D modeling environment, luminaire development is turned into an interactive process, allowing for real-time modifications and immediate feedback on the light distribution. Since the simulation results converge to a physically plausible solution that can be imported as a representation of a luminaire into a light-planning software, our work contributes to combining the two former decoupled workflows of luminaire development and lighting design, reducing the overall production time and cost for luminaire manufacturers.
Pixel Cache Light Tracing
(The Eurographics Association, 2017) Jendersie, Johannes; Rohmer, Kai; Brüll, Felix; Grosch, Thorsten; Matthias Hullin and Reinhard Klein and Thomas Schultz and Angela Yao
In this paper, we introduce Pixel Cache Light Tracing, which is a new low-noise combination of eye-path and light-path tracing. In the first pass, eye-path vertices are distributed from the observer and stored in a hit point map analogous to progressive photon mapping. In the second pass, photons are traced from the light source and projected to the image as well as gathered by the hit point map. We combine the paths from both sampling strategies in a deterministic way without multiple importance sampling, such that the final result is consistent and free from firefly artifacts. In many practical cases, this combination leads to sharper caustics and reduced noise when compared to alternative techniques at equal time. Further, the simplicity of the path combination strategy is predestined for GPU-based implementations and requires less memory than a comparable photon mapping implementation. In addition, we provide a fast, parallel and lean hash map implementation for both photon and hit point queries.
Semantic-Aware Image Smoothing
(The Eurographics Association, 2017) Li, Weihao; Jafari, Omid Hosseini; Rother, Carsten; Matthias Hullin and Reinhard Klein and Thomas Schultz and Angela Yao
Structure-preserving image smoothing aims to extract semantically meaningful image structure from texture, which is one of the fundamental problems in computer vision and graphics. However, it is still not clear how to define this concept. On the other hand, semantic image labeling has achieved significant progress recently and has been widely used in many computer vision tasks. In this paper, we present an interesting observation, i.e. high-level semantic image labeling information can provide a meaningful structure prior naturally. Based on this observation, we propose a simple and yet effective method, which we term semantic smoothing, by exploiting the semantic information to accomplish semantically structure-preserving image smoothing. We show that our approach outperforms the state-of-the-art approaches in texture removal by considering the semantic information for structure preservation. Also, we apply our approach to three applications: detail enhancement, edge detection, and image segmentation, and we demonstrate the effectiveness of our semantic smoothing method on these problems.
Star Convex Cuts with Encoding Swaps for Fast Whole-Spine Vertebrae Segmentation in MRI
(The Eurographics Association, 2017) Rak, Marko; Tönnies, Klaus D.; Matthias Hullin and Reinhard Klein and Thomas Schultz and Angela Yao
We propose an automatic approach for fast vertebral body segmentation in three-dimensional magnetic resonance images of the whole spine. Previous works are limited to the lower thoracolumbar section and often take minutes to compute, which can be problematic in clinical routine or for data sets with numerous subjects. We address these limitations by a graph cut formulation. Our formulation involves appearance and shape information as well as star-convexity constraints to ensure a topologically correct segmentation for each vertebra. For close targets such as adjacent vertebrae, implementing star-convexity without fusing targets (naive binary formulations) or increasing run time/loosing optimality guarantees (multi-label formulations) is challenging. We provide a solution based on encoding swaps, which preserve optimality and ensure topological correctness between vertebrae. We validated our approach on two data sets. The first contains T1- and T2-weighted whole-spine images of 64 subjects. The second comprises 23 T2-weighted thoracolumbar images and is publicly available. Our results are competitive to previous works (or better) at a fraction of the run time. We yielded Dice coefficients of 85:1 +/- 4:4% and 89:7 +/- 2:3% with run times of 1:65 +/- 0:28 s and 2:73 +/- 0:36 s per vertebra on consumer hardware.
Structure-aware Stylization of Mountainous Terrains
(The Eurographics Association, 2017) Kratt, Julian; Eisenkeil, Ferdinand; Spicker, Marc; Wang, Yunhai; Weiskopf, Daniel; Deussen, Oliver; Matthias Hullin and Reinhard Klein and Thomas Schultz and Angela Yao
We present a method for the stylization of mountainous terrains that allows creating abstract representations in different rendering styles. Our method consists of two major components: structure-aware terrain filtering and streamline-based hatching. For a given input terrain we compute different Levels-of-Detail (LoD) according to a crest line oriented importance measure and then filter each LoD accordingly. We generate flow fields for each LoD and compute streamlines to direct the production of hatching lines. The combination of crest and silhouette lines with streamline-based hatching allows us to create a variety of styles in different Levels-of-Detail. We evaluate our method using several terrains and demonstrate the effectiveness of our method by composing a number of different illustration styles.
Template-Based 3D Non-Rigid Shape Estimation from Monocular Image Sequences
(The Eurographics Association, 2017) Kausch, Lisa; Hilsmann, Anna; Eisert, Peter; Matthias Hullin and Reinhard Klein and Thomas Schultz and Angela Yao
This paper addresses the problem of reconstructing non-rigid 3D geometries from temporal image sequences captured with only a single camera under full perspective projection. Without the knowledge of a shape deformation model, this task is severly under-constrained, because multiple shape configurations can produce the same image projections. The challenge remains even if a template 3D model of the static, un-deformed state is available, because the depth along the line of sight is unkown. Often, this is handled by assuming an orthographic camera model. In contrast, we address a full perspective camera model. Also, our reconstruction is not limited to the model parts that are visible in the current image, but deformation is estimated for the entire template across the temporal sequence. In a first step, we compute a template of the geometry in un-deformed pose, assuming that the object was captured while being static. Next, the object starts to deform while being captured by a single camera, and the non-rigid shape is reconstructed sequentially by estimating the camera position and the deformations with respect to the template model. Our objective minimization function combines image data and temporal consistency information, and constrains the deformation space by a rotation-invariant volumetric graph Laplacian and as-rigid-as-possible constraints defined on the tesselation of the template model. The method is evaluated on synthetic and real data, including different object classes, thereby concentrating on the class of articulated deformations.
Temporal Focus+Context for Clusters in Particle Data
(The Eurographics Association, 2017) Staib, Joachim; Grottel, Sebastian; Gumhold, Stefan; Matthias Hullin and Reinhard Klein and Thomas Schultz and Angela Yao
Time-dependent particle-based simulations are typically carried out by direct calculation of interactions between particles over time. The investigation of higher order effects of particle clusters helps understanding the system's dynamic. Existing methods for particle data analysis either rely on animation, where only one time step is visible, or abstraction, which is giving up on visualizing the data in its spatial domain. Inspired from illustrative techniques, we present an interactive focus+context visualization, based on flow ribbons, that combines both approaches. Our method jointly shows one time step in detail, as well as an abstract contextual visualization of past and future dynamics in one image. It allows to assess the time evolution of various cluster attributes around the current temporal focus. We show the usefulness of the approach on two exemplary case studies.
User-study Based Optimization of Fast and Accurate Mahalanobis Brushing in Scatterplots
(The Eurographics Association, 2017) Fan, Chaoran; Hauser, Helwig; Matthias Hullin and Reinhard Klein and Thomas Schultz and Angela Yao
Brushing is at the heart of most modern visual analytics solutions with coordinated, multiple views and effective brushing is crucial for swift and efficient processes in data exploration and analysis. Given a certain data subset that the user wishes to brush in a data visualization, traditional brushes are usually either accurate (like the lasso) or fast (e.g., a simple geometry like a rectangle or circle). In this paper, we now present a new, fast and accurate brushing technique for scatterplots, based on the Mahalanobis brush, which we have extended and then optimized using data from a user study. We explain the principal, sketchbased model of our new brushing technique (based on a simple click-and-drag interaction), the details of the user study and the related parameter optimization, as well as a quantitative evaluation, considering efficiency, accuracy, and also a comparison with the original Mahalanobis brush.
Visualization of Cardiac Blood Flow Using Anisotropic Ambient Occlusion for Lines
(The Eurographics Association, 2017) Köhler, Benjamin; Grothoff, Matthias; Gutberlet, Matthias; Preim, Bernhard; Matthias Hullin and Reinhard Klein and Thomas Schultz and Angela Yao
Ambient occlusion (AO) for lines (LineAO) was introduced by Eichelbaum et al. [EHS13] as an adaption of screen-space AO to static line bundles, such as white brain matter fiber tracts derived from diffusion tensor imaging (DTI). In this paper, we further adapt the LineAO technique to dynamic scenes, in particular the animation of blood flow-representing pathlines that were integrated in cardiac 4D phase-contrast magnetic resonance imaging (PC-MRI) data. 4D PC-MRI is a non-invasive technique that allows to acquire time-resolved blood flow velocity data in all three spatial dimensions, i.e., a 4D vector field of one heart beat. Our main extension is a line alignment factor that reduces the AO-induced darkening if nearby lines have similar screen-space tangents. We further enhance the perception of homogeneous flow by incorporating depth-dependent halos. Our technique facilitates the quicker identification of prominent flow structures while showing the full flow context.
Visualization of Neural Network Predictions for Weather Forecasting
(The Eurographics Association, 2017) Roesch, Isabelle; Günther, Tobias; Matthias Hullin and Reinhard Klein and Thomas Schultz and Angela Yao
Recurrent neural networks are prime candidates for learning relationships and evolutions in multi-dimensional time series data. The performance of such a network is judged by the loss function, which is aggregated into a single scalar value that decreases during successful training. Observing only this number hides the variation that occurs within the typically large training and testing data sets. Understanding these variations is of highest importance to adjust hyperparameters of the network, such as the number of neurons, number of layers or even to adjust the training set to include more representative examples. In this paper, we design a comprehensive and interactive system that allows to study the output of recurrent neural networks on both the complete training data as well as the testing data. We follow a coarse-to-fine strategy, providing overviews of annual, monthly and daily patterns in the time series and directly support a comparison of different hyperparameter settings. We applied our method to a recurrent convolutional neural network that was trained and tested on 25 years of climate data to forecast meteorological attributes, such as temperature, pressure and wind speed. The presented visualization system helped us to quickly assess, adjust and improve the network design.

Browse

Browsing VMV17 by Title

Results Per Page

Sort Options