VMV2021

Permanent URI for this collection

https://diglib.eg.org/handle/10.2312/2633087

Browse

Now showing 1 - 15 of 15

Axes Bundling and Brushing in Star Coordinates
(The Eurographics Association, 2021) Rave, Hennes; Molchanov, Vladimir; Linsen, Lars; Andres, Bjoern and Campen, Marcel and Sedlmair, Michael
Visual analysis of multidimensional data commonly involves dimensionality reduction to project the data samples into a lowerdimensional visual space. Star coordinates (SC) provide a means to explore the multidimensional data distribution by interactively changing the linear projection matrix. While SC have the advantages of being intuitive, allowing for relating the data samples to their original dimensions, having low computation costs, and scaling well with the number of data samples, they have the disadvantages of not scaling well to larger number of dimensions and being restricted to linear projections. We address these short-comings by introducing novel SC interactions. First, interactive bundling of axes is proposed to reduce the number of dimensions. While bundles are fully customizable, the bundling interactions are supported by visualizations of correlation matrices and hierarchical axes clustering dendrograms. Second, we enhance classical region brushing in SC projections with axes brushing, which allows for multidimensional cluster selection, even if two (separable) clusters are projected to the same area of the visible space. Axes brushing is supported by visualizing 1D histograms of data distributions along the SC axes. Our brushing interactions alleviate the restriction of SC to linear projections. The integration of histograms into SC also eases other interactions such as moving axes to change the projection matrix. A user study evaluates how analysis tasks for labeled and unlabeled multidimensional data can benefit from our extensions.
Capturing Anisotropic SVBRDFs
(The Eurographics Association, 2021) Kaltheuner, Julian; Bode, Lukas; Klein, Reinhard; Andres, Bjoern and Campen, Marcel and Sedlmair, Michael
In this work, we adapt and improve recent isotropic material estimation efforts to estimate spatially varying anisotropic materials with an additional Fresnel term using a variable set of input images and are able to handle any resolution. We combine an initial estimation network with an auto-encoder to fine-tune the decoding of latent embedded appearance parameters on the input images to produce finely detailed SVBRDFs. For this purpose, the training must be adapted so that the determination is possible on the basis of a small number of images that still capture as much reflective behavior of materials as possible. The resulting appearance parameters are capable of capturing and reconstructing complex spatially varying features in detail, but place increased demands on the input images.
Context Aware Exemplar-based Image Inpainting using Irregular Patches
(The Eurographics Association, 2021) Fotsing, Cedrique; Cunningham, Douglas; Andres, Bjoern and Campen, Marcel and Sedlmair, Michael
We propose a new exemplar-based image inpainting method in this paper. Our method is based on the Criminisi pipeline. We focused on three main stages of the pipeline; calculation of priorities, construction of patches, and the search for the best match. To assign a high priority to patches constructed from the edge pixels, we use the ability of segmentation algorithms to divide an image into different texture blocks. The patches built from pixels located at the border between several texture blocks receive a high priority. Unlike most patch-based image inpainting methods which use regular patches (rectangle, square), the shape and size of our patches depend on the textural composition around the original pixel. The patches are built using a region growing principle in the different texture blocs around the original pixel. The search for the best match is done contextually. We search for the best match of the patch with the highest priority in a similar environment to its neighborhood around the target zone. The method is simple and easy to implement. The experiments show that our method obtains more plausible results than the basic method of Criminisi and its improved version Amoeba in most cases.
CoSi: Visual Comparison of Similarities in High-Dimensional Data Ensembles
(The Eurographics Association, 2021) Heim, Anja; Gröller, Eduard; Heinzl, Christoph; Andres, Bjoern and Campen, Marcel and Sedlmair, Michael
Comparative analysis of multivariate datasets, e.g. of advanced materials regarding the characteristics of internal structures (fibers, pores, etc.), is of crucial importance in various scientific disciplines. Currently domain experts in materials science mostly rely on sequential comparison of data using juxtaposition. Our work assists domain experts to perform detailed comparative analyses of large ensemble data in materials science applications. For this purpose, we developed a comparative visualization framework, that includes a tabular overview and three detailed visualization techniques to provide a holistic view on the similarities in the ensemble. We demonstrate the applicability of our framework on two specific usage scenarios and verify its techniques using a qualitative user study with 12 material experts. The insights gained from our work represent a significant advancement in the field of comparative material analysis of high-dimensional data. Our framework provides experts with a novel perspective on the data and eliminates the need for time-consuming sequential exploration of numerical data.
EMCA: Explorer of Monte Carlo based Algorithms
(The Eurographics Association, 2021) Ruppert, Lukas; Kreisl, Christoph; Blank, Nils; Herholz, Sebastian; Lensch, Hendrik P. A.; Andres, Bjoern and Campen, Marcel and Sedlmair, Michael
Debugging or analyzing the performance of global illumination algorithms is a challenging task due to the complex path-scene interaction and numerous places where errors and programming bugs can occur. We present a novel, lightweight visualization tool to aid in the understanding of global illumination and the debugging of rendering frameworks. The tool provides detailed information about intersections and light transport paths. Users can add arbitrary data of their choosing to each intersection, based on their specific demands. Aggregate plots allow users to quickly discover and select outliers for further inspection across the globally linked visualization views. That information is further coupled with 3D visualization of the scene where additional aggregated information on the surfaces can be inspected in false colors. These include 3D heat maps such as the density of intersections as well as more advanced colorings such as a diffuse transport approximation computed from local irradiance samples and diffuse material approximations. The necessary data for the 3D coloring is collected as a side-product of quickly rendering the image at low sample counts without significantly slowing down the rendering process. It requires almost no precomputation and very little storage compared to point cloud-based approaches. We present several use cases of how novices and advanced rendering researchers can leverage the presented tool to speed up their research.
FERMIUM: A Framework for Real-time Procedural Point Cloud Animation and Morphing
(The Eurographics Association, 2021) Wegen, Ole; Böttger, Florence; Döllner, Jürgen; Trapp, Matthias; Andres, Bjoern and Campen, Marcel and Sedlmair, Michael
This paper presents a framework for generating real-time procedural animations and morphing of 3D point clouds. Point clouds or point-based geometry of varying density can easily be acquired using LiDAR cameras or modern smartphones with LiDAR sensors. This raises the question how this raw data can directly be used in the creative industry to create novel digital content using animations. For this purpose, we describe a framework that enables the implementation and combination of animation effects for point clouds. It takes advantage of graphics hardware capabilities and enables the processing of complex datasets comprising up to millions of points. In addition, we compare and evaluate implementation variants for the subsequent morphing of multiple 3D point clouds.
GPU-Parallel Constant-Time Limit Evaluation of Catmull-Clark Solids
(The Eurographics Association, 2021) Besler, Sebastian; Altenhofen, Christian; Stork, André; Fellner, Dieter W.; Andres, Bjoern and Campen, Marcel and Sedlmair, Michael
Subdivision solids, such as Catmull-Clark (CC) solids, are versatile volumetric representation schemes that can be employed for geometric modeling, physically based simulation, and multi-material additive manufacturing. With volumetric limit evaluation still being the performance bottleneck for these applications, we present a massively parallel approach to Altenhofen et al.'s constant-time limit evaluation method for CC solids. Our algorithm exploits the computational power of modern GPUs, while maintaining the mathematical concepts of Altenhofen et al.'s method. Distributing the computations for a single cell across multiple streaming multiprocessors (SMs) increases the utilization of the GPU's resources compared to straightforward parallelization. Specialized compute kernels for different topological configurations optimize shared memory usage and memory access. Our hybrid approach dynamically chooses the best kernel based on the topology and the evaluation parameters, resulting in speedups of between 5.75x and 61.58x compared to a CPU-parallel implementation of Altenhofen et al.'s method.
LayoutExOmizer: Interactive Exploration and Optimization of 2D Data Layouts
(The Eurographics Association, 2021) Schader, Philipp; Beckmann, Raphael; Graner, Lukas; Bernard, Jürgen; Andres, Bjoern and Campen, Marcel and Sedlmair, Michael
Reducing the overlap of data points in 2D visualizations while preserving original positions is a challenging task. Traditionally, hand-crafted solutions have been proposed while more recently layout algorithms with a high degree of automation have been introduced. However, with a continuous parameter space, the number of alternative solutions is virtually infinite. So which one is best? This assessment can depend on many factors, coined by subjective human judgment as well as quantitative quality measures. Our approach follows the idea to have both humans and algorithms in control, to combine the strengths of both. We propose LayoutExOmizer, which stands for Layout Explorer and Optimizer. It is a visual analytics approach that guides users in finding meaningful solutions. LayoutExOmizer supports users in generating a preferred layout by discovering a corresponding set of input parameters. This parameter search is supported by visual interfaces (1) to directly steer the parameters of the layout optimization, (2) to assess the quality of layouts using quality measures, (3) to relate input and out space, and (4) to filter layouts by their quality. We demonstrate the usefulness of our approach in two usage scenarios with different quality measures, including the full set of Scagnostics measures.
Real-Time Curvature-aware Re-Parametrization and Tessellation of Bézier Surfaces
(The Eurographics Association, 2021) Buchenau, Christoph; Guthe, Michael; Andres, Bjoern and Campen, Marcel and Sedlmair, Michael
Interactive tessellation of parametric surfaces has many applications in both engineering and entertainment computing. The most common primitives are bi-cubic Bézier patches which are, among others, an intermediate representation of subdivision surfaces for rendering. The current state-of-the-art employs hardware tessellation where a uniform subdivison pattern is used per patch. If the curvature varies strongly over a patch, this results in an over-tessellation of flat areas. Based on the observation that the second derivative changes linearly over the patch, we show that it is possible to reparameterize the patches such that the tessellation adapts to the curvature. This way, we reduce the number of primitives by an average of 15% for the same error bound.
Real-Time Gaussian-Product Subdivision on the GPU
(The Eurographics Association, 2021) Komar, Alexander; Preiner, Reinhold; Andres, Bjoern and Campen, Marcel and Sedlmair, Michael
We propose a real-time technique for rendering Gaussian-Product subdivision surfaces. This is achieved by our real-time subdivision pipeline, able to accept the base of Gaussian-Product subdivision, that is, a covariance mesh, which extends regular base meshes by storing additional 3x3 covariance matrices per vertex. Our technique evaluates the non-linear limit subdivision surface by computing B-spline patches embedded in a 9-dimensional dual space, where the subdivision scheme becomes linear. We construct and evaluate these B-spline patches using real-time tessellation capabilities of current GPUs. We analyzed the performance of our technique on all supported subdivision levels, and provide an analysis of its visual quality and geometric accuracy.
Refinable Multi-sided Caps for Bi-quadratic Splines
(The Eurographics Association, 2021) Karciauskas, Kestutis; Peters, Jörg; Andres, Bjoern and Campen, Marcel and Sedlmair, Michael
Subdivision surfaces based on bi-quadratic splines have a control net, the DS-net, whose irregularities are n-sided facets. To date their limit shape is poor due to a small footprint of the refinement rules and the difficulty of controlling shape at the center irregularity. By contrast, a control net where vertices are surrounded by n quadrilateral faces, a CC-net, admits higher-quality subdivision and finite polynomial constructions. It would therefore be convenient to leverage these constructions to fill holes in a C1 bi-quadratic spline complex. In principle the switch in layout from a control net with central n-sided facet to one with a central irregular point is easy: just apply one step of Catmull-Clark refinement. The challenge, however, is to define the transition between the bi-quadratic bulk and the n-sided cap construction to be of sufficiently good shape to not destroy the advantage of higher-quality algorithms. This challenge is addressed here by explicit formulas for conversion from a DS-net to a CC-net.
SuBloNet: Sparse Super Block Networks for Large Scale Volumetric Fusion
(The Eurographics Association, 2021) Rückert, Darius; Stamminger, Marc; Andres, Bjoern and Campen, Marcel and Sedlmair, Michael
Training and inference of convolutional neural networks (CNNs) on truncated signed distance fields (TSDFs) is a challenging task. Large parts of the scene are usually empty, which makes dense implementations inefficient in terms of memory consumption and compute throughput. However, due to the truncation distance, non-zero values are grouped around the surface creating small dense blocks inside the large empty space. We show that this structure can be exploited by storing the TSDF in a block sparse tensor and then decomposing it into rectilinear super blocks. A super block is a dense 3d cuboid of variable size and can be processed by conventional CNNs. We analyze the rectilinear decomposition and present a formulation for computing the bandwidth-optimal solution given a specific network architecture. However, this solution is NP-complete, therefore we also a present a heuristic approach for fast training and inference tasks. We verify the effectiveness of SuBloNet and report a speedup of 4x towards dense implementations and 1.7x towards state-of-the-art sparse implementations. Using the super block architecture, we show that recurrent volumetric fusion is now possible on large scale scenes. Such a systems is able to reconstruct high-quality surfaces from few noisy depth images.
Visual Comparison of Multi-label Classification Results
(The Eurographics Association, 2021) Krause, Cedric; Agarwal, Shivam; Ghoniem, Mohammad; Beck, Fabian; Andres, Bjoern and Campen, Marcel and Sedlmair, Michael
In multi-label classification, we do not only want to analyze individual data items but also the relationships between the assigned labels. Employing different sources and algorithms, the label assignments differ. We need to understand these differences to identify shared and conflicting assignments. We propose a visualization technique that addresses these challenges. In graphs, we present the labels for any classification result as nodes and the pairwise overlaps of labels as links between them. These graphs are juxtaposed for the different results and can be diffed graphically. Clustering techniques are used to further analyze similarities between labels or classification results, respectively. We demonstrate our prototype in two application examples from the machine learning domain.
Visualizing Temporal-Thematic Patterns in Text Collections
(The Eurographics Association, 2021) Knabben, Moritz; Baumann, Martin; Blascheck, Tanja; Ertl, Thomas; Koch, Steffen; Andres, Bjoern and Campen, Marcel and Sedlmair, Michael
Visualizing the temporal evolution of texts is relevant for many domains that seek to gain insight from text repositories. However, existing visualization methods for text collections do not show fine-grained temporal-thematic patterns. Therefore, we developed and analyzed a new visualization method that aims at uncovering such patterns. Specifically, we project texts to one dimension, which allows positioning texts in a 2D diagram of projection space and time. For projection, we employed two manifold learning algorithms: the self-organizing map (SOM) and UMAP. To assess the utility of our method, we experimented with real-world datasets and discuss the resulting visualizations. We find our method facilitates relating patterns and extracting associated texts beyond what is possible with previous techniques. We also conducted interviews with historians to show that our prototypical system supports domain experts in their analysis tasks.
VMV 2021: Frontmatter
(The Eurographics Association, 2021) Andres, Bjoern; Campen, Marcel; Sedlmair, Michael; Andres, Bjoern and Campen, Marcel and Sedlmair, Michael

Browse

Browsing VMV2021 by Title

Results Per Page

Sort Options