Eurographics Digital Library
Not a member yet
17121 research outputs found
Sort by
Narrative Medical Visualization to Communicate Vision Restoration to Patients: A Case Study
No full textNarrative medical visualization aims to make complex medical information accessible to a broad audience by combining storytelling with visual techniques. This approach bridges the communication gap between clinicians and patients, promoting patient collaboration and long-term recovery. Understanding the concept of therapy and its significance fosters a sense of ownership and empowerment, leading to increased motivation and engagement. However, clinicians face the challenge of providing patients with educational materials that explain complex medical mechanisms in a way that patients can understand and relate to their lifestyle. This case study focuses on an animation video prototype that demonstrates a therapy concept for patients with low vision, aiming to ensure that patients are well-informed and motivated to engage in the treatment process. An initial evaluation with patients showed a positive experience with the supplemental material and added value to the consultant's explanation.Eurographics Workshop on Visual Computing for Biology and MedicineSession
Real-Time Rendering of Algebraic Surfaces
No full textWe investigate the problem of robust and real-time rendering of algebraic surfaces. We show that expressing the intersection of the ray and the algebraic surface as a single univariate polynomial is not robust in practice, comparing results between monomial, Bernstein, Lagrange, and Chebyshev basis fits. We show that fitting multiple polynomials over subintervals, such as a unit length subdivision of the ray extent within the region of interest, improves robustness at a negligible performance cost.Eurographics 2025 - PostersPoster
Efficient GPU-Based Real-Time Rendering of Curved Geometry for Large-Scale Applications
No full textThis paper presents an efficient and practical approach to GPU-based real-time generation, simulation and rendering of approximated curved geometry. We focus on how this approach can be embedded into a resource-intensive application, such as a large-scale real-time strategy game, which requires a significantly large number of instances of these temporally persistent primitives, which we call ribbons. A key step in our technique is minimizing data transfer between the CPU and GPU, resulting in our usage of transient resources and employment of parallel data recycling methods, which map well to a SIMD architecture. Furthermore, we detail optimizations such as batching ribbon instances by material for rendering, grouping segments together for geometry construction and early culling of control points.Computer Graphics and Visual Computing (CGVC)Geometry, Rendering, Animatio
Multisensor 3D Documentation of the Palaeolithic Complex of the Railway Trench (Sierra de Atapuerca, Burgos, Spain)
No full textWe present the first comprehensive, high-resolution three-dimensional (3D) documentation of the Railway Trench at the Sierra de Atapuerca (Burgos, Spain) -one of Europe's most important palaeoarchaeological complexes- by integrating multisensor geomatic techniques. Although Atapuerca's rich fossil and lithic record has been studied since the late 1970s, detailed spatial data have remained limited to excavations. To overcome this, our study deploys a unified geospatial framework combining GNSS-RTK and total station surveying to support the application of Terrestrial Laser Scanning and Unmanned Aerial Vehicles photogrammetry within the 550 m-long trench and its immediate surroundings (50 ha). By delivering an unprecedentedly detailed and accurate digital twin (sub-decimetre resolution) of the palaeolithic context, digitalisation aims to enhance conservation, monitoring, virtual dissemination and future excavation planning. The results shown here demonstrate the potential of integrated multisensor workflows in Palaeolithic archaeology and are intended to safeguard and interpret the complex landscape of the Sierra de Atapuerca in future studies.Digital HeritagePoster
No More Shading Languages: Compiling C++ to Vulkan Shaders
No full textGraphics APIs have traditionally relied on shading languages, however, these languages have a number of fundamental defects and limitations. By contrast, GPU compute platforms offer powerful, feature-rich languages suitable for heterogeneous compute. We propose reframing shading languages as embedded domain-specific languages, layered on top of a more general language like C++, doing away with traditional limitations on pointers, functions, and recursion, to the benefit of programmability. This represents a significant compilation challenge because the limitations of shaders are reflected in their lower-level representations. We present the Vcc compiler, which allows conventional C and C++ code to run as Vulkan shaders. Our compiler is complemented by a simple shading library and exposes GPU particulars as intrinsics and annotations. We evaluate the performance of our compiler using a selection of benchmarks, including a real-time path tracer, achieving competitive performance compared to their native CUDA counterparts.High-Performance Graphics - Symposium PapersGraphics Simulators, Systems and Compiler
Optimizing Free-Form Grid Shells with Reclaimed Elements under Inventory Constraints
No full textWe propose a method for designing 3D architectural free-form surfaces, represented as grid shells with beams sourced from inventories of reclaimed elements from dismantled buildings. In inventory-constrained design, the reused elements must be paired with elements in the target design. Traditional solutions to this assignment problem often result in cuts and material waste or geometric distortions that affect the surface aesthetics and buildability. Our method for inventory-constrained assisted design blends the traditional assignment problem with differentiable geometry optimization to reduce cut-off waste while preserving the design intent. Additionally, we extend our approach to incorporate strain energy minimization for structural efficiency. We design differentiable losses that account for inventory, geometry, and structural constraints, and streamline them into a complete pipeline, demonstrated through several case studies. Our approach enables the reuse of existing elements for new designs, reducing the need for sourcing new materials and disposing of waste. Consequently, it can serve as an initial step towards mitigating the significant environmental impact of the construction sector.Computer Graphics ForumBuilt for Reality: Analyzing, Crafting and Fabricating Structures44
Joint Gaussian Deformation in Triangle-Deformed Space for High-Fidelity Head Avatars
No full textCreating 3D human heads with mesoscale details and high-fidelity animation from monocular or sparse multi-view videos is challenging. While 3D Gaussian splatting (3DGS) has brought significant benefits into this task, due to its powerful representation ability and rendering speed, existing works still face several issues, including inaccurate and blurry deformation, and lack of detailed appearance, due to difficulties in complex deformation representation and unreasonable Gaussian placement. In this paper, we propose a joint Gaussian deformation method by decoupling the complex deformation into two simpler deformations, incorporating a learnable displacement map-guided Gaussian-triangle binding and a neural-based deformation refinement, improving the fidelity of animation and details of reconstructed head avatars. However, renderings of reconstructed head avatars at unseen views still show artifacts, due to overfitting on sparse input views. To address this issue, we leverage synthesized pseudo views rendered with fitted textured 3DMMs as priors to initialize Gaussians, which helps maintain a consistent and realistic appearance across various views. As a result, our method outperforms existing state-of-the-art approaches with about 4.3 dB PSNR in novel-view synthesis and about 0.9 dB PSNR in self-reenactment on multi-view video datasets. Our method also preserves high-frequency details, exhibits more accurate deformations, and significantly reduces artifacts in unseen views.Eurographics Symposium on RenderingGaussian
LEAD: Latent Realignment for Human Motion Diffusion
No full textOur goal is to generate realistic human motion from natural language. Modern methods often face a trade-off between model expressiveness and text-to-motion (T2M) alignment. Some align text and motion latent spaces but sacrifice expressiveness; others rely on diffusion models producing impressive motions but lacking semantic meaning in their latent space. This may compromise realism, diversity and applicability. Here, we address this by combining latent diffusion with a realignment mechanism, producing a novel, semantically structured space that encodes the semantics of language. Leveraging this capability, we introduce the task of textual motion inversion to capture novel motion concepts from a few examples. For motion synthesis, we evaluate LEAD on HumanML3D and KIT-ML and show comparable performance to the state-of-the-art in terms of realism, diversity and textmotion consistency. Our qualitative analysis and user study reveal that our synthesised motions are sharper, more human-like and comply better with the text compared to modern methods. For motion textual inversion (MTI), our method demonstrates improvements in capturing out-of-distribution characteristics in comparison to traditional VAEs.Computer Graphics ForumOriginal Article44
Multimodal 3D Few-Shot Classification via Gaussian Mixture Discriminant Analysis
No full textWhile pre-trained 3D vision-language models are becoming increasingly available, there remains a lack of frameworks that can effectively harness their capabilities for few-shot classification. In this work, we propose PointGMDA, a training-free framework that combines Gaussian Mixture Models (GMMs) with Gaussian Discriminant Analysis (GDA) to perform robust classification using only a few labeled point cloud samples. Our method estimatesGMMparameters per class from support data and computes mixture-weighted prototypes, which are then used in GDA with a shared covariance matrix to construct decision boundaries. This formulation allows us to model intra-class variability more expressively than traditional single-prototype approaches, while maintaining analytical tractability. To incorporate semantic priors, we integrate CLIP-style textual prompts and fuse predictions from geometric and textual modalities through a hybrid scoring strategy. We further introduce PointGMDA-T, a lightweight attention-guided refinement module that learns residuals for fast feature adaptation, improving robustness under distribution shift. Extensive experiments on ModelNet40 and ScanObjectNN demonstrate that PointGMDA outperforms strong baselines across a variety of few-shot settings, with consistent gains under both training-free and fine-tuned conditions. These results highlight the effectiveness and generality of our probabilistic modeling and multimodal adaptation framework. Our code is publicly available at https://github.com/djzgroup/PointGMDA.Computer Graphics ForumCreating and Processing Point Clouds44
XEventNet: Extreme Weather Event Prediction using Convolutional Neural Networks and In Situ Visualization
No full textExtreme weather phenomena such as cyclones, torrential rainfall, snow storms, flash floods and landslides pose serious threat to living beings and property all over the world. An accurate and early prediction system for these extreme events may minimize the loss of life and property. However, this requires an online prediction system integrated with the weather simulation model for faster prediction such that low I/O bandwidth does not hinder performance. We present an in situ framework, XEventNet, that integrates weather simulation, deep learning-based prediction, and visualization. XEventNet predicts extreme events at real-time while the simulation is running using a Convolutional Neural Network (CNN). XEventNet is trained and tested on 400 events (extreme and non-extreme). Data is streamed online from XEventNet simulation processes to prediction processes for parallel inference. XEventNet uses the prediction values with high confidence to selectively transfer sub-domains of the large parent simulation domain. We use ADIOS2 for parallel data transfers via memory between groups of processes. This helps in timely prediction and visualization of critical weather events despite large volume of simulation data. We performed weather simulations at 9 km resolutions, thereby producing gigabytes of data per time step. XEventNet is able to classify four extreme events at real-time and visualize the same. We achieved an average prediction accuracy of 90.25% for all extreme events using a single CNN model. We ran weather simulations on up to 512 processes and parallel predictions on up to 64 processes, thereby streaming gigabytes of data in parallel within seconds. This was possible due to efficient data transfer and process mapping. Furthermore, our selective data transfer for visualization resulted in more than 70% reduction in data size, thereby improving the end-to-end simulation-prediction-visualization times.Eurographics Symposium on Parallel Graphics and VisualizationPaper