Eurographics Digital Library
Not a member yet
17121 research outputs found
Sort by
Towards a Better Evaluation of 3D CVML Algorithms: Immersive Debugging of a Localization Model
No full textAs advancements in robotics, autonomous driving, and spatial computing continue to unfold, a growing number of Computer Vision and Machine Learning (CVML) algorithms are incorporating three-dimensional data into their frameworks. Debugging these 3D CVML models often requires going beyond traditional performance evaluation methods, necessitating a deeper understanding of an algorithm's behavior within its spatio-temporal context. However, the lack of appropriate visualization tools presents a significant obstacle to effectively exploring 3D data and spatial features in relation to key performance indicators (KPIs). To address this challenge, we explore the application of Immersive Analytics (IA) methodologies to enhance the debugging process of 3D CVML models. Through in-depth interviews with eight CVML engineers, we identify common tasks and challenges faced during the development of spatial algorithms, and establish a set of design principles for creating tools tailored to spatial model evaluation. Building on these insights, we propose a novel immersive analytics system for debugging an indoor localization algorithm. The system is built using web technologies and integrates WebXR to enable fluid transitions across the reality-virtuality continuum. We conduct a qualitative study with six CVML engineers using our system on Apple Vision Pro, observing their analytical workflow as they debug an indoor localization sequence. We discuss the advantages of employing immersive analytics in the model evaluation workflow, emphasizing the role of seamlessly integrating 2D and 3D visualizations across varying levels of immersion to facilitate more effective model assessment. Finally, we reflect on the implementation trade-offs and discuss the generalizability of our findings for future efforts in immersive 3D CVML model debugging.Computer Graphics ForumEvaluation and Guidanc
3D Garments: Reconstructing Topologically Correct Geometry and High-Quality Texture from Two Garment Images
No full textWe present a fully integrated pipeline for generating topologically correct 3D meshes and high-fidelity textures of fashion garments. Our geometry reconstruction module takes two input images and employs a semi-signed distance field representation with shifted generalized winding numbers in a deep-learning framework to produce accurate, non-watertight meshes. To create realistic, high-resolution textures (up to 4K) that closely match the input, we combine diffusion-based inpainting with a differentiable renderer, further enhancing the quality through normal-guided projection to minimize projection distortions in the texture image. Our results demonstrate both precise geometry and richly detailed textures. In addition, we are making a portion of our high-quality training dataset publicly available, consisting of 250 lower-garment triangulated meshes with 4K textures.Eurographics 2025 - Short PapersShort Paper
OpenLIME: An open and flexible web framework for creating and exploring complex multi-layered relightable image models
No full textWe introduce OpenLIME (Open Layered IMage Explorer), an open, scalable, and flexible framework for creating web-based interactive tools to annotate and inspect large multi-layered and multi-channel standard and relightable image models. Adaptive image management and display use a data-flow approach, where images from sources of any size are efficiently streamed into screen-sized buffers that can be processed and combined using customizable WebGL shaders. The framework natively supports multispectral images, Bidirectional Reflectance Distribution Function (BRDF), and Reflectance Transformation Imaging (RTI) datasets and can be extended to accommodate other multi-channel raster datasets, such as neural representations. Multi-layer and multi-faceted visualizations are achieved through opacity adjustments, blending modes, and interactive lenses. The released library provides a set of pre-configured layers, facilitating the rapid deployment of web-based datasets and kiosk applications. Its responsive user interface is compatible with desktop, mobile, and general multitouch environments, while its modular architecture allows for extensive customization, making it adaptable to diverse annotation and visualization needs. The paper illustrates the framework's design and discusses specific use cases, including the inspection of RTI models, the integration of novel relightable image formats, archaeological data documentation and annotation, and standalone museum application creation and deployment. The main components of the framework are released as open source.Digital HeritageXR Platforms and Frameworks for Cultural Engagemen
Frequency-Guided Self-Supervised Wind-Driven Garment Animation Simulation
No full textRealistic simulation of garment deformation under coupled multi-physical fields remains a critical challenge in 3d animation, due to diverse properties of fabric and complex interactions with external forces. Existing methods mainly focus on humandriven garment animation and offer dynamic environmental factors such as wind fields. Moreover, supervised learning methods suffer from strong data dependency and limited generalization. We propose FWGNet, a self-supervised training framework based on graph neural networks (GNNs) that models the physical interaction between garments, the human body and wind as an unified physical system. The framework is trained using differentiable physics-based constraints. A core component of FWGNet is a wind feature encoder that utilizes wavelet transforms to project wind velocity sequences into the frequency domain, enabling the network to effectively capture multi-scale turbulence effects on fabric behavior. To eliminate dependence on large datasets, we introduce physics-informed loss functions that incorporate gravitational potential energy, aerodynamic wind forces, and fabric deformation constraints. Experiments demonstrate that our approach produces highly realistic visual effects, such as detailed wrinkle formation and fabric fluttering under dynamic wind conditions. Quantitative evaluations across physical metrics confirm that FWGNet achieves a strong balance between physical accuracy and visual realism, particularly in complex scenarios involving coupled physical interactions.Pacific Graphics Conference Papers, Posters, and DemosDigital Clothin
Transkribus, ChatGPT and EVT. From manuscripts to Digital Twins Workshop
No full textData, Big Data, Textual Analysis, and the Semantic Web are among the core concepts underpinning the digital turn. Everything becomes processable, and most objects are increasingly ''virtualized,'' shifting our experience of them away from their physical dimension. Words are transformed into data-entities, comparable to ''second-order abstractions'' (i.e., numbers) (Frege 2023; 1980), and as such they can be analyzed through algorithms capable of enhancing our knowledge. While this is certainly applicable to printed texts (which can be converted into machinereadable digital texts via OCR) and to ''born-digital'' content, the scenario changes when we consider the vast archival heritage composed of handwritten documents from various historical periods, characterized by highly diverse linguistic structures and forms of language-texts which currently resist large-scale computational analysis.Digital HeritageWorkshop
Combining CNN Feature Extraction and Kolmogorov-Arnold Networks Regression for Procedural 3D Shape Generation
No full textGenerating 3D objects with complex, nonlinear shapes directly from images is still an open research area. To address this problem, several state-of-the-art methods use Deep Learning (DL) to predict a set of parameters from images, which are then used to generate the 3D geometry, leveraging the characteristics of procedural modeling. Recently, Kolmogorov-Arnold Networks (KANs) have emerged as an alternative to traditional Multilayer Perceptrons (MLPs) in DL, and have been successfully integrated into architectures such as Convolutional Neural Networks (CNNs), Graph Neural Networks, and Transformers. In this work, we propose a DL architecture consisting of a hybrid CNN-KAN network for parametric 3D model generation from images. The model combines the ability of KANs to capture complex nonlinear functions with the strong visual feature extraction capabilities of CNNs. The method is evaluated using both quantitative error metrics and qualitative visualizations comparing predicted shapes with ground truth, and its performance is compared against a more standard CNN-MLP architecture.Smart Tools and Applications in Graphics - Eurographics Italian Chapter ConferenceLearning-based Algorithm
The Persistence of Agency within the Virtual World
No full textThis paper explores player agency in video games to consider the impact of visual persistence within the virtual environment. Drawing from philosophical, psychological, and game design theory, we argue that a player's perceived impact on the virtual world, manifested through visual change can strengthen engagement and immersion. We support this with examples from commercial games. We propose that reactive environmental design, supported by approaches like agency informing techniques and player modeling systems, can reinforce a player's sense of agency through persistent visual feedback. This paper argues a direction for future research into this novel research space.Computer Graphics and Visual Computing (CGVC)Short Papers Session: Games and Graphic
Spatial Body Augmentation for Destructive Interaction in Giant Avatar Embodiment using a Wearable Force Feedback Device
No full textThis study examined the effects of force presentation on physicality in a VR environment using a giant avatar. Focusing on a sense of body ownership (SoB) and a sense of agency (SoA), we designed an interaction in which the user's upper limb movements are synchronized with the arm movements of a giant avatar to destroy a building using a wearable force-sensing device. The device combines a magneto-viscous fluid brake and a vibration motor, and is capable of presenting inertial forces and a sense of collision in response to joint acceleration. Experimental results showed that the device improved the sense of weight and the sense of destruction while maintaining the sense of body ownership and the sense of agency.ICAT-EGVE 2025 - International Conference on Artificial Reality and Telexistence and Eurographics Symposium on Virtual Environments - Posters and DemosPosters and Demo
X-ray simulations with gVirtualXray in medicine and life sciences
No full textgVirtualXray (gVXR) is a programming interface framework to simulate realistic X-ray projections in realtime on graphics processing units (GPUs). It solves the Beer-Lambert law (attenuation law) using a deterministic X-ray simulation algorithm based on 3D computer graphics, namely rasterisation. Implemented as multi-pass rendering makes it more computationally optimal than the ray-tracing technique, which is a brute-force and straightforward approach to simulate X-ray images. Although written in C++ using OpenGL and its shading language (GLSL) to leverage the GPU, gVXR is available for other programming languages such as Python. Extensive validation studies, including comparisons with Monte Carlo simulations and real experimental data, have confirmed the accuracy of gVXR's simulations. gVXR was initially used in medical virtual reality (VR) for training purposes. It was then used in medical physics, and high-throughput data applications including mathematical optimisation and machine learning (ML). Micro-imaging studies on the C. elegans biological model are also reported.EuroVis 2025 - Dirk Bartz Prize3rd Priz
Monte Carlo Methods for 2D Flow Visualization
No full textIn this paper, we investigate how recent advances from the computer graphics literature can be applied to improve the visualization of two-dimensional vector fields. To this end, we propose two different approaches that both start from a set of evenly-spaced streamlines. The first approach avoids the need for contrast normalization, which is usually required for LIC approaches. For this, the image synthesis is phrased as a diffusion problem by placing double-sided Dirichlet boundary conditions along the streamlines. The diffusion problem is formally modeled as a linear elliptic partial differential equation, which is solved stochastically using a variant of the walk-on-spheres algorithm in order to achieve anti-aliased results. The second approach leverages human's perception of shape to convey flow patterns. For this, we lift the streamlines into the third dimension and generate visual contrast among adjacent streamlines by means of ambient occlusion. To synthesize the images, we apply a physically based material model and employ a Monte Carlo renderer to simulate the light transport.EuroVis 2025 - Short PapersTechniques and Tool