Eurographics Digital Library
Not a member yet
17121 research outputs found
Sort by
Learning Transformation-Isomorphic Latent Space for Accurate Hand Pose Estimation
No full textVision-based regression tasks, such as hand pose estimation, have achieved higher accuracy and faster convergence through representation learning. However, existing representation learning methods often encounter the following issues: the high semantic level of features extracted from images is inadequate for regressing low-level information, and the extracted features include task-irrelevant information, reducing their compactness and interfering with regression tasks. To address these challenges, we propose TI-Net, a highly versatile visual Network backbone designed to construct a Transformation Isomorphic latent space. Specifically, we employ linear transformations to model geometric transformations in the latent space and ensure that TI-Net aligns them with those in the image space. This ensures that the latent features capture compact, low-level information beneficial for pose estimation tasks. We evaluated TI-Net on the hand pose estimation task to demonstrate the network's superiority. On the DexYCB dataset, TI-Net achieved a 10% improvement in the PA-MPJPE metric compared to specialized state-of-the-art (SOTA) hand pose estimation methods. Our code is available at https://github.com/Mine268/TI-Net.Pacific Graphics Conference Papers, Posters, and DemosDetecting & Estimating from image
Reconstructing Gladiator Combat: A Multisensory Virtual Reality Training Environment
No full textThis study presents a research project focused on designing, implementing, and evaluating a multisensory virtual environment to simulate gladiatorial training. The aim is to analyze how immersive experiences impact the acquisition and refinement of technical skills in armed singular dueling. Conducted collaboratively by teams in virtual reality, biomechanics, and history, the project developed a historically contextualized environment centered on the provocator, a specific gladiator type. The virtual environment allows users to train in typical offensive maneuvers, offering a testbed for hypotheses about Roman combat and the effects of external conditions on performance. It serves as both a historical reconstruction tool and an experimental platform for studying ancient martial techniques. Built on rigorous historical and visual research, it uses motion capture technology to accurately recreate combat sequences, enhancing the authenticity and educational value of the simulation. A key contribution of this work lies in advancing the study of gladiatorial techniques, an area often distorted by popular culture. By integrating passive haptic and auditory feedback, the environment enhances sensory immersion, contributing to a deeper and more accurate understanding of gladiatorial practices. This multisensory approach not only supports the preservation of ancient techniques but also sheds light on the physical and cognitive demands faced by historical fighters. Ultimately, this research bridges disciplines-combining historical scholarship, biomechanics, and virtual reality-to offer an innovative way of exploring Roman gladiatorial training. The findings may inform broader discussions on the role of immersive technologies in skill development and historical interpretation within virtual environments.Digital HeritageReconstructing the Pas
VideoMat: Extracting PBR Materials from Video Diffusion Models
No full textWe leverage finetuned video diffusion models, intrinsic decomposition of videos, and physically-based differentiable rendering to generate high quality materials for 3D models given a text prompt or a single image. We condition a video diffusion model to respect the input geometry and lighting condition. This model produces multiple views of a given 3D model with coherent material properties. Secondly, we use a recent model to extract intrinsics (base color, roughness, metallic) from the generated video. Finally, we use the intrinsics alongside the generated video in a differentiable path tracer to robustly extract PBR materials directly compatible with common content creation tools.Computer Graphics ForumAppearance Modelling44
Impact of Visual, Auditory, and Mixed Interfaces on Human-Robot Collaboration in Multi-Robot Environments
No full textIn the field of Human Robot Collaboration (HRC) research, many studies have explored the use of visual and/or auditory cues as robot caution interfaces. However, many of these studies have focused on interfaces, such as displays of a single robot's future position or hazardous areas, without validating them in complex environments where multiple robots operate simultaneously and users need to perceive and respond to multiple robots at once. An increase in the number of robots can exceed human cognitive limits, potentially leading to a decrease in safety and operational efficiency. To achieve safe and work efficient HRC in environments with multiple robots, we proposed a design for auditory and visual augmented reality interfaces to help workers be aware of multiple robots. We evaluated both single-modal and multi-modal interfaces under varying numbers of robots in the environment to explore how user perception and safety are affected. We conducted a comparative evaluation using multiple metrics, including the number of collisions, the closest distance to a robot, interface response time, task completion time, and subjective measures. Although multi-modal interfaces can reduce the average number of collisions by approximately 19%- 49% compared to single-modal interfaces, and generally outperform them, their relative advantage diminished as the number of robots increased. This may be attributed to the physical limitations of the environment, where avoiding multiple robots simultaneously becomes inherently difficult, thereby reducing the impact of interface design on user performance.ICAT-EGVE 2025 - International Conference on Artificial Reality and Telexistence and Eurographics Symposium on Virtual EnvironmentsSoun
Diseño metodológico para la generación de imágenes multiresolución etiquetadas
No full textLa generación de conjuntos de datos multiescala etiquetados es fundamental para aplicaciones en teledetección y análisis de imágenes satelitales, especialmente en tareas como la clasificación de terrenos, la detección de cambios y la gestión de recursos naturales debido a la poca resolución con la que trabajan y lo difícil que esto hace el poder extraer conocimiento de estas. Este trabajo en progreso presenta el diseño de una metodología para la generación automática de imágenes de alta, media y baja resolución, incorporando para cada pixel una etiqueta semántica extraída a partir de las imágenes de mayor resolución espacial. Los conjuntos de datos resultantes son necesarios para el desarrollo de modelos de inteligencia artificial dirigidos a la comprensión de escenarios reales y clasificación de entidades. Aunque los resultados preliminares son prometedores, el trabajo se encuentra en una fase temprana, y se están explorando mejoras adicionales en la integración de características contextuales y la validación en diferentes escenarios de uso. Este trabajo en curso presenta un alto potencial para la generación de nuevos conjuntos de datos etiquetados más precisos y robustos para aplicaciones relacionadas con la simulación de la apariencia de materiales y otras soluciones relacionadas con teledetección.Spanish Computer Graphics Conference (CEIG)Short Paper
Using Virtual Worlds in Communicating Archaeology
No full textThe Museum of Cultural History at the University of Oslo, Norway, has a long history of digitizing cultural heritage in 3D. Applications vary from exhibitions and public outreach to academic research, university and elementary school teaching.To truly engage our audiences, we are forced to think creatively. This poster presents examples from recent projects at the museum that have been exploring ways to facilitate people to look closer and use digital 3D models as a gateway to further exploration.Digital HeritagePoster
Generating Color Schemes for your Digital Media & Data Visualizations
No full textThis tutorial provides an overview of the basics of color theory and shows how to use Generative AI tools, like OpenAI ChatGPT, Google Gemini, Microsoft Copilot and DeepSeek, to expand your data color scheme choices. You explore how to build your own colormaps by transforming color harmonies into data color schemes. This half day course is intended for a broad audience of individuals interested in understanding, applying, and building color schemes for data visualization.Eurographics 2025 - TutorialsTutorial
A Practical Inverse Rendering Strategy for Enhanced Albedo Estimation for Cultural Heritage Model Reconstruction
No full textWe present a practical single-image framework to address uncontrolled global and local illumination effects in flash photography for improved albedo estimation and color projection onto 3D cultural heritage models. Our approach leverages an inverse rendering pipeline to process a single registered flash photograph and models ambient illumination due to environmental reflections and local interreflections. By compensating for direct and indirect light contributions, we recover a more reliable albedo signal for color projection onto the 3D model. We validate our method through extensive evaluations on two synthetic datasets and real-world acquisitions in conservation and museum settings, demonstrating its effectiveness in improving photometric accuracy and support for relighting, and proper integration of optimized color data into existing 3D models.Digital HeritageDigitization and Technolog
Splatshop: Efficiently Editing Large Gaussian Splat Models
No full textWe present Splatshop, a highly optimized toolbox for interactive editing (selection, deletion, painting, transformation, . . . ) of 3D Gaussian Splatting models. Utilizing a comprehensive collection of heuristic approaches, we carefully balance between exact and fast rendering to enable precise editing without sacrificing real-time performance. Our experiments confirm that Splatshop achieves these goals for scenes with up to 100 million primitives. We also show how our proposed pipeline can be extended for use with head-mounted displays. As such, Splatshop is the first VR-capable editor for large-scale 3D Gaussian Splatting models and a step towards a ''Photoshop for Gaussian Splatting.''Computer Graphics ForumSplats and Points44
Spacewalker: Traversing Representation Spaces for Fast Interactive Exploration and Annotation of Unstructured Data
No full textIn industries such as healthcare, finance, and manufacturing, analysis of unstructured textual data presents significant challenges for analysis and decision making. Uncovering patterns within large-scale corpora and understanding their semantic impact is critical, but depends on domain experts or resource-intensive manual reviews. In response, we introduce Spacewalker, an interactive tool designed to analyze, explore, and annotate data across multiple modalities. It allows users to extract data representations, visualize them in low-dimensional spaces and traverse large datasets either exploratorily or by querying regions of interest. We evaluated Spacewalker through extensive studies, assessing its efficacy in improving data integrity verification and annotation. We show that Spacewalker reduces time and effort compared to traditional methods. The code of this work is publicly available on https://github.com/TIO-IKIM/Spacewalker.Machine Learning Methods in Visualisation for Big DataPaper