Eurographics Digital Library
Not a member yet
17121 research outputs found
Sort by
Uncertainty-Aware Adjustment via Learnable Coefficients for Detailed 3D Reconstruction of Clothed Humans from Single Images
No full textAlthough single-image 3D human reconstruction has made significant progress in recent years, few of the current state-of-theart methods can accurately restore the appearance and geometric details of loose clothing. To achieve high-quality reconstruction of a human body wearing loose clothing, we propose a learnable dynamic adjustment framework that integrates side-view features and the uncertainty of the parametric human body model to adaptively regulate its reliability based on the clothing type. Specifically, we first adopt the Vision Transformer model as an encoder to capture the image features of the input image, and then employ SMPL-X to decouple the side-view body features. Secondly, to reduce the limitations imposed by the regularization of the parametric model, particularly for loose garments, we introduce a learnable coefficient to reduce the reliance on SMPLX. This strategy effectively accommodates the large deformations caused by loose clothing, thereby accurately expressing the posture and clothing in the image. To evaluate the effectiveness, we validate our method on the public CLOTH4D and Cape datasets, and the experimental results demonstrate better performance compared to existing approaches. The code is available at https://github.com/yyd0613/CoRe-Human.Computer Graphics ForumDigital Human44
Real-Time Rendering of Old Glass Panes
No full textMany cultural heritage projects involve the 3D reconstruction of old glass buildings. Old glass, however, has many defects, due to the artisanal manufacturing techniques of the time, such as crown or cylinder blowing. The presence of these defects has a number of visual consequences, manifesting itself in the deviation of light passing through or reflecting off the glass panes. The appearance of these old glasses, the lighting produced through their surface, and the vision of the world perceived through them, is thus very different from what is perceived through contemporary industrial glass, and therefore has a considerable impact on the rendering that will be produced in their presence. However, setting up interactive virtual tours that take into account these old glasses is proving complex, as the materials available in commercial 3D rendering engines are unable to faithfully reproduce the lighting effects produced. In this paper, we propose a precise, real-time rendering of the surface and volume defects (bubbles, chords) present in some old glass and their impact on the appearance of the world perceived through these panes. Our approach is based on ray tracing, which not only interacts with the material's geometric defects, but also takes into account the curvature of light as the glass's refractive index varies.Digital HeritageVisual Archives and Historical Imagery in V
Personalized Cultural Heritage Recommendation System For Cognitive Exploration Levels
No full textThis study addresses the limitations of current digital cultural heritage platforms, which rely heavily on keyword searches and static categorization. These conventional approaches restrict users' ability to engage in exploratory and personalized experiences. To address this issue, we propose a recommendation system that provides two adaptive exploration paths-facet-based and semantic-linked-customized to align with users' cognitive levels. The system analyzes real-time behavioral data to generate personalized artifact recommendations, which are then presented through an individualized visual report called My Taste Report, built upon knowledge graph structures. Utilizing 195,441 artifact records from the National Museum of Korea, the system employs Transformer-based semantic similarity algorithms in combination with cultural heritage-specific named entity recognition (NER) techniques. Importantly, the system is designed for modular integration, allowing it to enhance existing cultural heritage portals without requiring a structural overhaul.Digital HeritagePoster
GATE: Geometry-Aware Trained Encoding
No full textThe encoding of input parameters is one of the fundamental building blocks of neural network algorithms. Its goal is to map the input data to a higher-dimensional space [RBA*19], typically supported by trained feature vectors [MESK22]. The mapping is crucial for the efficiency and approximation quality of neural networks. We propose a novel geometry-aware encoding called GATE that stores feature vectors on the surface of triangular meshes. Our encoding is suitable for neural rendering-related algorithms, for example, neural radiance caching [MRNK21]. It also avoids limitations of previous hash-based encoding schemes, such as hash collisions, selection of resolution versus scene size, and divergent memory access. Our approach decouples feature vector density from geometry density using mesh colors [YKH10], while allowing for finer control over neural network training and adaptive level-of-detail.High-Performance Graphics - Symposium PapersNeural Textures and Encoding
Bin-VBSR: Variable Block Size Binned Block-Compressed Sparse Row for Efficient GPU-Accelerated Finite Element Analysis
No full textWe present Binned Variable Block Compressed Sparse Row (Bin-VBSR), a novel GPU-optimized sparse matrix data structure and associated sparse matrix-vector multiplication algorithm for matrices with variable-size dense blocks. This includes a novel approach to handling long rows in the Binned Compressed Sparse Row (Bin-CSR) family of GPU-optimized sparse matrix data structures. We demonstrate speedups of up to 9.9× over Bin-BCSR* and extend its data compression advantages over compressed sparse row (CSR) to variable block size, resulting in an improvement of up to 50%.Vision, Modeling, and VisualizationGeometry, Simulation, and Optimizatio
Investigating the Ways in Which Mobile Phone Images with Open-Source Data Can Be Used to Create an Augmented Virtual Environment (AVE)
No full textThis paper presents the development of an interactive system for constructing Augmented Virtual Environments (AVEs) by fusing mobile phone images with open-source geospatial data. By integrating 2D image data with 3D models derived from sources such as OpenStreetMap (OSM) and Digital Terrain Models (DTM), the proposed system generates immersive environments that enhance situational context. The system leverages Python for data processing and Unity for 3D visualization, interconnected via UDP-based two-way communication. Preliminary user evaluation demonstrates that the resulting AVEs accurately represent real-world scenes and improve users' contextual understanding. Key challenges addressed include projector calibration, precise model construction from heterogeneous data, and object detection for dynamic scene representation.Computer Graphics and Visual Computing (CGVC)Short Papers Session: Visualisatio
Text-Guided Interactive Scene Synthesis with Scene Prior Guidance
No full text3D scene synthesis using natural language instructions has become a popular direction in computer graphics, with significant progress made by data-driven generative models recently. However, previous methods have mainly focused on one-time scene generation, lacking the interactive capability to generate, update, or correct scenes according to user instructions. To overcome this limitation, this paper focuses on text-guided interactive scene synthesis. First, we introduce the SceneMod dataset, which comprises 168k paired scenes with textual descriptions of the modifications. To support the interactive scene synthesis task, we propose a two-stage diffusion generative model that integrates scene-prior guidance into the denoising process to explicitly enforce physical constraints and foster more realistic scenes. Experimental results demonstrate that our approach outperforms baseline methods in text-guided scene synthesis tasks. Our system expands the scope of data-driven scene synthesis tasks and provides a novel, more flexible tool for users and designers in 3D scene generation. Code and dataset are available at https://github.com/bshfang/SceneMod.Computer Graphics ForumShape It Til You Make It: Programs for 3D Synthesis44
StoneVerse: Models and Methods in Cultural Heritage. The Open-Science Platform for Reproducible Modelling of Stone Decay
No full textThe preservation of cultural heritage is increasingly challenged by environmental factors such as weathering and pollutant exposure, which accelerate the deterioration of historic buildings, statues, and architectural elements. Understanding and mitigating these degradation processes require advanced modelling techniques and open access to experimental data. However, until now, a unified platform integrating state-of-the-art simulation models with a structured repository for data sharing is still lacking. In this work, we introduce StoneVerse, a web-based platform designed to support the study of porous material degradation in built heritage. Developed in alignment with the FAIR principles -Findability, Accessibility, Interoperability, and Reusability-, StoneVerse serves as a collaborative environment where researchers can both share experimental data and access simulation algorithms for predicting damage caused by water infiltration. To the best of our knowledge, it is the first digital platform of its kind, providing open-source computational tools for monitoring stone deterioration alongside a cohesive framework for data publication and dissemination. StoneVerse is developed within the European project ''Humanities and cultural Heritage Italian Open Science Cloud'' - H2IOSC - NextGenerationEU.Digital HeritageH2IOSC Project Developmen
A Multimodal Personality Prediction Framework based on Adaptive Graph Transformer Network and Multi-task Learning
No full textMultimodal personality analysis targets accurately detecting personality traits by incorporating related multimodal information. However, existing methods focus on unimodal features while overlooking the bimodal association features crucial for this interdisciplinary task. Therefore, we propose a multimodal personality prediction framework based on an adaptive graph transformer network and multi-task learning. Firstly, we utilize pre-trained models to learn specific representations from different modalities. Here, we employ pre-trained multimodal models' encoders as the backbones of the modality-specific extraction methods to mine unimodal features. Specifically, we introduce a novel adaptive graph transformer network to mine personalityrelated bimodal association features. This network effectively learns higher-order temporal dependencies based on relational graphs and emphasizes more significant features. Furthermore, we utilize a multimodal channel attention residual fusion module to obtain the fused features, and we propose a multimodal and unimodal joint learning regression head to learn and predict scores for personality traits. We design a multi-task loss function to enhance the robustness and accuracy of personality prediction. Experimental results on the two benchmark datasets demonstrate the effectiveness of our framework, which outperforms the state-of-the-art methods. The code is available at https://github.com/RongquanWang/PPF-AGTNMTL.Computer Graphics ForumFix it in Post: Image and Video Synthesis and Analysis44
Under Fire Heritage of Ukraine: some insights from early damage assessment activity
No full textCultural heritage sites have been increasingly destroyed within conflict areas around the world, both as collateral damage of action of bombing and shelling and as result of deliberate targeting. Qualitative and quantitative assessments of the damage that conflicts are inflicting to the heritage of countries involved in bombings and shelling have been conducted by archaeologists both on the ground and with the help of aerial datasets. In this paper we would like to present the initiative of a pilot assessment of the damage perpetrated to monuments and heritage locations in Ukraine during the Russian invasion of the country in March 2022. In this case, the aerial and ground assessments were conducted during the ongoing conflict and the paper will elucidate the positive and negative implications that this had on the investigation. The pilot study was a collaboration between a group of Ukrainian archaeologists from various institutions, the University College London, and the Global Heritage Fund, where team members with different expertise joined forces to documents the damage to Ukrainian heritage in the cities of Kharkiv and Chernihiv. An agreement with Planet Labs allowed the team to have access to weekly satellite coverage of the two cities, thus providing for a detail tracking of the patterns of destruction occurring during the initial phases of the military invasion. Meanwhile, a team of archaeologists conducted a ground damage assessment of heritage buildings and monuments within the two urban areas. The combination of the two sets of data, aerial monitoring and ground assessment, led to a more complete picture of the overall destruction and also allowed to clarify whether the heritage assets affected by bombing and shelling were the result of collateral damage or were indeed deliberately targeted. The results show that in Kharkiv the most affected monuments are located in the areas of the city that were primarily hit, whereas in Chernihiv they were specifically targeted in order to allegedly strike the heritage and identity of the Ukrainian people.Digital HeritageDigital Tools for Monitoring Heritage at Ris