Eurographics Digital Library
Not a member yet
17121 research outputs found
Sort by
Approximating Procedural Models of 3D Shapes with Neural Networks
No full textProcedural modeling is a popular technique for 3D content creation and offers a number of advantages over alternative techniques for modeling 3D shapes. However, given a procedural model, predicting the procedural parameters of existing data provided in different modalities can be challenging. This is because the data may be in a different representation than the one generated by the procedural model, and procedural models are usually not invertible, nor are they differentiable. In this paper, we address these limitations and introduce an invertible and differentiable representation for procedural models. We approximate parameterized procedures with a neural network architecture NNProc that learns both the forward and inverse mapping of the procedural model by aligning the latent spaces of shape parameters and shapes. The network is trained in a manner that is agnostic to the inner workings of the procedural model, implying that models implemented in different languages or systems can be used. We demonstrate how the proposed representation can be used for both forward and inverse procedural modeling. Moreover, we show how NNProc can be used in conjunction with optimization for applications such as shape reconstruction from an image or a 3D Gaussian Splatting.Computer Graphics ForumShape It Til You Make It: Programs for 3D Synthesis44
Uni-IR: One Stage is Enough for Ambiguity-Reduced Inverse Rendering
No full textInverse rendering aims to decompose an image into geometry, materials, and lighting. Recently, Neural Radiance Fields (NeRF) based inverse rendering has significantly advanced, bridging the gap between NeRF-based models and conventional rendering engines. Existing methods typically adopt a two-stage optimization approach, beginning with volume rendering for geometry reconstruction, followed by physically based rendering (PBR) for materials and lighting estimation. However, the inherent ambiguity between materials and lighting during PBR, along with the suboptimal nature of geometry reconstruction by volume rendering, compromises the outcomes. To address these challenges, we introduce Uni-IR, a unified framework that imposes mutual constraints to alleviate ambiguity by integrating volume rendering and physically based rendering. Specifically, we employ a physically-based volume rendering (PBVR) approach that incorporates PBR concepts into volume rendering, directly facilitating connections with materials and lighting, in addition to geometry. Both rendering methods are utilized simultaneously during optimization, imposing mutual constraints and optimizing geometry, materials, and lighting synergistically. By employing a carefully designed unified representation for both lighting and materials, Uni-IR achieves high-quality geometry reconstruction, materials, and lighting estimation across various object types.Pacific Graphics Conference Papers, Posters, and DemosRendering & Inverse Renderin
Perin del Vaga, His Workshop and Patterns of Fresco Painting in the Farnese Tower cycle through Multiple Non-Invasive Analyses
No full textThe aim of this work, carried out within the framework of the ENACTING ARTISTIC RESEARCH (EAR) project, is to presents preliminary results from ongoing analyses conducted on a 16th-century fresco cycle originally located in the Tower of Pope Paul III Farnese and now housed at the Academy of Fine Arts in Rome. The artworks are being investigated through an integrated approach combining art, science, and non-invasive diagnostic techniques, thanks to the collaboration between the Academy of Fine Arts in Rome and the National Institute of Nuclear Physics of Roma Tre. To address the numerous diagnostic questions, concerning the uncertain attribution of the artworks, the multiplicity of execution techniques, and the complex conservation history, several analytical methods have been employed, including multispectral imaging, digital microscopy and X-ray fluorescence (XRF).Digital HeritageAnalysing and Documenting the Creation Process, Evolution and Contex
CMI-MTL: Cross-Mamba interaction based multi-task learning for medical visual question answering
No full textMedical visual question answering (Med-VQA) is a crucial multimodal task in clinical decision support and telemedicine. Recent self-attention based methods struggle to effectively handle cross-modal semantic alignments between vision and language. Moreover, classification-based methods rely on predefined answer sets. Treating this task as a simple classification problem may make it unable to adapt to the diversity of free-form answers and overlook the detailed semantic information of free-form answers. In order to tackle these challenges, we introduce a Cross-Mamba Interaction based Multi-Task Learning (CMI-MTL) framework that learns cross-modal feature representations from images and texts. CMI-MTL comprises three key modules: fine-grained visual-text feature alignment (FVTA), cross-modal interleaved feature representation (CIFR), and free-form answer-enhanced multi-task learning (FFAE). FVTA extracts the most relevant regions in image-text pairs through fine-grained visual-text feature alignment. CIFR captures cross-modal sequential interactions via cross-modal interleaved feature representation. FFAE leverages auxiliary knowledge from open-ended questions through free-form answerenhanced multi-task learning, improving the model's capability for open-ended Med-VQA. Experimental results show that CMI-MTL outperforms the existing state-of-the-art methods on three Med-VQA datasets: VQA-RAD, SLAKE, and OVQA. Furthermore, we conduct more interpretability experiments to prove the effectiveness. The code is publicly available at https://github.com/BioMedIA-repo/CMI-MTL.Pacific Graphics Conference Papers, Posters, and DemosVisualizatio
Uncovering Relations in High-Dimensional Behavioral Data of Drosophila Melanogaster
No full textUnderstanding how behaviour changes under genetic or experimental conditions is a key challenge in behavioural neuroscience. High-throughput tracking enables the collection of high-dimensional datasets describing locomotion, posture, and stimulus orientation in Drosophila melanogaster larvae (fruit fly). However, exploring relations across numerous dimensions remains challenging. We present a Visual Analytics system that integrates coordinated views, type-aware relation metrics, and hierarchical clustering to support relation discovery and validation in behavioural data. The system was initially developed based on prior experience and refined through evaluation with domain experts to address key analysis tasks, including grouping dimensions, exploring behavioural patterns, and validating hypotheses. We demonstrate how it supports both confirmatory and exploratory workflows, enabling users to confirm known effects and uncover novel patterns-such as an unexpected correlation between head-casting behaviour and locomotion speed. This work highlights how tailored visual analysis can advance behavioural research.Vision, Modeling, and VisualizationVisualization, Visual Analytics, and V
Microglia Cell Segmentation Using a Hand-crafted Method Capable of Handling High Noise Levels in Image Data
No full textMicroglia are the resident immune cells in the brain and spinal cord, playing a crucial role in various pathological processes. Accurate segmentation of microglia is the first and most critical step in the analysis of their morphology, which serves as a labelfree, primary indicator of microglial phenotype. In this work, we present a fully automated microglia segmentation method that is capable of reliably detecting and segmenting microglia from surrounding tissue, even under challenging conditions with substantial tissue-caused background noise in the image data. Our method incorporates several novel approaches, including a highly effective way to remove background noise while preserving microglial structures and an approach for filtering out microglial structures without an associated cell nucleus. We compared our microglia cell segmentation method with three well-known segmentation approaches reported in previous work on microglial morphology. The methods were applied to 20 fluorescence microscopy images of the spinal cord containing hundreds of microglia, for which a manually segmented ground truth segmentation has been obtained. We show that our proposed method clearly outperforms the previous methods.Eurographics Workshop on Visual Computing for Biology and MedicineSession
PrismBreak: Exploration of Multi-Dimensional Mixture Models
No full textIn data science, visual data exploration becomes increasingly more challenging due to the continued rapid increase of data dimensionality and data sizes. To manage complexity, two orthogonal approaches are commonly used in practice: First, data is frequently clustered in high-dimensional space by fitting mixture models composed of normal distributions or Student t-distributions. Second, dimensionality reduction is employed to embed high-dimensional point clouds in a two- or threedimensional space. Those algorithms determine the spatial arrangement in low-dimensional space without further user interaction. This leaves little room for a guided exploration and data analysis. In this paper, we propose a novel visualization system for the effective exploration and construction of potential subspaces onto which mixture models can be projected. The subspaces are spanned linearly via basis vectors, for which a vast number of basis vector combinations is theoretically imaginable. Our system guides the user step-by-step through the selection process by letting users choose one basis vector at a time. To guide the process, multiple choices are pre-visualized at once on a multi-faceted prism. In addition to the qualitative visualization of the distributions, multiple quantitative metrics are calculated by which subspaces can be compared and reordered, including variance, sparsity, and visibility. Further, a bookmarking tool lets users record and compare different basis vector combinations. The usability of the system is evaluated by data scientists and is tested on several high-dimensional data sets.Computer Graphics ForumDimensionality Reduction and High-Dimensional Dat
ASDGen: A Shape Dataset Generator using a Simulated CAD Process
No full textNeural networks have shown great promise in 3D applications like shape analysis, object recognition, and design optimization. Machine learning methods depend on high-quality, structured datasets. While databases exist for general 3D shapes, there is a lack of databases tailored for subdivision surface representations. To address this, we introduce ASDGen, an algorithm to generate quadrilateral meshes through a sequence of CAD operations arbitrarily applied to an initial user-defined seed mesh. The resulting meshes are guaranteed to be manifold and can serve as control meshes for generating Catmull-Clark subdivision surfaces. The algorithm may be employed to generate large sets of synthetic shape data represented as quadrilateral meshes of varying degree of refinement, along with all CAD operations applied to a seed mesh to create the shape. The resulting data is ideal to be employed for data-driven analysis of subdivision surfaces. In addition to the shape-data generator, we provide a robust pipeline for extracting various differential shape properties as metadata, e.g. curvature and complexity measures, and for converting these meshes into signed distance fields. We generate a sample dataset of Catmull-Clark subdivision shapes which we make publicly available together with the generator. To demonstrate the potential of ASDGen, present two learning-based applications: a neural network model trained to predict mesh complexity and a prediction of maximum curvature points from the signed distance field of the shape. Our work lays the groundwork for a new class of learning problems rooted in CAD-inspired geometry, and provides both the tools and data necessary to support further research in this domain.Smart Tools and Applications in Graphics - Eurographics Italian Chapter ConferenceDataset
Theoretical Model Validation of the Multisensory Role on Subjective Realism, Presence and Involvement in Immersive Virtual Reality
No full textWith the consistent adoption of iVR and growing research on the topic, it becomes fundamental to understand how the perception of Realism plays a role in the potential of iVR. This work puts forwards a hypothesis-driven theoretical model of how the perception of each multisensory stimulus (Visual, Audio, Haptic and Scent) is related to the perception of Realism of the whole experience (Subjective Realism) and, in turn, how this Subjective Realism is related to Involvement and Presence. The model was validated using a sample of 216 subjects in a multisensory iVR experience. The results indicated a good model fit and provided evidence on how the perception of Realism of Visual, Audio and Scent individually is linked to Subjective Realism. Furthermore, the results demonstrate strong evidence that Subjective Realism is strongly associated with Involvement and Presence. These results put forwards a validated questionnaire for the perception of Realism of different aspects of the virtual experience and a robust theoretical model on the interconnections of these constructs. We provide empirical evidence that can be used to optimise iVR systems for Presence, Involvement and Subjective Realism, thereby enhancing the effectiveness of iVR experiences and opening new research avenues.Computer Graphics ForumMajor Revision from Eurographics Conference44
Eigenvalue Blending for Projected Newton
No full textWe propose a novel method to filter eigenvalues for projected Newton. Central to our method is blending the clamped and absolute eigenvalues to adaptively compute the modified Hessian matrix. To determine the blending coefficients, we rely on (1) a key observation and (2) an objective function descent constraint. The observation is that if the quadratic form defined by the Hessian matrix maps the descent direction to a negative real number, the decrease in the objective function is limited. The constraint is that our eigenvalue filtering leads to more reduction in objective function than the absolute eigenvalue filtering [CLL∗24] in the case of second-order Taylor approximation. Our eigenvalue blending is easy to implement and leads to fewer optimization iterations than the state-of-the-art eigenvalue filtering methods.Computer Graphics ForumSimulating Complex Systems: Turbulent, Crowded, and Shattered44