GRAPP 2025 Abstracts

Area 1 - Geometry and Modeling
Full Papers Short Papers
Area 2 - Rendering
Full Papers Short Papers
Area 3 - Animation and Simulation
Full Papers Short Papers
Area 4 - Interactive Environments
Full Papers Short Papers

Area 1 - Geometry and Modeling

Full Papers
Paper Nr: 44
Title:

Ground Awareness in Deep Learning for Large Outdoor Point Cloud Segmentation
Authors:

Kevin Qiu, Dimitri Bulatov and Dorota Iwaszczuk
Abstract: This paper presents an analysis of utilizing elevation data to aid outdoor point cloud semantic segmentation through existing machine-learning networks in remote sensing, specifically in urban, built-up areas. In dense outdoor point clouds, the receptive field of a machine learning model may be too small to accurately determine the surroundings and context of a point. By computing Digital Terrain Models (DTMs) from the point clouds, we extract the relative elevation feature, which is the vertical distance from the terrain to a point. RandLA-Net is employed for efficient semantic segmentation of large-scale point clouds. We assess its performance across three diverse outdoor datasets captured with varying sensor technologies and sensor locations. Integration of relative elevation data leads to consistent performance improvements across all three datasets, most notably in the Hessigheim dataset, with an increase of 3.7 percentage points in average F1 score from 72.35% to 76.01%, by establishing long-range dependencies between ground and objects. We also explore additional local features such as planarity, normal vectors, and 2D features, but their efficacy varied based on the characteristics of the point cloud. Ultimately, this study underscores the important role of the non-local relative elevation feature for semantic segmentation of point clouds in remote sensing applications.
Download
Paper Nr: 141
Title:

Efficient Modeling of 3D Epithelial Cell Structure Dynamics via Backbone Spreads
Authors:

Javier Buceta, Stefan Funke and Sabine Storandt
Abstract: An important means to study the morphogenesis of epithelial cell structures is faithful modeling and simulation of the underlying cell dynamics. In particular, changes in cell shapes and cell neighborhoods have to be captured to gain understanding of organ development and other complex processes. A typical 2D tissue model is cell-center based Voronoi tessellation. However, since epithelial cells form curved layers in three-dimensional space, a 2D model cannot encompass all relevant aspects. In this paper, we provide a formal 3D model for epithelial cell structures based on the notion of backbone spreads and study its geometric properties. Based thereupon, we devise a modified version of the well-known Metropolis-Hastings algorithm to find cell tissue configurations that minimize a given energy function. We prove that our new algorithm is very efficient from a theoretical perspective and also demonstrate its good performance in practice on the example of tubular epithelia. Furthermore, we show that a rich set of cell shapes and connectivity structures emerge in our model, and we analyze their frequency of occurrence with respect to the model parameters.
Download
Paper Nr: 236
Title:

Single-Exemplar Lighting Style Transfer via Emissive Texture Synthesis and Optimization
Authors:

Pierre Ecormier-Nocca, Lukas Lipp, Annalena Ulschmid, David Hahn and Michael Wimmer
Abstract: Lighting is a key component in how scenes are perceived. However, many interior lighting situations are currently either handcrafted by expert designers, or simply consist of basic regular arrangements of luminaires, such as to reach uniform lighting at a predefined brightness. Our method aims to bring more interesting lighting configurations to various scenes in a semi-automatic manner designed for fast prototyping by non-expert users. Starting from a single photograph of a lighting configuration, we allow users to quickly copy and adapt a lighting style to any 3D scene. Combining image analysis, texture synthesis, and light parameter optimization, we produce a lighting design for the target 3D scene matching the input image. We validate via a user study that our results successfully transfer the desired lighting style more accurately and realistically than state-of-the-art generic style transfer methods. Furthermore, we investigate the behaviour of our method under potential alternative choices in an ablation study.
Download
Paper Nr: 304
Title:

Curvy: A Parametric Cross-Section Based Surface Reconstruction
Authors:

Aradhya N. Mathur, Apoorv Khattar and Ojaswa Sharma
Abstract: In this work, we present a novel approach for reconstructing shape point clouds using planar sparse cross-sections with the help of generative modeling. We present unique challenges pertaining to the representation and reconstruction in this problem setting. Most methods in the classical literature lack the ability to generalize based on object class and employ complex mathematical machinery to reconstruct reliable surfaces. We present a simple learnable approach to generate a large number of points from a small number of input cross-sections over a large dataset. We use a compact parametric polyline representation using adaptive splitting to represent the cross-sections and perform learning using a Graph Neural Network to reconstruct the underlying shape in an adaptive manner reducing the dependence on the number of cross-sections provided. Project page: https://graphics-research-group.github.io/curvy/.
Download
Short Papers
Paper Nr: 75
Title:

Fast Approximate Symmetry Plane Computation as a Density Peak of Candidates
Authors:

Alex König and Libor Váša
Abstract: Symmetry is a common characteristic exhibited by both natural and man-made objects. This property can be used in various applications in computer vision and computer graphics. There are various types of symmetries, amongst the most prominent belong reflection symmetries and rotation symmetries. In this paper, a method focusing on the fast detection of approximate reflection symmetry of a 3D point cloud with respect to a plane is proposed. The method is based on the creation of a set of candidates that are represented as rigid transformations, and have assigned weights, reflecting the estimated quality of the candidate. The final symmetry plane corresponds to a density peak in the transformation space. The method is demonstrated to be able to find symmetry planes in various objects in 3D, with its main benefit being the speed of the computation.
Download
Paper Nr: 79
Title:

Geometry and Texture Completion of Partially Scanned 3D Objects Through Material Segmentation
Authors:

Jelle Vermandere, Maarten Bassier and Maarten Vergauwen
Abstract: This work aims to improve the geometry and texture completion of partially scanned 3D objects in indoor environments through the integration of a novel material prediction step. Completing segmented objects from these environments remains a significant challenge due to high occlusion levels and texture variance. State-of-the-art techniques in this field typically follow a two-step process, addressing geometry completion first, followed by texture completion. Although recent advancements have significantly improved geometry completion, texture completion continues to focus primarily on correcting minor defects or generating textures from scratch. This work highlights key limitations in existing completion techniques, such as the lack of material awareness, inadequate methods for fine detailing, and the limited availability of textured 3D object datasets. To address these gaps, a novel completion pipeline is proposed, enhancing both the geometry and texture completion processes. Experimental results demonstrate that the proposed method produces clearer material boundaries, particularly on scanned objects, and generalizes effectively even with synthetic training data.
Download
Paper Nr: 161
Title:

Efficient Tiling of Point Features to 3D Tiles with Discrete LOD
Authors:

Samuel Rundel and Raffaele De Amicis
Abstract: 3D Tiles were developed to visualize geospatial data and deliver high-quality, interactive visualizations to users over the Internet. However, there is a lack of direct methods to generate 3D Tiles from point features. This paper introduces a novel method for generating high-fidelity, very large-scale 3D Tilesets directly from geospatial point features. Our approach consistently produces 3D Tiles with well-defined Level of Detail (LOD) and handles any quantity or type of feature without restrictions. Additionally, it allows for partial updates of the Tileset in response to data changes, improving the efficiency of visualization. Our paper provides a thorough comparison of our procedure with existing methods, demonstrating its advantages and effectiveness.
Download
Paper Nr: 418
Title:

Manipulating Gloss of Real Objects Under Omnidirectional Lighting
Authors:

Yuki Miyoshi, Ryo Kawahara and Takahiro Okabe
Abstract: Manipulating the appearances of real-world objects by using active illumination is useful for XR. In this paper, we propose a method for manipulating the gloss of real objects observed by our naked eyes under omnidirectional lighting environments. Our proposed method makes use of the fact that specular reflection components are sensitive to the polarization state of the incident light and the high-frequency components of the illumination environment, while diffuse reflection components are insensitive to them. Specifically, our method optimizes the polarization angles and intensities of incident lighting environments for manipulating the gloss of real objects. We build a lighting system by using a dome screen and two pairs of a projector and a transmissive LC panel for controlling both the polarization angles and high-frequency components of incident lighting environments. We conduct a number of experiments, and show that our method achieves the gloss manipulation without using the geometric and photometric properties of an object of interest.
Download
Paper Nr: 80
Title:

Finding the Shear Reflection Symmetry Plane in a 3D Point Cloud
Authors:

Vítek Poór, Ivana Kolingerová and Damjan Strnad
Abstract: Many objects, namely man-made ones, show signs of various types of symmetry. The most common type perceived by humans is reflection symmetry to some plane. When detecting the symmetry for geometric models, the existing algorithms look for orthogonal reflection symmetry. However, the models can be sheared, therefore, algorithms detecting shear reflection symmetry would be useful. In this paper, we propose an algorithm for detecting the plane of shear reflection symmetry in a 3D point cloud on condition that the shear was done in one of the coordinate axes.
Download
Paper Nr: 177
Title:

AI-Accelerated Viewshed Computation for High-Resolution Elevation Models
Authors:

Cédric Schwencke, Dominik Stütz and Dimitri Bulatov
Abstract: Viewshed computation, essential for visibility analysis in GIS applications, involves determining visible areas from a given point using the digital terrain model (DTM) and digital surface model (DSM). The traditional methods, though accurate, can be computationally intensive, especially with increasing search distances and high-resolution elevation DSMs. This paper introduces a novel approach leveraging neural networks to estimate the farthest visible point (FVP). At this point the viewshed computation could be aborted, which significantly reducing computation time without compromising accuracy. The proposed method employs a fully connected neural network trained on varied terrain profiles, achieving over 99% accuracy in visibility predictions while reducing the required amount of computations by more than 90%. This approach demonstrates substantial performance gains, making it suitable for applications requiring fast visibility analysis.
Download
Paper Nr: 235
Title:

Second Order Differential Properties of Tensor Product Fractal Surfaces
Authors:

Clement Poull, Christian Gentil, Celine Roudet, Lucie Druoton and Michaël Roy
Abstract: Many domains require non-smooth surface geometries: industry with quality control or CAD, computer graphics with geometric texture generation or terrain synthesis... Fractal models like the Iterated Function Systems (IFS) model are capable of generating self-similar multiscale objects, allowing the generation of a large variety of surfaces with non-standard geometries. Preceding works on IFS have demonstrated how to compute and control pseudo-tangents (defined by two different directions for the right and left tangents at each point) everywhere on these nowhere differentiable geometries. The second-order differential form, that provides even more control possibilities, was only proposed for fractal curves via the introduction of the Differential Characteristic Function (DCF). In this paper, we introduce the Surface Differential Characteristic Function (SDCF), an analytical form that helps characterising and analysing the differential properties (tangents and curvatures) of tensor product non-differentiable surfaces. We use the SDCF to compute the pseudo-curvatures for surfaces generated by tensor products of IFS.
Download
Paper Nr: 247
Title:

Multiple Multi-Modal AI for Semantic Annotations of 3D Spatial Data
Authors:

Lee Kent, Hermenegildo Solheiro and Keisuke Toyoda
Abstract: 3D reconstruction of physical environments presents significant challenges, particularly when it comes to the semantic interpretation of these spaces, which often requires human input. This paper introduces a novel process that leverages multiple AI models trained on 2D images to automatically interpret and semantically annotate 3D spaces. Using a game engine as an intermediary, the process facilitates the integration of various 3D formats with 2D-trained AI models, enabling the capture and reprojection of semantic annotations back into the 3D space. A representative 3D scene is employed to evaluate the system’s performance, achieving an object identification accuracy of 87% alongside successful semantic annotation. By offloading semantic annotation tasks to external 2D AI, this approach reduces the computational burden on edge devices, enabling dynamic updates to the system’s internal knowledge base. This methodology enhances the scalability of spatial AI, providing a more comprehensive understanding of 3D reconstructed environments and improving the feasibility of real-time, AI-driven reasoning in spatial applications.
Download
Paper Nr: 271
Title:

Symmetry Detection and Symmetrization in Cellular Automata
Authors:

Vít Gregor and Ivana Kolingerová
Abstract: Symmetry is the important property of many geometric objects. Our work analyzes the symmetry in two-dimensional objects created using the cellular automata in the context of the initial configurations and rules of the automata. Symmetry of basic geometric shapes, such as circles, rectangles, and curves, mapped into the cells of the automaton, is analyzed in this paper. Also, the symmetry of random objects is analyzed. This paper also describes the method for centroidal symmetry detection and axial symmetry detection in the cellular automata, and it also brings the approach of using the cellular automata for object symmetrization by comparing it with objects in the library of symmetrical objects.
Download
Paper Nr: 324
Title:

MCPro: A Procedural Method for Topologically Correct Isosurface Extraction Based on Marching Cubes
Authors:

Julian Stahl and Roberto Grosso
Abstract: In this work, we present an innovative procedural algorithm designed for extracting isosurfaces from scalar data within hexahedral grids. The geometry and the topological features of the intersection of a level set of the trilinear interpolant with the faces and the interior of a reference unit cell are analyzed. Ambiguities are solved without the help of lookup tables, generating a topologically correct triangulation of the level set within the cell that is consistent across cell boundaries. The algorithm is based on constructing and triangulating a polygonal halfedge data structure that includes contours and critical points of the trilinear interpolant. Our algorithm is capable of handling many singular cases that were not solved by previous methods. The efficacy and correctness of the algorithm were tested on a variety of academic and praxis-relevant CT and MRI datasets.
Download
Paper Nr: 360
Title:

Reconstruction of the Botanical Trees from Single-View Images Using Signed Distance Functions
Authors:

Anna Semrau and Dariusz Sawicki
Abstract: Fractal modeling methods revolutionized the approach to modeling objects in computer graphics. They made it possible to create natural objects, where mathematical description using traditional Euclidean Geometry is difficult or even impossible. Plants, especially trees, are a perfect example of this type of objects. However, fractals, due to their properties, create application problems, related to the generation process. One of the new attempts to solve these problems is modeling 3D objects from single-view images. Signed distance functions (SDF), when combined with ray marching, represent a traditional yet underutilized technique that demonstrates significant potential for fractal modeling. The aim of the article is to analyze the possibility of using this method to increase the efficiency of tree modeling, focusing on the 3D reconstruction. We have proposed a novel method for reconstructing the 3D geometry of botanical trees from single-view photographs, which is based on the SDF.
Download
Paper Nr: 396
Title:

Lightweight Transformer Occupancy Networks for 3D Virtual Object Reconstruction
Authors:

Claudia Melis Tonti and Irene Amerini
Abstract: The increasing demand for edge devices highlights the necessity for modern technologies to be adaptable to general-purpose hardware. Specifically, in fields like augmented reality, virtual reality, and computer graphics, reconstructing 3D objects from sparse point clouds is highly computationally intensive, presenting challenges for execution on embedded devices. In previous works, the speed of 3D mesh generation has been prioritized with respect to preserving a high level of detail. Our focus in this work is to enhance the speed of the inference in order to get closer to real-time mesh generation.
Download

Area 2 - Rendering

Full Papers
Paper Nr: 274
Title:

Backface Distance Fields: Relaxing Signed Distance Fields
Authors:

Róbert Bán, Csaba Bálint and Gábor Valasek
Abstract: We propose backface distance functions, an implicit volume representation that improves the convergence rate of sphere tracing. We employ the closest signed distances to backfacing surface points, introducing a relaxed representation of signed distance functions. The backface and signed distance functions coincide within the volume. For external points, we prove that a backface distance-sized step is the largest direction-independent step along a ray that does not pass through the volume boundary more than once. We show analytic and discrete realizations of our concept. We present a discrete backface distance field generation method to construct exact and approximate fields from triangular meshes and procedural implicit scenes. We employ generation-time processing and correction steps in the discrete case to ensure robust surface visualization in combination with GPU filtering. We validate the proposed discrete and analytic representations empirically as well by comparing their performance to basic, relaxed, and enhanced sphere tracing and demonstrate that it generally outperforms the other methods.
Download
Short Papers
Paper Nr: 51
Title:

Ray-LUT: A Lookup-Based Method for Camera Lens Simulation in Real-Time Using Ray Tracing
Authors:

Jan Honsbrok, Sina Mostafawy, Jens Herder and Alina Huldtgren
Abstract: Lens systems have a major influence on the image due to effects such as depth of field or optical aberrations. The only method to simulate these effects precisely is to trace rays through an actual lens system. This provides accurate results, but only with high computational effort. To speed up the ray tracing through the lens system, various acceleration methods have been developed, requiring considerable precomputations. We present a new method based on the Realistic Camera by Kolb et. al.. Instead of tracing each ray through the lens system, the rays are precomputed once and stored in a lookup table. In contrast to other methods, our method is simple, and does not require substantial preprocessing upfront. We can simulate complex effects such as chromatic aberrations accurately in real-time, regardless the number of lens surfaces in the system. Our method achieves the same performance as state-of-the-art methods like Polynomial Optics, while maintaining the same number of samples per pixel.
Download
Paper Nr: 185
Title:

Forge: Extending Anvil for Visual Evaluation of Rendering Pipelines
Authors:

Kevin Napoli, Keith Bugeja and Sandro Spina
Abstract: This paper introduces Forge, an extension of Anvil, aimed at enhancing evaluation processes in computer graphics pipelines. Forge addresses critical challenges in rendering systems, such as ensuring consistent configurations, minimising human error, and increasing reproducibility of experimental results. By decoupling evaluation logic from rendering engines, Forge facilitates seamless comparisons across different systems without manual configuration. The framework’s architecture supports decentralised evaluations, enabling operations across diverse environments and platforms. This flexibility allows for both local and remote evaluations, making Forge adaptable for a broad range of research applications, from small-scale experiments to comprehensive distributed rendering evaluations. Through case studies, this paper demonstrates Forge’s effectiveness in verifying rendering techniques, comparing performance, and aiding development of new algorithms, thereby providing a robust solution for accurate and reliable comparative studies in the field of computer graphics.
Download
Paper Nr: 239
Title:

Deep Image-Based Adaptive BRDF Measure
Authors:

Wen Cao
Abstract: Efficient and accurate measurement of the bi-directional reflectance distribution function (BRDF) plays a key role in realistic image rendering. However, obtaining the reflectance properties of a material is both time-consuming and challenging. This paper presents a novel iterative method for minimizing the number of samples required for high quality BRDF capture using a gonio-reflectometer setup. The method is a two-step approach, where the first step takes an image of the physical material as input and uses a lightweight neural network to estimate the parameters of an analytic BRDF model. The second step adaptive sample the measurements using the estimated BRDF model and an image loss to maximize the BRDF representation accuracy. This approach significantly accelerates the measurement process while maintaining a high level of accuracy and fidelity in the BRDF representation.
Download
Paper Nr: 241
Title:

ARPΔ: Hardware-Accelerated Ray-Traced Photon Differentials
Authors:

Adrian De Barro, Keith Bugeja and Sandro Spina
Abstract: Photon mapping is a widely used rendering technique that provides biased but consistent global illumination through particle and radius-based density estimation. In this work, we enhance photon mapping by integrating photon differentials, representing each photon as a beam connected to its neighbours. Our proposed method, ARPΔ, dynamically adjusts bandwidth by leveraging changes in both the photon’s position and direction, allowing for adaptive control based on the photon’s path through the scene. Additionally, ARPΔ combines multiple photon differential strategies to enable efficient global illumination on ray tracing hardware, seamlessly transitioning to progressive photon mapping in highly anisotropic conditions. Experimental results demonstrate that ARPΔ achieves image quality comparable to state-of-the-art photon mapping techniques, validating its effectiveness in producing high-fidelity renders.
Download
Paper Nr: 340
Title:

Parallel Axis Split Tasks for Bounding Volume Construction with OpenMP
Authors:

Gustaf Waldemarson and Michael Doggett
Abstract: Many algorithms in computer graphics make use of acceleration structures such as Bounding Volume Hierarchies (BVHs) to speed up performance critical tasks, such as collision detection or ray-tracing. However, while the typical algorithms for constructing BVHs are relatively simple, actually implementing them for performance critical systems is still challenging. Further, to construct them as quickly as possible, it is also desirable to parallelize the process. To that end, parallelization APIs such as OpenMP® can be leveraged to greatly simplify this matter. However, BVH construction is not a trivially parallelizable problem. Thus, in this paper we propose a method of using OpenMP® tasking to further parallelize the spatial splitting algorithm and thus improve construction performance. We evaluate the proposed way and compare it with other ways of using OpenMP®, finding that some of these work well to improve the construction time by between 3 and 5 times on an 8-core machine with a minimal amount of work and negligible quality reduction of the final BVH.
Download

Area 3 - Animation and Simulation

Full Papers
Paper Nr: 36
Title:

Tutrace: Editable Animated Vector Shape from Video
Authors:

Loïc Vital
Abstract: We present a new video vectorization technique for converting a sequence of binary images into an animated vector shape. Our approach offers the benefit of producing an output that can be directly used in a compositing software to apply manual edits and corrections, which is often a mandatory constraint for rotoscoping artists in the VFX industry. Our process initially builds a frame-by-frame vectorization of the input, finds correspondences between vertices at different frames, and extracts an animated vector shape from the induced graph structure. Although the presented method is completely automatic, the general approach is flexible and offers several controls to adjust the fidelity vs. simplicity trade-off in the generated output.
Download
Paper Nr: 65
Title:

Oasis: A Real-Time Hydraulic and Aeolian Erosion Simulation with Dynamic Vegetation
Authors:

Alexander Maximilian Nilles, Lars Günther and Stefan Müller
Abstract: We present a novel real-time combined simulation for aeolian erosion, hydraulic erosion and vegetation, capable of transforming barren deserts with sand dunes into lush forest landscapes and vice versa using simple user interaction. Existing aeolian and hydraulic erosion methods are extended and unified using a moisture model on a layered heightmap, supporting bedrock, soil and sand as terrain materials. Vegetation uses a 3D radius-based model and is efficiently rasterized to a 2D density map via a split-Gaussian model, inhibiting erosion. Abiotic factors such as moisture, terrain slope, surface water and illumination are considered in vegetation growth and vegetation can spread radially as well as with the wind. Each plant considers the position and size of all neighboring plants as a biotic growth factor, made possible through a set of uniform grids of varying resolutions. The user can freely model different plant species by defining their ecological niche and adaptability to changes in terrain elevation and competition with other plants. Even underwater vegetation is possible. Interspecies competition can be defined freely using a competition matrix. The resulting method runs in real-time at a terrain resolution of 20482 with 2,000,000.00 plants.
Download
Paper Nr: 113
Title:

Evaluation of Body Parts Representations in Motion Reconstruction
Authors:

Philippe de Clermont Gallerande, Quentin Avril, Philippe Henri Gosselin, Ferran Argelaguet and Ludovic Hoyet
Abstract: Acquiring, encoding, transmitting, decoding, and displaying motion signals is an essential challenge in our new world of interconnected immersive applications (XR, online games etc.). In addition to being potentially disturbed by multiple factors (e.g., signal noise, latency, packet loss), this motion data should be modifiable and customizable to fit the needs of specific applications. Simultaneously, several approaches have successfully proposed to explicitly integrate the semantics of the human body in a deep learning framework by separating it into smaller parts. We propose to use such an approach to obtain a robust streamed animation data. Specifically, we create and train several neural networks on the motion of different body parts independently from each other. We further compare the performances of several body decompositions using multiple objective reconstruction metrics. Eventually, we show that this Body Parts approach brings new opportunities compared to a compact one, such as a perfectly partitioned and more interpretable motion data, while obtaining comparable reconstruction results.
Download
Paper Nr: 124
Title:

Galerkin Enhanced Graph-Based FEM for Interactive Fracture and Sculpting Applications
Authors:

Avirup Mandal, Parag Chaudhuri and Subhasis Chaudhuri
Abstract: Physically-based fracture and cutting simulations are rarely incorporated in interactive graphics systems because the required computation stifles the speed of interaction. We enhance a physically based method for object deformation and fracture by using multigrid approximations to expedite the full dynamics solve of the system. Our method combines a Galerkin multigrid approach with the graph-based Finite Element Method so that remeshing-free fracture and cutting simulation can be done by solving the system dynamics on a hierarchy of coarse to fine meshes while accumulating residual error that is fully resolved only at the coarsest level. We demonstrate the effectiveness of our algorithm by using it to develop an interactive virtual sculpting framework that enables users to shape object meshes in a physically consistent manner. We compare our method with other state-of-the-art virtual 3D object editing solutions to show that our method can provide better physically consistent solutions at interactive speeds.
Download
Paper Nr: 127
Title:

Fast Detection of Jitter Artifacts in Human Motion Capture Models
Authors:

Mateusz Pawłowicz, Witold Alda and Krzysztof Boryczko
Abstract: Motion capture is the standard when it comes to acquiring detailed motion data for animations. The method is used for high-quality productions in many industries, such as filmmaking and game development. The quality of the outcome and the time needed to achieve it are incomparable with the keyframe-based manual method. However, the motion capture data sometimes gets corrupted, which results in animation artifacts that make it unrealistic and unpleasant to watch. An example of such an artifact is a jitter, which can be defined as the rapid and chaotic movement of a joint. In this work, we focus on detecting the jitter in animation sequences created using motion capture systems. To achieve that, here is proposed a multilevel analysis framework that consists of two metrics: Movement Dynamics Clutter (MDC) and Movement Dynamics Clutter Spectrum Strength (MDCSS). The former measures the dynamics of a joint, while the latter metric allows the classification of a sequence of frames as a jitter. The framework was evaluated on popular datasets to analyze the properties of the metrics. The results of our experiments revealed that two of the popular animation datasets, LAFAN1 and Human3.6M, contain instances of jitter, which was not known before inspection with our method.
Download
Paper Nr: 312
Title:

Enhancing Sketch Animation: Text-to-Video Diffusion Models with Temporal Consistency and Rigidity Constraints
Authors:

Gaurav Rai and Ojaswa Sharma
Abstract: Animating hand-drawn sketches using traditional tools is challenging and complex. Sketches provide a visual basis for explanations, and animating these sketches offers an experience of real-time scenarios. We propose an approach for animating a given input sketch based on a descriptive text prompt. Our method utilizes a parametric representation of the sketch’s strokes. Unlike previous methods, which struggle to estimate smooth and accurate motion and often fail to preserve the sketch’s topology, we leverage a pre-trained text-to-video diffusion model with SDS loss to guide the motion of the sketch’s strokes. We introduce length-area (LA) regularization to ensure temporal consistency by accurately estimating the smooth displacement of control points across the frame sequence. Additionally, to preserve shape and avoid topology changes, we apply a shape-preserving As-Rigid-As-Possible (ARAP) loss to maintain sketch rigidity. Our method surpasses state-of-the-art performance in both quantitative and qualitative evaluations. https://graphics-research-group.github.io/ESA/.
Download
Short Papers
Paper Nr: 286
Title:

Beyond the Ragdoll: Designing a Safe-Falling Controller for Physically Simulated Characters
Authors:

Lovro Boban and Mirko Sužnjević
Abstract: In this paper we propose the design for a procedural, physically simulated animation system that produces safe falling animations using reinforcement learning. A character controller is trained to minimize external and internal forces on the humanoid character's body after it is pushed backwards on a flat surface. We design a questionnaire and conduct a user study that compares the reinforcement learning approach with a motion capture approach using subjective ratings on various aspects of the character's movement. Our findings, based on a sample size of (n = 25), indicate that users prefer the motion capture approach on 6 out of 8 aspects, but prefer the reinforcement learning approach on the aspect of reactivity.
Download
Paper Nr: 344
Title:

Diffusion Transformer Framework for Speech-Driven Stylized Gesture Generation
Authors:

Nada Elmasry, Yanbo Cheng and Yingying Wang
Abstract: Gestures are a vital component of human expression, playing a pivotal role in conveying information and emotions. Generating co-speech gestures remains challenging in human-computer interaction due to the intricate relationship between speech and gestures. While recent advances in learning-based methodologies have shown some progress, they still encounter limitations, as a lack of diversity and a mismatch between generated gestures and the semantic and emotional context of speech, impacting the effectiveness of communication. In this work, we propose a novel gesture generation framework that takes speech audio and a target style gesture example as inputs, automatically synthesizing new gesture performances that align with the speech in the desired style. Specifically, our framework comprises four main components: a dual-stream audio encoder, a gesture-style encoder, a cross-attention modality fusion module, and a latent diffusion generation module. The dual-stream audio encoder and gesture style encoder extract diverse modality embeddings from audio and motion inputs; the cross-attention fusion module maps the multi-modal embeddings into a unified latent space, and the diffusion module produces expressive and stylized gestures. The results demonstrate the exceptional performance of our method in generating natural and diversified gestures that accurately and coherently convey the intended information, surpassing the benchmarks established by traditional methods. Finally, we discuss future directions for our research.
Download
Paper Nr: 348
Title:

Semi-Automatic Generation of Rotoscope Animation Using SAM and k-means Clustering
Authors:

Mizuki Sakakibara and Tomokazu Ishikawa
Abstract: This paper proposes a novel method for automating the rotoscoping process in anime production by combining SAM (Segment Anything Model) and k-means clustering. Traditional rotoscoping, which involves manually tracing live-action footage, is time-consuming and labor-intensive. Our method automatically generates line drawings and coloring regions suitable for anime production workflows through three main steps: line drawing creation using SAM2, shadow region generation using k-means clustering, and finishing with color design. Experimental results from 134 participants showed that our method achieved significantly higher ratings in both “rotoscope-likeness” and “anime-likeness” compared to existing methods, particularly in depicting complex human movements and details. The method also enables hierarchical editing of animation materials and efficient color application across multiple frames, making it more suitable for commercial anime production pipelines than existing style transfer approaches. While the current implementation has limitations regarding segmentation accuracy and line drawing detail, it represents a significant step toward automating and streamlining the anime production process.
Download
Paper Nr: 327
Title:

Generative Narrative-Driven Game Mechanics for Procedural Driving Simulators
Authors:

Nelson Bilber Rodrigues, António Coelho and Rosaldo J. F. Rossetti
Abstract: Driving simulators are essential tools for training, education, research, and scientific experimentation. However, the diversity and quality of virtual environments in simulations is limited by the specialized human resources availability for authoring the content, leading to repetitive scenarios and low complexity of real-world scenes. This work introduces a pipeline that can process text-based narratives outlining driving experiments to procedurally generate dynamic traffic simulation scenarios. The solution uses Retrieval-Augmented Generation alongside local open-source Large Language Models to analyse unstructured textual information and produce a knowledge graph that encapsulates the world scene described in the experiment. Additionally, a context-based formal grammar is generated through inverse procedural modelling, reflecting the game mechanics related to the interactions among the world entities in the virtual environment supported by CARLA driving simulator. The proposed pipeline aims to simplify the generation of virtual environments for traffic simulation based on descriptions from scientific experiment, even for users without expertise in computer graphics.
Download

Area 4 - Interactive Environments

Full Papers
Paper Nr: 222
Title:

AI-Informed Interactive Task Guidance in Augmented Reality
Authors:

Viacheslav Tekaev and Raffaele de Amicis
Abstract: This paper presents a proof of concept for an augmented reality (AR) and artificial intelligence (AI)-powered task guidance system, demonstrated through the task of opening a door handle. The system integrates an AR frontend, deployed on an Oculus Quest Pro, with an AI backend that combines computer vision for real-time object detection and tracking, and natural language processing (NLP) for dynamic user interaction. Objects such as door handles are identified using YOLOv8-seg, and their 3D positions are calculated to align with the user’s environment, ensuring accurate task guidance. The AI backend supports local and cloud processing, maintaining performance even without internet connectivity. The system provides adaptive feedback, adjusting guidance based on user actions, such as correcting improper rotation of a knob. Real-time communication between components is achieved via WebSocket, minimizing latency. Technical challenges like tracking accuracy, latency, and synchronization are addressed through calibration and stress testing under vary-ing conditions. The study emphasizes the system’s adaptability to complex scenarios, offering error-handling mechanisms and smooth interaction through AR overlays. This proof of concept highlights the potential of AR-AI integration for task guidance in diverse applications.
Download
Paper Nr: 313
Title:

Exploring Seated Locomotion Techniques in Virtual Reality for People with Limited Mobility
Authors:

Marlene Huber, Simon Kloiber, Annalena Ulschmid, Agata Marta Soccini, Alessandro Clocchiatti, Hannes Kaufmann and Katharina Krösl
Abstract: Virtual reality (VR) is often designed as a standing experience, excluding individuals with limited mobility. Given that a significant portion of the population experiences lower-body mobility restrictions, accessible VR locomotion must accommodate users without requiring lower-body movement. To build a comprehensive understanding of suitable locomotion techniques (LTs) for this demographic, it is crucial to evaluate the feasibility of various approaches in virtual environments (VEs). As a starting point, we present our evaluation approach and a user study on the feasibility and potential of selected LTs for accessible seated locomotion in VR. Our findings indicate that common LTs can be adapted for seated stationary VR. Teleportation-based techniques, in particular, stand out as viable options for accessible locomotion. Although our simulated wheelchair was less popular with non-disabled participants, it was well-received by wheelchair users and shows promise as an intuitive LT for this target group.
Download
Short Papers
Paper Nr: 87
Title:

A Method for Standardizing Eye-Tracking and Behavioral Data in Real and Virtual Environments
Authors:

Maxime Dumonteil, Marc J.-M. Macé, Valérie Gouranton, Théophane Nicolas and Ronan Gaugne
Abstract: This paper introduces a methodology for generating standardized and comparable eye-tracking and behavioral data across multiple modalities, in real and virtual environments. Our approach handles data collected using different devices, thereby enabling a comprehensive comparison between different modalities: a real environment, a virtual one using an immersive room setup, and another virtual environment using head-mounted displays. The versatility of this methodology is illustrated through an archaeological case study, in which the gaze patterns and behavioral responses of participants are analyzed while they interact with artifacts. However, this methodology is applicable to broader research areas involving eye tracking and behavior in mixed environments. By explaining a workflow for the preparation, data acquisition, and post-processing of data, our approach enables the generation of 3D eye-tracking and behavioral data. Subsequently, our presentation is accompanied by examples of metrics and visualization that are relevant in such a comparison study, providing insights into cross-modal behavioral and gaze pattern analysis.
Download
Paper Nr: 134
Title:

Lightweight Visualisation for Vortex Tracking in Airflow Acquisition
Authors:

Nicolas Courilleau, Louis-Wilhelm Raban-Schürmann, Daniel Meneveaux, Kamel Abed-Meraim and Anas Sakout
Abstract: Ventilation systems are spread in most buildings and housing. They provide control on air quality, while provisioning acceptable thermal conditions. However, slotted plates that direct air jets often produce acoustic disturbances with a self-sustained tone. Understanding and controlling the phenomenon requires complex experimentation with expensive setups. This article proposes an open source, web-based, interactive visualisation system dedicated to the observation and analysis of recorded high frequency captures (3kHz or more) of air streams, charged with thin oil particles. It relies on the acquired images and estimated vector fields, coming from industrial existing systems. Our goal is to visualise the main flow parameters, such as speed, gradients, directions, vortex tracking with path prediction, and vortex frequencies. The obtained results highlight interesting phenomena that illustrate sounds production. They can be employed by physicists to understand, explain, and control the generated acoustics.
Download
Paper Nr: 140
Title:

A Whole New World: Can Virtual Reality Help to Understand Non-Euclidean Geometries?
Authors:

Maé Mavromatis, Ronan Gaugne, Rémi Coulon and Valérie Gouranton
Abstract: With the democratisation of digital technologies, new pedagogical approaches are emerging that leverage these innovative media to enhance student engagement and promote different ways of learning. This article compares three learning modalities—slides, screen, and VR—in terms of knowledge acquired, time spent, and usability. The slides modality involves an illustrated slide presentation, the screen modality uses an onscreen simulation with navigation, and the VR modality shows the same simulation in virtual reality with a Head-Mounted Display (HMD). In this study, we investigated the impact of these modalities on students’ understanding of the essential properties of the unintuitive non-Euclidean geometries S3 and H3. All three modalities helped participants improve their answers to the mathematics questionnaire, though further research is needed to fully exploit the unique benefits of virtual reality.
Download
Paper Nr: 195
Title:

A Proposed Immersive Digital Twin Architecture for Automated Guided Vehicles Integrating Virtual Reality and Gesture Control
Authors:

Mokhtar Ba Wahal, Maram Selsabila Mahmoudi, Ahmed Bahssain, Ikrame Rekabi, Abdelhalim Saeed, Mohamad Alzarif, Mohamed Ellethy, Neven Elsayed, Mohamed Abdelsalam and Tamer ElBatt
Abstract: Digital Twins (DTs) are virtual replicas of physical assets, facilitating a better understanding of complex Cyber-Physical Systems (CPSs) through bidirectional communication. As CPS grows in complexity, the need for enhanced visualization and interaction becomes essential. This paper presents a framework for integrating virtual reality (VR) with a Dockerized private cloud to minimize communication latency between digital and physical assets, improving real-time communication. The integration, based on the Robot Operating System (ROS), leverages its modularity and extensive libraries to streamline robotic control and system scalability. Key innovations include a proximity heat map surrounding the digital asset for enhanced situational awareness and VR-based hand gesture control for intuitive interaction. The framework was tested using TurtleBot3 and a 5-degree-of-freedom robotic arm, with user studies comparing these techniques to traditional web-based control methods. Our results demonstrate the efficacy of the proposed VR and private cloud integration, providing a promising approach to advance Human-Robot Interaction (HRI).
Download
Paper Nr: 350
Title:

Blending Realities: Accessible Mixed Reality for Tremor Rehabilitation in Parkinson’s Disease
Authors:

Xinjun Li and Zhenhong Lei
Abstract: This position paper presents a novel MR-based hand motion assistance device designed for individuals with Parkinson’s disease (PD)-related tremor disorders. We argue that the integration of ergonomic hardware design with adaptive MR software can significantly enhance the efficacy of tremor rehabilitation, improve patient engagement, and lead to superior functional outcomes. Our approach combines a smartphone-based MR system with an ergonomic physical support device, leveraging advanced spatial computing, computer vision algorithms, and haptic feedback technologies. This innovative solution addresses both immediate stabilization needs and long-term motor skill improvement, potentially revolutionizing home-based rehabilitation for millions of PD patients worldwide. By seamlessly blending virtual and real-world elements, our system creates immersive, interactive, and personalized therapeutic experiences that overcome the limitations of traditional rehabilitation methods. The paper discusses the design research methodology, comparative analysis with existing approaches, and the potential impact of this technology on PD rehabilitation.
Download
Paper Nr: 386
Title:

Reinforcement Learning-Enhanced Procedural Generation for Dynamic Narrative-Driven AR Experiences
Authors:

Aniruddha Srinivas Joshi
Abstract: Procedural Content Generation (PCG) is widely used to create scalable and diverse environments in games. However, existing methods, such as the Wave Function Collapse (WFC) algorithm, are often limited to static scenarios and lack the adaptability required for dynamic, narrative-driven applications, particularly in augmented reality (AR) games. This paper presents a reinforcement learning-enhanced WFC framework designed for mobile AR environments. By integrating environment-specific rules and dynamic tile weight adjustments informed by reinforcement learning (RL), the proposed method generates maps that are both contextually coherent and responsive to gameplay needs. Comparative evaluations and user studies demonstrate that the framework achieves superior map quality and delivers immersive experiences, making it well-suited for narrative-driven AR games. Additionally, the method holds promise for broader applications in education, simulation training, and immersive extended reality (XR) experiences, where dynamic and adaptive environments are critical.
Download
Paper Nr: 392
Title:

A Virtual Reality Prototype for Evaluating Emotion Induced by Poetry and Ambient Electronic Music
Authors:

Irene Fondón and María Luz Montesinos
Abstract: Music and poetry evoke emotion. The neural correlates of emotion are not well known but constitute a growing research field in neuroscience. Using brain imaging techniques, emotion elicited by music has been addressed by several laboratories. However, most studies are focused on classical music. Surprisingly, ambient electronic music, which possesses the aesthetic characteristics prone to evoke auto-memories and emotion, has not been evaluated, and only a few studies are available regarding poetry. In the last years, virtual reality (VR) has become a powerful tool to elicit and enhance emotions, since this technology immerses the user in a three-dimensional environment, increasing the sense of presence. In this work we describe a VR prototype which combines ambient electronic music and poetry, as a preliminary step in our project to address the neural correlates of emotion and self-awareness evoked by these arts.
Download
Paper Nr: 397
Title:

Integrating Generative AI in Architectural Education: A Comparative Study of Traditional, Stock LLMs, and Custom Tools
Authors:

Abdelrahman Aly, Abdelsamie Elazazy and Nada Sharaf
Abstract: The rapid development of generative artificial intelligence (AI) is transforming architectural education by reshaping creativity, technical skills, and problem-solving approaches. This paper presents a comparative analysis of traditional methods, general-purpose AI tools like ChatGPT and Midjourney, and a custom-built Architecture AI Tool (ArchAI) tailored to the needs of architectural education. The study highlights the strengths and limitations of each approach, focusing on their impact on creativity, efficiency, and educational outcomes. The findings reveal that while general-purpose AI tools enhance accessibility and ideation, their domain-specific applications are limited. In contrast, custom AI solutions, integrated with architectural principles and tailored datasets, offer significant advantages by automating design tasks, providing real-time feedback, and fostering innovative learning experiences. This work underscores the need for a balanced integration of generative AI to optimize learning outcomes and prepare students for professional practice.
Download
Paper Nr: 405
Title:

VIRTA-Yoke: A Virtual-Integrated Poka-Yoke System for Error Prevention and Operator Training in Manufacturing Processes
Authors:

Guillermo Leale, Baltazar Cortina and Rodrigo D’Andrea
Abstract: This paper introduces VIRTA-Yoke (VIRTA: Virtual Integrated Reliability and Training Assistant), a proof-of-concept virtual Poka-Yoke platform developed to increase reliability and efficiency in manufacturing processes for aluminum engine components. In contrast to traditional mechanical Poka-Yoke systems that require custom fixtures for each part, VIRTA-Yoke employs a low-cost webcam and a virtual reality (VR) headset to guide operators through each assembly step, verify correct placement in specific control areas, and provide real-time feedback when deviations occur. The system uses a convolutional neural network (ConvNet) to detect errors in coil insert placements. This information appears on the VR headset, minimizing operator distraction, optimizing operation times, and improving process adherence. In addition, the VR headset serves as a training environment, allowing new personnel to learn assembly procedures through a virtual component model before working on the factory facility. Preliminary tests indicate an accuracy exceeding 90% in overall defect detection, suggesting that VIRTA-Yoke is a scalable, cost-effective method for streamlining quality control, improving operator training, and eliminating the need for multiple custom mechanical fixtures across a wide range of parts.
Download
Paper Nr: 223
Title:

Design and Development of an Interactive and Intelligent Wood Harvester Machine Operator Simulator
Authors:

Barve Pranjali Ramesh, Ian Backus and Raffaele de Amicis
Abstract: This paper presents the design and development of a virtual reality (VR) forestry harvester simulator, optimized for the untethered Meta Quest. The simulator offers an immersive training environment where users can practice essential harvester operations such as navigation, tree processing, and control of the harvester head and boom. A comprehensive functional evaluation was conducted using 20 black-box test cases to ensure the simulator functions as intended, with testing performed in both standalone and PC-tethered configurations. The results confirmed the simulator’s reliability, highlighting differences in responsiveness and graphical performance across configurations. With the portability and accessibility of the Meta Quest, the system provides a flexible, cost-effective solution suitable for both training and educational applications. Future work will focus on evaluating the usability of the system and validating its effectiveness within formal educational settings by integrating the simulator into forestry curricula.
Download