GRAPP 2023 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: 7
Title:

Local Reflectional Symmetry Detection in Point Clouds Using a Simple PCA-Based Shape Descriptor
Authors:

Lukáš Hruda, Ivana Kolingerová and David Podgorelec
Abstract: Symmetry is a commonly occurring feature in real world objects and its knowledge can be useful in various applications. Different types of symmetries exist but we only consider the reflectional symmetry which is probably the most common one. Multiple methods exist that aim to find the global reflectional symmetry of a given 3D object and although this task on its own is not easy, finding symmetries of objects that are merely parts of larger scenes is much more difficult. Such symmetries are often called local symmetries and they commonly occur in real world 3D scans of whole scenes or larger areas. In this paper we propose a simple PCA-based local shape descriptor that can be easily used for potential symmetric point matching in 3D point clouds and, building on previous work, we present a new method for detecting local reflectional symmetries in 3D point clouds which combines the PCA descriptor point matching with the density peak location algorithm. We show the results of our method for several real 3D scanned scenes and demonstrate its computational efficiency and robustness to noise.
Download
Paper Nr: 17
Title:

Multiclass Texture Synthesis Using Generative Adversarial Networks
Authors:

Maroš Kollár, Lukas Hudec and Wanda Benesova
Abstract: Generative adversarial networks as a tool for generating content are currently one of the most popular methods for content synthesis. Despite its popularity, multiple solutions suffer from the drawback of a shortage of generality. It means that trained models can usually synthesize only one specific kind of output. The usual synthesis approach for generating N different texture species requires training N models with changing training data. However, few solutions explore the synthesis of multiple types of textures. In our work, we present an alternative approach for multiclass texture synthesis. We focus on the synthesis of realistic natural non-stationary textures. Our solution divides textures into classes based on the objects they represent and allows users to control the class of synthesized textures and their appearance. Thanks to the controllable selections from latent space, we also explore possibilities of creating transitions between classes of trained textures for potential better usage in applications where texture synthesis is required.
Download
Paper Nr: 22
Title:

Shape Morphing as a Minimal Path in the Graph of Cubified Shapes
Authors:

Raphaël Groscot and Laurent D. Cohen
Abstract: The systematic study of morphings for non parametric shapes suffers from ambiguities in defining good general morphings, such as the trade-off between plausibility and smoothness, above all under large topology changes. In the recent years, only neural networks have offered a generic solution, using their latent space as a shape prior. But these models are optimized for single shape reconstruction, giving little control on the generated morphings. In this paper, we show how qualitatively similar results can be achieved when replacing neural networks with a set of carefully crafted components: a style-content separation method via the fitting of a Deformable Voxel Grid, a similarity metric adapted to the extracted content, and a formulation of morphings as minimal paths in a graph. While forgoing the automatic learning of a generative model, we still achieve similar morphing capabilities. We performed various evaluations, quantitative analysis on the robustness of our proposed method and on the quality of the results, and demonstrate the usefulness of each component. Finally, we provide guidance on how manual intervention can improve quality. This is indeed possible since, unlike neural networks, each component in our method is interpretable.
Download
Short Papers
Paper Nr: 15
Title:

Accurate Cutting of MSDM-Based Hybrid Surface Meshes
Authors:

Thomas Kniplitsch, Wolfgang Fenz and Christoph Anthes
Abstract: The mass-spring-damper model (MSDM) is a popular method for the physics simulation of surface meshes. Cutting such meshes requires consideration of various contradicting factors: accurate cut representation, maintaining material properties (given by the MSDM geometry) and simulation cost. A hybrid mesh approach partially decouples physics simulation mesh from render mesh by allowing partially rendered physics simulation elements. This paper presents a cutting method for hybrid surface meshes which provides accurate cut representation and maintains MSDM element geometry of cut areas while keeping simulation costs at a competitive level. Additionally, auxiliary data structures, suitable for independent usage, are presented. The bounding box ternary tree is a space partitioning data structure for storing volumetric objects. It subdivides space along an axis-aligned separation plane at each tree level, partitioning objects into below, above and intersecting. A point clustering data structure for efficient retrieval of all points within a given distance is also presented.
Download
Paper Nr: 21
Title:

Dense Point-to-Point Correspondences Between Genus-Zero Shapes Using Cubic Mapping and Horn-Schunck Optical Flow
Authors:

Pejman Hashemibakhtiar, Thierry Cresson, Jacques De Guise and Carlos Vázquez
Abstract: Establishing correspondences is a fundamental and essential task in computer graphics for further processing of shapes. We have proposed an important modification to an existing method to remove several large matching errors in specific regions. The method uses the unit sphere and the regular spherical grid as parameterization spaces to perform registration and obtain the matching map between two three-dimensional genus-zero shapes, considering non-rigid and non-isometric deformations. Although the unit sphere is a suitable parameterization space for rigid alignment, mapping the sphere to a regular spherical grid for non-rigid registration makes the process unstable since it is not a distance-preserving projection. Therefore, it produces large detachments on the grid and for several regions. Replacing the regular spherical grid mapping with Cubic mapping results in smooth displacement and locality for all corresponding vertices on each cube face. Due to our enhancement, the Optical Flow faces a smooth flow field in the non-rigid registration process. Our modification results in the elimination of matches with significant normalized geodesic error and an increase in the accuracy of the correspondence map, compared to the base method and other recently published approaches.
Download
Paper Nr: 24
Title:

Topological Data Structure: The Fast Marching Example
Authors:

Sofian Toujja, Thierry Bay, Hakim Belhaouari and Laurent Fuchs
Abstract: This article lies in the field of front propagation algorithms on a surface represented by triangle meshes. An implementation of the fast marching algorithm using a topological structure, the generalized maps or g-maps, as the data structure of the mesh is presented. G-maps have the advantage of allowing to store and retrieve information related to the neighborhood of a cell. In this article, the necessary knowledge about generalized maps and the fast marching method are reviewed in order to facilitate the understanding of the proposed implementation and the benefits brought by g-maps as underlying data structure. Then some various applications of this implementation are presented.
Download
Paper Nr: 25
Title:

Computerised Muscle Modelling and Simulation for Interactive Applications
Authors:

Martin Cervenka, Ondrej Havlicek, Josef Kohout and Libor Váša
Abstract: The main challenges of collision detection and handling in muscle modelling are demonstrated. Then, a collision handling technique is tested, exploiting the issue of muscle penetrating the bone in some circumstances, mainly when the movement is too rapid or the displacement of the bone is too high. Our approach also detects the problem, using Discregrid to see the immediate direction change towards the penetrated bone. Some alternatives to the described PBD (Position-Based dynamics) technique are presented: PBD with As-Rigid-As-Possible modification and radial basis function approach.
Download
Paper Nr: 27
Title:

Analysis of Wettability Model Using Adhesional and Spreading Works
Authors:

Nobuhiko Mukai, Takuya Natsume, Masamichi Oishi and Marie Oshima
Abstract: We have developed a new method of wettability, which is a feature for a liquid to keep the contact angle formed between a liquid and a solid body. Conventional models required the contact angle in advance for simulations, which angle can be measured by physical experiments. On the other hand, our new model does not need the contact angle and forms the shape of liquid on a solid body by considering adhesional and spreading works. We demonstrated that the proposed method was able to represent wettability by simulations without contact angles. This paper evaluates the proposed method by investigating the drop time of the liquid extruded from a thin tube.
Download
Paper Nr: 1
Title:

Automatic Reconstruction of Roof Overhangs for 3D City Models
Authors:

Steffen Goebbels and Regina Pohle-Fröhlich
Abstract: Most current 3D city models, created automatically from cadastral and remote sensing data and represented in CityGML, do not include roof overhangs, although these overhangs are very characteristic for the appearance of buildings. This paper describes an algorithm that procedurally adds such overhangs. When a CityGML model is textured, the size of the overhangs is determined by recognizing overhangs in facade textures. In this case, the method only needs an already existing model in CityGML representation. Alternatively, if an additional point cloud (e.g., from airborne laser scanning) is available, this cloud can be utilized to calculate the overhang sizes. We compare the results of both methods.
Download

Area 2 - Rendering

Full Papers
Paper Nr: 34
Title:

Sampling-Distribution-Based Evaluation for Monte Carlo Rendering
Authors:

Christian Freude, Hiroyuki Sakai, Károly Zsolnai-Fehér and Michael Wimmer
Abstract: In this paper, we investigate the application of per-pixel difference metrics for evaluating Monte Carlo (MC) rendering techniques. In particular, we propose to take the sampling distribution of the mean (SDM) into account for this purpose. We establish the theoretical background and analyze other per-pixel difference metrics, such as the absolute deviation (AD) and the mean squared error (MSE) in relation to the SDM. Based on insights from this analysis, we propose a new, alternative, and particularly easy-to-use approach, which builds on the SDM and facilitates meaningful comparisons of MC rendering techniques on a per-pixel basis. In order to demonstrate the usefulness of our approach, we compare it to commonly used metrics based on a variety of images computed with different rendering techniques. Our evaluation reveals limitations of commonly used metrics, in particular regarding the detection of differences between renderings that might be difficult to detect otherwise—this circumstance is particularly apparent in comparison to the MSE calculated for each pixel. Our results indicate the potential of SDM-based approaches to reveal differences between MC renderers that might be caused by conceptual or implementation-related issues. Thus, we understand our approach as a way to facilitate the development and evaluation of rendering techniques.
Download
Short Papers
Paper Nr: 5
Title:

Optimal Activation Function for Anisotropic BRDF Modeling
Authors:

Stanislav Mikeš and Michal Haindl
Abstract: We present simple and fast neural anisotropic Bidirectional Reflectance Distribution Function (NN-BRDF) efficient models, capable of accurately estimating unmeasured combinations of illumination and viewing angles from sparse Bidirectional Texture Function (BTF) measurement of neighboring points in the illumination/viewing hemisphere. Our models are optimized for the best-performing activation function from nineteen widely used nonlinear functions and can be directly used in rendering. We demonstrate that the activation function significantly influences the modeling precision. The models enable us to reach significant time and cost-saving in not trivial and costly BTF measurements while maintaining acceptably low modeling error. The presented models learn well, even from only three percent of the original BTF measurements, and we can prove this by precise evaluation of the modeling error, which is smaller than the errors of alternative analytical BRDF models.
Download
Paper Nr: 12
Title:

Deep Interactive Volume Exploration Through Pre-Trained 3D CNN and Active Learning
Authors:

Marwa Salhi, Riadh Ksantini and Belhassen Zouari
Abstract: Direct volume rendering (DVR) is a powerful technique for visualizing 3D images. Though, generating high-quality efficient rendering results is still a challenging task because of the complexity of volumetric datasets. This paper introduces a direct volume rendering framework based on 3D CNN and active learning. First, a pre-trained 3D CNN was developed to extract deep features while minimizing the loss of information. Then, the 3D CNN was incorporated into the proposed image-centric system to generate a transfer function for DVR. The method employs active learning by involving incremental classification along with user interaction. The interactive process is simple, and the rendering result is generated in real-time. We conducted extensive experiments on many volumetric datasets achieving qualitative and quantitative results outperforming state-of-the-art approaches.
Download

Area 3 - Animation and Simulation

Full Papers
Paper Nr: 10
Title:

Unifying Human Motion Synthesis and Style Transfer with Denoising Diffusion Probabilistic Models
Authors:

Ziyi Chang, Edmund C. Findlay, Haozheng Zhang and Hubert P. H. Shum
Abstract: Generating realistic motions for digital humans is a core but challenging part of computer animations and games, as human motions are both diverse in content and rich in styles. While the latest deep learning approaches have made significant advancements in this domain, they mostly consider motion synthesis and style manipulation as two separate problems. This is mainly due to the challenge of learning both motion contents that account for the inter-class behaviour and styles that account for the intra-class behaviour effectively in a common representation. To tackle this challenge, we propose a denoising diffusion probabilistic model solution for styled motion synthesis. As diffusion models have a high capacity brought by the injection of stochasticity, we can represent both inter-class motion content and intra-class style behaviour in the same latent. This results in an integrated, end-to-end trained pipeline that facilitates the generation of optimal motion and exploration of content-style coupled latent space. To achieve high-quality results, we design a multi-task architecture of diffusion model that strategically generates aspects of human motions for local guidance. We also design adversarial and physical regulations for global guidance. We demonstrate superior performance with quantitative and qualitative results and validate the effectiveness of our multi-task architecture.
Download
Short Papers
Paper Nr: 2
Title:

Real-Time Physics-Based Mesh Deformation with Haptic Feedback and Material Anisotropy
Authors:

Avirup Mandal, Parag Chaudhuri and Subhasis Chaudhuri
Abstract: We present a real-time, physics-based framework to simulate porous, deformable materials and interactive tools with haptic feedback that can reshape them. In order to allow the material to be moulded nonhomogeneously, we propose an algorithm to change the material properties of the object depending on its water content. To enable stable visual and haptic feedback at interactive rates, we implement a multi-resolution, multi-timescale solution. We test our model for physical consistency, accuracy, interactivity and appeal through a user study and quantitative performance evaluation.
Download
Paper Nr: 26
Title:

Development of a Realistic Crowd Simulation Environment for Fine-Grained Validation of People Tracking Methods
Authors:

Paweł Foszner, Agnieszka Szczęsna, Luca Ciampi, Nicola Messina, Adam Cygan, Bartosz Bizoń, Michał Cogiel, Dominik Golba, Elżbieta Macioszek and Michał Staniszewski
Abstract: Generally, crowd datasets can be collected or generated from real or synthetic sources. Real data is generated by using infrastructure-based sensors (such as static cameras or other sensors). The use of simulation tools can significantly reduce the time required to generate scenario-specific crowd datasets, facilitate data-driven research, and next build functional machine learning models. The main goal of this work was to develop an extension of crowd simulation (named CrowdSim2) and prove its usability in the application of people-tracking algorithms. The simulator is developed using the very popular Unity 3D engine with particular emphasis on the aspects of realism in the environment, weather conditions, traffic, and the movement and models of individual agents. Finally, three methods of tracking were used to validate generated dataset: IOU-Tracker, Deep-Sort, and Deep-TAMA.
Download
Paper Nr: 29
Title:

Colour-Field Based Particle Categorization for Residual Stress Detection and Reduction in Solid SPH Simulations
Authors:

Gizem Kayar
Abstract: Residual stress remains in an object even in the absence of external forces or thermal pressure, which, in turn, may cause significant plastic deformations. In case the residual stress creates unwanted effects on the material and so is undesirable, an efficient solution is necessary to track and eliminate this stress. Smoothed Particle Hydrodynamics has been extensively used in solid mechanics simulations and the inherent colour-field generation approach is a promising tracker for the residual stress. In this paper, we propose a way to use the colour-field approach for eliminating the residual stress and prevent the undesirable premature failure of solid objects.
Download

Area 4 - Interactive Environments

Full Papers
Paper Nr: 4
Title:

Improved Directional Guidance with Transparent AR Displays
Authors:

Felix P. Strobel and Voicu Popescu
Abstract: In a popular form of augmented reality (AR), the scene is captured with the back-facing camera of a handheld phone or tablet, and annotations are overlaid onto the live video stream. However, the annotations are not integrated into the user’s field of view, and the user is left with the challenging task of translating the annotations from the display to the real world. This challenge can be alleviated by modifying the video frame to approximate what the user would see in the absence of the display, making the display seem transparent. This paper demonstrates a robust transparent display implementation using only the back-facing camera of a tablet, which was tested extensively over a variety of complex real world scenes. A user study shows that the transparent AR display lets users locate annotations in the real world significantly more accurately than when using a conventional AR display.
Download
Paper Nr: 6
Title:

Automatic Prediction of 3D Checkpoints for Technical Gesture Learning in Virtual Environments
Authors:

Noura Joudieh, Djadja Jean Delest Djadja, Ludovic Hamon and Sébastien George
Abstract: Nowadays, Virtual Learning Environments (VLE) dedicated to learning gestures are more and more used in sports, surgery, and in every domain where accurate and complex technical skills are required. Indeed, one can learn from the observation and imitation of a recorded task, performed by the teacher, through a 3D virtual avatar. In addition, the student’s performance can be automatically compared to that of the teacher by considering kinematic, dynamic, or geometric properties. The motions of the body parts or the manipulated objects can be considered as a whole, or temporally and spatially decomposed into a set of ordered steps, to make the learning process easier. In this context, CheckPoints (CPs) i.e. simple 3D shapes acting as “visible landmarks”, with which a body part or an object must go through, can help in the definition of those steps. However, manually setting CPs can be a tedious task especially when they are numerous. In this paper, we propose a machine learning-based system that predicts the number and the 3D position of CPs, given some demonstrations of the task to learn in the VLE. The underlying pipeline used two models: (a) the “window model” predicts the temporal parts of the demonstrated motion that may hold a CP and (b) the “position model” predicts the 3D position of the CP for each predicted part from (a). The pipeline is applied to three learning activities: (i) glass manipulation (ii), geometric shapes drawing and (iii), a dilution process in biology. For each activity, the F1-score is equal to or higher than 70% for the “window model”, while the Normalized Root Mean Squared Error (NRMSE) is below 0.07 for the “position model”.
Download
Paper Nr: 16
Title:

Experimental Setup and Protocol for Creating an EEG-signal Database for Emotion Analysis Using Virtual Reality Scenarios
Authors:

Elías M. Valderrama, Auxiliadora S. Vega, Iván Durán-Díaz, Juan A. Becerra and Irene F. García
Abstract: Automatic emotion recognition systems aim to identify human emotions from physiological signals, voice, facial expression or even physical activity. Among these types of signals, the usefulness of signals from electroencephalography (EEG) should be highlighted. However, there are few publicly accessible EEG databases in which the induction of emotions is performed through virtual reality (VR) scenarios. Recent studies have shown that VR has great potential to evoke emotions in an effective and natural way within a laboratory environment. This work describes an experimental setup developed for the acquisition of EEG signals in which the induction of emotions is performed through a VR environment. Participants are introduced to the VR environment via head-mounted displays (HMD) and 14-channel EEG signals are collected. The experiments carried out with 12 participants (5 male and 7 female) are also detailed, with promising results, which allow us to think about the future development of our own dataset.
Download
Paper Nr: 31
Title:

Real-Time Volume Editing on Low-Power Virtual Reality Devices
Authors:

Iordanis Evangelou, Anastasios Gkaravelis and Georgios Papaioannou
Abstract: The advent of consumer-grade, low-power, untethered virtual reality devices has spurred the creation of numerous applications, with important implications to training, socialisation, education and entertainment. However, such devices are typically based on modified mobile architectures and processing units, offering limited capabilities in terms of geometry and shading throughput, compared to their desktop counterparts. In this work we provide insights on how to implement two combined and particularly challenging tasks on such a platform, those of real-time volume editing and physically-based rendering. We implement and showcase our techniques in the context of a virtual sculpting edutainment application, intended for mass deployment at a virtual reality exhibition centre.
Download
Paper Nr: 35
Title:

Mobile Augmented Reality for Analysis of Solar Radiation on Facades
Authors:

Carolina Meireles, Maria Beatriz Carmo, Ana Paula Cláudio, António Ferreira, Ana Paula Afonso, Paula Redweik, Cristina Catita, Miguel Centeno Brito and Daniel Soares
Abstract: The recent developments of mobile devices have enhanced the possibilities of applications of Augmented Reality (AR), namely, providing data visualization in situ. The application prototype presented in this paper, SolAR, designed for Android tablet devices, allows the user to extract financial and energy feedback from a photovoltaic system, simulating its placement on facades. Such a solution can either serve as a support tool for technicians and researchers in the area or it can be useful for the average user, contributing to the dissemination of the use of renewable energies. SolAR provides a view of the real world augmented with graphical representations of aggregated irradiation data drawn over the facades of buildings. Starting from the previous work, this paper presents the various additions made, particularly the possibility of adding matrices of photovoltaic (PV) modules to several facades of a building and the possibility of obtaining contextual data through a web service. A user study was carried out with 32 volunteers. It revealed that the participants were able to successfully place the PV modules to acquire the best energy efficiency and that the relevance of the new functionalities implemented as well as the usability of the application was positively assessed.
Download
Short Papers
Paper Nr: 13
Title:

An Immersive Feedback Framework for Scanning Probe Microscopy
Authors:

Denis Heitkamp, Jaccomo Lorenz, Maximilian Koall, Philipp Rahe and Philipp Lensing
Abstract: In this paper we introduce an application for analyzing datasets obtained by scanning probe microscopy (SPM). Datasets obtained by such microscopes are typically depicted by two-dimensional images where the measured quantity (typically forces or electric current) is represented by pixel intensities of a rasterized image. Recording several images of this kind with one parameter being in- or decremented before recording the next image results in three-dimensional datasets. A conventional two-dimensional representation of such data by visualizing an axis-aligned slice cutting through the 3D data seems insufficient, since only a fraction of the available data can be examined at once. To improve the understanding of the measured data we propose utilizing a haptic device with four different real-time haptic models (collision, force, vibration and viscosity) in order to reinterpret nano surfaces in an intuitive way. This intuition is furthermore improved by virtually scaling the nano data to normal sized surfaces perceived through a Head Mounted Display (HMD). This stereoscopic visualization is real-time capable while providing different rendering techniques for 3D (volumetric) and 2D datasets. This combination of appealing real-time rendering in conjunction with a direct haptic feedback creates an immersive experience, which has the potential to improve efficiency while examining SPM data.
Download
Paper Nr: 32
Title:

Cartesian Robot Controlling with Sense Gloves and Virtual Control Buttons: Development of a 3D Mixed Reality Application
Authors:

Turhan Civelek and Arnulph Fuhrmann
Abstract: In this paper, we present a cartesian robot controlled by a mixed reality interface that includes virtual buttons, virtual gloves, and a drill. The mixed reality interface, into which the virtual reality input/output devices and sense gloves have been integrated, enables the control of the cartesian robot. The data transfer between the mixed reality interface and the cartesian robot is implemented via Arduino kit. The cartesian robot moves in 6 axes simultaneously with the movement of the sense gloves and the touch of the virtual buttons. This study aims to perform remote-controlled and task-oriented screwing and unscrewing of a bolt using a cartesian robot with a mixed reality interfaces.
Download