Ing. David Sedláček, Ph.D.

The presentation of clothing collection objects in museums brings specific constraints and requirements. Textile objects are particularly sensitive to mechanical handling and prolonged exposure to light. At the same time, it is interesting to see these objects in motion as they were worn. In this paper, we propose a new method of presenting digital 3D models of clothing collection objects that allows us to display these objects in motion. The presentation is handled by two real-time applications. Virtual Wardrobe, designed especially for enthusiasts, aims at simulating the movement of the fabric as closely as possible, with the possibility of zooming in on details and showing clothing in a historicizing environment. Virtual Mirror, aimed towards the general public, allows visitors to virtually put on selected collection objects. The Virtual Mirror can display up to three visitors simultaneously and performs semiautomatic calibration to their character types. The main innovation of the proposed presentation of historical clothing objects is presentation of their detailed 3D reconstruction in motion in real time on dimensionally appropriate models of figures. This makes the solution different from the optimized models used in computer games and the non-real-time rendering used in the film industry. Both applications were subjected to a long-term usage study and a questionnaire study to determine visitor acceptance, which was positive. We summarize the experience of deploying both applications in practice at the museum exhibition.

Illusive perspectives (forced perspectives) have been used in art and architecture to create certain space perceptions (e.g., let built space appear larger or deeper), space illusions (e.g., let persons or objects in certain positions appear larger or distorted) and artificial effects (e.g., nudge the visitor to stand at a certain position, i.e., forced perspective). To induce these spatial visual illusions, built spaces like hallways have been constructed such that visual cues induce increased depth if perceived by the visitor from a certain position. In our study, our aim was to determine how the properties of a distorted hallway (e.g., angle of the walls, amount of distortion) influence the perceived illusory space. We created a simplifiedVR model of an illusive hallway inspired by a historical example designed by Italian architect Francesco Borromini, and conducted a qualitative pilot study.

Urban data visualization plays a vital role in sustainable city evolution. Visual media enable efficient communication, which is the cornerstone of any development. This paper presents a design and implementation of a modular urban data visualization system. The modular design allows extending the visualization system, which helps answer questions as they arise during problem exploration. The paper explores the role of visualization platforms in urban development. Technical topics such as possible geometry representations and ways to process large geospatial datasets are discussed. A simple and extensible styling language is proposed to enable visualization customization based on the object metadata. The system is available as two Python/C++ packages. The first package focuses on data processing, utilizing spatial and temporal acceleration data structures, while the second encapsulates a WebGL-based visualization application. An iterative qualitative user study validated the proposed solution's performance and accessibility.

Článek se věnuje fenoménu XR technologie v kontextu použití na hradech a zámcích. Cílem článku je seznámit audienci s XR technologií a diskutovat základní aspekty jejího využití na hradech a zámcích, včetně detailnějšího popisu praktických zkušeností a implementací autorů v průmyslových, zábavních i muzejních realizacích. XR technologie se již etablovala nejen v domácnostech, ale i ve školách, kde přispívá k porozumění probírané látky i k prohlubování znalostí. Hlavním přínosem článku má být osvěta audience a přimění k využívání XR technologie k oživení stávajících instalací.

Základní důvody, které často vedou muzea k 3D digitalizaci, jsou využití pro vědu a výzkum, ochranu a restaurování či prezentaci v online prostředí. V tomto případě má 3D digitalizace zaručeně své místo, protože odbourává problémy s opotřebením předmětů a díky 3D prohlížení je možné s objekty jakkoliv otáčet, přibližovat si je či poslat kolegovi na druhý konec světa. Méně už je řešena otázka uchování takto vzniklých dat, která jsou často objemnější a ve specifických formátech a souborech, narozdíl od výstupů běžné digitalizace. Kromě zmiňovaných tradičních způsobů mají 3D digitalizáty velký potenciál využití v oblasti 3D tisku, produktovém designu, virtuální a rozšířené realitě, vzdělávání a zejména kreativních průmyslů.

To increase the interest of young people in the technological and scientific aspects of space research, the European Space Education Resources Office (ESERO) has decided to develop a set of games with space themes. Our group was selected to design and implement a game in virtual reality. In this paper, we describe the main design principles, interaction and navigation techniques, and technological details of the final game titled Moon Base.

This paper presents a review of various techniques for improving the performance of neural networks on segmentation task using 3D convolutions and voxel grids – we provide comparison of network with and without max pooling, weighting, masking out the segmentation results, and oversampling results for imbalanced training dataset. We also present changes to 3D U-net architecture that give better results than the standard implementation. Although there are many out-performing architectures using different data input, we show, that although the voxel grids that serve as an input to the 3D U-net, have limits to what they can express, they do not reach their full potential.

Unmanned aerial vehicles (UAVs or drones) are nowadays commonly used for data acquisition, for monitoring dangerous places and for free time activities. Given their mobility and fast and often chaotic movement, following and selecting them can be quite hard. In cooperation with Agent Technology Center and their UAV simulator our work was focused on creation of a visualisation for the operator to control drones and plan missions for them.

Older adults gradually lose their physical and psychical abilities, and they have to get accustomed to this fact. This situation causes certain discomfort for these people and question arises how to make their life easier to minimize the discomfort. One of the possible solutions is to provide them with training by means of which they can acquire new skills that may ideally compensate their inabilities to a satisfactory extent. In this paper, we will discuss possibilities how to employ new information technologies allowing to simulate the virtual interaction of the person in various environments. The use of these technologies also allows personalization, i.e., taking into account the abilities of individual users that suffer from a particular handicap. Moreover, these technologies make possible to train people for new situations in new environments in advance, before they occur there physically, for instance, to train them for operating new devices in the household, etc. The goal of this paper is to attract the attention of people who work in institutions where care for older adults is provided to new methods that in principle can increase the comfort of older adults together with allowing caregivers to increase quality and efficiency of their work. The technology cannot replace human caregivers, but it should provide them with new possibilities in the process when older adults are treated.

The role of human in the Industrie 4.0 vision is still considered as irreplaceable. Therefore, user interfaces of cyber-physical systems involved in the production automation need to be well designed and taking into consideration the industrial application requirements. With the advances in augmented and virtual reality data visualization and novel interaction techniques like mid-air gestures, these approaches seem to be suitable for integration into the industry environment. This paper describes the implementation of an augmented reality application for smart glasses with mid-air gestures and smart phone with touch interaction to compare and evaluate the usage of such interfaces in a production cell comprising an industrial robot.

3D rekonstrukce se snaží převést objekty reálného světa do trojrozměrné digitální reprezentace. Metoda pomáhá například architektům, archeologům nebo pracovníkům muzeí snažícím se zachovat kulturní dědictví příštím generacím a zároveň je zpřístupnit laické veřejnosti. Významnou událostí v této oblasti byla 3D digitalizace papírového Langweilova modelu staré Prahy.

The paper describes a project entitled TABLETARIUM – Tablets in Science Uducation, currently implemented in the Czech Republic to pilot and evaluate different forms of use of tablets in scho ol teaching. Within the project we are preparing 18 original tablet apps (for physics, chemistry and biology), using different types of design and all of the main educational tools (e.g. Augmented Reality in its interactive as well as passive form, active GPS locations, student collaboration, and many interactive tasks). As an important part, the programming development stage also involves creating a common platform environment to share data/results within the classroom, working on the generation of final reports, and testing connectivity of the apps with different measurement probes and other devices (microscopes, micro-cameras).

A large data set of input RAW photos for 3D reconstruction of historical paper model of Prague is presented in this paper. The data acquisition process and its organisation are described in detail together with color and camera calibration algorithms used. Moreover, a complete reconstruction work-flow together with several cultural heritage applications are briefly presented. The subject of this 3D reconstruction has been the Langweil model of Prague, originally created in the first half of 19th century. Three parts of this model with different characteristics are afforded for computer scientists through the internet.

A tool for user driven 3D reconstruction is presented in this poster together with ideas of tool extension for a work in immersive environment with multi-user editing possibility. The main purpose of the tool is cultural heritage and architecture reconstruction with direct visualization.

An intuitive image-based 3D reconstruction tool based on inaccurate user strokes is presented in this paper. The combination of fast image segmentation method together with user knowledge about reconstructed scene forms a novel low-polygonal editor suitable for architecture reconstruction. The user interaction is minimized thanks to propagation of strokes among input photographs. The final model geometry is created by innovative algorithm. The input to the tool is a set of calibrated photographs together with a sparse pointcloud. The output is a structured low-poly 3D model.

focused on procedural techniques like L systems or shape grammars and their usability in procedural architecture creation. Then, we introduce essential articles on procedural modelling of cities and we also include ideas from papers specialized in generation of buildings, facades, architectural structures and related city components. Finally, we focus on interesting computer vision techniques in architecture reconstruction and their connection to formal languages. Our aim is to demonstrate that the procedural city modelling needs to integrate specially modified computer vision algorithms in order to achieve a higher degree of realism and a wider acceptance in various professional application areas.

The challenges for current state-of-the-art 3D reconstruction algorithms covered in the Langweil model of Prague data set are summarized in this poster. The model properties are presented together with problems which bring.

This paper presents a novel interactive approach for adding depth information into hand-drawn cartoon images and animations. In comparison to previous depth assignment techniques our solution requires minimal user effort and enables creation of consistent pop-ups in a matter of seconds. Inspired by perceptual studies we formulate a custom tailored optimization framework that tries to mimic the way that a human reconstructs depth information from a single image. Its key advantage is that it completely avoids inputs requiring knowledge of absolute depth and instead uses a set of sparse depth (in)equalities that are much easier to specify. Since these constraints lead to a solution based on quadratic programming that is time consuming to evaluate we propose a simple approximative algorithm yielding similar results with much lower computational overhead. We demonstrate its usefulness in the context of a cartoon animation production pipeline.

The reconstruction of 3D objects from a point-cloud is based on sufficient separation of the points representing objects of interest from the points of other, unwanted objects. This operation called segmentation is discussed in this paper. We present an interactive unstructured point-cloud segmentation based on graph cut method where the cost function is derived from euclidean distance of point-cloud points. The graph topology and direct 3D point-cloud segmentation are the novel parts of our work. The segmentation is presented on real application, the terrain reconstruction of a complex miniature paper model, the Langweil model of Prague.

In this paper, we introduce a light-weight and robust tracking technique based on color balls. The algorithm builds on a variant of randomized Hough transform and is optimized for a use in real-time applications like on low-cost Augmented Reality (AR) systems. With just one conventional color camera our approach provides the ability to determine the 3D position of several color balls at interactive frame rates on a common PC workstation. It is fast enough to be easily combined with another real-time tracking engine. In contrast to popular tracking techniques based on recognition of planar fiducial markers it offers robustness to partial occlusion, which eases handling and manipulation. Furthermore, while using balls as markers a proper haptic feedback and visual metaphor is provided. The exemplary use of our technique in the context of two AR applications indicates the effectiveness of the proposed approach.

We present a solution for a variant of the 2D containment problem, to create a texture atlas. The texture atlas is an efficient color representation for a given VRML scene. It contains a composition of several texture images stored in one file. The containment problem is the question how to place a set of shapes into a "container shape" without overlapping, while minimizing the area of the container. In our case we place a set of texture images of arbitrary shape into a rectangle. We use a heuristic search algorithm based on Minkowski operators to solve the 2D translational containment problem. This particular problem is called packing problem and is known to be NP-complete. We present its practical implementation and show results measured on several test scenes.