No description available.

3D Object Classification Using Neural Networks Bc. Miroslav Krabec Classification of 3D objects is of great interest in the field of artificial intelligence. There are numerous approaches using artificial neural networks to address this problem. They differ mainly in the representation of the 3D model used as input and the network architecture. The goal of this thesis is to explore and test these approaches on publicly available datasets and subject them to independent comparison, which has not so far appeared in the literature. We provide a unified framework allowing to convert the data from common 3D formats. We train and test ten different network on the ModelNet40 and ShapeNetCore datasets. All the networks performed reasonably well in our tests, but we were generally unable to achieve the accuracies reported in the original papers. We suspect this could be due to extensive, albeit unreported, hyperparameter tuning by the authors of the original papers, suggesting this issue would benefit from further research. 1

Virtual reality (VR) technology provides clinicians, therapists, and researchers with new opportunities to observe, assess, and train behavior in realistic yet well-controlled environments. However, VR also comes with a number of challenges. For example, compared to more abstract experiments and tests on 2D computer screens, VR-based tasks are more complex to create, which can make it more expensive and time-consuming. One way to overcome these challenges is to create, standardize, and validate VR content and to make it openly available for researchers and clinicians. Here we introduce the OpenVirtualObjects (OVO), a set of 124 realistic 3D household objects that people encounter and use in their everyday lives. The objects were rated by 34 younger and 25 older adults for recognizability, familiarity, details (i.e., visual complexity), contact, and usage (i.e., frequency of usage in daily life). All participants also named and categorized the objects. We provide the data and the experiment- and analysis code online.With OVO, we hope to facilitate VR-based research and clinical applications. Easy and free availability of standardized and validated 3D objects can support systematic VR-based studies and the development of VR-based diagnostics and therapeutic tools.

In this thesis, we study object ranking in mobile 3D visual search. The conventional methods of object ranking achieve ranking results based on the appearance of objects in images captured by mobile devices while ignoring the underlying 3D geometric information. Thus, we propose to use the method of mobile 3D visual search to improve the ranking by using the underlying 3D geometry of the objects. We develop an algorithm of fast 3D geometric verication to re-rank the objects at low computational complexity. In that scene, the geometry of the objects such as round corners, sharp edges, or planar surfaces as well as the appearance of objects will be considered for 3D object ranking. On the other hand, we also investigate flaws of conventional vocabulary trees and improve the ranking results by introducing a credibility value to the TF-IDF scheme. By combining novel vocabulary trees and fast 3D geometric verification, we can improve the recall-datarate performance as well as the subjective ranking results for mobile 3D visual search.

Цель: разработка модели сегментации трёхмерных объектов на основе методологии машинного обучения. Объект: процессы сегментации трёхмерных объектов, представленных облаком точек. Методы: проведение исследование моделей сегментации трёхмерных объектов на основании датасета ScanNet с оценкой точности на основании метрики Average Intersection over Union (avgloU). Результаты: в ходе работы проведено сравнение и выявлены наиболее точные и производительные сочетания внутренней структуры обрабатываемых данных и архитектуры моделей, которые являются самыми перспективными для дальнейших исследований. наилучшие результаты показала библиотек машинного обучения Point Transformer со значением метрики avgIoU, равной 0,794. Полученные результаты будут использованы для дальнейшей работы над методами обработкой данных, поиском и настройкой моделей машинного обучения для задачи сегментации 3D-объектов для достижения лучшей точности и производительности. / Objective: development of a three-dimensional object segmentation model based on machine learning methodology. Object: segmentation processes of three-dimensional objects represented by a point cloud. Methods: conducting a study of three-dimensional object segmentation models based on the ScanNet dataset with accuracy assessment based on the Average Intersection over Union (avgloU) metric. Results: during the work, a comparison was made and the most accurate and productive combinations of the internal structure of the processed data and the architecture of the models were identified, which are the most promising for further research. The best results were shown by the Point Transformer machine learning library with an avgIoU metric value of 0.794. The obtained results will be used for further work on data processing methods, searching and tuning machine learning models for the task of segmenting 3D objects to achieve better accuracy and performance.

No description available.

The diploma thesis deals with the issue of distinguishing and denominating of geometric models by children aged five to six years. The aim of the work is to find out to what extent children are currently able to distinguish the world of the plane and the world of space and to what extent these two worlds merge for children. There are three hypotheses. H1: "Children are able to identify at least half of objects from a set of spatial bodies selected from a construction set.", hypothesis H2: "More than a half of children are confused naming planar and spatial objects." and hypothesis H3: "When identifying some of the objects familiar to children, they come up with their own names for them." To verify them, quantitative research was used, which also used audio recording to get data and to be able analyze the observed phenomena. The research confirmed all of those three these hypotheses. Two thousand two hundred data were analyzed. The work on a sample of hundred children shows what the level of object distinction and naming them is currently among preschool children in the Pilsen region. It opens up a number of other areas that need to be further addressed and lots of questions to be answered. These include mixing the world of space and the plane. KEYWORDS Space; plane; identification of 3D objects;...

The aim of the master thesis is to provide visitors of various types of objects - typically galleries, museums, etc. with additional information about exhibited objects with emphasis on visual display in augmented reality. It includes an analysis of already existing solutions, design and implementation of the entire system needed for deployment. From the point of view of the content creator (administrator), the system represents a complex solution enabling the creation of an interactive walk - scanning 3D objects, scanning 2D objects and adding content to the given objects. On the other hand, the tool for visitors (users) who are part of an interactive walk, provides additional information about objects primarily located near the user, which are automatically detected by the camera on the mobile phone. Solution includes augmented reality, which is implemented using ARKit technology, so the final application is built on the iOS platform. The work addresses the issue of detection of 3D objects and their subsequent recognition, along with the way to work with this information, how to store it and then use it for purposes of this thesis. In the final solution, emphasis is placed on the simplicity of the usage (guide marks, hints...) and overall user experience.

Pontificia Universidad Católica de Valparaíso

De nos jours, l'économie d'énergie est devenue un enjeu crucial. Les bâtiments des différents pays ont été identifiés comme étant une source importante de perte énergétique. De ce constat a émergé le projet RIDER (Reasearch for IT Driven EneRgy efficiency). Ce projet a pour objectif de développer un système d'information innovant permettant d'optimiser la consommation énergétique d'un bâtiment ou d'un groupe de bâtiments. Ce système est basé sur des composants logiciels, notamment une solution générale de modélisation du bâtiment, une solution de fouille de données, une solution de visualisation. Chacun de ces composants est destiné à améliorer le modèle de données de RIDER. Dans ce manuscrit, nous nous intéressons à la partie visualisation et nous proposons donc une solution d'amélioration du modèle par cette méthode. Dans ces travaux, nous allons présenter les solutions que nous avons mises en place pour modéliser le bâtiment ; pour ce faire nous avons utilisé une solution à base de particules dont la valeur est interpolée par rapport aux différents capteurs du bâtiment. Nous présentons également les différentes solutions mises en place pour visualiser les données et les méthodes d'interactions pour améliorer le modèle du bâtiment. Enfin, notre dernière partie présente les résultats de notre solution au travers de deux jeux de données. / Energy efficiencies are became a major issue. Building from any country have been identified as gap of energy, building are not enough insulated and energy loss by this struc- ture represent a major part of energy expenditure. RIDER has emerged from this viewpoint, RIDER for Research for IT Driven EneRgy efficiency. This project has goal to develop a new kind of IT system to optimize energy consumption of buildings. This system is based on a component paradigm, which is composed by a pivot model, a data warehouse with a data mining approach and a visualization tool. These two last components are developed to improve content of pivot model.In this manuscript, our focus was on the visualization part of the project. This manuscript is composed in two parts: state of the arts and contributions. Basic notions, a visualization chapter and a visual analytics chapter compose the state of the art. In the contribution part, we present data model used in this project, visualization proposed and we conclude with two experimentations on real data.