Master's Thesis
Refine
Year of publication
Document Type
- Master's Thesis (46) (remove)
Keywords
- Augmented Reality (3)
- virtual reality (3)
- 360 Grad (1)
- 360 degree (1)
- 3D-Scan (1)
- 3D-Visualisierung (1)
- AR (1)
- Ausrichtungswerkzeug (1)
- BRDF (1)
- Breaking (1)
- Bruchsimulation (1)
- Computergrafik (1)
- Computergraphik (1)
- Computervisualistik (1)
- DTI (1)
- Datenschutz (1)
- Diffusionsbildgebung (1)
- Distanzfeld (1)
- Echtzeit (1)
- Eyetracking (1)
- Fiber Tracking (1)
- Fracturing (1)
- GPU (1)
- Gebrauchstauglichkeit (1)
- Gehirn (1)
- Gonioreflectometer (1)
- Herausforderung (1)
- IceCube (1)
- Implizite Oberfläche (1)
- Indirekte Beleuchtung (1)
- Informatik (1)
- Maschinelles Lernen (1)
- Material Point Method (1)
- Matting (1)
- MeVisLab (1)
- Mitral Valve (1)
- Mitralklappe (1)
- Mixed Reality (1)
- Motivation (1)
- Natural Image Matting (1)
- Neutino (1)
- Physiksimulation (1)
- Planspiel 2.0 (1)
- Point Rendering (1)
- Position Based Dynamics (1)
- Punktrendering (1)
- Raumplanung (1)
- Raytracing (1)
- Robust Principal Component Analysis (1)
- Sand (1)
- Schnee (1)
- Segmentation (1)
- Segmentierung (1)
- Shattering (1)
- Simulation (1)
- Spaß (1)
- Sprechtherapie (1)
- Statistical Shape Model (1)
- Tracking (1)
- Ultraschall (1)
- Ultrasound (1)
- Unterricht (1)
- Zerbrechen (1)
- computer games (1)
- flow (1)
- hybride App (1)
- image warping (1)
- interaktiv (1)
- leap motion (1)
- machine learning (1)
- performance optimization (1)
- ray tracing (1)
- rtx (1)
- stereoscopic rendering (1)
- tracking (1)
- volume rendering (1)
- vulkan (1)
- warp divergence (1)
Institute
- Institut für Computervisualistik (46) (remove)
Constituent parsing attempts to extract syntactic structure from a sentence. These parsing systems are helpful in many NLP applications such as grammar checking, question answering, and information extraction. This thesis work is about implementing a constituent parser for German language using neural networks. Over the past, recurrent neural networks have been used in building a parser and also many NLP applications. In this, self-attention neural network modules are used intensively to understand sentences effectively. With multilayered self-attention networks, constituent parsing achieves 93.68% F1 score. This is improved even further by using both character and word embeddings as a representation of the input. An F1 score of 94.10% was the best achieved by constituent parser using only the dataset provided. With the help of external datasets such as German Wikipedia, pre-trained ELMo models are used along with self-attention networks achieving 95.87% F1 score.
In recent years head mounted displays (HMD) and their abilities to create virtual realities comparable with the real world moved more into the focus of press coverage and consumers. The reason for this lies in constant improvements in available computing power, miniaturisation of components as well as the constantly shrinking power consumption. These trends originate in the general technical progress driven by advancements made in smartphone sector. This gives more people than ever access to the required components to create these virtual realities. However at the same time there is only limited research which uses the current generation of HMDs especially when comparing the virtual and real world against each other. The approach of this thesis is to look into the process of navigating both real and virtual spaces while using modern hardware and software. One of the key areas are the spatial and peripheral perception without which it would be difficult to navigate a given space. The influence of prior real and virtual experiences on these will be another key aspect. The final area of focus is the influence on the emotional state and how it compares to the real world. To research these influences a experiment using the Oculus Rift DK2 HMD will be held in which subjects will be guided through a real space as well as a virtual model of it. Data will be gather in a quantitative manner by using surveys. Finally, the findings will be discussed based on a statistical evaluation. During these tests the different perception of distances and room size will the compared and how they change based on the current reality. Furthermore, the influence of prior spatial activities both in the real and the virtual world will looked into. Lastly, it will be checked how real these virtual worlds are and if they are sufficiently sophisticated to trigger the same emotional responses as the real world.
Simulation of fractures
(2014)
Real-time computing often avoids the simulation of fractures due to its complexity. The field of engineering science provides methods to create these simulations to improve games and other applications. Steadily rising computer capacities allow suitable simulations on a real-time basis and make this aspect increasingly interesting. The topic and aim of this research is to simulate fractures of stiff bodies. The primary objective is the physical plausibility and performance of the application. This thesis analyses the potential of computer science to realize the simulation of fractures.
Three existing as well as one self-created were implemented and analysed. The works "Real time simulation of deformation and Fracture of stiff material" from Müller et al., "real time simulation of Brittle Fracture using Modal analysis" from Glondu et al. and "Fast and Controllable simulation of the Shattering of Brittle Objects" from Smith et al. form the basis of this thesis. The introduced methods use different computation of forces and fractures. The developed procedure uses the idea of generating secondary breaks. The approaches were implemented based on the Bullet physics-engine. The results of the work show that physically based breaks are realizable on a real-time basis.
The analysis of the physical methods demonstrates that their performance mainly depends on the constitution of the used objects. This thesis shows that the further investigation of this topic can discover new possibilities. The improvement of the realism in virtual worlds can be achieved by executing physically plausible methods.
Since the invention of U-net architecture in 2015, convolutional networks based on its encoder-decoder approach significantly improved results in image analysis challenges. It has been proven that such architectures can also be successfully applied in different domains by winning numerous championships in recent years. Also, the transfer learning technique created an opportunity to push state-of-the-art benchmarks to a higher level. Using this approach is beneficial for the medical domain, as collecting datasets is generally a difficult and expensive process.
In this thesis, we address the task of semantic segmentation with Deep Learning and make three main contributions and release experimental results that have practical value for medical imaging.
First, we evaluate the performance of four neural network architectures on the dataset of the cervical spine MRI scans. Second, we use transfer learning from models trained on the Imagenet dataset and compare it to randomly initialized networks. Third, we evaluate models trained on the bias field corrected and raw MRI data. All code to reproduce results is publicly available online.
With the emergence of current generation head-mounted displays (HMDs), virtual reality (VR) is regaining much interest in the field of medical imaging and diagnosis. Room-scale exploration of CT or MRI data in virtual reality feels like an intuitive application. However in VR retaining a high frame rate is more critical than for conventional user interaction seated in front of a screen. There is strong scientific evidence suggesting that low frame rates and high latency have a strong influence on the appearance of cybersickness. This thesis explores two practical approaches to overcome the high computational cost of volume rendering for virtual reality. One lies within the exploitation of coherency properties of the especially costly stereoscopic rendering setup. The main contribution is the development and evaluation of a novel acceleration technique for stereoscopic GPU ray casting. Additionally, an asynchronous rendering approach is pursued to minimize the amount of latency in the system. A selection of image warping techniques has been implemented and evaluated methodically, assessing the applicability for VR volume rendering.
Today, augmented reality is becoming more and more important in several areas like industrial sectors, medicine, or tourism. This gain of importance can easily be explained by its powerful extension of real world content. Therefore, augmented realty became a way to explain and enhance the real world information. Yet, to create a system which can enhance a scene with additional information, the relation between the system and the real world must be known. In order to establish this relationship a commonly used method is optical tracking. The system calculates its relation to the real world from camera images. To do so, a reference which is known is needed in the scene to serve as an orientation. Today, this is mostly a 2D-marker or a 2D-texture. These are placed in the real world scenery to serve as a reference. But, this is an intrusion in the scene. That is why it is desirable that the system works without such an additional aid. An strategy without manipulating the scene is object-tracking. In this approach, any object from the scene can be used as a reference for the system. As an object is far more complex than a marker, it is harder for the system to establish its relationship with the real world. That is why most methods for 3D-object-tracking reduce the object by not using the whole object as reference. The focus of this thesis is to research how a whole object can be used as a reference in a way that either the system or the camera can be moved in any 360 degree angle around the object without loosing the relation to the real world. As a basis the augmented reality framework, the so called VisionLib, is used. Extensions to this system for 360 degree tracking are implemented in different ways and analyzed in the scope of this work. Also, the different extensions are compared. The best results were achieved by improving the reinitialization process. With this extension, current camera images of the scene are given to the system. With the hek of these images, the system can calculate the relation to the real world faster in case the relation went missing.