Wir suchen ständig interessierte Studenten für Bachelor- und Masterarbeiten, als Hiwis oder für andere Forschungsarbeiten in unseren Forschungsgebieten. Bitte schreiben Sie eine Email an Prof. Deussen und machen sie ein Treffen aus. 

Exzellenzcluster Collective Behaviour

Unsere Gruppe ist Teil des neuen Exzellenzclusters der Universität Konstanz "Centre for the Advanced Study of Collective Behaviour" (https://www.exc.uni-konstanz.de/collective-behaviour/), einer interdisziplinären Initiative zur Erforschung des kollektiven Verhaltens zwischen der Universität Konstanz und dem Max-Planck-Institut für Ornithologie.

Hier bieten wir Bachelor- und Masterarbeiten (sowie HiWi-Jobs) in den folgenden Bereichen an:
- 3D-Rekonstruktion und Objektklassifizierung mit Fokus auf Unterwasser-Ökosysteme
- Vereinfachung von Trajektorien aus Bewegungsdaten
- Einsatz von maschinellem Lernen zur Verfolgung von Identitäten in Bewegungsdaten

Dies ist keine vollständige Liste, sondern soll Ihnen einen Überblick über die Themen innerhalb dieses Clusters geben. Wenn Sie daran interessiert sind, in einem dieser Themenbereiche in einer interdisziplinären Gruppe zu arbeiten, schreiben Sie bitte eine E-Mail an oliver.deussen@uni-konstanz.de

Generalizing Alpha Shapes

Master Project / Thesis

Alpha shapes can be used to describe the shape of a point set and to find the corresponding contour of this shape. But what happens when the underlying set is not made from points, but consists of circles with varying radius. How does this influence the resulting shape and how can such a shape be computed efficiently?

For this project you will:

  • Extend the alpha shape algorithm to work on implicit objects
  • Research the implications on computational complexity
  • Implement a prototype of your algorithm

 Contact: Jochen Görtler

Dynamic Stippling

Bachelor / Master

We developed a novel algorithm for Stippling that works by splitting and merging Voronoi cells. It is simple, fast, and produces high quality point sets. You can find more details about it on our project page here. However, there is still room for improvements, both with regard to speed and quality.

For this project you will:

  • Get familiar with our algorithmic idea
  • Identify bottlenecks with regard to convergence
  • Build upon our algorithm (or create a novel) to improve performance and/or output quality

 Contact: Marc Spicker

Abstract representation of procedural content

Bachelor / Master

There are many different methods for the procedural generation of content, which can be created (semi-)automatically by generative techniques. For future projects, state-of-the-art generation methods for terrain, urban structures and plants are needed, and some kind of query language to retrieve information from the algorithm without generating a surface description.

For this project you will:

- Implement a generation method
- Identify the limitation regarding to level-of-detail etc.
- Enhance the algorithm to provide a query language

Contact: Jens Metzner

e-David Painting Robot

Various Projects available

The e-David painting robot is under ongoing development and both its works and the machine are frequently exhibited in prominent locations. In order to further the development of the machine several topics can be addressed in a Bachelor's or Master's thesis:

  • Implementation of new painting processes, based on our feedback mechanism.
  • Robotic brush and paint handling for optimizing tool use during the painting process.
  • Hardware and driver development for new robots, to allow their use as e-David machines.
  • Exploring the applicability of machine learning techniques to generate painting commands for the robot.

Some of these projects can involve working with hardware and electronics. Others are mostly about graphics and image processing. The focus of your thesis can be set based upon your interests in these topics.

If you are interested in art, robotics and the real world implementation of computer graphics techniques, please contact Marvin Gülzow.

Planned Projects:

  1. Painting Simulation and Visualization: You will implement a C++ module for the e-David control software, which allows system users to preview robot movements for different use cases. A technical view should display movements in 3D, so newly developed stroke techniques can be analyzed for issues like collisions. An artistic view should show a simulated painting, based on a set of brush commands.
  2. High Speed Portraiture: You will implement a tool which uses face-detection algorithms to produce a sketch of a human face. The main goal is to create a recognizable sketch using a minimal amount of strokes, so the robot is able to produce a recognizable portrait in less than 5 minutes. Ideally, strokes will be ordered such that the most important features are painted first.
  3. Scene Decomposition for Robotic Painting: Using existing image abstraction techniques you will decompose arbitray images into shapes and features. The resulting system should produce a hierarchy of shapes and surfaces, which can then be transformed into strokes for the robot.

Linguistic Insight Generation using Text Visualization

Bachelor / Master

In order to get deeper insights into how we use language to ask questions and summarize information, we need techniques to analyze different linguistic features (e.g. POS tags or knowledge representations), and show them in interactive visualizations. Using the visualization allows users to get insights about the text they are analyzing and find new relationships between features. As these features are usually high-dimensional, dimensional reduction techniques such as PCA or t-SNE are needed to create suitable visualizations.

For this project you will:

  • Implement a text visualization using linguistic features
  • Compare different level-of-detail and projection techniques
  • Incorporate side-by-side view for multiple languages

If you are interested in text visualization and linguistic theory,
please contact: Rebecca Kehlbeck