Portfolio Chronicle

Volume MMXXI No. 2 1.30 €

After two days: Four witches and wizards members of inner circle

Two magicians in a labyrinth. One is casting a protective spell while being attacked by a red monster.
The magicians have to cooperate to solve challenging puzzles. Photo: S. Höffner

At the Global Game Jam in January 2021, Dmi­try Ale­xan­drovsky and Se­bas­ti­an Höf­fner submitted the magic game “Initiation”.

In the high-paced multiplayer action game, up to four magicians have to cooperate to solve all tasks required to become a member of the inner circle. After each solved task, the magicians each lose one ability. Thus, the magicians have to cooperate more closely and use their powers together as the game progresses. Possible tasks include puzzle solving and epic fights. The game is developed in Unity and the multiplayer functionality is built using Mirror.

Design choices for questionnaires in VR: World-anchoring and laser pointers take the lead

A virtual reality user shooting with a bow at some ballons.
Sebastian Höffner built the VR archery game. Photo: S. Putze

At the CHI'20, Ale­xan­drovsky, Putze et al. published their evaluation on design choices for questionnaires for VR user studies. Their main result: placing a questionnaire at a fixed location inside the virtual world and using laser pointers to interact with it works best.

Sebastian Höffner programmed the user study and took part in the data analysis and paper writing process.

Ale­xan­drovsky, Putze et al. (2020): Examining Design Choices of Questionnaires in VR User Studies. CHI'20. DOI: 10.1145/3313831.3376260

Introductory lecture to Python programming was a great success

A still of a video of Sebastian Höffner delivering a lecture on regular expressions.
Sebastian Höffner delivering a lecture on regex. Photo: UOS

The course Basic Programming in Python during the summer term 2017 at the Osnabrück University was taught by Sebastian Höffner and Aline Vilks. The course was designed for master students from disciplines who had no previous introduction to programming but required it for their studies in Osnabrück.

It started with the classic “Hello world!”, variables and functions, and finished with practical recommendations for packages and how to learn more. All-time favorites of the students were the walk-in practice sessions, where Sebastian Höffner and Aline Vilks offered an open practice session to discuss homework exercises and team up to solve them together. The format was well received, and eventually students did not only discuss the Python homework but also how to apply their new skills to other subjects, such as Machine Learning.

For the final submission, the students formed groupes and solved individual problems. These ranged from flash-card vocabulary training tools for Arabian languages over computer games to data analysis tools.

New Machine Learning exercises

A colored world map.
A map colored with a self-organized colormap. Photo: S. Höffner

For the summer term 2016, the Machine Learning tutors created exercise sheets. The exercise sheets accompanied Prof. G. Heidemann's lecture.

The topics ranged from concept learning over clustering algorithms towards neural network architectures. For each week, the tutors designed a set of exercises which allowed to discuss the new topics, learn new tricks, and transfer the new algorithms and concepts to further applications.

The tutors Sebastian Höffner, Alexander Höreth, and Andrea Suckro developed most content and ensured a high standard for the Jupyter notebooks. All exercises, including extensive solutions, are hosted on GitHub.

One of the student's favorite exercises was implementing a self-organizing map. When successful, it could be used organize countries by some features such as the GDP or the adult literacy rate. By organizing a colormap the same way, a map could be colored such that similar countries were similarly colored. Other highlights were the implementation of a neural network from scratch, and an introduction to reinforcement learning with Q-learning.

Gaze tracking using webcams

Gaze tracking is typically performed with expensive eye-tracking hardware in controlled laboratory conditions. There are notably exceptions such as the Pupil Labs glasses for mobile eye tracking or the tobii devices for consumer products.

In his master thesis, Sebastian Höffner worked on a model-based gaze tracking solution which works with common webcams. This could make gaze tracking more affordable and readily available for many devices right away. The results still need work: pupil tracking works great using an algorithm developed by Tristan Hume, but gaze tracking is still not perfect.

Some boost could be achieved using modern infrared cameras on mobile phones, or using a more accurate head model.

Best cross media school magazine

In 2011, the BDZV honored the school magazine INSIDE with the special award “Crossmedial”. Chief editor Tobias Rischen was delighted to receive the award: “This is totally unexpected, but it's extremely motivating to see this kind of appreciation for our hard work!” In 2009, he founded the INSIDE together with Sebastian Höffner, who was responsible for the layout and website administration.

The duo and their team of about ten up-and-coming journalists increased the initial circulation of their school magazine from 200 to 500 copies within a year. Shortly after, they won their first award in 2010: the second place at the school magazine contest WoRtWiTz of the MHMK. The high quality layout and articles convinced the jury.

Robots! Lasers!

No, they are not attacking. In his bachelor thesis in 2014, Sebastian Höffner implemented a particle filter for probabilistic robot localization in 3D. To localize itself in a known map, the robot uses a laser scanner and compares sampled poses in the map with its sensor data. The algorithm, implemented in C++ for ROS, was very slow and did not converge well.

Years later, Sebastian Höffner learned that one solution to localization in 3D is to use more sensors to resolve ambiguities. In an interview, he stated that “it is great that the topic is still relevant.”

Free and open source
at github.com/shoeffner

Looking for talent?
Connect now!

24/7 IT support

Call 0000-0002-1609-6049
for science!