From eye-tracking calculators to AI action video games: Master AI Showcase Evening 2023

At the end of the 3rd semester students of the master program AI Engineering showed the final presentations of their projects. Afterwards, one winning group each from the category “Artificial Intelligence” and “Game” was selected.

In the winter semester 2021/2022, the former master program Game Engineering and Simulation was further developed to AI Engineering. Those students who had chosen the specialization “Game Engineering” in their studies developed a computer game with an AI part in their project, while all others looked for topics around “AI Technologies” as well as, for example, digital image processing or natural language processing.

The spectrum of this year’s projects at the “Best of Class” award ranged from a calculator that works exclusively with the help of eye control, to a 3rd person action computer game with an AI-generated environment, to an automatic assembly machine for computer chips.
After the successful playtests, all projects could be extensively tested and discussed on site at the Showcase Evening. At the end of the evening, the jury, Bernhard Knapp (Master AI program director), Alexander Nimmervoll (Master AI game deepening manager), Dietmar Hauser (Senior Software Engineer at Epic Games), Martin Filipp, (Managing Director Mi’pu’mi Games) as well as a large number of other editors and external company cooperation partners, were finally ready to announce the two winning projects.

Winning projects: Youtube title generator and horror shooter

In the category “AI” the team with Arapsih Güngör, Philipp Jonas, Marvin Kosmider, and Daniel Rajs won with their project “Content CreAItors”. The function of generating a matching Youtube title from subtitles of a video also inspired the project’s supervisor Mina Schütz. “Humans Never Change” is the name of the winning project in the “Game” category with Philipp Andert, Philip Ewert and Stephan Steidl and supervisor Jürgen Konrad. In this horror shooter, players wake up on an abandoned ship and are pursued by AI-controlled alien life forms that exhibit complex behavior.
“Once again, the projects were of very high quality. It was therefore extremely difficult to select winners. The decision was very close. Congratulations to all the teams on their great work,” said program director Bernhard Knapp.

For students of the master’s program AI Engineering, far-reaching and promising fields of application are in sight. Artificial intelligence is considered one of the key concepts of digitalization, which will affect many areas of our daily lives. People are already being addressed by intelligent bots on the World Wide Web, the first cars are driving themselves, and intelligent household appliances are finding their way into our homes. Gamers play against smart computer opponents in automatically generated artificial worlds, doctors are supported by software in image analysis, texts (e.g. chatGPT) and images (e.g. Stable Diffusion) are generated by AIs.

Team: Florian Federhofer, Jonas Wendel, Sünje Winteler
Supervisor: Julian Breddy
Category: Game

Description: DeadSapling is a 3rd person action tower defense video game with intelligent AI controlled enemies, developed by IceCreamConnection. It belongs to the genre of tower defense games, but breaks away from the traditional top-down view and takes the player into a dynamic, action-packed 3D world.

Team: Christian Clemenz, Maximilian Deutsch, Alexander Graf, Florentin Rieger, Peter Schlagnitweit
Supervisor: Christoph Quas
Category: Game

Description: šanû is a roguelike game developed in Unreal Engine 5 in which the enemy AI adapts to the player’s behavior through machine learning. Due to the adaptive nature of the AI, the player is forced to change weapons and combat strategies on a regular basis. The story is set in an ancient Mesopotamian temple from which the player must escape in the guise of the goddess Tiamat. During the escape, the goddess must fight enemies that become more adaptable with each encounter.

Team: Philipp Andert, Philip Ewert, Stephan Steidl
Supervisor: Jürgen Konrad
Category: Game

Description: In the horror shooter “Humans Never Change”, players wake up on an abandoned ship, the “Amber Hawk”, and fight for a way to regain their freedom. Pursued by alien intelligent life forms lurking in the shadows – through mazes of corridors and locked doors – they witness what dooms the Amber Hawks to its eerie existence.

Team: Arapsih Güngör, Philipp Jonas, Marvin Kosmider, Daniel Rajs
Supervisor: Mina Schütz
Category: AI

Description: The goal of this project is to train a Natural Language Model to generate an appropriate YouTube title from a summary of a text. The focus was limited to the topic of edutainment. The summary is generated from the subtitles of the video, which allows the whole process to be automated.

Team: Thomas Gruber, Christina Obereigner, Daniel Pötscher, Alexander Zulechner
Supervisor: Bernhard Knapp
Category: AI

Description: The project dealt with the development of a 2D platform game in which 4 artificial intelligences compete against each other. For this purpose, a dynamic level was created, which is built from different hand-crafted level segments. Parameters were added for training the AIs to randomly create levels for training. Challenges during the semester were used to determine the best AI. Human players can now compete against the AI in predefined levels.

Team: Daniel Bachus, Eric Eckstein, Michael Hermann-Hubler, Marie-Lena Müller, Richard Schultheis
Supervisor: Patrick Link
Category: AI

Description: For many people with disabilities, things we take for granted are often an impossibility, such as controlling a computer with a mouse. With the help of software, a glance is enough to make the use of PCs accessible. Here, the mouse does not have to be controlled by hand. A calculator app serves as an example, with which one can add, subtract, multiply and divide – all with the help of the gaze.

Team: Matthias Chory, Alexander Dickbauer, Antonia Langer
Supervisor: Sharwin Rezagholi
Category: AI

Description: The project “Language Sloth” is a native application that deals with speech synthesis and translation of texts and speech recordings. Speech synthesis is the artificial generation of the human speaking voice. The application contains two functions. The first function is Speech-to-Speech, where the user can record speech using the application and select which other language the recorded text should be translated into. The result can then be played back directly in the application. The second function is a text-to-text translator. As the name suggests, the user can enter a text into the application and it will be translated into the desired language.

Team: Marius Hochwald, Michael Kranl, Daniel Krottendorfer
Supervisor: Bernhard Knapp
Category: AI

Description: Current pick and place machines do their job very well, but the big drawback is their price. Instead of buying a machine for thousands of dollars, this project focuses on training an AI model that predicts the position and rotation of electrical components, such as computer chips on an image. The model is trained using only synthetic data. The AI detects in real time which components are currently being captured by the camera. The accuracy of the component recognition is displayed on the screen as a percentage for each component.