Dynamic Object Interaction with a Boston Dynamics Spot Robot
Prof. Dr. Sören Pirk - Christian-Albrechts-University of Kiel
The goal of the project is to build a dynamic mobile manipulation platform by integrating a robotic arm with the Boston Dynamics Spot robot. We aim to enable the robot to autonomously navigate and interact with its environment. The system is supposed to perform tasks like pointing at landmarks or picking objects. The project may also incorporate language-based interaction, allowing users to control or query the robot through natural language commands.
Semiformal AI Prompting: Eliminating AI Guesswork
Daniel Dietrich - Typefox
Based on an example web application generator, we’ll explore how structured syntax produces more precise results compared to ambiguous natural language prompts. Use a playground to create full-stack web applications of any kind while experimenting with the DSL approach. There’s room for thinking about advanced applications that go beyond web app generation...
Look'n'Learn AI: Interactive Visual Inspection Assistant
Johannes Köpnick -
We're building "Look'n'Learn AI," an intuitive visual inspection tool that learns from user feedback (OK/NOT OK). It uses local, open-source AI to analyze images/video, allowing users to teach the system their specific quality criteria through simple interactions and chat. Our goal is to create a self-improving visual assistant that adapts to various inspection tasks without needing pre-labeled datasets.
Coin Raiders: Looting Lara’s Purse for Science
Christopher Hansen - Christian-Albrechts-University of Kiel
Ancient coins are tiny time machines — packed with rich history, powerful imagery, and hidden stories. But here’s the kicker: despite decades of research, it's still an open challenge to automatically describe and locate what's on these coins — be it emperors, animals, gods, weapons, or cryptic inscriptions. Build a system that not only tells the story of each coin but also shows where each element is — from portraits to legends. You can go wild with segmentation models, vision-language models, or multi-modal transformers. We’re open to fresh, unconventional ideas!
Fine-Tuning Whisper for Sign Language Subtitling
Steffen Brandt - opencampus.sh
This project fine-tunes a video-adapted Whisper model on sign language data using Google Cloud Vertex AI to generate accurate subtitles and enhance accessibility.
More Exciting Projects Coming Soon!
Your Name Here - Your Institution
Project descriptions will be added as participants submit their proposals. Stay tuned for innovative AI and machine learning projects!
