The Multimodal Deep Learning Project Seminar focuses on the hands-on development and evaluation of deep learning methods for multimodal data.
Students will work in teams of approximately three on realistic machine-learning projects involving video, audio, biosignals, or related multimodal applications.
Projects will consider practical challenges such as limited data, restricted computational resources, model generalization, and robust evaluation.
The seminar starts with a kick-off session on 13 October 2026, where the project organization, requirements, example projects, and timeline will be introduced.
The project phase includes literature review, implementation, experimentation, evaluation, and regular status updates.
Final deliverables include the project code on GitHub, documentation, a 6-page research paper, and a final presentation.
Final presentations consist of 20 minutes of presentation followed by 5 minutes of Q&A.

Kurs im HIS-LSF