Summer Semester 2026, TU München

Organiser: Roman Pflugfelder

Modern computer vision is largely based on neural network and machine learning models originally developed within the machine learning community. In this seminar, students study state-of-the-art computer vision models in their historical and methodological context, focusing on selected visual tasks aligned with their academic interests. The seminar emphasizes critical engagement with current research literature, student-led presentations and discussion, and the collaborative preparation of a scientific seminar paper.

Students are expected to be enrolled in a Master’s programme in Informatics or a related field and to have a solid foundation in core computer science concepts. The seminar builds on content from advanced Informatics electives (e.g. Computer Vision II/III, Introduction to Deep Learning, Advanced Deep Learning for Computer Vision); prior attendance of one or more of these courses is strongly recommended. Familiarity with academic literature, research methods, and citation standards will be developed in the seminar.

1. After completing the seminar, students will be able to independently break down a given neural network for machine vision into its fundamental components and building blocks, analyse their functions and interdependencies, and evaluate their influence on the overall model.
2. After completing the seminar, students will be able to identify the general structural features of a given neural model for machine vision and characterize their significance in the context of its application.
3. After completing the seminar, students will be able to independently research, critically evaluate and correctly cite scientific literature, systematically analyze central content and structure it in a comprehensible scientific elaboration, and present the results orally in an appropriate form.
4. After completing the seminar, students will have the necessary methodological and interdisciplinary skills to independently familiarize themselves with a challanging scientific topic in the field of computer science, to analyze relevant questions and to design and write an independent scientific seminar paper according to current standards, as well as to present the results in a structured manner and to discuss them in professional discourse.

Assignment to the lab is done via the matching system.

More information about the course can be found in TUM Online.

• 10.02.2026: Pre-seminar meeting (4 p.m. - 5 p.m., Zoom)
• 14.04.2026: Kick-off meeting (3 p.m. - 6 p.m., room 5609.02.023)
• 12.05.2026: Seminar (3 p.m. - 6 p.m., room 00.08.055)
• 02.06.2026: Seminar (3 p.m. - 6 p.m., room 00.08.055)
• 23.06.2026 - 24.06.2026 : Presentations (9 a.m. - 13 a.m., room 01.09.014)
• 14.07.2026: Seminar (4 p.m. - 5 p.m., Zoom)

Will be presented in the kick-off meeting and is based on the following criteria:

• Participation in the collaborative literature studies
• Your engagement to the discussions during the semester
• Presentation: structure, slides, clarity, depth, own understanding
• Coding examples in Jupyter notebooks (bonus)
• Contribution to a joint seminar paper which we will publish on arXiv: structure, clarity, originality, understanding