Zum Inhalt springen
Technische Universität München

Deep Representation Learning for Computer Vision and Beyond (DRL4CVB)

Overview

This course deals with the fundamentals of deep learning, where we investigate how to learn meaningful representations from not only RGB Images or Video but also Medical Imaging Data (Volumetric Images), Graphs, Language or even 3D Data. All these modalities can be considered under similar learning paradigms - Gradient Descent and Neural Networks - but require different strategies for model architectures and objective functions depending on the problem setting and data. Students are expected to work in teams on a project related to current state of the art research in these fields, answering such questions as:

  • “What is a good representation for this task or modality?”
  • “How can we measure the ‘goodness’ of a representation?”
  • “What objective function do we need to achieve this representation?”
  • “Self-Supervised vs. Supervised Learning, which one can we use here?” 

Pre-Course Meeting

The preliminary meeting will be held on 06.02.2026 14:00 - 14:30 on zoom:

tum-conf.zoom-x.de/j/64356101532

 

Registration

Register in the TUM matching system until 23.02.2026.

Please send a very short motivation letter, CV (optional) & transcript (mandatory) to drl4cvb(at)camp.cit.tum.de until 23.02.2026.

  • Why do you want to participate in this course in particular?
  • What background related to computer vision & machine learning do you have?

Prerequisites: 

  • Introduction to Deep Learning or Machine Learning
  • Practical experience with Deep Learning projects (bonus, not required)
  • Practical experience with one or more of the modalities (bonus, not required)

Recommended Reading

Practical Course: Deep Representation Learning for Computer Vision and Beyond (IN2106)

Lecturer (assistant)
Nassir Navab [L]
Feng Li
Felix Tristram
Magdalena Wysocki
Number 0000001754
Type practical training
Duration 6 SWS
Term Sommersemester 2026
Language of instruction English
Position within curricula See TUMonline
Dates See TUMonline

Dates

Admission information

See TUMonline
Note: ote: Please register in the matching system for the course registration (http://docmatching.in.tum.de). Keep in mind that your chances to be assigned to the course increase if you give it a higher rank in your choices. For further details about how the matching system works and its schedule please check the link provided.

Objectives

After successful completion of the Praktikum Students are able to:
- Work in a Team on a Deep Learning Project
- Understand what Representation Learning means, what a good representation is and how this relates to current research areas
- Present the outcome of their projects
- Write a short scientific report about the outcome of their projects

Description

This course deals with the fundamentals of deep learning, where we investigate how to learn meaningful representations from not only RGB Images or Video but also Medical Imaging Data (Volumetric Images), Graphs, Language or even 3D Data. All these modalities can be considered under similar learning paradigms - Gradient Descent and Neural Networks - but require different strategies for model architectures and objective functions depending on the problem setting and data. Students are expected to work in teams on a project related to current state of the art research in these fields, answering such questions as:
“What is a good representation for this task or modality?”
“How can we measure the ‘goodness’ of a representation?”
“What objective function do we need to achieve this representation?”
“Self-Supervised vs. Supervised Learning, which one can we use here?”

Prerequisites

Introduction to Deep Learning or Machine Learning
Practical experience with Deep Learning projects (bonus, not required)
Practical experience with one or more of the modalities (bonus, not required)

Teaching and learning methods

10 ECTS * 30h = 300h

Examination

Final Presentation, Project Report and Project Work (Meetings with Course Tutors, Code deliverables)

Recommended literature

Torralba, A., Isola, P., & Freeman, W. T. (2024). Foundations of computer vision. MIT Press. - Chapter 30: https://visionbook.mit.edu/representation_learning.html
Buchanan, S., Pai, D., Wang, P., & Ma, Y. (2025). Learning Deep Representations of Data Distributions. https://ma-lab-berkeley.github.io/deep-representation-learning-book/index.html

Links

To top