Winter Term 2025/26

This page contains automatically translated content.

All required information and all links to platforms for our courses are collected on this website. Please, do not write individual emails to the teachers, but use the online courses and platforms instead.

 

Bachelor:

Start of the course: For further information, see the Moodle course.

Content:

This course is offered in the form of a conference seminar. Similar to a scientific conference, participants submit their own conference papers, participate in the review of other papers, and meet at the end of the semester for a joint workshop where the results are presented and discussed. Thematically, this conference falls within the field of machine learning. The specific topics of the seminar papers will be announced by research assistants in the department and presented at the introductory event. This will take place at the beginning of the lecture period.

The introductory event is expected to take place in person. Current information on the course schedule can be found within the Moodle course.

If you are interested, please feel free to attend the introductory event (see Moodle).

Links:

Contact Person:

Start of the lecture:

20.10.2025 - 16:15

Start of exercise:

Group 1: 27.10.2025 08:00 AM
Group 2: 27.10.2025 09:00 AM
Group 3: 28.10.2025 08:00 AM
Group 4: 28.10.2025 09:00 AM
Group 5: 29.10.2025 08:00 AM
Group 6: 29.10.2025 09:00 AM 

 

The material of the lecture will be taught according to the teaching concept "Flipped Classroom" in the form of videos. Accompanying the videos, there will be weekly live sessions in which Prof. Sick will discuss further questions with you to deepen your understanding. In addition, there is a short summary of the material from the previous week and a preview of the material for the coming week.

The exercise sheets are provided every Monday. The solutions are uploaded one week later after the last exercise. Only presence exercises are offered in which the exercises can be discussed.

All further information and regular announcements regarding the lecture and the test for exam admittance can be found in the corresponding Moodle course starting from September 22nd.

 

Meeting Information

 

Contact information:

Name: Minh Tuan Pham

E-Mail: stochastik[at]uni-kassel[dot]de

 

Master:

Start of the course: For further information, see the Moodle course.

Content:

This course is offered in the form of a conference seminar. Similar to a scientific conference, participants submit their own conference papers, participate in the review of other papers, and meet at the end of the semester for a joint workshop where the results are presented and discussed. Thematically, this conference falls within the field of machine learning. The specific topics of the seminar papers will be announced by research assistants in the department and presented at the introductory event. This will take place at the beginning of the lecture period.

The introductory event is expected to take place in person. Current information on the course schedule can be found within the Moodle course.

If you are interested, please feel free to attend the introductory event (see Moodle).

Links:

Contact Person:

Deep Learning Lab: Computer vision for an autonomous ship cleaning robot

In the Deep Learning Lab (DLL), we deal with a real, highly topical research question: the development of a computer vision unit for an autonomous ship cleaning robot. The robot moves over the hull while the ship is moving ("in-transit"), classifies the fouling into different degrees of soiling (0-4) and creates a soiling map that stores the proportion of soiled areas.

Background: Soiled ship hulls are responsible for around 1% of global CO₂ emissions. Continuous cleaning could save 20-30% of fuel and greenhouse gas emissions, prevent the spread of invasive species and reduce the use of toxic antifouling paints, which contribute significantly to microplastic pollution. Such a cleaning robot is being developed at the University of Kassel - we are working on the computer vision part of this system in the Deep Learning Lab.

 

Contents and procedure

  • Kick-off meeting
    Introduction to the course, motivation and discussion with one of the developers of the cleaning robot.
  • Annotation work on the data set (individual work, exam admission)
    Creation and preparation of noisy labels for soiling classification.
  • Project work in groups
    Application of modern deep learning techniques in the field of computer vision on the previously annotated data sets. The subject area can be chosen according to interest, e.g:
    • Convolutional Neural Networks (CNNs)
    • Vision Transformer (ViT)
    • Transfer Learning and Fine-Tuning
    • Few-shot learning
    • Data augmentation and active learning
    • Explainable AI (XAI) for model interpretation
    • Noisy label learning
  • Documentation and final presentation
    Written elaboration of the results and presentation of the project work in plenary (in project groups).

Examination and grading

Scope: 6 CP. Grading is based on the following:

  • Written documentation 50 %
  • Submission of the developed project 30 %
  • Final presentation 20 %

Learning objectives

  • Understanding the challenges of computer vision in a real robotics application
  • Practical experience in building, training and evaluating deep learning models
  • Dealing with data set annotation, model validation and result interpretation
  • Teamwork, scientific documentation and presentation

Prerequisites

Students from different disciplines are welcome.
Helpful:

  • Basic knowledge of Python and machine learning
  • First experience with PyTorch or similar frameworks
  • Basic knowledge of linear algebra and statistics
  • Willingness to work in a team and independently familiarize yourself with new tools

Further information can be found here.

Information about the meeting

Contact persons:

Jens Decke

jens.decke[at]uni-kassel[dot]de>

 

Start of the lecture:

20.10.2025 - 14:15 - 15:45

Start of exercise:

05.11.2025 - 10:00 - 11:30

 

The lecture covers the foundations of pattern recognition from a probabilistic point of view. The following topics are discussed:

  • Basics (e.g., stochastics, model selection, curse of dimensionality, decision and information theory)
  • Distributions (e.g., multinomial, Dirichlet, Gaussian and Student distributions, nonparametric estimation)
  • Linear models for regression
  • Linear models for classification
  • Neural networks
  • Kernel methods

 

Meeting Information

  • room 0303c (IES Lab room)

 

Contact: Huseljic, Denis

E-Mail: dhuseljic[at]uni-kassel[dot]de