Computer Vision II

SoSe 2022

This course provides a more in-depth introduction into computer vision – the problem of making computers understand the visual world we live in. We will focus on probabilistic/learning approaches to computer vision from a variety of areas including 3D scene structure estimation from stereo images, image motion estimation (optical & scene flow), image restoration, semantic image segmentation, tracking of objects in video sequences, and more.


Course materials (Moodle) Link
TUCaN Link
Module no. 20-00-0401-iv
Event type integrated course (IV4, 6CP, 4SWS), elective
Being taught regularly, usually summer semesters
Time Lecture: Online recordings, Exercise: Wed. 13:30-15:10
First class See Moodle
Location Lecture: Online, Exercise: S202/C120
Lecturer Prof. Stefan Roth, Ph.D.
Assistants Samin Hamidi, Robin Hesse
Exam TBA


  • Computer vision as (probabilistic) inference
  • Robust estimation and modeling
  • Foundations of Bayesian networks and Markov random fields
  • Basic inference and learning methods in computer vision
  • Image restoration
  • Stereo
  • Optical flow
  • Bayesian tracking of (articulated) objects
  • Semantic segmentation
  • Current research topics

After successfully attending the course, students have developed a more in-depth understanding of computer vision. They formulate image and video analysis tasks as inference problems, taking challenges of real applications into account, e.g. regarding robustness. They solve the inference problem using discrete or continuous inference algorithms, and apply these to realistic imagery. They quantitatively evaluate the application specific results.

Can be taken for credit toward BSc / MSc Informatik, MSc Visual Computing, MSc Autonome Systeme, BSc / MSc Computational Engineering and others. Students from other departments, e.g. Mathematik, Elektrotechnik, IST, or Physik are welcome, though academic credit may need to be arranged.

It is recommended to have taken Visual Computing (formerly Introduction to Human Computer Systems) and Computer Vision I. Working knowledge in mathematics, in particular basics of probability and statistics is required.

Note: The course primarily aims at students who have already taken Computer Vision I. Motivated students who have not taken Computer Vision I can also take this class, provided they are willing to catch up on some basics.