Lecture: Markov Random Fields for Vision and Graphics


  • Lecturer(s):
  • Start: 11.04.2018
  • Dates: Wed. 12:30 (s.t.) - 14:00, SR U.027, Institut für Informatik
  • Course number: MA-INF 2117
  • Curriculum: Master
  • Effort: 6 CP
  • Exams: Aug. 8, 2018



This course addresses advanced topics for Markov Random Fields and their use in applications for vision and graphics.  We will cover advanced topics in inference and learning such as loopy belief propagation, MCMC sampling, graph cuts and move-making algorithms, dual decomposition and structured learning.  Applications discussed will include low and mid-level vision and graphics concepts such as optical flow and stereo depth, super-resolution, superpixels, texture synthesis, segmentation as well as higher-level concepts such as semantic segmentation and object detection.

It is recommended but not required to have taken Probabilistic Graphical Models (MA-INF 4315) as a pre-requisite for this course.  Those who have not taken Probabilistic Graphical Models should be comfortable with concepts in probability theory and optimization.


Please note that there is no exercise class on May 31. You can send your solutions for Ex. 2: Min-Cuts (Theoretical Exercises) via email. You will give your paper presentation on 07.06.2018 during the exercise class.

Please note that there is no lecture on 18.07.2018.


Additional Documents

Assignment Sheets

Exercise 1: BeliefProp
Assignment sheet  (PDF document, 155 KB)
Exercise 2: GraphCuts
Assignment sheet  (PDF document, 202 KB)
Exercise 3: SemanticSegmentation
Assignment sheet  (PDF document, 173 KB)
Exercise 4: PoseEstimation
Assignment sheet  (PDF document, 266 KB)