Probabilistische Graphische Modelle in der Computer Vision

What it is about

You learn how probabilistic graphical models (directed and undirected) are used to model and solve typical computer vision problems. At the end you can understand MRF/CRF models, select appropriate inference and learning methods, and apply them to tasks such as segmentation, pose estimation, stereo or object recognition.

What you will be able to do

• Understanding of directed and undirected graphical models (Bayesian networks, MRF, CRF)
• Mastery of parameter learning methods for MRF/CRF (including gradient-based methods and SGD)
• Knowledge of exact MAP inference methods (Belief Propagation on trees, sum-product, binary graph-cuts)
• Knowledge of approximate MAP inference methods (loopy BP, mean field, alpha-expansion/-beta-swap, LP relaxations)
• Ability to practically apply and implement inference and learning methods

What the module consists of

• VorlesungVermittlung des theoretischen Hintergrunds
• ÜbungenPraktische Aufgaben zur Vertiefung des Vorlesungsstoffs und zur Implementierung

Teaching method

• VorlesungErklärung des theoretischen Hintergrunds
• ÜbungsaufgabenAnwenden und Vertiefen der Vorlesungsinhalte durch praktische Probleme