Basics of Modern Image Analysis

Lecturer: Fred Hamprecht. Teaching Assistant: Carsten Haubold


In the last four years, the performance of computer vision systems has increased by leaps and bounds thanks to "deep" neural networks. In limited application domains, such systems now surpass the accuracy and speed of the average human.

This course will teach the foundation of such systems: more than half of the course will introduce in much detail the Fourier representation, characterization of linear filters, separability and steerability of filters, wavelets, etc.

In the remainder, we will study multi-layer perceptrons and how to train them, convolutional neural networks, the recently discovered tricks that allow the training of deep networks, and advanced architectures.

This course will be accompanied by extensive python exercises that let you apply and study these concepts yourselves. Former python experience is a plus, but not strictly required.

The grand goal of the course is to endow you with a reasonably detailed understanding of the foundations of a limited but important class of modern vision systems.

Venue and Time

The course is on Tuesdays from (note: new time) 14:15 through 16:00 in Mathematikon Bauteil B, Berliner Strasse 43, SR B128. The time for the exercises will be determined jointly. We start on Tuesday, April 19th, 2016.

The course is eligible as part of the MSc-Vertiefung Computational Physics. It is worth 4CP upon successful completion of the exercises. The course is taught in English.