Jump label

Service navigation

Main navigation

You are here:

Main content

Lecture Computer Vision, SS 2020

Lecture Format

Due to the challenging situation caused by the Covid-19 outbreak, we decided to hold the lecture online via video conferencing. We aim at keeping a traditional lecture format, allowing for the most interaction possible. Therefore, we will technically rely on Zoom. In order to participate, we recommend the installation of the Zoom client which is available for all common platforms. Meeting IDs for upcoming lectures will be sent via email to all registered students (see also below).

Registration

All students interested in participating in the Computer Vision lecture are kindly requested to register via the LSF system.

NOTE: Registration is required in order to attend the online lectures!

Contact

Questions regarding technical and administrative issues can be directed to Fabian Wolf.


Tutorials: to be announced (1 week block course after the end of the lecture period)

Introduction:

For the majority of living beeings vision is the most important perception mechanism for orienting themselves in the environment. Therefore, there exists a multitude of attempts to recreate this capability in artificial systems. In contrast to image processing techniques found in industrial applications the aim of such advanced systems for machine vision is to obtain a task-oriented interpretation of a complex scene with as few restrictions as possible concerning the context and the recording conditions.

In this lecture advanced techniques of machine vision are covered which to some extent are inspired by cognitive processes known from human visual perception. First, important aspects of imaging processes are introduced with an emphasis on the perception of colors. Afterwards, methods for the extraction of image primitives (e.g. regions and edges) and for the calculation of feature representations (e.g. texture, depth, or motion) are presented. Finally, the lecture focusses on visual perception processes at the boundary between image processing and scene interpretation. Several appearance based object recognition techniques will be covered, e.g., Bag-of-Features approaches, Eigenimages, and deep Convolutional Neural Networks (CNNs) which define the state-of-the art for many current computer vision problems.

The accompanying tutorials will give students the opportunity to deepen their knowledge of the theoretical concepts presented in the lecture by working on relevant practical problems.

Specialization Module (Vertiefungsmodul INF-MSc-502) for Master (Applied) Computer Science

Topical focus areas (Schwerpunktgebiete): 2 (..., Embedded Systems, ...), 7 (Intelligent Systems)

Bibliography:

Accompanying Materials: