Basic knowledge of pattern recognition as taught in the module Cognitive Systems is expected.

The lecture introduces the basics, as well as advanced aspects of deep learning methods and their application for a number of computer vision tasks. The following topics will be addressed in the lecture:

- Introduction to Deep Learning

- Convolutional Neural Networks (CNN): Background

- CNNs: basic architectures and learning algorithms

- Object Recognition with CNN

- Image Segmentation with CNN

- Recurrent Neural Networks

- Generating image descriptions (Image Captioning)

- Automatic question answering (Visual Question Answering)

- Generative Adversarial Networks (GAN) and their applications

- Deep Learning platforms and tools

The lecture is partially given in German and English.

In recent years tremendous progress has been made in analysing and understanding image and video content. The dominant approach in Computer Vision today are deep learning approaches, in particular the usage of Convolutional Neural Networks.

90 Stunden

Students should be able to grasp the underlying concepts in the field of deep learning and its various applications.

• Understand the theoretical basis of deep learning
• Understand the Convolutional Neural Networks (CNN)
• Develop basis for the concepts and algorithms used in building and training the CNNs.
• Able to apply deep learning in different computer vision applications

Die Erfolgskontrolle erfolgt in Form einer schriftlichen Prüfung (im Umfang von i.d.R. 60 Minuten) nach § 4 Abs. 2 Nr. 1 SPO.

Abhängig von der Teilnehmerzahl wird sechs Wochen vor der Prüfungsleistung angekündigt (§ 6 Abs. 3 SPO), ob die Erfolgskontrolle

• in Form einer mündlichen Prüfung nach § 4 Abs. 2 Nr. 2 SPO oder
• in Form einer schriftlichen Prüfung  nach § 4 Abs. 2 Nr. 1 SPO

stattfindet.