Development of Deep Vision Systems [DEEPVIS]
Computer vision systems try to mimic human capabilities of visual perception to support time-consuming and labor-intensive tasks like the recognition, localization, and tracking of critical objects. Nowadays, such systems increasingly rely on methods and tools from the field of machine learning to automatically extract useful information from images that can be utilized for decision support and business automation purposes.
This course provides the necessary fundamentals for the development of modern vision systems based on machine learning. The particular focus is on deep neural networks and their capabilities of automated feature learning. More specifically, we consider different types of network architectures, look at the steps of image labelling and data preparation, discuss crucial hyperparameters and evaluation criteria, and review other related aspects, such as 3D vision, hybrid intelligence, and explainable artificial intelligence.
The course has a strong practical focus. At the beginning of the semester, all fundamentals are provided in lecture sessions and hands-on exercises. Afterwards, students are encouraged to work (in groups) on real projects to apply the methods and concepts learned during the teaching sessions. The results are presented and discussed at the end of the semester.
This course is taught together with Prof. Zschech (Intelligent Information Systems). For details about the course, see StudOn.