Oral exam (in case of a large number of participants: written exam); 50% of the max. number of excercise points are necessary for being admitted to the exam

This class provides a broad overview of principles and algorithms for video analysis via classical machine learning techniques, convolutional networks, as well as transformer architectures. Specific topics include video feature representation, temporal representation in neural networks, as well as specific algorithms and techniques for the processing of video data. In parallel, we will discuss applications such as video clip classification, temporal video segmentation, spatio-temporal action detection, video context, spatio-temporal modeling of humans and objects, video summarization, semantic video segmentation, and many more, as well as state-of-the-art methods for video analysis in general.

Students gain an understanding of the theoretical and practical concepts needed for automated video analysis. After this course, students should be able to understand and apply the basic concepts of classification and temporal segmentation on video data, train video-based models, reproduce research results and conduct original research in this area.

-