After successful completion of the course, students are able to implement complex media analysis systems comprising of signal processing and machine learning. Implementation includes design, programming and evaluation based on ground truth.

IMPORTANT: Successful registration for the course requires successful completion of the Registration Assessment Test in the e-learning course! Please register via "Self-Enrollment" in the linked Tuwel course, read the assignment description and take the test before the deadline.

Classical (multimedia) data analysis with signal processing and machine learning (without Deep Learning) using audiovisual media as an example:  

• Low-level feature extraction from audiovisual media
• Semantic feature modeling
• Similarity modeling and feature classifciation
• Performance evaluation and statistical data analysis
• Examples of applications and advanced topics

The goal is for students to learn the classical methods of information retrieval that are still relevant today when, for example, training data is not sufficiently available or limitations in processing capacity arise (e.g., edge computing).

Similarity Modeling 2 deals with more complex methods than Similarity Modeling 1 and builds on this course. However, the organization is the same in both courses.

Central topic of the course is understanding the importance of transformation operations for digital media description and similarity measurement. The desired understanding is developed interactively in the lecture part and consolidated by a realistic exercise in the practical part of the course.

IMPORTANT! Due to the Corona pandemic the lecture will be replaced by a set of videos + summaries written by the students. All details can be found in due time in the TUWEL forum.

Pedagogic concept

• Frame of knowledge transfer with lecture block at the beginning an exam at the end of the lecture
• Exploration of lecture contents in a lab project in groups of 2-3 students
• Application of state of the art visualization and seminar methods for enabling student participation during the lecture
• Application of an open forum for knowledge exchange over groups during the lab course

ECTS Breakdown

Description                       ECTS  Hours
---------------------------------------------
Preparation                       0.04    1.0
Lecture                           0.32    8.0
Preparation of the Group Project  0.04    1.0
Group Project Work                1.88   47.0
Preparation of the Oral Exam      0.70   17.5
Oral Exam                         0.02    0.5
---------------------------------------------
Total                             3.00   75.0

Students are graded based on their performance in the practical part of the course as well as on thei participation in the lecture part.

H. Eidenberger: "Professional Media Understanding", Atpress, Vienna, 2012.

List of topics and links in the TUWEL forum.

Programming in Java and/or Python