AMIGO - Automatic Indexing of Lecture Footage

Abstract

Video content has become an integral part of e-learning programs, particularly since the COVID pandemic. It offers a richer context than text documents and allows a more individual learning experience meeting students' individual information needs, schedule, and speed. However, a key challenge lies in the fact that videos co-exists with lectures' other textual content, which requires students to alternate between documents and videos. Unfortunately, however, video is hard to navigate, and finding the desired bits of information can be tedious. To tackle this challenge, this talk introduces AMIGO, an educational video platform developed and hosted at RheinMain University of Applied Sciences. AMIGO automatically links videos with the slides displayed within them, which facilitates a text search, navigation and scrolling directly in the video. To do so, AMIGO requires no additional hardware (except a video camera). It uses a feature-based image analysis which localizes text documents (most often powerpoint slides) within the video and recognizes which slides is visible at each point in time. AMIGO covers screencasts as well as e-lectures. It has been in operation at Hochschule RheinMain, covers >1000 videos of >1200 hours covering >150 courses, and is being used by >3000 users.

Biography

Adrian Ulges is a professor at RheinMain University of Applied Sciences in Wiesbaden and an alumnus of TU Kaiserslautern / Germany (2009 PhD in computer science, 2005 diploma in computer science). He has been as a researcher with the German Research Center for Artificial Intelligence (DFKI) (2005-2012), and has worked with Google as an intern (2005, Mountain View) and as a visiting scientist (2011, Zurich). Adrian's research focuses on machine learning and multimedia analysis and has been awarded with a Google Research Award in 2010. His publication record includes over 50 peer-reviewed papers.