|
Department of
Computational
Perception
|
|
Multimedia Search and Retrieval
344.038, KV, 3h (4.5 ECTS), WS 2013/14
Assist.-Prof. Dr.
Markus Schedl
Univ.-Ass. Dr. Peter Knees
Time: Wednesday,
13:45-16:15
Start: Wednesday, Oct. 9, 2013
Location: K 001A
This class will be taught in English.
Class Objectives and Content
Students will learn the basics in
multimedia processing and analysis (feature extraction, feature
analysis, pattern discovery, information visualization), which are
required to understand and build multimedia information systems. We
will first cover the four most important types of multimedia data
(text, audio/music, image, video) individually, focusing on
particularities of each data source and on aspects of feature
extraction and data cleaning. After the first set of lectures, students
will have acquired a solid knowledge in state-of-the-art feature
extraction techniques. Hereafter, methods to analyze the extracted
features will be discussed, in particular, methods of data analysis for
pattern discovery and visualization. This
will be complemented by a practical project in which students will
apply the techniques learned.
The main topics covered include:
- Text features
- Audio and music features
- Image features
- Video features
- Linear and non-linear models for data analysis
- Data fusion of multimodal data
- Interfaces and browsing
- Search and retrieval in multimedia repositories
Practical Exercise
The task to be solved in the
practical exercise is to predict the most suitable music soundtrack for
a TV commercial, from a list of candidate songs. Given a TV commercial
(video), students are required to use a wide range of multimedia
features (text, music, image/video), experiment with fusion techniques,
and elaborate algorithms to select suitable music from a real-world
collection of 5000 songs. This is a challenging task, in which
multimodal information sources should be considered, which do not
trivially connect to each other.
The practical exercise will be carried out in groups of four. Information related to the task can be found
here, more details are given in the first lecture.
Schedule
- Oct 09, 2013: Introduction to Multimedia Retrieval and to Practical Exercise
- Oct 16, 2013: Text Features
- Oct 23, 2013: Audio/Music Features I
- Oct 30, 2013: Audio/Music Features II
- Nov 06, 2013: Image & Video Features
- Nov 13, 2013: Multimedia Data FusionNov 20.11.2013: Multimedia Data Analysis
- Nov 27, 2013: Interfaces for Browsing and Retrieval
- Dec 2013: Practical Exercise
- Jan 08, 2014: Introduction to Practical Exercise Evaluation
- Jan 15, 2014: Presentation of Results
- Jan 22, 2014: Written Exam
last edited by pk on Sep 11, 2013