Multimedia Data Management M Second cycle degree programme (LM) in - - PDF document

1

ALMA MATER STUDIORUM - UNIVERSITÀ DI BOLOGNA

Bologna, February 19th, 2020

Multimedia Data Management M

Second cycle degree programme (LM) in Computer Engineering University of Bologna

Course presentation

Academic Year 2019/2020

Home page: http://www-db.disi.unibo.it/courses/MDM/ Electronic version: 0.01.Presentation.pdf Electronic version: 0.01.Presentation-2p.pdf

1

2

§ Prof.ssa Ilaria Bartolini

Department of Computer Science and Engineering (DISI) University of Bologna Viale Risorgimento, 2, Bologna Multimedia Database Group

http://www-db.disi.unibo.it/MMDBGroup/

Datalab

http://www-db.disi.unibo.it/research/datalab/

Teacher

• I. Bartolini

Multimedia Data Management M

2

2

3

Contacts

• I. Bartolini

§ E-mail:

§ ilaria.bartolini@unibo.it

§ Telephone:

§ 051 20 93550

§ Web site:

§ http://www-db.disi.unibo.it/ibartolini/

§ Office hours:

§ on Monday, from 11:00 to 13:00, c/o Palazzina “gialla” DISI (close to the 2nd School entrance of “via Vallescura”)

§ please, send me an email for appointment first

Multimedia Data Management M

3

4

§ Name: “Multimedia Data Management M” § Credits: 8 § Teaching hours: 64 hours § Period: Semester II

§ February 19th 2020 – June 5th 2020

§ Teaching hours:

§ Wednesday – 14:00-17:00 – Room 2.7.A (beginning of the lesson at 14:15) § Friday – 9:00-11:00 – Room 1.4 (beginning of the lesson at 9:15)

General information

• I. Bartolini

Course calendar

Multimedia Data Management M

4

3

5

Learning outcomes

§ The course aims to provide the students with all necessary knowledge and skills to deal with the effective and efficient management of “non-conventional” data types, notably multimedia (MM) data (e.g., textual documents, still images, videos, sound, audio/visual streams, etc.); the final goal is to find, within very large collections, namely “Big Data”, those objects that are better suited to fulfill the information needs of non-expert users § We pay special attention to problems of MM data modeling/ representation, MM data retrieval models, and interaction paradigms between the user and the MM system (both for purposes of data presentation and exploration) § We first consider architectures of traditional (“standalone”) MM systems; then, we concentrate on more complex MM services, by primarily focusing on search engines, social networks and recommender systems

Course contents

• I. Bartolini

Multimedia Data Management M

5

6

Topics at a glance

§ Multimedia data and data types classification § Textual documents: the easiest case of multimedia data § Multimedia data content representations § Automatic techniques for MM data semantic annotations § Efficient and effective techniques for multimedia data retrieval, browsing, and visualization § Multimedia data on the Web !Each lecture is enriched by practical examples, use cases, demos, and exercises… J

Course contents

• I. Bartolini

Multimedia Data Management M

6

4

§ Facilitate and improve the “access” to very large multimedia data collections for general users, conjunctively exploiting:

§ low level features (e.g., color distribution of a video key frame) § semi-automatically provided annotations § “dedicated users” manually provided meta-data

Main goal in one pic!

§ Trimotore Fiat G212 § Data: 1947 § Collezione: Tema di cultura industriale § Tipologia: Immagine § Aereo, Motore, Ali § Das Cabinet des Dr. Caligari § Data: 1920 § Nazione: Germania § Regista: Robert Wiene § Genere: Horror § Espressionismo, Ipnosi, Sonnambulismo § La Gioconda § Sito: Museo Louvre, Parigi § Secolo: XVI § Autore: Leonardo da Vinci § Periodo: Rinascimento § Data: 1503 § Dipinto, Ritratto, Sorriso

Archivio Storico Fiat Cineteca Archivio Artistico

Models, Algorithms, Interfaces

7

• I. Bartolini

Multimedia Data Management M

7

8

§ Multimedia data and data types classification

§ MM data and applications § Srtuctured data § Semi-strucutred data § Unstructured data

§ Textual documents: the easiest case of multimedia data

§ Documents representation § Automatic indexing techniques, stemming, stoplist § Searches of Boolean type § Searches of phrases and for proximity § The vector space model: weighing techniques and ranking of the results

Detailed program (1)

• I. Bartolini

Multimedia Data Management M

8

5

9

§ Multimedia data content representations

§ MM data coding § MM data content representation

§ How to find multimedia data of interest

§ Description models for complex MM objects § Similarity measures for MM data content § MM database management systems

§ Efficient algorithms for MM data retrieval

§ MM query formulation paradigms § Sequential retrieval of MM data § Index-based retrieval of MM data

§ Automatic techniques for MM data annotations: filling the semantic gap

§ Traditional techniques § Graph-based solutions

Detailed program (2)

• I. Bartolini

Multimedia Data Management M

9

10 10

§ Browsing MM data collections

§ Browsing paradigms

§ MM data presentation

§ User interface design principles § Visualization paradigms § Dimensionality reduction techniques

§ Result accuracy

§ Quality of the results: reference metrics § User-system interaction and relevance feedback: how to improve the quality of the results

§ Multimedia Data on the Web

§ Web search engines: principles § Graph-based data: semantic Web and social networks § Web recommender systems: basics

Detailed program (3)

• I. Bartolini

Multimedia Data Management M

10

6

Course home page

Contents:

§ News § Copy of slides, extra material in PDF format § Bibliography& Useful links § Assessment methods § Exam sessions § Project work

§ Modalities/topics

http://www-db.disi.unibo.it/courses/MDM/ https://iol.unibo.it/course/view.php?id=44586#section-1

11

• I. Bartolini

Multimedia Data Management M

11

12 12

§ Education material provided by the teacher (copies of the slides used in the classroom, scientific literature, etc.) For any further additional information, recommended books are:

§ Candan, Sapino. “Data Management for Multimedia Retrieval”, Cambridge,

• 2010. ISBN: 978-0-521-88739-7

§ Zhang, Zhang. “Multimedia Data Mining: A Systematic Introduction to Concepts and Theory”, Chapman and Hall/CRC, 2008. ISBN: 9781584889663 § Chapman & Chapman. “Digital Multimedia”, Wiley & Sons Ltd, 2009. ISBN: 13 978-0-470-51216-6 § Colace, De Santo, Moscato, Picariello, Schreiber, Tanca. “Data Management for Pervasive Systems”, Springer, 2015. ISBN: 978-3-319- 20061-3

Readings/Bibliography

• I. Bartolini

Multimedia Data Management M

12

7

13 13

§ Most course lectures are in “traditional” classrooms and exploit the slides § Real use cases (+ relative demos) are also proposed and discussed based on open-source software libraries and frameworks

Teaching methods

• I. Bartolini

Multimedia Data Management M

13

14 14

§ The exam evaluation consists of an oral examination

§ To participate to the exam, interested students have to register themselves by exploiting the usual UniBO Web application, called AlmaEsami

§ The students can also arrange with the teacher a “Project work” of Multimedia Data Management M based on their own preferences or suggested topics… more details on this in few minutes… J

§ In this case, the oral examination can be taken conjunctly with the project presentation

Assessment methods

• I. Bartolini

Multimedia Data Management M

14

8

15 15

§ Six examination sessions per year divided as follows:

§ three sessions before the summer § starting from September, at the request of the students

Examination sessions

• I. Bartolini

Multimedia Data Management M

15

16 16

§ The project activity aims to apply the notions and skills acquired during the course by developing a project § The project consists in the concrete resolution of a problem concerning the management of multimedia data

§ The topic of the project can be proposed either by the teacher and the student

§ Usually a scientific paper is considered as the base for the development of a new multimedia data management algorithm and/or infrastructure and/or service and/or interface

§ however, everything that “looks interesting” is allowed to be candidate! J

§ Once you start your project work, the development stages are verified by the teacher with periodical meetings

Project work details (1)

• I. Bartolini

Multimedia Data Management M

16

9

17 17

§ The final evaluation consists of a Power Point presentation integrated by a project demo

§ According to students’ preferences, the project presentation can be taken conjunctly with the final course oral examination

§ During the development phase of the project, students can profitably use/extend existing software libraries, such as

§ Open-source Multimedia libraries and architectures § Multimedia frameworks, e.g.,

§ Windsurf, for the management of large image collections § Shiatsu, for the management of large video databases § RAM3S, for the real-time analysis of massive multimedia streams § … developed within the Multimedia Database Group @ DISI

(http://www-db.disi.unibo.it/MMDBGroup/)

§ Augmented reality environments for immersive 3D applications

Project work details (2)

• I. Bartolini

Multimedia Data Management M

17

ALMA MATER STUDIORUM - UNIVERSITÀ DI BOLOGNA

Questions?

Course presentation

Academic Year 2019/2020

18

• I. Bartolini

Multimedia Data Management M

18