Multimedia Data Management M Second cycle degree programme (LM) in - - PowerPoint PPT Presentation
ALMA MATER STUDIORUM - UNIVERSIT DI BOLOGNA Multimedia Data Management M Second cycle degree programme (LM) in Computer Engineering University of Bologna Course presentation Academic Year 2019/2020 Home page:
ALMA MATER STUDIORUM - UNIVERSITÀ DI BOLOGNA
Bologna, February 19th, 2020
Second cycle degree programme (LM) in Computer Engineering University of Bologna
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
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/
Multimedia Data Management M
3
§ 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
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)
Multimedia Data Management M
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
Multimedia Data Management M
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
Multimedia Data Management M
§ 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
§ 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
Multimedia Data Management M
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
Multimedia Data Management M
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
Multimedia Data Management M
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
Multimedia Data Management M
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
Multimedia Data Management M
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,
§ 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
Multimedia Data Management M
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
Multimedia Data Management M
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
Multimedia Data Management M
15 15
§ Six examination sessions per year divided as follows:
§ three sessions before the summer § starting from September, at the request of the students
Multimedia Data Management M
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
Multimedia Data Management M
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
Multimedia Data Management M
ALMA MATER STUDIORUM - UNIVERSITÀ DI BOLOGNA
Academic Year 2019/2020
18
Multimedia Data Management M