Video Information Retrieval Mark Ruzomberka IST 497 11/07/02.

34
Video Information Retrieval Mark Ruzomberka IST 497 11/07/02

Transcript of Video Information Retrieval Mark Ruzomberka IST 497 11/07/02.

Page 1: Video Information Retrieval Mark Ruzomberka IST 497 11/07/02.

Video Information Retrieval

Mark Ruzomberka

IST 497

11/07/02

Page 2: Video Information Retrieval Mark Ruzomberka IST 497 11/07/02.

Joke

Page 3: Video Information Retrieval Mark Ruzomberka IST 497 11/07/02.

Outline

What is Video Information Retrieval (VIR) ?Reasons VIR is necessaryTheoreticalWhere we are todayExamplesProblemsFuture Work Conclusion

Page 4: Video Information Retrieval Mark Ruzomberka IST 497 11/07/02.

What is Video Information Retrieval (VIR) ?

Recognition technologies Image Voice Text transcripts

Document retrieval technologies Topic segmentation Topic matching Text summarization

Presentation Technologies Combine Recognition and retrieval technologies

Result is an integrated application

Page 5: Video Information Retrieval Mark Ruzomberka IST 497 11/07/02.

VIR-Need, or Why do I care?

Consider the task of trying to find a five minute video clip of interest in a library of 1000 hour long tapes.

Consider the “go to the part where” problem

Page 6: Video Information Retrieval Mark Ruzomberka IST 497 11/07/02.

What do people want from IR

D-Lib Magazine’s asks:

“What do People want from Information Retrieval?”

# 8 Multimedia

Page 7: Video Information Retrieval Mark Ruzomberka IST 497 11/07/02.

Specificly, Reasons for Video IR

Reading is slow compared to your potential for understanding information

Humans think in pictures not words Reading is particularly slow on a computer screen Example: Daydreaming while some one is talking Reading a page in a book and not remembering what it was about

Page 8: Video Information Retrieval Mark Ruzomberka IST 497 11/07/02.

VIR makes for quicker human understanding.

Palm/Grafitti 25 Hand Writing 35-40 Typing 50-70 Speaking 135-175 Reading 200 Listening 400 - 500 Thinking 500+

•Video IR allows for faster access to information

Page 9: Video Information Retrieval Mark Ruzomberka IST 497 11/07/02.

Theoretical:

Think of the “Jetsons mail system”

You “talk” to the computer, Computer intelligently “talks” back to you

Page 10: Video Information Retrieval Mark Ruzomberka IST 497 11/07/02.

Where we are today

Two of Video Information Retrieval System are currently available:

Type One- keyword/text basedType Two- Content based

Page 11: Video Information Retrieval Mark Ruzomberka IST 497 11/07/02.

Type One- keyword/text based

•DVR- basic expansion of image IR, •not as interesting

Page 12: Video Information Retrieval Mark Ruzomberka IST 497 11/07/02.

Type Two- Content based

Video Mail Informedia MSR Video

Skimmer

Page 13: Video Information Retrieval Mark Ruzomberka IST 497 11/07/02.

Example: Video Mail

University of Cambridge 1994-1996

AT&T 1999

2000-project ended

Page 14: Video Information Retrieval Mark Ruzomberka IST 497 11/07/02.

Video Mail: Medusa network

Medusa multimedia environment at Olivetti Research Ltd. In Cambridge

It takes a modular approach unlike that of a pc or workstation Unified by a common interface to ATM network Devices plug directly into network and include:

Cameras Audio devices Networked frame buffers Processor farms Disk drives

Page 15: Video Information Retrieval Mark Ruzomberka IST 497 11/07/02.

Video Mail: Medusa Network

“The network is the computer” metaphor is used Solves storage and network speed problems Complicates expense problem

Page 16: Video Information Retrieval Mark Ruzomberka IST 497 11/07/02.

How it works-Overview

Page 17: Video Information Retrieval Mark Ruzomberka IST 497 11/07/02.

The Integrated Application

“narrow” by sender,date, time

Page 18: Video Information Retrieval Mark Ruzomberka IST 497 11/07/02.

Video Mail: Video Browser

Content is now being viewedKeywords are flagged

Page 19: Video Information Retrieval Mark Ruzomberka IST 497 11/07/02.

Video Mail: Video Browser

In the latest version “thumb-nailed” pictures of key frames replace color coded line of the search keyword

Page 20: Video Information Retrieval Mark Ruzomberka IST 497 11/07/02.

Informedia

The Informedia Digital Video Library Project automatically combines speech, image and natural language understanding to create a full-content searchable digital video library.

Page 21: Video Information Retrieval Mark Ruzomberka IST 497 11/07/02.

Informedia

Page 22: Video Information Retrieval Mark Ruzomberka IST 497 11/07/02.

Informedia: human factor issues

Interaction MotivationEffective usage modes

Commercial compressionVHS quality playback. Terabyte (1,000 gigabytes) of storage 1000 hours of video.

Page 23: Video Information Retrieval Mark Ruzomberka IST 497 11/07/02.

Problems

1.Human understanding

2.Spoken document retrieval

3.Poor video browsers

4.Expensive

5.Slow access to data

6.Large amounts of data

Page 24: Video Information Retrieval Mark Ruzomberka IST 497 11/07/02.

Microsoft Research (MSR) Video Skimmer

Page 25: Video Information Retrieval Mark Ruzomberka IST 497 11/07/02.

Microsoft Research (MSR) Video Skimmer

Enhanced Browser Controls: Time Compression Pause Removal Textual Indices:

TOC, Notes

Visual IndicesShot Boundary FramesTimeline Markers

Jump Control (Back/Next)

Page 27: Video Information Retrieval Mark Ruzomberka IST 497 11/07/02.

Problem: Expensive

Hard drive space expensive Video adds to problem

High bandwidth needs are also expensive

Year Drive Size Drive Cost Per MB/Cost

1956 5 megabytes 50,000.00 10,000.00

1980 26 megabytes 5,000.00 193.00

1985 10 megabytes 710.00 71.00

1989 40 megabytes 1,199.00 36.00

1995 1.2 gigabytes 680.00 68.60

2000 30.0 gigabytes 249.99 0.96

•http://www.littletechshoppe.com/ns1625/winchest.html

Page 28: Video Information Retrieval Mark Ruzomberka IST 497 11/07/02.

Problem: Slow Access to Data

Broadband still not available everywhereAvailability doesn’t mean acceptanceEspecially after dot com crash 2000

Page 29: Video Information Retrieval Mark Ruzomberka IST 497 11/07/02.

Problem: Large Amounts of Data

Current Systems use MPEG2Newer compression technologies

MPEG 4-DIVX -DVD QualityVideo consumes orders of magnitude

more storage than textMPEG 7 is on horizon

Page 30: Video Information Retrieval Mark Ruzomberka IST 497 11/07/02.

Future Work ?

Sky the limit ?Sci-Fi the limit ?

Hard Drive Space, Bandwidth are current limitations.

Page 31: Video Information Retrieval Mark Ruzomberka IST 497 11/07/02.

Conclusion

Not yet ready for prime timeStorage and Network Costs decreasingSuccess is in day to day usageSlowly Becoming Mainstream E.x.TivoProblems of “real world tests”

Idiot proof ATM and Medusa aren’t mainstream

Page 32: Video Information Retrieval Mark Ruzomberka IST 497 11/07/02.

Papers

Video Mail Retrieval Using Voice: Report on Keyword.. - Jones, Foote, Jones.. (1994) What do people want from Information Retrieval?. Croft, Bruce W. D-Lib Magazine. (1995) Video Skimming for Quick Browsing based on Audio and Image.. - Smith, Kanade (1995) The VISION digital video library (context) - Gauch, Li et al. – (1997) Informedia: News-on-Demand Multimedia Information.. - Hauptmann, Witbrock (1997) M.G. Christel and D.J. Martin, "Information Visualization within a Digital Video Library", J.

Intelligent Info. Systems 11(3), (1998), pp. 235-257 Browsing Digital Video. Li, Gupta, Sanocki et. Al.

Page 33: Video Information Retrieval Mark Ruzomberka IST 497 11/07/02.

Questions?

Page 34: Video Information Retrieval Mark Ruzomberka IST 497 11/07/02.

Joke?

"There are 10 types of people in the world...

those who understand binary and those who don't."