2003.09.23 - SLIDE 1IS 202 – FALL 2003 Lecture 10: Metadata for Media Prof. Ray Larson & Prof. Marc Davis UC Berkeley SIMS Tuesday and Thursday 10:30 am.

2003.09.23 - SLIDE 1IS 202 – FALL 2003

Lecture 10: Metadata for Media

Prof. Ray Larson & Prof. Marc Davis

UC Berkeley SIMS

Tuesday and Thursday 10:30 am - 12:00 pm

Fall 2003http://www.sims.berkeley.edu/academics/courses/is202/f03/

SIMS 202:

Information Organization

and Retrieval

2003.09.23 - SLIDE 2IS 202 – FALL 2003

Today’s Agenda

• Review of Last Time

• Metadata for Motion Pictures– Representing Video

– Current Approaches

– Media Streams

• Discussion Questions

• Action Items for Next Time

2003.09.23 - SLIDE 3IS 202 – FALL 2003

Today’s Agenda

• Review of Last Time

• Metadata for Motion Pictures– Representing Video

– Current Approaches

– Media Streams

• Discussion Questions

• Action Items for Next Time

2003.09.23 - SLIDE 4IS 202 – FALL 2003

The Media Opportunity

• Vastly more media will be produced• Without ways to manage it (metadata

creation and use) we lose the advantages of digital media

• Most current approaches are insufficient and perhaps misguided

• Great opportunity for innovation and invention

• Need interdisciplinary approaches to the problem

2003.09.23 - SLIDE 5IS 202 – FALL 2003

What is the Problem?

• Today people cannot easily find, edit, share, and reuse media

• Computers don’t understand media content– Media is opaque and data rich– We lack structured representations

• Without content representation (metadata), manipulating digital media will remain like word-processing with bitmaps

2003.09.23 - SLIDE 6IS 202 – FALL 2003

M E T A D A T AMETADATA

Traditional Media Production Chain

PRE-PRODUCTION POST-PRODUCTIONPRODUCTION DISTRIBUTION

Metadata-Centric Production Chain

2003.09.23 - SLIDE 7IS 202 – FALL 2003

Asset Retrieval and Reuse

Automated Media Production Process

Web Integration and

Streaming MediaServices

FlashGenerator

WAP

HTML Email

Print/PhysicalMedia

ActiveCapture

1Automatic

Editing3

Personalized/Customized

Delivery

4

Adaptive Media Engine

2 Annotationand Retrieval

Reusable Online Asset Database

Annotation ofMedia Assets

2003.09.23 - SLIDE 8IS 202 – FALL 2003

Technology Summary

• Media Streams provides a framework for creating metadata throughout the media production cycle to make media assets searchable and reusable

• Active Capture automates direction and cinematography using real-time audio-video analysis in an interactive control loop to create reusable media assets

• Adaptive Media uses adaptive media templates and automatic editing functions to mass customize and personalize media and thereby eliminate the need for editing on the part of end users

• Together, these technologies will automate, personalize, and speed up media production, distribution, and reuse

2003.09.23 - SLIDE 9IS 202 – FALL 2003

Active Capture

2003.09.23 - SLIDE 10IS 202 – FALL 2003

Active Capture: Reusable Shots

2003.09.23 - SLIDE 11IS 202 – FALL 2003

Marc Davis in Godzilla Scene

2003.09.23 - SLIDE 12IS 202 – FALL 2003

Evolution of Media Production

• Customized production– Skilled creation of one media product

• Mass production– Automatic replication of one media product

• Mass customization– Skilled creation of adaptive media templates– Automatic production of customized media

2003.09.23 - SLIDE 13IS 202 – FALL 2003

• Movies change from being static data to programs

• Shots are inputs to a program that computes new media based on content representation and functional dependency (US Patents 6,243,087 & 5,969,716)

Central Idea: Movies as Programs

Parser

Parser

Producer

Media

Media

Media

ContentRepresentation

ContentRepresentation

2003.09.23 - SLIDE 14IS 202 – FALL 2003

Today’s Agenda

• Review of Last Time

• Metadata for Motion Pictures– Representing Video

– Current Approaches

– Media Streams

• Discussion Questions

• Action Items for Next Time

2003.09.23 - SLIDE 15IS 202 – FALL 2003

Representing Video

• Streams vs. Clips

• Video syntax and semantics

• Ontological issues in video representation

2003.09.23 - SLIDE 16IS 202 – FALL 2003

Video is Temporal

Stream of 100 Frames of Video

A Clip from Frame 47 to Frame 68 with Descriptors

2003.09.23 - SLIDE 17IS 202 – FALL 2003

Streams vs. Clips

The Stream of 100 Frames of Video with 6 Annotations Resulting in ManyPossible Segmentations of the Stream

Stream of 100 Frames of Video

2003.09.23 - SLIDE 18IS 202 – FALL 2003

Stream-Based Representation

• Makes annotation pay off– The richer the annotation, the more numerous the

possible segmentations of the video stream

• Clips – Change from being fixed segmentations of the video

stream, to being the results of retrieval queries based on annotations of the video stream

• Annotations– Create representations which make clips, not

representations of clips

2003.09.23 - SLIDE 19IS 202 – FALL 2003

Video Syntax and Semantics

• The Kuleshov Effect

• Video has a dual semantics

– Sequence-independent invariant semantics of shots

– Sequence-dependent variable semantics of shots

2003.09.23 - SLIDE 20IS 202 – FALL 2003

Ontological Issues for Video

• Video plays with rules for identity and continuity

– Space

– Time

– Person

– Action

2003.09.23 - SLIDE 21IS 202 – FALL 2003

Space and Time: Actual vs. Inferable

• Actual Recorded Space and Time– GPS– Studio space and time

• Inferable Space and Time– Establishing shots– Cues and clues

2003.09.23 - SLIDE 22IS 202 – FALL 2003

Time: Temporal Durations

• Story (Fabula) Duration– Example: Brushing teeth in story world (5 minutes)

• Plot (Syuzhet) Duration– Example: Brushing teeth in plot world (1 minute: 6

steps of 10 seconds each)

• Screen Duration– Example: Brushing teeth (10 seconds: 2 shots of 5

seconds each)

2003.09.23 - SLIDE 23IS 202 – FALL 2003

Character and Continuity

• Identity of character is constructed through– Continuity of actor– Continuity of role

• Alternative continuities– Continuity of actor only– Continuity of role only

2003.09.23 - SLIDE 24IS 202 – FALL 2003

Representing Action

• Physically-based description for sequence-independent action semantics– Abstract vs. conventionalized descriptions– Temporally and spatially decomposable

actions and subactions

• Issues in describing sequence-dependent action semantics– Mental states (emotions vs. expressions)– Cultural differences (e.g., bowing vs. greeting)

2003.09.23 - SLIDE 25IS 202 – FALL 2003

“Cinematic” Actions

• Cinematic actions support the basic narrative structure of cinema– Reactions/Proactions

• Nodding, screaming, laughing, etc.

– Focus of Attention• Gazing, headturning, pointing, etc.

– Locomotion• Walking, running, etc.

• Cinematic actions can occur• Within the frame/shot boundary• Across the frame boundary• Across shot boundaries

2003.09.23 - SLIDE 26IS 202 – FALL 2003

Today’s Agenda

• Review of Last Time

• Metadata for Motion Pictures– Representing Video

– Current Approaches

– Media Streams

• Discussion Questions

• Action Items for Next Time

2003.09.23 - SLIDE 27IS 202 – FALL 2003

The Search for Solutions

• Current approaches to creating metadata don’t work– Signal-based analysis– Keywords– Natural language

• Need standardized metadata framework– Designed for video and rich media data– Human and machine readable and writable– Standardized and scaleable– Integrated into media capture, archiving, editing,

distribution, and reuse

2003.09.23 - SLIDE 28IS 202 – FALL 2003

Signal-Based Parsing

• Practical problem– Parsing unstructured, unknown video is very,

very hard

• Theoretical problem– Mismatch between percepts and concepts

2003.09.23 - SLIDE 29IS 202 – FALL 2003

Perceptual/Conceptual Issue

Clown Nose Red Sun

Similar Percepts / Dissimilar Concepts

2003.09.23 - SLIDE 30IS 202 – FALL 2003

Perceptual/Conceptual Issue

Car Car

Dissimilar Percepts / Similar Concepts

John Dillinger’s Timothy McVeigh’s

2003.09.23 - SLIDE 31IS 202 – FALL 2003

Signal-Based Parsing

• Effective and useful automatic parsing

– Video• Shot boundary detection• Camera motion analysis• Low level visual similarity• Feature tracking• Face detection

– Audio• Pause detection• Audio pattern matching• Simple speech recognition• Speech vs. music

detection

• Approaches to automated parsing

– At the point of capture, integrate the recording device, the environment, and agents in the environment into an interactive system

– After capture, use “human-in-the-loop” algorithms to leverage human and machine intelligence

2003.09.23 - SLIDE 32IS 202 – FALL 2003

Keywords vs. Semantic Descriptors

dog,biting,Steve

2003.09.23 - SLIDE 33IS 202 – FALL 2003

Keywords vs. Semantic Descriptors

dog,biting,Steve

2003.09.23 - SLIDE 34IS 202 – FALL 2003

Why Keywords Don’t Work

• Are not a semantic representation

• Do not describe relations between descriptors

• Do not describe temporal structure

• Do not converge

• Do not scale

2003.09.23 - SLIDE 35IS 202 – FALL 2003

Jack, an adult male police officer, while walking to the left, starts waving with his left arm, and then has a puzzled look on his face as he turns his head to the right; he then drops his facial expression and stops turning his head, immediately looks up, and then stops looking up after he stops waving but before he stops walking.

Natural Language vs. Visual Language

2003.09.23 - SLIDE 36IS 202 – FALL 2003

Natural Language vs. Visual Language

Jack, an adult male police officer, while walking to the left, starts waving with his left arm, and then has a puzzled look on his face as he turns his head to the right; he then drops his facial expression and stops turning his head, immediately looks up, and then stops looking up after he stops waving but before he stops walking.

2003.09.23 - SLIDE 37IS 202 – FALL 2003

Notation for Time-Based Media: Music

2003.09.23 - SLIDE 38IS 202 – FALL 2003

Visual Language Advantages

• A language designed as an accurate and readable representation of time-based media– For video, especially important for actions,

expressions, and spatial relations

• Enables Gestalt view and quick recognition of descriptors due to designed visual similarities

• Supports global use of annotations

2003.09.23 - SLIDE 39IS 202 – FALL 2003

Today’s Agenda

• Review of Last Time

• Metadata for Motion Pictures– Representing Video

– Current Approaches

– Media Streams

• Discussion Questions

• Action Items for Next Time

2003.09.23 - SLIDE 40IS 202 – FALL 2003

After Capture: Media Streams

2003.09.23 - SLIDE 41IS 202 – FALL 2003

Media Streams Features

• Key features– Stream-based representation (better segmentation)– Semantic indexing (what things are similar to)– Relational indexing (who is doing what to whom)– Temporal indexing (when things happen)– Iconic interface (designed visual language)– Universal annotation (standardized markup schema)

• Key benefits– More accurate annotation and retrieval– Global usability and standardization– Reuse of rich media according to content and structure

2003.09.23 - SLIDE 42IS 202 – FALL 2003

Media Streams GUI Components

• Media Time Line

• Icon Space– Icon Workshop– Icon Palette

2003.09.23 - SLIDE 43IS 202 – FALL 2003

Media Time Line

• Visualize video at multiple time scales

• Write and read multi-layered iconic annotations

• One interface for annotation, query, and composition

2003.09.23 - SLIDE 44IS 202 – FALL 2003

Media Time Line

2003.09.23 - SLIDE 45IS 202 – FALL 2003

Icon Space

• Icon Workshop– Utilize categories of video representation– Create iconic descriptors by compounding iconic

primitives– Extend set of iconic descriptors

• Icon Palette– Dynamically group related sets of iconic descriptors– Reuse descriptive effort of others– View and use query results

2003.09.23 - SLIDE 46IS 202 – FALL 2003

Icon Space

2003.09.23 - SLIDE 47IS 202 – FALL 2003

Icon Space: Icon Workshop

• General to specific (horizontal)– Cascading hierarchy of icons with increasing

specificity on subordinate levels

• Combinatorial (vertical)– Compounding of hierarchically organized

icons across multiple axes of description

2003.09.23 - SLIDE 48IS 202 – FALL 2003

Icon Space: Icon Workshop Detail

2003.09.23 - SLIDE 49IS 202 – FALL 2003

Icon Space: Icon Palette

• Dynamically group related sets of iconic descriptors

• Collect icon sentences

• Reuse descriptive effort of others

2003.09.23 - SLIDE 50IS 202 – FALL 2003

Icon Space: Icon Palette Detail

2003.09.23 - SLIDE 51IS 202 – FALL 2003

Video Retrieval In Media Streams

• Same interface for annotation and retrieval

• Assembles responses to queries as well as finds them

• Query responses use semantics to degrade gracefully

2003.09.23 - SLIDE 52IS 202 – FALL 2003

Media Streams Technologies

• Minimal video representation distinguishing syntax and semantics

• Iconic visual language for annotating and retrieving video content

• Retrieval-by-composition methods for repurposing video

2003.09.23 - SLIDE 53IS 202 – FALL 2003

Non-Technical Challenges

• Standardization of media metadata (MPEG-7)

• Broadband infrastructure and deployment

• Intellectual property and economic models for sharing and reuse of media assets

2003.09.23 - SLIDE 54IS 202 – FALL 2003

Today’s Agenda

• Review of Last Time

• Metadata for Motion Pictures– Representing Video

– Current Approaches

– Media Streams

• Discussion Questions

• Action Items for Next Time

2003.09.23 - SLIDE 55IS 202 – FALL 2003

Discussion Questions (Davis)

• John Snydal on Media Streams– What is the target audience of users (annotators/retrievers) for

Media Streams? In the article the following groups are mentioned:

• Content providers• Video editors• News teams• Documentary film makers• Film archives• Stock photo houses• Video archivists• Video producers• (international audience)• (illiterate and preliterate people)

– Is it possible that Media Streams could satisfy the needs, goals and requirements of all of these groups, or would it be more appropriate to develop separate, tailored applications for the unique needs of each group?

2003.09.23 - SLIDE 56IS 202 – FALL 2003

Discussion Questions (Davis)

• danah boyd on Media Streams– Icons require visual literacy. Icons are also

culturally constructed. Thus, for them to work as an information access bit, people must learn the visual language; it is not inherent. What are the social consequences of a system dependent on unfamiliar cues?

2003.09.23 - SLIDE 57IS 202 – FALL 2003

Discussion Questions (Davis)

• danah boyd on Media Streams– Films are constructed narratives. But most

commonplace storytelling is not. Even in a creative form, people often piece together found objects instead of finding objects to fit their story. (Think teenage girls making collages out of the latest YM.) Storytelling also happens around media far more than through media (i.e. telling a story about a picture rather than using a collection of pictures to tell a story). My guess is that this social phenomenon goes beyond the retrieval issues. Do you think that Media Streams would encourage new behavior regarding storytelling or will it only be useful for those with a constructed narrative in mind? Why (not)?

2003.09.23 - SLIDE 58IS 202 – FALL 2003

Discussion Questions (Davis)

• Jesse Mendelsohn on Media Streams– Media Streams does not allow iconic

descriptions of emotion or scene-interpretation. How would someone searching stock footage for a “suspenseful scene of two men beating each other” go about doing it? The actual sense of “suspense” and the act of “beating” cannot be iconified. Does this limit Media Streams' ability or is there a way around it within its capabilities as described?

2003.09.23 - SLIDE 59IS 202 – FALL 2003

Discussion Questions (Davis)

• Jesse Mendelsohn on Media Streams– In order for Media Streams to work well it

relies on a the availability of a very large and extensive resource of well-annotated video. Is the current annotation process too primitive and/or time consuming to allow Media Streams to work to its full potential? Will changing how Media Streams can be used to annotate video or changing video annotation methods in general make Media Streams more effective?

2003.09.23 - SLIDE 60IS 202 – FALL 2003

Today’s Agenda

• Review of Last Time

• Metadata for Motion Pictures– Representing Video

– Current Approaches

– Media Streams

• Discussion Questions

• Action Items for Next Time

2003.09.23 - SLIDE 61IS 202 – FALL 2003

Assignment 4.1

• Assignment 4.1

• Phone Metadata Design - Part 1– Due Oct 2

2003.09.23 - SLIDE 62IS 202 – FALL 2003

Next Time

• Database Design (RRL)

• Readings– Handouts in Class

• Database Modeling and Design -- Ch. 2 The ER Model - Basic Concepts (Teorey, T.J.)

• Logical Database Design and the Relational Model (F. R. McFadden, J. A. Hoffer)