Personal Video Manager: Managing and Mining Home Video Collections - PowerPoint PPT Presentation

Loading...

PPT – Personal Video Manager: Managing and Mining Home Video Collections PowerPoint presentation | free to download - id: 267e8-Njc4M



Loading


The Adobe Flash plugin is needed to view this content

Get the plugin now

View by Category
About This Presentation
Title:

Personal Video Manager: Managing and Mining Home Video Collections

Description:

Personal Video Manager: Managing and Mining Home Video Collections. Peng Wu and Pere Obrador ... metadata from personal videos: a multifold challenge ... – PowerPoint PPT presentation

Number of Views:81
Avg rating:3.0/5.0
Slides: 21
Provided by: pen2
Learn more at: http://www.hpl.hp.com
Category:

less

Write a Comment
User Comments (0)
Transcript and Presenter's Notes

Title: Personal Video Manager: Managing and Mining Home Video Collections


1
Personal Video Manager Managing and Mining Home
Video Collections
ISL
  • Peng Wu and Pere Obrador
  • Imaging Systems Laboratory, HP Labs
  • peng.wu_at_hp.com

2
Outline
  • Motivation
  • PVM solution
  • PVM managing functionality
  • Mining from consumer experience
  • Summary and future work

3
Outline
  • Motivation
  • PVM solution
  • PVM managing functionality
  • Mining from consumer experience
  • Summary and future work

4
Motivation
  • Video asset extraction
  • shot detection
  • summarization
  • skimming
  • keyframe extraction
  • Video asset management
  • standard mpeg-7, MPV
  • industrial practice Virage, MARVEL

consumer demands?
  • Personal Video Manager
  • Better understanding consumers needs
  • Better understanding technologys strength

5
Outline
  • Motivation
  • PVM Solution
  • PVM management functionalities
  • Mining from consumer experience
  • Summary and future work

6
PVM Solution Overview
  • PVM provides tools and technologies to help
    consumers to manage their video collections
  • PVM covers the major steps in consuming consumer
    videos

Content Ingestion
Content Consumption
Asset Collection
A station that can digitize videos from various
inputs and perform video format conversion
A comprehensive server-client architecture to
enable web based video asset annotation and
search/browse
7
PVM Solution Overview
PVM Management Application
Browsing/Searching Application
PVM Client
PVM Server
Input
Database
Metadata Schema
Video Storage
Digitization Station
Indexing/FTS
Mining Engine
8
Metadata and Metadata Schema
  • The Metadata Reusable and managed knowledge of
    personal videos for easy sharing and accessing
  • Metadata schema Make knowledge manageable

Schema Design
VideoSource
VideoInfo
Raw metadata format, length, date,
Editing metadata editors view of content
Easier sharing and accessing of content
9
Metadata Extraction
  • Collecting metadata from personal videos a
    multifold challenge
  • Metadata extraction in PVM


Automatic keyframe extraction, face detection,
shot detection,
Semi - automatic highlight,
Metadata accuracy
Consumers effort
Manual a UI with means to enable interaction
10
Outline
  • Motivation
  • PVM Solution
  • PVM management functionalities
  • Mining from consumer experience
  • Summary and future work

11
PVM Web Application Asset Collection
12
PVM Web Application Search/Browse
13
Outline
  • Motivation
  • PVM Solution
  • PVM management functionalities
  • Mining from consumer experience
  • Summary and future work

14
Mining from consumer experience
  • Setting the stage
  • 80 hours of home video content
  • 11 consumers
  • Learning from the experience
  • Directly interfacing with consumers
  • concern on content sensitivity (at different
    levels)
  • highlight important and laborious
  • event keyword vs. visual cues in querying
  • Data mining (analysis) on the metadata
  • analysis service directly operate on relational
    database
  • data analysis clustering and decision tree

15
Mining from consumer experience
  • An example

Clusters 61 total 1 zoom in (25) 2 zoom out
(21) 3 pan left (8) 4 pan right (7)
16
Mining from consumer experience
  • Mining VideoCameraMotion
  • 46 Camera zoom in/out vs. 15 camera pan
  • 15 camera pan 8 from left to right, 7 from right
    to left
  • pairing relationship (checking the time stamp)
  • dominant cluster left to right and then left
  • Mining VideoCmaeraMotion VideoAudioEvent
  • camera motion meaningful audio event (speech,
    singing, etc.)
  • Mining VideoKeyframe
  • 30 keyframes 15 with face, 7 with important
    signs (logo, cake, building, ...)
  • keyframe what is that?

17
Mining from consumer experience
  • Remarks
  • interfacing with consumers is important and
    rewarding
  • consumer feedback embeds rich information to
    guide research and development, but has not been
    well addressed
  • Understanding/analyzing experience requires
    technology development data mining multimedia
    modeling experience modeling

18
Outline
  • Motivation
  • PVM Solution
  • PVM management functionalities
  • Mining from consumer experience
  • Summary and future work

19
Summary and future work
  • Summary
  • PVM offers an environment to consumers for
    annotating and managing video collections.
  • PVM is an effort to develop systematical approach
    to better understand consumers demand
  • Future work
  • Incorporating richer capture/analysis metadata to
    enable better video capture/consumption
  • Exploring metadata mining
  • Deriving technologies from consumers demand
  • In consumer video/multimedia management,
    understanding consumers demands is one of the
    keys to truly prompt the whole RD community

20
HP logo
About PowerShow.com