Transcript Data mining on Web Video Auto Tagging

Data mining for Web
Video Auto Tagging
TRAN Hoang Tung
Outline
Data mining for Web Video Auto Tagging
Introducing myself
General Context
Explain what I want to do in my PhD thesis
Data mining
Applying data mining techniques into video tagging
Last year and future plan!
USTH Consortium - Toulouse 2011
2
10/19/11
About myself
My name is TRAN Hoang Tung
Arrival date: 25th Oct 2010
I’m working in Hubert Curien Laboratory, Jean Monnet
University and CNRS
City: Saint Etienne (close to Lyon and Grenoble)
My supervisors:
Francois Jacquenet (full professor)
Elisa Fromont (assistant professor)
Baptiste Jeudy (assistant professor)
Keywords: data mining, video tagging, video annotation
USTH Consortium - Toulouse 2011
3
10/19/11
USTH Consortium - Toulouse 2011
4
10/19/11
Web videos (Youtube…)
Current video search engines are text-based (title,
description, tags). Title & description are written by
each uploader (normally as a complete phase). Tags
are single words!!
However, tags are notoriously:
Incomplete (don’t fully represent such video)
Incorrect (spam, increase number of view)
Unranked (the most important tag is not the first tag)
My thesis goal: creating an auto-tagging system
which reduces above disadvantages of current tags.
USTH Consortium - Toulouse 2011
5
10/19/11
Data mining
Data mining is a field of computer science, and more
precisely of artificial intelligence. The goal is to
describe (very) large data in an informative way. For
example discovering patterns !
Example: Market-Basket Analysis
TID
items
Consider item set {bread, diaper}:
Support = 4/5 = 80%
1
{bread, milk}
2
{bread, diaper, beers, eggs}
3
{bread, diaper, beers, cola}
4
{bread, diaper, beers, milk}
with confidence:
5
{bread, diaper, cola, milk}
= s({bread,diaper,beers})/s({bread,diaper}) =3/4
USTH Consortium - Toulouse 2011
Consider association rule:
{bread, diaper} -> beers
6
10/19/11
Data mining & video tagging
Assumption: similar videos will have (with high probability)
similar tags.
Steps:
Compute similarities between videos based on patterns
Propagate the tags
USTH Consortium - Toulouse 2011
7
10/19/11
The Past and Future!
Last year:
Studying data mining
Followed a master course in Machine Learning
Participating to a winter school in Machine Learning
applied to image processing
Reading bibliography about data mining applied to
video analysis
Studying French ! (240 hours)
Next year:
Take a look at other types of patterns: discriminative
patterns, sequences, …
Try other techniques to convert video into binary form
Write something ! 
USTH Consortium - Toulouse 2011
8
10/19/11