[PPT] - Video Object Mining : Issues and Perspectives Jonathan Weber, S PowerPoint Presentation

SLIDE 1

Issues Video Mining Systems Perspectives Conclusion

Video Object Mining : Issues and Perspectives

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski

LSIIT,University of Strasbourg Ready Business System

September 23, 2010

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 1/13

SLIDE 2

Issues Video Mining Systems Perspectives Conclusion

1 Issues 2 Survey of video mining systems 3 Perspectives: Video Object Mining 4 Conclusion

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 2/13

SLIDE 3

Issues Video Mining Systems Perspectives Conclusion

Issues

High quantity of video data Low cost digital camera High storage capacity High-speed internet Ex : YouTube, Dailymotion, ... User needs content-based tools but Video meta-data are not sufficient Manual tagging of video is time consuming and error-prone There is a semantic gap between video data and what they represent

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 3/13

SLIDE 4

Issues Video Mining Systems Perspectives Conclusion

Issues

High quantity of video data Low cost digital camera High storage capacity High-speed internet Ex : YouTube, Dailymotion, ... User needs content-based tools but Video meta-data are not sufficient Manual tagging of video is time consuming and error-prone There is a semantic gap between video data and what they represent

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 3/13

SLIDE 5

Issues Video Mining Systems Perspectives Conclusion

Issues

High quantity of video data Low cost digital camera High storage capacity High-speed internet Ex : YouTube, Dailymotion, ... User needs content-based tools but Video meta-data are not sufficient Manual tagging of video is time consuming and error-prone There is a semantic gap between video data and what they represent

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 3/13

SLIDE 6

Issues Video Mining Systems Perspectives Conclusion

Issues

High quantity of video data Low cost digital camera High storage capacity High-speed internet Ex : YouTube, Dailymotion, ... How can we mine efficiently these complex data ? User needs content-based tools but Video meta-data are not sufficient Manual tagging of video is time consuming and error-prone There is a semantic gap between video data and what they represent

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 3/13

SLIDE 7

Issues Video Mining Systems Perspectives Conclusion

Issues

High quantity of video data Low cost digital camera High storage capacity High-speed internet Ex : YouTube, Dailymotion, ... How can we mine efficiently these complex data ? User needs content-based tools but Video meta-data are not sufficient Manual tagging of video is time consuming and error-prone There is a semantic gap between video data and what they represent

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 3/13

SLIDE 8

Issues Video Mining Systems Perspectives Conclusion

Issues

High quantity of video data Low cost digital camera High storage capacity High-speed internet Ex : YouTube, Dailymotion, ... How can we mine efficiently these complex data ? User needs content-based tools but Video meta-data are not sufficient Manual tagging of video is time consuming and error-prone There is a semantic gap between video data and what they represent

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 3/13

SLIDE 9

Issues Video Mining Systems Perspectives Conclusion

Issues

High quantity of video data Low cost digital camera High storage capacity High-speed internet Ex : YouTube, Dailymotion, ... How can we mine efficiently these complex data ? User needs content-based tools but Video meta-data are not sufficient Manual tagging of video is time consuming and error-prone There is a semantic gap between video data and what they represent

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 3/13

SLIDE 10

Issues Video Mining Systems Perspectives Conclusion

Issues

High quantity of video data Low cost digital camera High storage capacity High-speed internet Ex : YouTube, Dailymotion, ... How can we mine efficiently these complex data ? User needs content-based tools but Video meta-data are not sufficient Manual tagging of video is time consuming and error-prone There is a semantic gap between video data and what they represent

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 3/13

SLIDE 11

Issues Video Mining Systems Perspectives Conclusion

Issues

High quantity of video data Low cost digital camera High storage capacity High-speed internet Ex : YouTube, Dailymotion, ... How can we mine efficiently these complex data ? User needs content-based tools but Video meta-data are not sufficient Manual tagging of video is time consuming and error-prone There is a semantic gap between video data and what they represent = ⇒ Lack of semantic information

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 3/13

SLIDE 12

Issues Video Mining Systems Perspectives Conclusion

Issues

High quantity of video data Low cost digital camera High storage capacity High-speed internet Ex : YouTube, Dailymotion, ... How can we mine efficiently these complex data ? User needs content-based tools but Video meta-data are not sufficient Manual tagging of video is time consuming and error-prone There is a semantic gap between video data and what they represent = ⇒ Lack of semantic information How can we introduce semantic in the mining process ?

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 3/13

SLIDE 13

Issues Video Mining Systems Perspectives Conclusion

1 Issues 2 Survey of video mining systems 3 Perspectives: Video Object Mining 4 Conclusion

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 4/13

SLIDE 14

Issues Video Mining Systems Perspectives Conclusion

Video Mining Systems

Video Mining Process of extracting information/knowledge from large amounts of video data Many recent publications deal with video mining:

A. Anjulan and N. Canagarajah, Signal Processing: Image Communication, 2007
A. Anjulan and N. Canagarajah, IEEE ICIP, 2007
S. de Avila, A. da Luz, and A. de Araujo, IWSSIP, 2008
A. Basharat, Y. Zhai, and M. Shah, Computer Vision and Image Understanding, 2008
F. Chevalier, J.-P. Domenger, J. Benois-Pineau, and M. Delest, Pattern Recognition Letters, 2007
X. Gao, X. Li, J. Feng, and D. Tao, Pattern Recognition Letters, 2009
D. Liu and T. Chen, Computer Vision and Image Understanding, 2009
W. Ren and Y. Zhu, IIH-MSP, 2008
J. Sivic and A. Zisserman, Proceedings of the IEEE, 2008
L. F. Teixeira and L. Corte-Real, Pattern Recognition Letters, 2009
S. Zhai, B. Luo, J. Tang, and C.-Y. Zhang, ICIG, 2007

Study of existing systems Characterize the different systems Check if they use semantics

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 5/13

SLIDE 15

Issues Video Mining Systems Perspectives Conclusion

Video Mining Systems

Video Mining Process of extracting information/knowledge from large amounts of video data Many recent publications deal with video mining:

A. Anjulan and N. Canagarajah, Signal Processing: Image Communication, 2007
A. Anjulan and N. Canagarajah, IEEE ICIP, 2007
S. de Avila, A. da Luz, and A. de Araujo, IWSSIP, 2008
A. Basharat, Y. Zhai, and M. Shah, Computer Vision and Image Understanding, 2008
F. Chevalier, J.-P. Domenger, J. Benois-Pineau, and M. Delest, Pattern Recognition Letters, 2007
X. Gao, X. Li, J. Feng, and D. Tao, Pattern Recognition Letters, 2009
D. Liu and T. Chen, Computer Vision and Image Understanding, 2009
W. Ren and Y. Zhu, IIH-MSP, 2008
J. Sivic and A. Zisserman, Proceedings of the IEEE, 2008
L. F. Teixeira and L. Corte-Real, Pattern Recognition Letters, 2009
S. Zhai, B. Luo, J. Tang, and C.-Y. Zhang, ICIG, 2007

Study of existing systems Characterize the different systems Check if they use semantics

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 5/13

SLIDE 16

Issues Video Mining Systems Perspectives Conclusion

Video Mining Systems

Video Mining Process of extracting information/knowledge from large amounts of video data Many recent publications deal with video mining:

A. Anjulan and N. Canagarajah, Signal Processing: Image Communication, 2007
A. Anjulan and N. Canagarajah, IEEE ICIP, 2007
S. de Avila, A. da Luz, and A. de Araujo, IWSSIP, 2008
A. Basharat, Y. Zhai, and M. Shah, Computer Vision and Image Understanding, 2008
F. Chevalier, J.-P. Domenger, J. Benois-Pineau, and M. Delest, Pattern Recognition Letters, 2007
X. Gao, X. Li, J. Feng, and D. Tao, Pattern Recognition Letters, 2009
D. Liu and T. Chen, Computer Vision and Image Understanding, 2009
W. Ren and Y. Zhu, IIH-MSP, 2008
J. Sivic and A. Zisserman, Proceedings of the IEEE, 2008
L. F. Teixeira and L. Corte-Real, Pattern Recognition Letters, 2009
S. Zhai, B. Luo, J. Tang, and C.-Y. Zhang, ICIG, 2007

Study of existing systems Characterize the different systems Check if they use semantics

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 5/13

SLIDE 17

Issues Video Mining Systems Perspectives Conclusion

Video Mining Systems

Video Mining Process of extracting information/knowledge from large amounts of video data Many recent publications deal with video mining:

A. Anjulan and N. Canagarajah, Signal Processing: Image Communication, 2007
A. Anjulan and N. Canagarajah, IEEE ICIP, 2007
S. de Avila, A. da Luz, and A. de Araujo, IWSSIP, 2008
A. Basharat, Y. Zhai, and M. Shah, Computer Vision and Image Understanding, 2008
F. Chevalier, J.-P. Domenger, J. Benois-Pineau, and M. Delest, Pattern Recognition Letters, 2007
X. Gao, X. Li, J. Feng, and D. Tao, Pattern Recognition Letters, 2009
D. Liu and T. Chen, Computer Vision and Image Understanding, 2009
W. Ren and Y. Zhu, IIH-MSP, 2008
J. Sivic and A. Zisserman, Proceedings of the IEEE, 2008
L. F. Teixeira and L. Corte-Real, Pattern Recognition Letters, 2009
S. Zhai, B. Luo, J. Tang, and C.-Y. Zhang, ICIG, 2007

Study of existing systems Characterize the different systems Check if they use semantics

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 5/13

SLIDE 18

Issues Video Mining Systems Perspectives Conclusion

Essential properties to characterize video mining systems

Properties Objectives Elements Descriptors Scales Supervision levels

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 6/13

SLIDE 19

Issues Video Mining Systems Perspectives Conclusion

Essential properties to characterize video mining systems

Properties Objectives Elements Descriptors Scales Supervision levels Summarization

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 6/13

SLIDE 20

Issues Video Mining Systems Perspectives Conclusion

Essential properties to characterize video mining systems

Properties Objectives Elements Descriptors Scales Supervision levels Retrieval

+

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 6/13

SLIDE 21

Issues Video Mining Systems Perspectives Conclusion

Essential properties to characterize video mining systems

Properties Objectives Elements Descriptors Scales Supervision levels Classification

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 6/13

SLIDE 22

Issues Video Mining Systems Perspectives Conclusion

Essential properties to characterize video mining systems

Properties Objectives Elements Descriptors Scales Supervision levels

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 6/13

SLIDE 23

Issues Video Mining Systems Perspectives Conclusion

Essential properties to characterize video mining systems

Properties Objectives Elements Descriptors Scales Supervision levels Pixel pixel timeline

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 6/13

SLIDE 24

Issues Video Mining Systems Perspectives Conclusion

Essential properties to characterize video mining systems

Properties Objectives Elements Descriptors Scales Supervision levels Region timeline region

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 6/13

SLIDE 25

Issues Video Mining Systems Perspectives Conclusion

Essential properties to characterize video mining systems

Properties Objectives Elements Descriptors Scales Supervision levels Object

bject

timeline

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 6/13

SLIDE 26

Issues Video Mining Systems Perspectives Conclusion

Essential properties to characterize video mining systems

Properties Objectives Elements Descriptors Scales Supervision levels Frame frame timeline

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 6/13

SLIDE 27

Issues Video Mining Systems Perspectives Conclusion

Essential properties to characterize video mining systems

Properties Objectives Elements Descriptors Scales Supervision levels Shot shot timeline

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 6/13

SLIDE 28

Issues Video Mining Systems Perspectives Conclusion

Essential properties to characterize video mining systems

Properties Objectives Elements Descriptors Scales Supervision levels

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 6/13

SLIDE 29

Issues Video Mining Systems Perspectives Conclusion

Essential properties to characterize video mining systems

Properties Objectives Elements Descriptors Scales Supervision levels Motion timeline

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 6/13

SLIDE 30

Issues Video Mining Systems Perspectives Conclusion

Essential properties to characterize video mining systems

Properties Objectives Elements Descriptors Scales Supervision levels Color timeline

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 6/13

SLIDE 31

Issues Video Mining Systems Perspectives Conclusion

Essential properties to characterize video mining systems

Properties Objectives Elements Descriptors Scales Supervision levels Shape timeline

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 6/13

SLIDE 32

Issues Video Mining Systems Perspectives Conclusion

Essential properties to characterize video mining systems

Properties Objectives Elements Descriptors Scales Supervision levels

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 6/13

SLIDE 33

Issues Video Mining Systems Perspectives Conclusion

Essential properties to characterize video mining systems

Properties Objectives Elements Descriptors Scales Supervision levels Global

POLICE Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 6/13

SLIDE 34

Issues Video Mining Systems Perspectives Conclusion

Essential properties to characterize video mining systems

Properties Objectives Elements Descriptors Scales Supervision levels Grid

POLICE Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 6/13

SLIDE 35

Issues Video Mining Systems Perspectives Conclusion

Essential properties to characterize video mining systems

Properties Objectives Elements Descriptors Scales Supervision levels Region

POLICE

1 2

3 4 5 6

7 8 9 12 11 10 13 14 15 18 17 16 19 20 21 22 23

24

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 6/13

SLIDE 36

Issues Video Mining Systems Perspectives Conclusion

Essential properties to characterize video mining systems

Properties Objectives Elements Descriptors Scales Supervision levels Object

POLICE Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 6/13

SLIDE 37

Issues Video Mining Systems Perspectives Conclusion

Essential properties to characterize video mining systems

Properties Objectives Elements Descriptors Scales Supervision levels Interest point

POLICE Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 6/13

SLIDE 38

Issues Video Mining Systems Perspectives Conclusion

Essential properties to characterize video mining systems

Properties Objectives Elements Descriptors Scales Supervision levels Pixel

POLICE Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 6/13

SLIDE 39

Issues Video Mining Systems Perspectives Conclusion

Essential properties to characterize video mining systems

Properties Objectives Elements Descriptors Scales Supervision levels

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 6/13

SLIDE 40

Issues Video Mining Systems Perspectives Conclusion

Essential properties to characterize video mining systems

Properties Objectives Elements Descriptors Scales Supervision levels Null

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 6/13

SLIDE 41

Issues Video Mining Systems Perspectives Conclusion

Essential properties to characterize video mining systems

Properties Objectives Elements Descriptors Scales Supervision levels Supervised

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 6/13

SLIDE 42

Issues Video Mining Systems Perspectives Conclusion

Essential properties to characterize video mining systems

Properties Objectives Elements Descriptors Scales Supervision levels Semi-Supervised

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 6/13

SLIDE 43

Issues Video Mining Systems Perspectives Conclusion

Essential properties to characterize video mining systems

Properties Objectives Elements Descriptors Scales Supervision levels Parameters

153 0.98 23.2 25 0.01

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 6/13

SLIDE 44

Issues Video Mining Systems Perspectives Conclusion

Current usages of 11 existing systems (2007-2009)

Objectives Elements Descriptors Scales Supervision

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 7/13

SLIDE 45

Issues Video Mining Systems Perspectives Conclusion

Current usages of 11 existing systems (2007-2009)

Objectives

Retrieval Classification Summarization

6 3 2

Elements Descriptors Scales Supervision

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 7/13

SLIDE 46

Issues Video Mining Systems Perspectives Conclusion

Current usages of 11 existing systems (2007-2009)

Objectives

Retrieval Classification Summarization

6 3 2

Elements

Object Video Shot

5 5 1

Descriptors Scales Supervision

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 7/13

SLIDE 47

Issues Video Mining Systems Perspectives Conclusion

Current usages of 11 existing systems (2007-2009)

Objectives

Retrieval Classification Summarization

6 3 2

Elements

Object Video Shot

5 5 1

Descriptors

Texture Color Multi Motion

6 2 2 1

Scales Supervision

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 7/13

SLIDE 48

Issues Video Mining Systems Perspectives Conclusion

Current usages of 11 existing systems (2007-2009)

Objectives

Retrieval Classification Summarization

6 3 2

Elements

Object Video Shot

5 5 1

Descriptors

Texture Color Multi Motion

6 2 2 1

Scales

Region Global Grid Object

5 3 2 1

Supervision

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 7/13

SLIDE 49

Issues Video Mining Systems Perspectives Conclusion

Current usages of 11 existing systems (2007-2009)

Objectives

Retrieval Classification Summarization

6 3 2

Elements

Object Video Shot

5 5 1

Descriptors

Texture Color Multi Motion

6 2 2 1

Scales

Region Global Grid Object

5 3 2 1

Supervision

Parameters Supervised Null S-Supervised

9 1 1

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 7/13

SLIDE 50

Issues Video Mining Systems Perspectives Conclusion

1 Issues 2 Survey of video mining systems 3 Perspectives: Video Object Mining 4 Conclusion

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 8/13

SLIDE 51

Issues Video Mining Systems Perspectives Conclusion

Video Object Mining: A way of introducing semantics

Objectives

Retrieval Classification Summarization

6 3 2

Our approach for Video Object Mining Generic framework Object is the most semantic element Combination of object-oriented descriptors 2 types of scales : inner and context scale User relevance feedback brings semantics

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 9/13

SLIDE 52

Issues Video Mining Systems Perspectives Conclusion

Video Object Mining: A way of introducing semantics

Elements

Object Video Shot

5 5 1

Our approach for Video Object Mining Generic framework Object is the most semantic element Combination of object-oriented descriptors 2 types of scales : inner and context scale User relevance feedback brings semantics

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 9/13

SLIDE 53

Issues Video Mining Systems Perspectives Conclusion

Video Object Mining: A way of introducing semantics

Descriptors

Texture Color Multi Motion

6 2 2 1

Our approach for Video Object Mining Generic framework Object is the most semantic element Combination of object-oriented descriptors 2 types of scales : inner and context scale User relevance feedback brings semantics

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 9/13

SLIDE 54

Issues Video Mining Systems Perspectives Conclusion

Video Object Mining: A way of introducing semantics

Scales

Region Global Grid Object

5 3 2 1

Our approach for Video Object Mining Generic framework Object is the most semantic element Combination of object-oriented descriptors 2 types of scales : inner and context scale User relevance feedback brings semantics

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 9/13

SLIDE 55

Issues Video Mining Systems Perspectives Conclusion

Video Object Mining: A way of introducing semantics

Scales

=

inner context

Our approach for Video Object Mining Generic framework Object is the most semantic element Combination of object-oriented descriptors 2 types of scales : inner and context scale User relevance feedback brings semantics

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 9/13

SLIDE 56

Issues Video Mining Systems Perspectives Conclusion

Video Object Mining: A way of introducing semantics

Supervision level

Parameters Supervised Null S-Supervised

9 1 1

Our approach for Video Object Mining Generic framework Object is the most semantic element Combination of object-oriented descriptors 2 types of scales : inner and context scale User relevance feedback brings semantics

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 9/13

SLIDE 57

Issues Video Mining Systems Perspectives Conclusion

Our approach compared to related works

Objectives

Retrieval Classification Summarization

6 3 2

Elements

Object Video Shot

5 5 1

Descriptors

Texture Color Multi Motion

6 2 2 1

Scales

Region Global Grid Object

5 3 2 1

Supervision

Parameters Supervised Null S-Supervised

9 1 1

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 10/13

SLIDE 58

Issues Video Mining Systems Perspectives Conclusion

VOMF: Video Object Mining Framework

Video repository Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 11/13

SLIDE 59

Issues Video Mining Systems Perspectives Conclusion

VOMF: Video Object Mining Framework

Object segmentation

Video repository Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 11/13

SLIDE 60

Issues Video Mining Systems Perspectives Conclusion

VOMF: Video Object Mining Framework

Object segmentation Object evaluation

Video repository Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 11/13

SLIDE 61

Issues Video Mining Systems Perspectives Conclusion

VOMF: Video Object Mining Framework

Object segmentation Object evaluation

Video repository Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 11/13

SLIDE 62

Issues Video Mining Systems Perspectives Conclusion

VOMF: Video Object Mining Framework

Object segmentation Object evaluation

Video repository Objects Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 11/13

SLIDE 63

Issues Video Mining Systems Perspectives Conclusion

VOMF: Video Object Mining Framework

Object segmentation Object evaluation Object mining

Video repository Objects Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 11/13

SLIDE 64

Issues Video Mining Systems Perspectives Conclusion

VOMF: Video Object Mining Framework

Object segmentation Object evaluation Object mining Mining evaluation

Video repository Objects Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 11/13

SLIDE 65

Issues Video Mining Systems Perspectives Conclusion

VOMF: Video Object Mining Framework

Object segmentation Object evaluation Object mining Mining evaluation

Video repository Objects Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 11/13

SLIDE 66

Issues Video Mining Systems Perspectives Conclusion

VOMF: Video Object Mining Framework

Object segmentation Object evaluation Object mining Mining evaluation

Video repository Mining result Objects Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 11/13

SLIDE 67

Issues Video Mining Systems Perspectives Conclusion

1 Issues 2 Survey of video mining systems 3 Perspectives: Video Object Mining 4 Conclusion

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 12/13

SLIDE 68

Issues Video Mining Systems Perspectives Conclusion

Conclusion

Issues

Massive video repositories need new data mining schemes Semantics have to be considered to achieve user goals These are still open problems

Perspectives

Video Object Mining has to be explored Semi-supervised learning limits user’s workload A new generic framework has been proposed This framework is currently being implemented

Difficulties

Object segmentation Interactive mining (involving user relevance feedback)

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 13/13

SLIDE 69

Issues Video Mining Systems Perspectives Conclusion

Conclusion

Issues

Massive video repositories need new data mining schemes Semantics have to be considered to achieve user goals These are still open problems

Perspectives

Video Object Mining has to be explored Semi-supervised learning limits user’s workload A new generic framework has been proposed This framework is currently being implemented

Difficulties

Object segmentation Interactive mining (involving user relevance feedback)

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 13/13

SLIDE 70

Issues Video Mining Systems Perspectives Conclusion

Conclusion

Issues

Massive video repositories need new data mining schemes Semantics have to be considered to achieve user goals These are still open problems

Perspectives

Video Object Mining has to be explored Semi-supervised learning limits user’s workload A new generic framework has been proposed This framework is currently being implemented

Difficulties

Object segmentation Interactive mining (involving user relevance feedback)

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 13/13

SLIDE 71

Issues Video Mining Systems Perspectives Conclusion

Conclusion

Issues

Massive video repositories need new data mining schemes Semantics have to be considered to achieve user goals These are still open problems

Perspectives

Video Object Mining has to be explored Semi-supervised learning limits user’s workload A new generic framework has been proposed This framework is currently being implemented

Difficulties

Object segmentation Interactive mining (involving user relevance feedback)

Jonathan Weber, S´ ebastien Lef` evre, Pierre Gan¸ carski Video Object Mining - 13/13