Content Analysis
Overview
We have been working on semantic content analysis for media such as photos
and videos. Such an analysis will enable more efficient and effective
utiliziation of massive media, for example, search, organization, summarization,
presentation and retrieval.
We have been invloved in the effort of
TREC Video Retrieval
Evaluation benchmark
sponsored by NIST.
In the year of 2007, we participated in two tasks,
high-level concept detection and
rushes video summarization.
Specifically, we have developed an efficient post-processing approach
to improve the MAP measurement for the high-level concept detection task.
This method was published by IEEE Trasactions on Multimedia in February 2008.
The paper title is
Association and Temporal Rule
Mining for Post-Processing of Semantic Concept Detetion in video.
In ACM Multimedia 2008, we proposed a unified framework,
multi-cue fusion,
to further improve MAP by integrating contextual correlation and temporal dependency simultaneously.
The average performance gain on MAP is around 30% for two popular benchmarks, VIREO-374 and Columbia-374.
We have also worked on developing algorithms to automatically segment a home
video into video segments based on events.
As a first attempt, we focused on church wedding videos. The initial
results was published in MIR 2007. The paper's title is
Semantic-Event Based
Analysis and Segmentation of Wedding Ceremony Videos.
A more complete version of this work has been published as an IEEE TCSVT article in November 2008.
Recently, we have made efforts on collecting benchmarks for
Region-of-Interest (ROI) of images using a collaborative game,
called PhotoShoot.
With the collected benchmark, we can compare exisiting ROI algorithms
quantatively. The preliminary results are published in
a CVPR 2009 paper.
We have also developed methods for learning landmark ontology from geotagged
images of Flickr and Wikipedia article. The resultant ontology could be
used in appliations such as automatic tag suggestion and content-relevant
image re-ranking.
The paper is published in MMM 2010.
Publications
- Learning Landmarks
by Exploting Social Media
-
Chia-Kai Liang,
Yu-Ting Hsieh, Tien-Jung Chuang, Yin Wang,
Ming-Fang Weng
Yung-Yu Chuang
- MMM 2010
-
A Collaborative
Benchmark for Region of Interest Detection Algorithms
-
Tz-Huan Huang,
Kai-Yin Cheng,
Yung-Yu Chuang
- IEEE CVPR 2009
- Multi-Cue Fusion
for Semantic Video Indexing
-
Ming-Fang Weng,
Yung-Yu Chuang
- ACM Multimedia 2008
- Association and Temporal
Rule Mining for Post-Processing of Semantic Concept Detection in Video
-
Ken-Hao Liu,
Ming-Fang Weng,
Chi-Tao Tseng,
Yung-Yu Chuang,
Ming-Syan Chen
- IEEE Transactions on Multimedia, Feburary 2008
- Semantic Analysis
for Automatic Event Recognition and Segmentation of Wedding Ceremony Videos
-
Wen-Huang Cheng,
Yung-Yu Chuang,
Yin-Tzu Lin,
Chi-Chang Hsieh,
Shao-Yen Fang,
Bing-Yu Chen,
Ja-Ling Wu
- IEEE Transactions on Circuits and Systems for Video Technology, November 2008
- Semantic-Event Based Analysis and Segmentation of Wedding Ceremony Videos
-
Wen-Huang Cheng,
Yung-Yu Chuang,
Bing-Yu Chen,
Ja-Ling Wu,
Shao-Yen Fang,
Yin-Tzu Lin,
Chi-Chang Hsieh,
Chen-Ming Pan,
Wei-Ta Chu,
Min-Chun Tien
- Proceedings of the 9th ACM SIGMM International Workshop on Multimedia Information Retrieval 2007
- NTU TRECVID-2007 Fast Rushes Summarization System
-
Chen-Ming Pan,
Yung-Yu Chuang,
Winston H. Hsu
- Proceedings of TRECVID BBC Rushes Summarization Workshop 2007
Support
This research is supported by:
- NTU98R0062-04
- NSC97-2622-E-002-010-CC2
- NSC96-2622-E-002-002
- NTU95R0062-AE00-02
cyy -a-t- csie.ntu.edu.tw