Special Sessions:

The MIR workshop this year will include two special sessions. They were selected from submitted session proposals based on the Call for Special Sessions.

Special Session 1: Semantic Indexing of Consumer and Web Videos

Consumer and Web media presents some of the last frontiers for automated content analysis and indexing, which is needed for managing the explosive amount of user-generated or Web-based media content. Associated information like EXIF metadata or user tags is useful but often insufficient for describing the content shown in the images and videos. Recently, exciting progresses have been seen in developing large benchmark data sets (like TRECVID and ImageClef), ontologies (like LSCOM), and classification models for several domains such as broadcast news. However, solutions suitable for user-generated videos and user-centric usage scenarios are still missing. We see a timely opportunity for organizing a special session to bring together active researchers and practitioners from both academia and industry to share recent progresses and stimulate innovations in this exciting nascent area.

Organizers:

Shih-Fu Chang, Columbia University (USA)
Dan Ellis, Columbia University (USA)
Alex Loui, Kodak Research Labs (USA)
Jiebo Luo, Kodak Research Labs (USA)

Papers:

1. Kodak Consumer Video Benchmark Data Set: Concept Definition and Annotation
Alexander C. Loui, Jiebo Luo, Shih-Fu Chang, Dan Ellis, Wei Jiang, Lyndon Kennedy, Keansub Lee, Akira Yanagawa

2. Large-Scale Multimodal Semantic Concept Detection for Consumer Video
Shih-Fu Chang, Dan Ellis, Wei Jiang, Keansub Lee, Akira Yanagawa, Alexander C. Loui, Jiebo Luo

3. Multi-Modal Web Video Categorization
Linjun Yang, Jiemin Liu, Tao Mei, Xiaokang Yang, Xian-Sheng Hua

4. Watch What I Watch: Using Community Activity To Understand Content
David A. Shamma, Ryan Shaw, Peter L. Shafton, Yiming Liu

5. Learning People Annotation from the Web via Consistency Learning
Jay Yagnik, Atiq Islam

6. ITEMS: Intelligent Travel Experience Management System
Chih-Chieh Liu, Chun-Hsiang Huang, Wei-Ta Chu, and Ja-Ling Wu

Special Session 2: Personalized Multimedia Information Retrieval

The explosion of multimedia content in databases, broadcasts, streaming media, etc. has generated new requirements for more effective access to these global information repositories. Content extraction, indexing, and retrieval of multimedia data continue to be one of the most challenging and fastest-growing research areas. A consequence of the growing consumer demand for multimedia information is that sophisticated technology is needed for representing, modeling, indexing, and retrieving multimedia data. In particular, we need robust techniques to index/retrieve and compress multimedia information, new scalable browsing algorithms allowing access to very large multimedia databases, and semantic visual interfaces integrating the above components into unified multimedia browsing and retrieval systems.

Consider the following futuristic scenario:

John Citizen lives in Brussels, holds a degree in economics and works for a multinational company dealing in oil imports. He enjoys travel with emphasis on warm Mediterranean sites with good swimming and fishing. When watching TV his primary interest is international politics, particularly European. During a recent armed conflict he wanted to have different perspectives on the war, including both relevant historical material as well as future projections from commentators.

When he returns home from work, a personalized interactive multimedia program is ready for him, created automatically from various multimedia segments taken from diverse sources including multimedia news feeds, digital libraries, and collected analyst commentaries. The program includes different perspectives on the events, discussions and analyses appropriate for a university graduate. The video program is production quality, including segment transitions and music. Sections of the program allow him to interactively explore information and knowledge of particular relevance to him, namely the impact of war on oil prices in various countries (his business interest), and the potential impact on tourism and accommodation prices across the Mediterranean next summer. Some presentations may be synchronized with a map display, accessible interactively. Johns behavior and dialogue with the display are logged together with a record of the information presented to allow the system to accumulate and update his state of knowledge and discern his interests in order to better serve him in the future. When John is away from home for business or leisure, he can receive the same personalized information on his mobile device, with a focus on information reflecting the neighborhood of his current Mediterranean location.

This vision has many consequences from both societal and economic perspectives. Societal consequences impact on personal information, continuous education, e-government, tourism, leisure, etc. The economic infrastructure of information provision and dissemination will also change, as new actors enter the market (mainly providers of the new services to be offered), and existing actors radically change their mode of operation (for example introducing new paradigm for the production of the TV programs of the future) and business models.

The aim of this special session is to address this hot topic in Multimedia Information Retrieval. We are considering several aspects of this problem. On one hand, the user will want to have a personalized access to his image/video collections and this can be achieved by providing intuitive and natural browsing capabilities and customized features. Furthermore, the system is required to perform user profiling and to adapt the existing parameters of the system to the user needs and interest. On the other hand, it is also important to consider the devices and applications in which this technology is going to be deployed. Mobile media is a high growth area but the state-of-the-art technologies are lagging behind the consumers expectations. The presentations in this session are addressing precisely these three important aspects.

Organizers:

Nicu Sebe, University of Amsterdam (The Netherlands)
Qi Tian, University of Texas at San Antonio (USA)

Papers:

1. Personalized Multimedia Retrieval: The New Trend?
Nicu Sebe, Qi Tian

2. Browsing visual collections using graphs
Marcel Worring, Ork de Rooij and Ton van Rijn

3. Personalized Retrieval of Sports Video
Yifan Zhang, Xiaoyu Zhang, Changsheng Xu, Hanqing Lu

4. Evaluating the Implicit Feedback Models for Adaptive Video Retrieval
Frank Hopfgartner, Joemon Jose