Skip to content

Research

interests

professional directions

- I want to improve the way people search for video content.  Right now, there’s a gap between just looking for pictures with captions or tags and really searching video for its contents and getting what you want.  There’s a bigger gap in doing this interactively -- i.e. not a page-by-page traversal to find the things you want.  I hope to fill the gap in a couple of ways: provide cross-domain, state-of-the-art automatic detectors and push automatic search methods into an intuitive interactive environment.

recreational endeavors

- I believe that machine learning and some form of pattern recognition or statistical analysis and classification will drive my future free endeavors. In particular, I am interested in an automated system detecting some degree of salience or a general pattern through digital content (audio, video, or image) or loose textual description. I have participated in research endeavors in both multimedia and natural language processing and am always excited to discover and implement new approaches to non-traditional problems. My non-academic interests include hiking, skiing, running, photography, good entertainment, and always exploring new cities and countries -- and now I forsee scubadiving in my future as well. Don’t hesitate to contact me if you have a question or interesting discussion.

research projects

While projects in the lab seldom make it outside of a paper, I can point you to the search systems that I've helped to create...

  • CuZero: Setting the stage for my thesis work, CuZero is the culmination of prior interactive search work.  CuZero attempts to solve two problems: difficulty in formulating a query and difficulty in browsing many queries at a time.  The first problem is resolved by issuing joint human-computer queries that guides a user to the best concepts or query parameters for their search.  The second problem involves a new image browsing strategy that simultaneously allows a user to view many different query permutations by constructing a visual grid for fast re-weighting and inspection.  Following in the steps of CuVid, CuZero logs user interaction and has the potential to further automate the process of search and inspection of new images in a large database.
  • Visual Islands: A work called Visual Islands with Cheng-Chih Yang, a recent masters graduate at Columbia, involved changing the way we inspect results and was graciously given the best oral paper award at CIVR 2008.  Our assertion was that with the incorporation of semantic or other meta-data, we could offer the user an alternative layout for images that would allow faster comprehension and identification of related images.  We also proposed a method that allowed non-linear browsing of results by finding relevance images and hopping to that page of results directly.
  • CuVid (Semantics): A joint work with Lyndon Kennedy updated CuVid (with semantics) and involved the full exposure to our concept library in different ways.  We incorporated several semantic search options (i.e. lexical similarity, etc.), options to find related items by dominant concepts (published as a joint work with AT&T above), and options to recommend filters using these dominant concepts.  These filters often work best to reduce the amount of results you inspect, so they really only add specificity.  For example, starting a search with "soccer" can be further reduced with the concepts "field", "outdoor", "fan", or "athlete".  This filtering can be performed individually, or together as a single process.  This paper was later incorporated into a book called "Semantic Computing" which is currently in press (as of Fall 2008).
  • CuVid: My initial work in the lab was CuVid which first incorporated multi-modal search in what has become a standard fashion.  The neat thing is that behind the scenes, I record user interaction and (used to) offer the option to dynamically rerank your results with our published CueX reranking algorithm.

publications

Copyright Notice: This material is presented to ensure timely dissemination of scholarly and technical work. Copyright and all rights therein are retained by authors or by other copyright holders. All persons copying this information are expected to adhere to the terms and constraints invoked by each author's copyright. In most cases, these works may not be reposted without the explicit permission of the copyright holder.

  1. David Gibbon, Zhu Liu, Eric Zavesky, DeDe Paul, Deborah Swayne, Rittwik Jana, Behzad Shahraray, “Combining Content Analysis of Television Programs with Audience Measurement.” In 9th Annual IEEE Consumer Communications and Networking Conference, January 2012.
  2. Zhu Liu, Eric Zavesky, Ning Zhou, Behzad Shahraray. “AT&T Research at TRECVID 2011”. In NIST TRECVID Workshop, Gaithersburg, MD, December 2011.
  3. David Gibbon, Andrea Basso, Lee Begeja, Zhu Liu, Bernard Renger, Behzad Shahraray, Eric Zavesky, “TV Content Analysis for Multiscreen Interactive Browsing and Retrieval.” In TV Content Analysis, CRC Press, March 08, 2012.
  4. Eric Zavesky, "LipActs: Efficient Representations for Visual Speakers." IEEE International Conference on Multimedia and Expo, July 2011.
  5. Mattia Broilo, Eric Zavesky, Andrea Basso, Francesco G. B. De Natale, "Unsupervised Event Segmentation of News Content with Multimodal Cues". AIEMPro’10 at ACM Multimedia,  2010.
  6. Zhu Liu, Eric Zavesky, Behzad Shahraray, Neela Sawant, "AT&T RESEARCH AT TRECVID 2010". 2010.
  7. E. Zavesky, S.-F Chang. "CuZero: Embracing the Frontier of Interactive Visual Search for Informed Users." In ACM Multimedia Information Retrieval, Vancouver, British Columbia, Canada, October 2008. details
  8. Zhu Liu, Eric Zavesky, Behzad Shahraray, David Gibbon, and Andrea Basso. "Brief and High-Interest Video Summary Generation: Evaluating the AT&T Labs Rushes Summarizations." In ACM Multimedia Information Retrieval, Vancouver, British Columbia, Canada, October 2008.
  9. W. Jiang, E. Zavesky, S.-F. Chang, A. Loui. "Cross-Domain Learning Methods for High-Level Visual Concept Classification." In IEEE International Conference on Image Processing, San Diego, California, U.S.A, October 2008. details
  10. Zhu Liu, Eric Zavesky, Behzad Shahraray, David Gibbon, Andrea Basso. "Searching Videos in Visual Semantic Spaces" in Semantic Computing, Phillip Sheu, Heather Yu, C. V. Ramamoorthy, Arvind K. Joshi, and Lotfi A. Zadeh (Eds.), IEEE Press/Wiley, 2008.
  11. E. Zavesky, S.-F. Chang, C.-C. Yang. "Visual Islands: Intuitive Browsing of Visual Search Results." In ACM International Conference on Image and Video Retrieval, Pages 617-626, Niagara Falls, Canada, July 2008. details
  12. Shih-Fu Chang, Wei Jiang, Akira Yanagawa, Eric Zavesky. "Columbia University TRECVID2007 High-Level Feature Extraction." In NIST TRECVID Workshop, Gaithersburg, MD, November 2007. details
  13. Eric Zavesky, Zhu Liu, David Gibbon, Behzad Shahraray. "Searching Visual Semantic Spaces with Concept Filters." In IEEE International Conference on Semantic Computing, Irvine, California, September 2007. details
  14. Zhu Liu, David Gibbon, Eric Zavesky, Behzad Shahraray, Patrick Haffner. "A Fast, Comprehensive Shot Boundary Determination System." In IEEE International Conference on Multimedia and Expo, Pages 1487-1490, Beijing, China, July 2007. details
  15. Shih-Fu Chang, Lyndon Kennedy, Eric Zavesky. "Columbia University's Semantic Video Search Engine." In ACM International Conference on Image and Video Retrieval, Amsterdam, Netherlands, July 2007. details
  16. Shih-Fu Chang, Winston, Wei Jiang, Lyndon Kennedy, Dong Xu, Akira Yanagawa, Eric Zavesky. "Columbia University TRECVID-2006 Video Search and High-Level Feature Extraction." In NIST TRECVID Workshop, Gaithersburg, MD, November 2006. details
  17. Shih-Fu Chang, Winston Hsu, Lyndon Kennedy, Lexing Xie, Akira Yanagawa, Eric Zavesky, Dongqing Zhang. "Columbia University TRECVID-2005 Video Search and High-Level Feature Extraction." In NIST TRECVID workshop, Gaithersburg, MD, November 2005. details