CS 595I

Graduate Research Seminar on Computer Vision

Winter Quarter 2006

Instructor: Matthew Turk

Day and Time: Monday 4:30-6:00pm

Location: CS Conference Room (Engr. I, 2114)

Enrollment Code: 75457

Each week we will discuss a current topic in computer vision, focusing on a recent journal article (and perhaps supporting material). Everyone is responsible for reading the main paper and coming prepared to discuss it. The job of the presenters for a given week is to present the paper by:

• Clearly describing the problem at hand

• Explaining why it's interesting and relevant (why does it matter)

• Briefly discussing the historical context (what had been done before in the area)

• Presenting the research and its results

• Critiquing the paper (positives, negatives)

(In fact, I would suggest structuring the presentation in this way, with five main sections: The Problem, Relevance, Background, The Paper, and Critique.)

This may involve reading other papers and finding other background material to help understand the work. The presenters should post helpful information to the web site (if any) in advance so others can refer to it.

Further comments on paper presentations:

Don't think of this as merely "presenting a paper." Rather, think of this as "presenting your understanding of the main problem, approaches, and ideas that the paper covers." It's not necessary to cover every detail in the paper, and the presentation should not just regurgitate the paper (first they did this, then they did this, then....). Refer back to the bullet points above, starting with "Clearly describing the problem at hand." Be sure to present the "forest" and not only the "trees."

Also, it's always a good idea to ask questions of the audience (e.g., "So why did they choose to do this rather than that?") to keep people engaged.

The mailing list for the seminar (called CV595) can be found at http://lists.cs.ucsb.edu/mailman/listinfo/cv595.

Announcements

• The Scharstein/Szeliski stereo paper is now in the group of papers

Schedule

Everyone should read the Primary paper for each week. The Secondary material is optional.

 

Date	Topic	Primary	Secondary	Presenters
Jan 9	Introduction			Matthew
Jan 16	No meeting (holiday)			None
Jan 23	Image-based rendering	Fitzgibbon	Gortler Adelson	Shane and Rogerio
Jan 30	Video segmentation	Wang	Li	Ya and Hang-Bong
Feb 6	Spacetime stereo	Davis	Zhang	Steve and Jason
Feb 13	Stereo matching methods	Sun-Zheng	Boykov Tappen	Longbin, Joriz, and Haiying
Feb 20	Superresolution	Shechtman	Ben-Ezra Baker	Taehee, JieJun, and Brynjar
Feb 27	Activity analysis	Hongeng	Oliver	Justin, Chris, and Jae Sik
Mar 6	Generalized PCA	Vidal	PCA tutorial	Chang-Ming, Chao-I, and Ping-Hsun
Mar 13	Object detection	Viola1	Viola2	Sehyung, Mei-chen, and Qiang

Participants

Papers (PDF files)

Links of Interest