COMP 790-096: Computer Vision and the Web
Fall 2007, Tuesdays 3:30-4:30, SN 115
Instructor: Svetlana Lazebnik (lazebnik -at- cs.unc.edu)
Quick links:
presentation schedule,
reading list
Introduction
Over the last few years, we have seen an explosion in the sheer amount of
image and video data available to us over the Internet. The number of images
indexed by Google and Yahoo is growing exponentially, and has currently reached
several billion. But the revolution is not only technological, it is also
cultural, giving rise to the phenomenon popularly referred to as Web 2.0.
Part of this phenomenon is the emergence of digital communities like Flickr and
YouTube that enable users to upload, tag, and share images and videos with
millions of other users.
The wealth of images on the Internet is beginning to revolutionize computer
vision. Researchers in object recognition are already taking advantage of the
Internet for dataset collection and automatic discovery of object categories.
But apart from simply using the Internet as a source of data, computer vision
research can play a significant role in helping people to navigate the chaotic
sea of visual information. More and more exciting Web-related research ideas
are showing up in the latest vision and graphics literature. These ideas include
organizing large image collections based on semantics or 3D geometry, using the
content of these collections to synthesize new pictures through computational
photography techniques such as image completion, or enabling users to interact
with photos in novel ways, such as creating 3D-popups from their vacation
snapshots. The purpose of the course is to get acquainted with these directions
and to speculate about promising avenues for future research and, more generally,
the future role of computer vision in the Web 2.0 revolution (and beyond).
This course is set up for variable credit. The basic format (for one unit) will
consist of a weekly paper reading group. Discussions will emphasize the big
picture and conceptual issues, so there are no formal technical prerequisites.
Anyone with an ability to understand (at a high level) papers from recent computer
vision and graphics conferences can benefit from the course. Interested students
can choose to do additional readings, a report, or a project for higher credit.
Schedule
| Date
| Topic
| Presenter
|
| August 21
| Show and tell: PowerPoint, links in html format
| Lana Lazebnik
|
| August 28
| Photo Tourism / PhotoSynth
| Brian Clipp
|
| September 4
| Automatic Photo Pop-up
| David Gallup
|
| September 11
| Photo Clip Art,
Image Completion
| Brian Eastwood
|
| September 18
| AutoCollage,
Content-Aware Image Resizing
| Stephen Guy
|
| September 25
| Photo Quality Assessment
| Zhimin Ren
|
| October 2
| Names and Faces in the News,
Animals on the Web
| Josh Markwordt
|
| October 9
| Visual Category Filter,
Learning from Google's Image Search
| Peter Lincoln
|
| October 16
| ICCV 2007 (week off)
|
|
| October 23
| ICCV Recap (part I)
| Lana Lazebnik
|
| October 30
| ICCV Recap (part II)
| Lana Lazebnik
|
| November 6
| LabelMe,
Tiny Images
| Miranda Steed
|
| November 13
| ESP Game,
Peekaboom
| Ryan Schubert
|
| November 20
| Show and Tell
| Everybody
|
| November 27
| Project presentations
| Nick, Sashi, Marc
|
| December 4
| Project presentations
| Xiaowei, Ram, Jake, Rahul
|
| December 11
| Final project reports due by the end of the day
|
|
Reading List*
*Starred papers are not on the presentation list. But feel free to read them for your own enlightenment or to incorporate them into your presentations if they are closely related to your main topic.
From 2D to 3D
Computational photography
-
Image completion using millions of photographs.
James Hays and Alexei Efros.
ACM Transactions on Graphics (SIGGRAPH Proceedings), 2007.
-
Photo Clip Art.
Jean-Francois Lalonde, Derek Hoiem, Alexei Efros, Carsten Rother, John Winn, Antonio Criminisi.
ACM Transactions on Graphics (SIGGRAPH Proceedings), 2007.
-
AutoCollage.
Carsten Rother, Lucas Bordeaux, Youssef Hamadi, Andrew Blake.
ACM Transactions on Graphics (SIGGRAPH Proceedings), 2006.
-
Seam Carving for Content-Aware Image Resizing.
Shai Avidan and Ariel Shamir.
ACM Transactions on Graphics (SIGGRAPH Proceedings), 2007.
Pictures and words
-
*Matching words and pictures.
Kobus Barnard, Pinar Duygulu, Nando de Freitas, David Forsyth, David Blei, and Michael I. Jordan.
Journal of Machine Learning Research, Vol 3, pp 1107-1135, 2003.
-
*Mutual information of words and pictures.
Kobus Barnard and Keiji Yanai.
Information Theory and Applications Inaugural Workshop, 2006.
-
Names and Faces in the News.
Tamara L. Berg, Alexander C. Berg, Jaety Edwards, Michael Maire, Ryan White, Yee Whye Teh, Erik Learned-Miller, David A. Forsyth.
Computer Vision and Pattern Recognition (CVPR) 2004.
-
Animals on the Web.
Tamara L. Berg, David A. Forsyth.
Computer Vision and Pattern Recognition (CVPR) 2006.
Learning visual models from Google image search
Dataset collection
Learning basic image properties
Indexing and retrieval in large datasets
|
