Here I some computer vision papers I found interesting at NIPS 2009 in Vancouver.
[pdf][bib] Unsupervised Detection of Regions of Interest Using Iterative Link Analysis (NIPS 2009)
Gunhee Kim, Antonio Torralba
[pdf][bib] Region-based Segmentation and Object Detection (NIPS 2009)
Stephen Gould, Tianshi Gao, Daphne Koller
[pdf][bib] Segmenting Scenes by Matching Image Composites (NIPS 2009)
Bryan Russell, Alyosha Efros, Josef Sivic, Bill Freeman, Andrew Zisserman
[pdf][bib] Beyond Categories: The Visual Memex Model for Reasoning About Object Relationships (NIPS 2009)
Tomasz Malisiewicz, Alyosha Efros
Great blog Tomasz! I'm wondering if you have any comments about what makes a computer vision paper suitable for NIPS. Does it have to be biologically motivated, contain machine learning, both, or something else altogether? Is there a high representation of computer vision papers/people at NIPS?ReplyDelete
Regarding computer vision at NIPS, I can only provide you with my own insights and not a detailed answer. First of all, regarding the problem of object recognition it is hard to find an approach that doesn't use machine learning. To be quite honest, my own notion of what is machine learning and what isn't has grown soft over the years. I think any approach which involves a good amount of optimization is of interest to the NIPS community. Sometimes I can't draw the line between "optimization" and "machine learning." There are still a good amount of vision researchers at NIPS, attending as well as reviewing. For example, Antonio Torralba is an area chair and he's definitely a hardcore vision person. I think it is very possible to publish a vision paper at NIPS (because it was a good paper and the vision reviewers liked it), but that doesn't mean that every person at NIPS will understand and/or be influenced by your approach.ReplyDelete
If you end up doing something non-standard such as set up your own problem and present your own algorithm then people at NIPS might be be as excited as the CVPR community. If you end up advancing the state of the art on some standard dataset with your approach then it is easier to get the NIPS crowd excited because they'll be able to jump into your problem by downloading the dataset/features used.
While a majority of computer vision papers will be of interest to some people at NIPS, there are some papers that would easily get into CVPR but the NIPS crowd will find un-interesting. NIPS likes math, and I think a paper without mathematics is not going to get much interest at NIPS.
In conclusion, there are still many good vision researchers attending NIPS and a good vision paper will definitely get noticed at NIPS. However, when talking about your work during a poster presentation you might have to go back a few steps and explain some of your assumptions to a less vision-inclined researcher/student.
I have a bunch of student friends studying machine learning and I think NIPS would be a great target venue for a collaboration between myself (from the vision side) and them (from the learning side). That way we could both represent the two sides of the collaboration: why it is important from the vision-side of things and why it is important from the learning side.
Your idea is very excellent.I am a phD student in China, and I am also interesting in ML. If i could talk with using the e-mail or skype ?ReplyDelete
If you have any serious inquiries, you can email me.ReplyDelete