Term Paper - Due Friday, April 16th.


As stated in the Syllabus, one requirement of this course is to write a concise summary of your findings regarding some real-world computer vision task. The motivation for the assignment is two fold. First is to give you practice writing a professional survey paper. Second, it will provide you an opportunity to dive relatively deeply into a specific topic having to do with the automatic interpreation of images. You are intentionally being given very little additional focus in terms of specific topics, since part of the excercise is to think through what you find more and less interesting. If you are looking for something simple to jog your curiosity about possible topics, consider the lists of topics commonly included in 'call for papers' associated with major computer visoin conferences. Just for example, here is the call for papers for CVPR 2010.

My goal is to let students write about any topic they choose, but at the same time, I wish to avoid having the entire class pick one topic. So, formally, you need to email me your tentative topic choice and get my approval, and if I seeone topic being seriously overworked, I will try to assist in suggesting alternatives.

Some Specifics

Here are some things to strive for in your paper. First, start off by clearly stating the commonly agreed upon definition of the problem you have chosen to study. Also notice that it is assumed that your going to organize your research around a problem, rather than for example, a technique. Second, work to organize approaches to solving the problem into categories. Survey papers you find in the literature can be extremely helpful for suggesting taxonomies. Finally, this is a short paper, so don't get over extended. Just as it would be a mistake to write about only two published papers, if you are starting to try to make sense of more than ten, you are probably over reaching.

Here are some things to please avoid. Many researchers early in their careers mistake a list with an essay. In other words, avoid falling into the style where you simply list summaries of papers in sequence without regard to connections, common aspects, distinctions, etc. It is fine for your working notes to be of the form: "Bill said apples are red. Ann said apples are green. Jim said banannas are typically yellow. etc.". But by the time you have a paper ready for someone else to read, it has to come up a level of abtraction and be more of the form" "Fruit that grows on trees comes in a variety of colors, as observed by Bill, Ann and Jim. Specific differences involve trees, in the case of Apples, versus overgrown grass stalks in the case of bananas."


Your paper should run between 10 and 12 pages double spaced. Please obtain my permission before exceeding this length guideline. You must have a bibliography, and if you are unsure about format, ask.

You will also be expected to present a short, roughly ten minute, summary of your paper in class. This presentation, while short, should be professional with suitable graphics (read as PowerPoint or equivalent).

You will submit your papers electronically through email. Keep in mind that I typically convert your papers to audio and listen to them before reading them. In general, this should not alter your approach. That said, try to avoid writing in a mathematically dense style unless you discuss it first with me. I might also, in some cases, come back to get your assistance in converting your paper to plain text. Finally, if you want a copy of the audio version, feel free to ask.

Ross 3/12/2010