Final Project
Please post your questions to our Google Group page.Summary of Deadlines:
* Mar. 4: form project group
* Mar. 6: final project group / discuss with Jian & Keqin your project ideas.
* Mar. 13: project proposal due
* Mar. 27: literature review due
* Apr. 8 - May 7: meeting with Jian once a week to get detailed comments on your project
* Apr. 8: project proposal presentation (including the literature / goals / contributions / expected results)
* Apr. 24: prototype demo, and plan for validating your results due
* May. 7: project demo (should have validated your results with a few people)
* May. 12: project paper due
What your project proposal should include:
A brief statement of the topic , including an assessment of its practical relevance and its potential contribution to the state of the art in interaction or visualization research.
A plan of work , including 3-4 milestones (proposal, literature review, preliminary implementation, verification (involving human subject), final implementation) and estimate of the time that will be required to meet each of the deadlines above. Clearly specify hat you intend to have done by the end point.
I encourage those of you who are engaged in an ongoing research project to define a topic what will complement your research efforts!
Possible types of final projects:
A thorough literature review of computer graphics research or its application areas, e.g., mesh morphology measurement, fluid flow simulations, motion data analysis; graph visualization in bioinformatics; review of semiology in computer graphics; Review of math visualizations. This could involve further reading on one of the topics touched upon in class. Generally, the review should include about 70 conference or journal papers. ACM Computer Survey Journal sets an example what we are looking for.
An implementation of three particularly interesting and useful algorithms that are big enough to be called a semester project, for example, hyperbolic tree map for bioinformatics data visualization; Non-photorealistic renderings of a brain diffusion tensor imaging; photorealistic rendering of a brain diffusion tensor imaging, photorealistic rendering of time-varying data; visualization of time-varying data. This type of study may or may not involve human subjects depending on the contributions and goals of your project.
An empirical evaluation of the effectiveness and efficacy (what other metrics you can think of, learnability?) of alternative graphics approaches for a particular application domain of use. You must choose a domain that you are familiar with or you can find an expert (e.g., faculty on campus or guest lecturers from our class) to assist you. One way is to pick several designs or algorithms from our class and show how we can use them to improve graphical design and why certain techniques outperforms the other for certain types of tasks of domain experts' interests.
A web-based toolbox that can be run on the web , that demonstrate an important graphics concepts in math, physics, or your domain of studies. For example, how good and bad choices of visual or graphical representations to reveal structures in some complex time-varying datasets, or how good and bad choices of these encodings to reveal 3D structure. The interface needs to be interactive. Similar to the previous type, you must recognize experts who can evaluate your design.
I encourage you to pick a topic above. If you feel you are running out of ideas, I will assist you to find a suitable topic, for example, either work on the data we have (some biology datasets) or work on the data you have to use graphics to help you answer your research questions.
Example topics:
Representing uncertainty in simulations - demonstrate some techniques that seems to be particularly well-suited for representing varying levels of confidence. Why are these techniques more successful than others that have been, or could have been tried?
Graph visualizations - graphs have broad applicability in biology and math. place graph algorithms in Bertin’s language or design a good graph visualization approach to fit your project needs.
Ranking visual encoding in 3D - we know a lot of 2D, but not 3D. We don’t know if the classical Bertin taxonomy can be expended to 3D and if so, if the MacKinlay and Cleveland orders are going to hold or not for a set of 3D tasks.
Lighting effects - Without lights, we will be living in the dark. Lighting in graphics generates shading effects. There are six different shadings, compare and study which one is most effective and how the effectiveness is affected by luminance.
AI algorithm visualization - pick a complex enough AI algorithm that is commonly difficult to be understood by students and visualize the algorithm to help understanding.
Interaction techniques - pick a non-conventional input device , e.g., kinect, and design a set of gestures to operate a series of windows on a large format display.
Tips for performing a literature review:
Your literature survey should include at least 10 relevant articles.
To perform your literature survey, first make sure you understand the scope of your chosen topic. Talk to Jian or Keqin if you need help getting started. Make a list of some terminology related to your topic that might serve as good google search terms. Once you're ready to search, you can either start with Google or start with more specific search engines that are more likely to return relevant results. These include the ACM digital library, IEEExplore (you have access on campus or via vpn), and Citeseer (all articles are free). Try several different versions of search strings, and search both titles and abstracts if possible. Browse the results looking for papers directly related to your topic. Once you have identified 2-3 "core" papers and read them, you can broaden your search in a couple of ways. You can look at the references of a core paper and follow the most relevant ones (searching backward in time). You can also search for other publications by the same author(s) on the same general topic. Finally, using some of the tools (esp. Citeseer) you can find out which papers have referenced your core paper and look at those (searching forward in time). After all this, you can do a Google search to uncover anything interesting (e.g. non-photorealistic rendering of brain imaging) that the other tools might have missed.
Once you have identified all the papers/sites that relate to your topic, read all the abstracts (at least). For those that are the most interesting or relevant, read the entire paper. Then organize your papers into categories – these categories will serve as the outline for the related work section of your final report. Finally, ask yourself, "What important research questions remain in this area that have not yet been adequately addressed?" Some of this will come from the future work sections of the papers you read, but you will also need to think deeply about the subject yourself. The bottom line is that you want to solve a problem that people care about your answers.
As you put together your annotated bibliography, you may format the references in any way as long as they are consistent. Include as much bibliographic information as you can (including page numbers, volume/issue numbers for journals, publishers, etc.) for each reference. You may reference URLs if that is the only option, but most of your references should be published articles.
At this time you may also want to draft the related work section of your final report. As you summarize the existing research, do not simply list papers or projects and describe their content. Rather, you need to provide a readable summary of the topic that shows how researchers have addressed the topic and how their approaches and results are similar or different. Bring out the relationships between different papers/projects. And show the limitations of the existing research (limitations that you will hopefully address). Support your arguments with as many citations as you can, but do not simply make the sections a list of citations. Here are two examples, one showing the style that I want, and one demonstrating a poor, "laundry list" style.
An example of good literature survey style (from a paper on comparisons of 3D displays): Note how the literature is divided into several categories, how the author makes several points of his own, and how he constructs an argument demonstrating the limitations of the existing research.
|
Many authors have noted the importance of studying the differences between displays and the effects of displays on users, applications, and tasks [e.g. 7, 16]. Few, however, have provided empirical evidence of these effects. One type of display comparison study found in the literature is a comparison of desktop and immersive displays for a particular task or application [e.g. 1, 11]. These studies attempt to demonstrate the effects of immersion, as opposed to the effects of a particular type of display. A second type of experiment compares the value of multiple VE displays for common tasks [e.g. 25, 28]. This is closer to the intent of our work, but is not explicitly focused on 3D interaction. A few studies have looked at the effects of particular display characteristics on interaction performance or usability. For example, Arthur [2] studied the effect of field of view in an HMD on performance in searching and walking tasks. The prior research most similar to ours involves studies that compare users' behavior and performance when interacting with VEs using different displays. Kjeldskov [15] reports an ambitious study on the usability of 40 common 3D interaction techniques in a semi-immersive curved display and a fully-immersive surround-screen display. He found qualitative differences in the usability of particular techniques between displays, but no quantitative data was collected. Our own prior work [3] did demonstrate a statistically significant difference in users' behavior between an HMD and a CAVE during a navigation task. |
An example of poor literature survey style (using the same references as the example above): Note how there is no organization to this writing (the paragraphs don't indicate different categories or themes), how the author doesn't analyze any of the literature, and how he simply lists the existing projects and papers.
|
There are several existing examples of display comparison experiments in the literature. Brooks [7] said that such experiments were important. One group compared a desktop display to a CAVE for an oil-drilling application [11]. Arthur [2] studied the effect of field of view in an HMD on performance in searching and walking tasks. Another study [25] looked at five different displays for construction-related tasks. Bowman et al. [3] looked at users' preferences for real and virtual turning in HMDs and CAVEs. A CAVE and a semi-immersive curved display were compared by Kjeldskov [15], and he used over 40 different 3D interaction techniques with the displays. Military applications on different displays were compared by Swan and his colleagues [28]. A comparison between a CAVE and a monitor has also been performed for a statistical analysis application [1]. A SIGGRAPH panel considered the relative advantages and disadvantages of HMDs and surround-screen displays [16]. |
Project proposal:
Project proposals should be simple and to the point. One to two pages of single-spaced, 10-point font should be sufficient; proposals should not exceed three pages. Format your paper using ACM SIGGRAPH / Trans. on Graphics format or IEEE Trans. on Computer Graphics and Visualization format.
Mandatory: talk to Jian between now and Mar. 8th about your project ideas. And she will assign some readings tightly coupled to your work to help get you started.
I also believe it is better to do a group project - a group of two people would be the best. Maximum 3 in a group is allowed, and your group is expected to produce more than those of 2.
The proposal should include, but is not limited to, the following:
* Title that clearly indicates the topic of the proposal
* Brief abstract (3-4 sentences) giving a summary of your proposal
* Names and contact information for all students working jointly on the project
* Introduction and motivation/rationale for the project (Why are you doing this project? Why is it important?)
* Research questions you will be addressing
* List of your preliminary hypotheses
* Hardware and software you will use to complete your implementation (displays, input devices, toolkits, languages, etc.)
* Tasks and metrics you might use to test your hypotheses
* Potential impact of the project (How might the results of your project extend the state-of-the-art in graphics or visualization, or enhance a faculty member's research, impact a particular application area, etc.?)