Project Ideas

Daily weather data are collected at ground stations and sampled to a 1 sq. km. for the entire United States. The data are made available in a scientific format called NetCDF, which is essentially a hierarchical format capable of storing an arbitrary number of nested arrays. While suitable for storage and targeted analysis, NetCDF is not easily amenable to spatial query and visualization.

This is a proposal to convert the NetCDF format to a spatial database format, and then write routines and build the interface that would allow spatial selection of weather data and display them on a map as geo-referenced image.

The project would focus on creating a web-based working prototype for a specific part of the country. The prototype would be scalable to the entire country.

Plus: I already have the data; the spatial routines are pretty well established; the scientific contribution factor is high.

Minus: There is not much “new” in this proposal even though the innovation quotient in terms of improving a process is pretty high.

(updated 4/13/2010)

There are merely 5 weeks left in the semester (the downside of having break left). And I need to figure out how to have a project for a pretty diverse set of students.

There are 3 different kinds of projects. If there were more time, we would do all three:

  1. Design – take a particular data set, and create a visualization of it. Doing this involves understanding the design choices in making an effective visualization. It may require a non-trivial amount of “data scraping” to gather the necessary data and get it into the form you need to make the visualization. Note: if you choose a design focused project, you will need to provide some discussion to let us know that there is enough “real work” involved.
  2. Implementation / Methodology – create a tool for visualizing a type of data, preferably by exploring the use of a particular method (or developing a new method). Such a project might try to develop a new technical method, or integrate a known method in a new way.
  3. Survey – read several papers about a topic, to create a summary of what is already known. A good survey doesn’t just summarize the papers: it also provides some value in describing their relationship (for example, how to choose amongst the methods surveyed, or describing how a set of ideas might be applied).

Note: that any project is likely to be a mix of all 3. Even if you are doing a design, you’ll want to look at the existing literature (survey), and there might be some implementation involved (for example to make an interactive tool for exploring the data set).

For this class, your project must have some aspect of all 3. You can choose one or two areas to focus on (for example, you might focus on design, in which case we’ll expect a lot of work in the design area, but at least some survey (reading) and implementation.

You may choose

For your project, you can pick (if you don’t pick, we’ll pick for you):

  • The topic. You might center your project around a particular data set you are interested in, a particular technique that you want to read a lot about, an idea for a method that you might want to explore.
  • The focus. You might focus on design, implementation/methodolgy, or survey. Note: that you still must do some of the other 2.

For any project you must:

  • Do enough reading. For any project, you must read at least 5 things and write summaries of them. Based on your topic, I will give you 1 or 2 to start, but its your job to find more. You should make a list of things that you should have read (if you had more time) as well, and pick the most important things from that list. If your project is survey focused, you should read at least 10 things, and have your writeup “add value” to the combined summary (as described above under Survey).
  • Create a visualization of a particular data set. If your project is implementation, you should show it off on some example data. If your project is survey, you should create some visualization that illustrates what you’ve learned. (in these cases, the visualization might be a static picture). If your project is design focused, you need to create some well thought out visualization that actually solves the problems of some real data set, and have your writeup discuss the demands of the problems, how your solution addresses it, as well as some discussion of the design space (what alternatives did you consider that were not as good, compare your choices with other things you could have done, …)
  • Create a writeup: your writeup must be in the format of a Eurovis poster paper IEEE Visualization poster paper and 4 pages, not 2 pages. The writeup must be submitted as PDF document. The format is described here, including Word and LaTeX templates. Note: you do not need to provide the “Descriptors” or “Keywords.”
  • Give a final presentation: the exact format of this is to be determined. But you might be expected to give a 10-15 minute presentation to the class, or we might have a poster session, or …
  • Submit all artifacts (programs, paper summaries, design documents, …)
  • Submit regular progress reports. Your grade grade will be determined not only by the final product, but also the documentation you create along the way (like submitting plans and progress reports along the way).

Some deadlines:

  • Wednesday, April 7th, noon: email the instructor and the TA with your topic, and some idea of what you are going to do. The unusual (Wednesday) deadline will give us some time to think about your project so we can help you develop a plan, and hopefully suggest some things to start reading. If you haven’t picked something specific, send us some rough ideas so we can help you pick something. If you don’t pick something by this deadline, we will pick something for you. If you are planning to work with a partner, please indicate that in the email.
  • Thursday, April 8th, class: Project kickoff. We’ll use the class to have a discussion of project topics. We will either have some 1-on-1 discussions, or schedule some for each project. In this time slot, we’ll hopefully transform rough ideas submitted  by email into more fleshed out project ideas.
  • Friday, April 16th: Wednesday, April 14th:Project plan. You must send a project plan (by email to the instructor and TA) you must post a project plan to the course web. Details are posted here. You may be asked to revise the plan based on our feedback of the initial version.
  • Friday, April 23rd: Progress report: You must email the TA and instructor with a description of what you’ve done so far. You must send at least 2 paper summaries, and hopefully some initial results.
  • Friday, April 30th: Progress report: This progress report should include a list of the papers (but not the actual summaries) you’ve read, a description of what you have done so far, some initial images, and a description of what you expect to have at the final deadline.
  • Thursday, May 6th: 11am (e.g. before the last class): everything due (final writeups, all “artifacts”, etc). Extensions until the exam week are available by specific request.

The idea is that you’ll spend the first week trying to define a project idea, and developing a plan. We will work with you to do this.

Some ideas or sources of ideas:

  1. I have created a blog category called “project ideas” that I will post project ideas. Some of the things listed will be there.
  2. The Design for America challenge is a great source of ideas.
  3. The Visual Analytics Benchmark Repository has a number of data sets with example problems to try. Some of these were given as content challenge problems.
  4. You could try your hand at the 2010 VAST challenge. We could even consider assembling a team to compete in the contest.
  5. You can look at the projects done in the Harvard CS171 class for inspiration.
  6. The Visualization conference posters are a good source of ideas for small projects (since many of them are student/class projects). I will make the poster papers available to the class.
  7. You can pick an interesting visualization tool and try to learn about it and give an example of how its used. For example, there is a public version of Tableau. Or you might try building something with a toolkit like protovis, flare, or prefuse.
  8. You can pick a topic we aren’t going to cover enough in class: you might learn about (and implement) some graph layout algorithms, learn about the variety of approaches for doing focus+context interactions, …
  9. You could explore the depiction of uncertainty and error in visualization, and try to develop ways to show uncertainty (statistical properties) in different kinds of visualizations. (e.g.: error bars are easy to add to a bar chart, but how do you use it to show the variance on a graph or a volume visualization).
  10. You could try to find ways to use optimization to choose sets of colors that are perceptually distinct, but also follow proper “color harmonies” (these can actually be defined mathematically).
  11. You could define and run some perceptual experiments to understand how effective different methods are.
  12. You could take some hard data set that we have access to. We have long time-series of accelerometers and foot switches tied to walking, molecular shape data, try to display large-scale protein sequence matching computations, …
  13. Personal multi-media collections are a particularly ripe topic for visualization, requiring a mix of design and implementation methodology. How can I keep track of my large collection of pictures? How can I figure out which of the pictures on my disk I’ve already uploaded to Flickr? How can I keep track that there are multiple versions (one picture is a cropped version of another, or is just a different file format)? You can try to take on the multi-media problem of figuring out the equivalences, or just assume that you get the answers from some place.
  14. You could think about/develop collaborative tools (either for online communities, or small groups to work together) for other vis problems (even if the problems are well-solved for a single person).
  15. You could flip through the Vis proceedings and find papers that describe interesting techniques that you might want to try. (often the techniques in papers are much easier to reproduce than to develop in the first place)

The Design for America contest is challenging the community to come up with information visualizations of some available government data.

The first 3 challenges (in the “Data Visualization” category) might make good final projects. (info on that coming soon)