Projects can be done individually on in pairs.

Deadlines

• September 9: Send me email stating whom you will work with (or that you will work on the project alone), and describing the project in one or two paragraphs. This email counts for 10 percent of your project grade.
• September 23: Send me email with your project plan, describing in detail what you plan to do and how. Your plan should be in Adobe PDF (preferred) or MS Word, and should include a problem description, a brief discussion of relevant literature (1-5 papers), a description of your deliverables, a schedule of work, a discussion of possible pitfalls and difficulties, and workarounds for them. Typical length of a project plan is 2-4 single-spaced pages. This plan counts for 30 percent of your project grade.
• November 6: Send me email with a draft of your project report. This can be a PDF or MS Word file, or a URL. In the latter case, make sure that all your links work. I cannot grade what I cannot access. A typical project report is 5-10 pages of single-spaced text and figures. Good form (including grammar and proper citations) counts as well. The project report draft counts for 30 percent of your project grade.
• November 25: Send me email with your final project report in the same format as your report draft (includng good form and grammar!). What counts is not whether the project was successful, but how much you show that you learned from it. The main differences between the draft and final version of the project report is that the latter may incorporate my feedback based on your draft. The final project report counts for 30 percent of your project grade.

Project Examples

The following project examples are intended to give you an idea of the scope of a typical project. You may choose to do one of the examples, but coming up with your own ideas is preferable. Variations on the project examples below can be obtained by replacing, say, "stereo" with "recognition," and so forth.

• Implement a face detection algorithm and report its performance on images of various sizes (see the resources page).
• Download existing code for the detection of pedestrians and discuss its performance on a few video sequences acquired in the vision lab, or downloaded from the web.
• Write code that recognizes handwritten digits, and test it on a subset of the MNIST database (see the resources page).
• Find a shortcoming of an existing recognition algorithm and attempt to fix it.
• Write a web crawler to mine images and labels from SnapMyLife (see the resources page), or from a similar web site. Organize the resulting images into directories for future work.

in all cases, the most important part of the final project report is a discussion of the results. Given your time constraints, you should not expect success with the more open-ended projects. However, do not be afraid to take risks: a good discussion of a failed attempt makes for a great project report.