A typical class project might be implementing and evaluating an algorithm from a research paper. The choice of projects can be very open-ended; ideally you can incorporate their own research. But your class project must be about new things you have done this semester; you can't use results you have developed in previous semesters. People can work in groups of size 2 to 3, though we encourage students to work in larger groups because the projects tend to be more substantial.


Project Proposal: 10% of project grade (due by 11:59 pm on proposal due date listed on schedule)
The proposal should be a PDF that includes a title, your name(s), roughly 2 paragraphs describing the project idea, and a list of 1-3 relevant papers. If you will be collecting your own images, describe any special experimental procedure (restriction of illumination conditions, the use of projectors, etc.). If we do not have the facilities here, you may be able to email authors of a paper for their collected data. If you will be using a standard image dataset, include a pointer to it.

Poster Presentation: 30% of project grade (due on date of your team's presentation on schedule)
The presentation day schedule will be released during the week before the presentation. You will be required to come during your poster session and encouraged to attend to see your fellow students' projects on the other day. See Piazza for note with list of guidelines.

Final Writeup: 60% of project grade (due by 11:59pm on writeup due-date listed in schedule)
You must prepare a writeup using the given CVPR format. There is a strict page limit of 8 pages including references. A single writeup is required per project. For groups with more than a single person, the writeup should include a division of labor between each person. You should include some setup for your problem (motivation, related work), as well as the work you did (methods, results). Focus on what you found did and did not work as well as discussion of why. We're looking to see what you have accomplished during this project, what you learned about vision and your application, and evidence of the effort you put into the project.

Poster Guidelines

Posters should fit within 40 x 30 inches (landscape preferred).

Be sure to use high resolution / vector graphics (the poster will be much larger than the size of your computer screen).

Choose large enough fonts. You can print subsections of your poster out on standard computer paper to get an idea of how large your font is after printing.

White space is your friend! Your goal is to convey information in a way that someone else can digest, not just throw results onto a slide. Less is usually better, and margins help the viewer organize and process your poster.

A good rule of thumb for poster content is: 50% Images/graphics; 30% text; 20% whitespace. Full sentences should be rare; paragraphs should be extremely rare.

Use only 2-3 fonts, and use them consistently (one for headings, one for text, one for captions). Keep formatting consistent between sections.

Use only 2-3 main colors, and keep them consistent. There are many websites that will help you choose color schemes that look good (a quick Google search will get you there). Red on blue never looks good; there are many other rules of thumb for colors you can avoid by picking a standard color scheme.

Practice explaining your poster to someone. Reformat and add images where you find yourself struggling to explain.

These are examples of generally good posters (each has their own issues; many do not use enough whitespace). They may be a good place to start: Example 1. Example 2. Example with guidelines.


You are strongly encouraged to find a project aligned with their own research interests. Here are some pointers to interesting papers in the context of the material we have talked about.

Research papers from CVPR.

Research papers from ICCV.

Research papers from ECCV.

Research papers from IJCV.

Previous class from UCI.

Similar class from UT Austin, CMU, MIT, UNC, NYU, UIUC.

List of datasets.

List project ideas by Serge Belongie.

Computer Vision Homepage with links to software and databases.