Introduction

Since there are a large variety of fascinating problems in computer vision research, you may want to have a more profound understanding of the ones that you are extremely interested in. The course projects provide you opportunities to pursue your interests, rather than confining you at the course materials that are introduced in class.

It is a good idea for you to treat your course project as a small research project, since it is very likely that you will encounter some open research challenges during the project. Keep this in mind: ultimate solutions may NOT exist for these projects, although you may be aware of some existing work. It is you that pursue it in a deeper sense. You have to think over something new, i.e., something that are different from the existing work. Since we treat the course projects as research projects, we DONOT appreciate those projects that just repeat what people have done. You have to have something of your own.

During the last week of the quarter, we'll have a mini workshop, in which you will present your work to all of us. And we may select one to win the best project! In addition to your talk, you need to hand in your 30-page project report before your receive the credit.

Possible Course Projects

• Face Detection
You are given a single image frame. Can you detect all the faces inside the image? You may want to consider the frontal view faces first, then you can pursue profile-view faces, and partially occluded faces later on. You may study the following three papers first:
[1] Henry Rowley, Shumeet Baluja and Takeo Kanade, "Neural Network-based Face Detection", IEEE trans. on PAMI, Jan. 1998
[2] Paul Viola and Michael Jones, "Rapid Object Detection using a Boosted Cascade of Simple Features", IEEE conf. on Computer Vision and Pattern Recognition (CVPR), 2001
[3] K. Sung and T. Poggio, "Example-based Learning for View-based Human Face Detection", IEEE T-PAMI, 1998
[4] Ming-Hsuan Yang, David Kriegman, and Narendra Ahuja, "Detecting Faces in Images: A Survey", IEEE Transactions on Pattern Analysis and Machine Intelligence (PAMI), vol. 24, no. 1, pp. 34-58, 2002

You may also want to visit http://vision.ai.uiuc.edu/mhyang/face-detection-survey.html
 

• Face Recognition
You are given a cropped picture of the face of your classmates. Can your program recognize who s/he is? I really want to see you show such a demo in class! You may study the following first:
[1] Matthew Turk and Alex Pentland, "Face Recognition Using Eigenfaces", CVPR'91
[2] Peter Belhumeur, Joao Hespanhua and David Kriegman, "Eigenfaces vs. Fisherfaces: Recognition Using Class Specific Linear Projection", IEEE T-PAMI, July, 1997
• Head Tracking
Can you program track a head no matter how and where it moves?
• Background Maintenance
Can you figure out a background model that enpowers efficinet foreground object segmentation? The idea is simple, but there are some unexpected challenges in practice, such as dynamic background, lighting changes, moving camera, etc. Read the following first:
[1] K. Toyama and et al. "Wallflower: Principles and Practices of Background Maintenance", ICCV'99
• Dominant Motion Detection
When you view a video sequence, you may find that there may exist a prominant moving object. We in general call its motion dominant motion, that roughly characterizes the motion features of the video. Dominant motion may refer to the motion of the largest target, or the camera motion itself.
• Video Event Detection
Video events are high level. They refer to a consisit concept associated with a video sequence. You may want to detect a high level video event from low-level visual features, such as motion, color, etc. This is an open problem. Let's see if you can find a good solution. You may want to read the following as an initial step:
[1] L. Zelnik-Manor and M. Irani, "Event-based Analysis of Video", CVPR'01.
• Video Texture
You are given a short video clip (say 2 sc), e.g., a flying flag. Can your program automatically generate a much longer similar sequence? (as long as you want).
[1] Arno Schodl et al. "Video Texture", ACM SIGGRAPH, 2000
[2] G. Doretto et al, "Dynamic Texture", IEEE conf. on Computer Vision, 2001
• Image Mosaic
You have a set of images taken from different positions of the lakefront of Chicago. Can you stitch them together to make a wide angle panorama?
• View Morphing
You have two images taken from two side views of the face of a person, can you generate the images from the videws in between? Read this:
[1] S. Seitz and C. Dyer, "View Morphing", SIGGRAPH'96
• Image Impainting
Suppose there is a picture you like very much, but it get some stains accidentally. Can you get rid of these stains? Read these first:
[1] A. Perez and K. Toyama, "Object Removal by Examplar-based Inpainting", CVPR'2003
[2] C. Ballester et al. "A Variational Model for Filling-in Gray Level and Color Images", ICCV'2001

Back to course home