COMS 6998 Computational Photography
Assignment 1. Simple Applications for Face Detection
Due: March 4th, 2011
In this assignment, we will implement some simple computational photography
applications using a publically available face detector. Here
you can find the face detector in Matlab. It can run in Windows,
Linux, and on a Mac. For a given input image, it will detect faces;
each detected face is represented as a bounding box.
Application 1. Automatic Rotation of Photographs (5%)
When we take photographs, we often rotate the camera in order to
get a good field of view or frame. When we dump the photographs to
computers, we need to manually rotate them to the correct orientation.
The idea of this application is to develop an automatic method
for rotating photographs, assuming there are faces in the
photographs. In other words, if the rotation of the photograph is
correct, we will detect faces; if the rotation is wrong, we will
detect fewer or no faces. In this assignment, we only consider 4
possible rotations: 0, 90, 180, 270 degrees.
Here you can download a set of
200 photographs with faces.
Application 2. Portraits or Group Shots (5%)
Once we correct for the orientation of the photographs, we can also
use the face detector to do a simple classification. Use the same
dataset as above to classify the photographs as either
portrait (i.e., one face) or group shot (i.e., multiple faces).
Application 3. Simple Face Tracking in Videos (5%)
This is another simple application of the face detector, but here we consider
faces in videos.
Suppose we have a video clip, in which a person's face is moving.
Our goal is to track the face over the video. Here are some video clips.
- Using the face detector, find the face in each frame of the video.
- For some frames, if the face detector fails to find the
face, use interpolation to predict the
location and the size of the face for these frames. In other
words, you should use the detection results from past and future frames
in which the faces are detected
to predict the face in the current frame.
- One way to show your tracking result is to crop the
detected face from each frame and put it at the top left
corner. You can also plot the face trajectory
over time.
How to Submit Your Assignment?
You should create a webpage to show your results and
explain what you did. Do not modify the webpage after the due
date. Send an email to Oliver Cossairt <olivercossairt@gmail.com>
with the URL to your webpage - before the due date.