Homework 1 CS 803 , due February 15


Implement Eigenfaces for Recognition as described by M. Turk and A. Pentland in Journal of Cognitive Neuroscience, 3(1), 1991. Your code should consist of two main components.
  1. Build the eigenspace from the forty images given on the course web page (train). Determine a reasonably low-dimension to project onto by looking at the eigenvalues of the covariance matrix. Store the new low-dimensional representation of each of the 40 training images.

  2. Given an image, project this image onto the new low-dimensional space, and find the closest match from the training set constructed above. Test the recognition on the original forty training images as well as forty images labeled (test).
Submit hardcopies (paper copies) of your code, and email me a tar/zip file with all of your code. Also, submit a short report that describes the accuracy of the face recognition for both the training and testing stages. Homework is due on February 15, in class (no late submissions will be accepted).