CS365 Artificial Intelligence

Homework 2: Face Recognization

In this homework, you are required to learn openCV which is an open source library in C for computer vision. You are expected to read relevant portions from the openCV book by Bradski and Kaehler and implement the following functions on these two images:

Part I: Canny edge detector

Experiment on detecting edges using a Canny filter. Report the parameters that work best (are not too noisy and miss few relevant edges) for each image.

Part II: Face detection

Mark the faces in each image with a rectangular box. This is a standard function in Open CV. You will then need to compare your rectangles (for the second image) with a "groundtruth" image which will be uploaded shortly; the groundtruth will provide the best estimate for the face regions in terms of rectangles. You must match your rectangles with these and report your errors as a False Positive (FP) and a False Negative (FN) percentage.

Click on the image to get the large (12Mp) image which you need to work on.
Access the groundtruth image and data here : GroundTruth

Submission

The output should be available as "home.iitk.ac.in/~userid/365/hw2/".

Please create an index.html file showing your output images. A tarball of your source code should be linked from this file. You will most likely be asked to meet the TA and run your code on these or some other images.

Please keep all images / source code in the same directory and link them from your index.html file.

Due by Feb 1, Monday 00:00hrs.