Slow down!
So I can catch up!
Apple just released a new version of iLife, including an update to iPhoto called Faces. It detects and tags faces in your pictures with the person’s name. You can see it here: http://www.apple.com/ilife/iphoto/#faces
What’s frustrating is this was my semester project for a computer vision course I took in spring of 2008. My solution automatically detected and tagged recognized faces and prompted the user to tag new/unrecognized faces. The program continuously improved on a database of known faces to increase the chance of a positive detection and recognition of known faces (from different angles, lighting, facial hair, etc). To accomplish this I used a combination of the OpenCV computer vision toolkit for face detection (using Viola-Jones / Haar feature cascades)and the Identix SDK for identification statistics and eye locations (as registration coordinates when comparing different sized faces) when comparing facial regions of interest.
While I came up with the idea on my own, I realize it isn’t original. I’m sure Apple and others have been working on an implementation of this before I even had the idea. The real killer is Apple’s Facebook plugin. When I started working on the idea the main application I was envisioned was automatic tagging of your Facebook photos. Anyway, below are some pictures from my class demonstration presentation.

- Face and eye detection in a group of people

- More face and eye detection

- Calculated eye coordinates on a detected face

- Using eye coordinates as registration points for comparing two faces

- Scaling one face and aligning eye coordinates for feature/structure comparison

- A frame from my flowchart animation

- Another frame from my flowchart animation