Face Detection

In this project we are developing computer methods which automatically locate human faces in still photographs and video. Our goal is to develop algorithms that are accurate and computationally efficient. A secondary goal is to apply these techniques to detections of other objects such as cars, text, cells in biomedical images, and everyday objects in an office setting such as staplers, pens, and keyboards.

Our approach is to use statistical modeling to capture the variation in facial appearance. Currently, we use a set of models that each describe the statistical behavior of a group of wavelet coefficients.

Our face detector is widely considered to to the most accurate for frontal face detection and the only known method that works reliably for non-frontal faces images. See face detection demo for an on-line demonstration of this algorithm. In addition, we have developed the only known method for finding passenger cars in photographic images.

In September 2004, several Carnegie Mellon University researchers launched Pittsburgh Pattern Recognition, Inc. to commercialize this technology. The company’s founders include Dr. Henry Schneiderman, Professor Robert Lowe (Tepper School of Business), and Dr. Michael Nechyba (Robotics Institute, Ph.D).

Steve Milborrow at the University of Cape Town used this face detector in his Master’s project.