Hidden faces at the World Expo0

The failures and limitations of a certain approach or a technique can often lead to interesting outcomes. Face detection is a well studied machine vision problem but the solutions do not always return the expected results. So, what happens when a basic face detection algorithm is applied to buildings? Wonwei to find out.

With only few days left until the grand opening of the the 2010 World Expo in Shanghai, we recently spent time at the Expo site, using the opportunity to snap photos of the country pavilions being completed. Looking up at the pavilion structures we wondered what a machine or an algorithm would see. Back in the lab, we applied a standard Haar classifier to the unprocessed images straight out of the camera, using a frontal face model. The detected ‘faces’ and the path taken by the classifier are visualized in the following images.

No meaning, no concept, no expected results. Just finding out.

Made with openFrameworks.